Energy Efficient Lighting in Plant Factories: Addressing Utilance

Abstract

Vertical farming is considered to play a crucial role in future food supply. Until today, the high amount of electrical energy required for artificial lighting has been problematic in this context. Various possibilities for increasing efficiency through adapted lighting conditions have been and are being investigated. However, comparably little attention is paid to increasing utilance, i.e., the amount of photons that can effectively be used by the plant. In this work, a novel targeted lighting strategy is therefore proposed that allows for a dynamic adaptation of the luminaires’ light distribution to match the effective crop size at each stage of plant growth in a fully-automated manner. It is shown that the resulting utilance can significantly be increased compared to standard full-coverage lighting. Moreover, it is found that the proposed strategy is likely to consume less than half of the electrical energy usually required for the latter. An additional increase in system efficiency can be prognosticated and the potential energy savings are estimated based on assumptions of future LED generations derived from literature.

Full text

agronomy
Article
Energy Efficient Lighting in Plant Factories:
Addressing Utilance
Jens Balasus 1,* , Janis Blank 1, Sebastian Babilon 1,2 , Tim Hegemann 1and Tran Quoc Khanh 1


Citation: Balasus, J.; Blank, J.;
Babilon, S.; Hegemann, T.; Khanh,
T.Q. Energy Efficient Lighting in Plant
Factories: Addressing Utilance.
Agronomy 2021,11, 2570. https://
doi.org/10.3390/agronomy11122570
Academic Editor: Byoung
Ryong Jeong
Received: 31 October 2021
Accepted: 15 December 2021
Published: 17 December 2021
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2021 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
1Laboratory of Adaptive Lighting Systems and Visual Processing, Technical University of Darmstadt,
Hochschulstr. 4a, 64289 Darmstadt, Germany; janis.blank@stud.tu-darmstadt.de (J.B.);
babilon@lichttechnik.tu-darmstadt.de (S.B.); hegemann@lichttechnik.tu-darmstadt.de (T.H.);
khanh@lichttechnik.tu-darmstadt.de (T.Q.K.)
2Light and Health Research Center, Department of Population Health Science and Policy,
Icahn School of Medicine at Mount Sinai, One Gustave L. Levy Place, New York, NY 10029, USA
*Correspondence: balasus@lichttechnik.tu-darmstadt.de; Tel.: +49-6151-16-22882
Abstract:
Vertical farming is considered to play a crucial role in future food supply. Until today,
the high amount of electrical energy required for artificial lighting has been problematic in this
context. Various possibilities for increasing efficiency through adapted lighting conditions have
been and are being investigated. However, comparably little attention is paid to increasing utilance,
i.e., the amount of photons that can effectively be used by the plant. In this work, a novel targeted
lighting strategy is therefore proposed that allows for a dynamic adaptation of the luminaires’ light
distribution to match the effective crop size at each stage of plant growth in a fully-automated
manner. It is shown that the resulting utilance can significantly be increased compared to standard
full-coverage lighting. Moreover, it is found that the proposed strategy is likely to consume less than
half of the electrical energy usually required for the latter. An additional increase in system efficiency
can be prognosticated and the potential energy savings are estimated based on assumptions of future
LED generations derived from literature.
Keywords:
horticultural lighting; dynamic light distribution; targeted lighting; utilance; LED lighting
1. Introduction
Vertical indoor farming comes along with a lot of benefits ranging from a stable
year-round production, high productivity, and crop quality to higher yields per land
area and less impact on the environment due to the development of water-, fertilizer-,
and pesticide-saving cultivation methods [
1
–
5
]. The cultivation of crops in such closed
environments allows for the explicit and optimal control of environmental parameters
affecting plant growth [
6
], but at the same time requires natural daylight to be replaced
by artificial light sources to drive the process of photosynthesis [7–9], which considerably
increases the demand for (electrical) energy. In the earlier days, fluorescent tubes were
used for this purpose. However, starting from 2015, LED-based solutions have become the
state-of-the-art technology in horticultural lighting [10].
Even though modern LED packages and chip-on-board (COB) modules with high
efficiencies are used as light sources, the amount of electrical energy needed for production
is often stated to be the limiting factor for a broad scale application of vertical and indoor
farming technologies. The reported amounts of electrical energy typically needed for horti-
cultural lighting in plant factories vary between studies and range from
52 %
(calculated
with data from [
11
]) up to
80 %
[
12
] of the total energy required for an optimal plant growth.
To produce a single head of lettuce, this translates to an electrical energy consumption
of approximately 1.0–
1.6 kW h
that can be referred to lighting. Nonetheless, in terms of
overall energy consumption per produced kilogram dry weight, vertical indoor farms have
been shown to outperform even the most efficient greenhouses [
13
]. However, the general
Agronomy 2021,11, 2570. https://doi.org/10.3390/agronomy11122570 https://www.mdpi.com/journal/agronomy

Agronomy 2021,11, 2570 2 of 17
demand of such plant factories for purchased energy is much higher than in the latter case
and reduces corresponding profit margins.
In recent years, though, developments in horticultural lighting have introduced new
sustainability concepts intended to significantly reduce the electrical energy consumption
of modern indoor farming. In this context, a distinction can be made between such concepts
that aim at physiological aspects to improve plant growth and those that focus more on
the technical parts of lighting. Regarding the former, various studies have shown that
the growth of crops, plants, and specific plant parts can be optimized by adjusting the
spectral composition [
14
–
17
], the intensity[
18
–
20
], and the timing [
20
–
22
] of light exposure,
whereas most technical approaches aim at increasing the conversion rate from electrical
energy to usable optical radiation by optimizing the lighting fixtures including their light
output characteristics.
An important measure in this context is the so-called photosynthetic photon flux
efficacy (PPFE), which is defined as the number of photons within the spectral region of
photosynthetically active radiation (PAR) emitted by the luminaire divided by the electrical
energy required for their creation and is expressed in
µmol J−1
. The latest generation
of horticultural luminaries with either a blue/red or white/red LED mixture typically
yield conversion efficacies of
3.0 µmol J−1
and
2.78 µmol J−1
, respectively. However, based
on theoretical deliberations on physical limitations, Kusuma et al. [
23
] prognosticated
efficacies of
4.1 µmol J−1
for the blue/red and of
3.4 µmol J−1
for the white/red solutions
to be achievable with current LED technologies.
Besides increasing the PPFE of a luminaire, it is important to ensure that the emitted
photons can efficiently be used by the plant. Ideally, each photon generated from electrical
energy should reach its usable area, i.e., the plant leaves. Thus, the larger the ratio between
the photon flux reaching the plant leaves and the total flux emitted by all relevant light
sources, the higher the overall efficiency of the horticultural lighting system. In this
context, Lee et al. [
24
] were able to show that the use of dedicated LED-lens combinations
for targeted beam shaping may increase the lighting system’s efficiency with regard to
crop cultivation by almost a factor of four compared to using conventional fluorescent
luminaires. System efficiency in horticultural lighting, however, does not only depend on
the light sources’ emission characteristics but also on the absolute distance between the
luminaire and the plant leaves.
To obtain an estimate for the amount of usable photons, the canopy photon capture
efficiency (CPCE) has been introduced as a reliable measure in horticultural lighting and
is basically determined by the fraction of emitted photons reaching the plant
leaves [25].
As plant growth, in general, is not linear, the resulting canopy density (
=
∧
leaf area index)
is changing over time. Thus, in order to ensure a constantly high CPCE across all growth
phases, the distance between the lighting fixtures and the growing plants must be adjusted
continuously in case of a steady illumination [
26
]. This adjustment process, however,
requires additional human interaction or at least the implementation of a dedicated (elec-
tro)mechanical system, both leading to increments in working hours and/or electrical
power consumption.
Instead of using a steady illumination in combination with a continuous distance
adjustment, a better and more sophisticated way of ensuring a constant CPCE is to adapt
the lighting fixtures’ spatial light distributions to comply with the changing canopy density.
Even though such targeted lighting strategies have been shown to significantly increase
the energy-biomass conversion efficiency [
27
,
28
], i.e., the produced biomass per unit of
electrical energy, previous implementations and proof-of-concepts found in the literature
still require manual adjustments over time for achieving optimal results. For this reason,
the purpose of the current work is to present a novel, fully-automated lighting strategy
that, by making use of a specifically designed luminaire, automatically adapts the light
distribution to match the effective leaf surface area at each stage of plant growth in order
to minimize the wasting of optical radiation.

Agronomy 2021,11, 2570 3 of 17
2. Materials and Methods
2.1. Dynamic Utilance
According to CIE S 017/E:2020 [
29
], the quantity “utilance” gives the ratio between
the luminous flux received by a certain reference surface and the sum of the individual
output fluxes of the luminaires of the lighting installation. Thus, the larger this ratio, the
higher the overall efficiency of the lighting system. Note that for horticultural applications,
it is common to use photon-based instead of photometric quantities. Hence, with the
standardized definition of utilance referring only to the latter, a new photon-based utilance
measure ηuis introduced in this work. The corresponding equation reads
ηu=φp,u
∑φp, (1)
where
φp,u
is the photon flux received by the reference surface and
∑φp
is the sum of
photon fluxes emitted by the luminaires of the lighting system.
For a typical indoor lighting application, the reference surface could for example
be the working area defined by an office desk. In case of horticultural lighting, though,
the most suitable choice for the reference surface is the area covered by the plant leaves,
which will also be denoted as usable area (UA) in the following. During the seedling
stage, the relevant leaf area covers only a small percentage of the cultivating area, but it
continuously increases over time throughout the various growth phases, which therefore
results in a progressive expansion of the UA. Thus, assuming a fixed non-adaptable lighting
installation illuminating the entire cultivating area, which is usually the standard for
existing commercial indoor farms, utilance and consequently the overall system efficiency
are quite low during the seedling stage and initial growth phases. With the expansion
of the leaf area over time, both the utilance and the overall system efficiency naturally
increase—but still in a non-optimal manner. If, on the other hand, a lighting system is used
that is capable of adapting its light distribution to match the leaf surface area at any stage
of plant growth, which will minimize the wasting of optical radiation without the need of
adjusting the distance between plants and luminaires, an optimal system efficiency can be
achieved. To get an overview of previous approaches, the following section summarizes
the existing work on such targeted horticultural lighting solutions that can be found in
the literature.
2.2. Previous Work on Targeted Horticultural Lighting
In 2014, Poulet et al. [
27
] proposed a novel luminaire setup for horticultural lighting
intended to reach higher efficiencies in lettuce growth for future human missions to Mars
and other long-distant space destinations requiring bioregenerative life-support systems.
For this purpose, they used four identical custom-made LED arrays consisting of individu-
ally addressable red (
λmax =630 nm
) and blue (
λmax =455 nm
) LEDs that were arranged
in a two-by-two layout for the illumination of a dedicated growth chamber. As part of their
conducted growth experiments, they compared the energy-biomass conversion efficiency
of a targeted lighting strategy to the efficiency of a traditional full-coverage lighting. While
for the latter, all LEDs were energized during the photoperiod of a crop production cycle,
the former strategy, by manual adjustment, aimed at energizing only those LEDs that were
located directly above individual lettuce heads, i.e., for the targeted lighting strategy, the
LEDs were turned on relative to plant position and changing plant size pre-determined by
visual estimation.
After 21 days of plant growth, in which the same lighting schedule of increasing
photosynthetic photon flux density (PPFD) at leaf level (see Figure 7 of Poulet et al. [
27
])
was applied for all test conditions, a total number of 16 test lettuces were harvested
in each case to compare their corresponding dry weights and calculate the respective
energy-biomass conversion efficiencies obtained for the different lighting strategies. It was
found that the conversion efficiency of the targeted lighting strategy was almost twice as
large as the efficiency determined for the full-coverage method (1.61 vs.
0.86 g kW−1h−1
),

Agronomy 2021,11, 2570 4 of 17
where the total energy consumption of the former was less than half of the latter (
9.6 kW h
vs.
26.6 kW h
). However, even though the conversion efficiency could significantly be
increased by the targeted lighting approach, there was no positive effect on the total
amount of produced edible biomass. On the contrary, the full-coverage treatment results
in a considerably larger edible dry weight of
18.06 g
compared to only
13.59 g
obtained
for its targeted counterpart. As discussed by Poulet et al. [
27
], this difference in terms of
dry weight can be explained by the additional amount of “wasted” light getting reflected
from the polystyrene growth surface used for crop spacing onto nearby plant leaves, which
eventually increased the total number of absorbed photons and lead to the approximately
25 %
higher crop productivity for the full-coverage compared to the targeted lighting
strategy, where only a negligibly small amount of light fell onto and got reflected from the
surface area between neighboring plants. At the same time, though, significantly less light
was wasted. In summary, Poulet et al. [
27
] stated that, despite the observed slight reduction
in productivity, targeted lighting strategies in general are capable of enhancing energy
savings and provide an appropriate and sustainable approach for modern crop cultivation.
A similar, yet different approach for targeted horticultural lighting was discussed
by Li et al. [
28
], who used a zoom lens setup to adjust the emitted light distribution of
the lighting system to precisely illuminate the plants’ canopy area. Their custom-made
lighting system comprised 16 multi-chip LEDs, each of which was equipped with a convex
lens and consisted of two red LED chips (
λmax =630 nm
) and a single blue LED chip
(
λmax =460 nm
). The combined LED-lens units were arranged in a four-by-four array,
where each LED entity was located right above an individual plant seedling (butterhead
lettuce) at a distance of
30 cm
. An array of 16 Fresnel lenses fixed in a horizontal plane frame
structure was mounted directly below the multi-chip LED array and could be manually
moved up and down along the vertical direction to change the sizes of the illumination
spots on the cultivation board in order to adapt them to match the dimensions of the plant
canopy at each growth stage. Corresponding growth curves were recorded over the course
of a 25 days growth period after which the test plants were harvested to measure relevant
growth parameters. A second, full-coverage setup was used as the corresponding control
condition, where the illumination was provided by a conventional LED arrangement of four
custom-made panels of red (
λmax =630 nm
) and blue (
λmax =460 nm
) LEDs placed at a
horizontal distance of
30 cm
above the seedlings. In both cases, the same lighting scheme of
increasing PPFD levels was applied using a fixed 8:1 red-vs-blue ratio (for further details see
“Light Treatments” section of Li et al. [
28
]). Comparing both illumination strategies, it was
found that the targeted zoom lens setup saved more than
52 %
of the electrical energy that
was required for the full-coverage treatment (23.73 vs.
49.51 kW h m−2
), which resulted in a
55.6 %
larger overall biomass conversion efficiency for the former in comparison to the latter
(94.38 vs.
60.64 g kW−1h−1
). However, similar to the findings of Poulet et al. [
27
], there
was again a significant decrease in absolute fresh (dry) leaf weight from
45.23 ±5.69 g
(1.89
±
0.45 g) for the full-coverage treatment to 37.68
±
2.04 g (
1.31 ±0.21
g) for the
targeted zoom lens solution, which was complemented by a reduction in the corresponding
photosynthesis activity. Li et al. argued that the enhanced photosynthesis and the higher
yield in plant growth obtained for the full-coverage treatment might be a result of the
observed higher temperatures of the growth environment [
30
–
32
] in combination with a
smaller leaf temperature leading to the formation of thicker leaves [
33
] due to an increased
transpiration rate [34,35] and, thus, a better leaf cooling.
Despite the observed reduction of the absolute yield in plant growth, the study results
of Poulet et al. and Li et al. both confirmed that, in contrast to conventional lighting
strategies, higher efficiencies are achievable by applying targeted setups. Even though the
reported dry weights decreased by 25% and 30% when comparing the corresponding yields
obtained for targeted versus conventional lighting, the former clearly outperformed the
latter in terms of biomass conversion efficiency, where an increase of 46.6% and 55.6% could
be observed in both studies. Thus, from the valid comparisons of both lighting strategies
performed by Poulet et al. and Li et al., it must be concluded that targeted lighting provides

Agronomy 2021,11, 2570 5 of 17
an excellent alternative for the overall improvement of lighting efficiencies and energy
savings, especially with regard to vertical indoor farming applications. However, so far, the
existing targeted lighting strategies still require manual adjustments that demand for an
additional (daily) human interaction and, consequently, appear unsuitable for large-scale
implementations. To overcome this drawback, the present work proposes a novel, fully-
automated approach that, by making use of a specifically designed luminaire with object
detection capabilities, automatically adapts the emitted light distribution to comply with
the canopy area at each stage of plant growth without the need for further intervention on
behalf of the operator.
2.3. System Design
An overview of the different system components is given in Figure 1. Each luminaire
of the scalable lighting system consists of a certain number of LED segments in an annular
arrangement that can be controlled individually via suitable LED drivers to vary the
corresponding light distribution. A so-called reference luminaire is used to detect and
measure the canopy area and crop size during the growth phase. For this purpose, it is
equipped with an 8-megapixel RGB Raspberry Pi camera (Camera Module 2, Raspberry
Pi Foundation, Cambridge, UK), whose captured images are analyzed in real-time using
dedicated software running on a Raspberry Pi micro-computer (Raspberry Pi 3 Model B+,
Raspberry Pi Foundation, Cambridge, UK). From this analysis, the crop size at each growth
stage is determined so that the LED segments can be adjusted accordingly to adapt the
luminaires’ light distributions to match the crops’ canopy area as good as possible. Each
luminaire is mounted at a height of
0.4 m
above the cultivation board resulting in a mean
PPFD at leaf level of
200 µmol m−2s−1
. A more detailed discussion on the different system
components is provided as part of the following subsections.
LED-driver
targeted light
targeted light
reference luminaire
software
luminaire
LED-driver
LED-array
LED setting
camera
picture
software workow
LED-array
Figure 1.
Overview of the proposed horticultural lighting system indicating the different system
components and their interactions.
2.3.1. Number and Size of Illumination Segments
When viewed from above, the canopy surface areas of growing crops such as lettuce
can be approximated by circles of changing diameter. To adapt the luminaires’ light
distributions to match these UAs as good as possible, it therefore seems expedient to

Agronomy 2021,11, 2570 6 of 17
combine their individual LED segments in an annular arrangement of changing diameter,
too. Ideally, an infinite number of LED segments would allow for a precise adjustment of
the resulting illumination spots to be in accordance with the UAs throughout all growth
stages. However, for real-world applications, a trade-off must be found between the
luminaires’ complexity and the potential energy savings in comparison to a full-coverage
lighting strategy.
In order to find such an optimal trade-off LED configuration, a clear distinction should
be made in the following between “LED segments” on the one hand and “illumination
segments” on the other hand. Whereas the former is a feature of the luminaire and, as
indicated above, denotes the annular arrangement of LEDs that belong together by being
connected in series (see Section 2.3.2), the latter denotes the correspondingly illuminated
areas on the cultivation surface and, thus, defines the relevant UAs. To find an adequate
set of illumination segments in terms of their number and sizes, lettuce growth data taken
from Li et al. [
28
] are used for optimization. Based on the corresponding results, an optimal
luminaire LED configuration can eventually be determined. As the optimization criterion,
a photon loss function, which gives the number of photons
Nloss
that do not hit the plant
leaves, can be defined accordingly and is given by:
Nloss =Ztharvest
tseed
Ep,mean ·(Aspot(t)−Aleaf (t))dt, (2)
where
Ep,mean
is the mean photon flux density,
Aspot
is the spot area of the illumination,
Aleaf
is the projected plant leaf area (i.e., the UA), and
[tseed
,
tharvest]
is the time interval
that has elapsed from planting the seedlings to harvesting the grown crops.
In the work of Li et al. [
28
], lettuce growth data were collected once a day over a
24 days period so that the integral of Equation (2) translates to a summation of the form
Nloss =Ep,mean ·π
24
∑
t=1
·(r2
spot(t)−r2
leaf(t)), (3)
where
rspot
and
rleaf
denote the radii of the illumination spot and the lettuce leaf area,
respectively. For this calculation, the growth curve of lettuce determined by Li et al. [
28
] is
used as
rleaf(t)
. Note that in the present case the latter is approximated by the maximal leaf
length observed for the specific growth stage. The radius of the illumination spot, on the
other hand, is constrained to be always equal or larger than the maximal leaf length value.
The relative energy savings in comparison to full-coverage lighting denoted by
ζ
is then
given by
ζ=1−Nloss
Ncoverage , (4)
where
Ncoverage
is the number of photons required for a full-coverage treatment, which can
be estimated from Equation
(3)
by setting
rspot(t)
to
150 mm
(based on the maximal lettuce
leaf length observed from Li et al.’s growth experiments) and
rleaf(t)
to 0. Equation
(4)
eventually needs to be maximized by finding an optimal illumination segment configura-
tion. Using Python, the potential energy savings were analyzed for different arrangements
of 1 to 5 illumination segments. In each case, the illumination segments’ radii were varied
between 24 mm and 150 mm in 1 mm increments to calculate all possible combinations. It
was further assumed that all photons created by the luminaire’s LED segments are received
by the corresponding illumation segments on the cultivation surface so that the optimiza-
tion can be performed in a technologically independent manner, i.e., no information on
the directional emission characteristics of the LED segments was required to estimate the
maximal potential energy savings compared to full-coverage lighting. The most efficient
combinations of segment radii for different numbers of illumination segments are finally
summarized in Table 1.

Agronomy 2021,11, 2570 7 of 17
Table 1.
Relative energy savings in comparison to full-coverage lighting for different illumination segment configurations
on the cultivation surface. The relative savings were calculated by assuming that all photons created by the luminaire’s LED
segments are received by the corresponding illumination segments on the cultivation surface.
Segment Number, Optimal Radius in mm Relative Savings in %
# of Segments 1 2 3 4 5
two 92 150 - - - 60.1
three 62 116 150 - - 79
four 53 92 119 150 - 87.6
five 41 62 92 119 150 90.9
Figure 2correspondingly depicts the relative maximal energy savings as a function of
the number of illumination segments. It can be seen that for more than three illumination
segments, the increase in energy savings flattens considerably. Thus, as a trade-off be-
tween the luminaires’ complexity and the expected energy savings potentials, using three
illumination segments of
62 mm
,
116 mm
and
150 mm
appears to be an adequate choice
for achieving sufficiently good system performance while still keeping its architectural
demands to a manageable level. Finally, it should be noted that, considering real-world se-
tups, different LED emission characteristics and arrangements of course change the amount
of photons received by each of the so-determined illumination segments or relevant UAs,
leading in general to deviating and slightly reduced degrees of energy savings in practice.
However, extensive pre-tests and simulations with different emission characteristics and
LED arrangements during the development of the luminaire prototype have revealed that
the overall best results in terms of UA coverage can be achieved for the proposed optical
design and system components outlined in the following sections.
1 2 3 4 5
0
20
40
60
80
100
amount of illumination segments
Relative savings in %
Figure 2.
Relative maximal energy savings by using targeted lighting in comparison to full-coverage
treatment as a function of illumination segments.
2.3.2. LED Segments
The annular LED segments, as a key feature of the proposed horticultural luminaire,
represent individually addressable luminaire elements that are intended to provide a tar-
geted illumination for crop cultivation by providing photons to the previously determined
adequate set of illumination segments defining the relevant UAs. The corresponding
LED segments can be dimmed separately so that the resulting photon flux densities can
be adjusted accordingly and harmonized between different UAs. For this purpose, each
LED segment comprises a certain number of white mid-power LEDs of the LUXEON
SunPlus 2835 Line (Lumileds, Amsterdam, The Netherlands) that were connected in series.
Their arrangement on a custom-designed printed circuit board (PCB), which also serves
as the luminaire’s heat sink, is illustrated in Figure 3a. Integrated LED driver electronics
(MAX16822A, with the recommended peripheral components) ensure direct control of the
LED segments via a microcomputer, i.e., a Raspberry Pi, without the need for a separate
wire-connected driver component. The reference luminaire further provides an additional
camera mount close to the center of its PCB. This mount can hold a small Raspberry Pi

Agronomy 2021,11, 2570 8 of 17
camera module for taking time-resolved images of the growing plants’ canopy surface area
for an automatic crop size detection (see Section 2.5).
Figure 3.
(
a
) Overview of the arrangement of the test luminaire’s illumination segments/relevant
UAs and LED segments. LEDs of the same color are connected in series. (
b
) Picture of the real
luminaire’s pcb.
Regarding the luminaire’s spectral design aspects, the selected LEDs show distinct
peaks at
450 nm
and
650 nm
of their emitted light spectrum to increase the overall PPFD
and, thus, are specifically tailored for horticultural applications. The relative spectral
radiance distribution of the used LEDs is given in Figure 4.
400 450 500 550 600 650 700 750 800
0
0.2
0.4
0.6
0.8
1
Wavelenght in nm
Relative spectral radiance
Figure 4. Relative spectral radiance distribution of the used LUXEON SunPlus 2835 LEDs.
Note that using only one type of LEDs additionally ensures spectral homogeneity
on illuminated surfaces. To further increase utilance and system efficiency, each LED is
equipped with a HEIDI-RS total internal reflection lens (LEDil Oy, Salo, Finland) with a
beam angle of
8°
, see Figure 3b. This helps to reduce the overall photon loss by shaping
the luminaire’s light emission to achieve a sharply delineated illumination of the UAs so
that less photons get wasted because of a too broad and diffuse light distribution.

Agronomy 2021,11, 2570 9 of 17
2.4. Metrics to Quantify Homogeneity of the PPFD Distribution
Using the proposed LED-lens combinations might lead to an increase of light spots
in the different UAs. Several uniformity metrics are therefore introduced to quantify the
corresponding illumination homogeneity in order to confirm the suitability of the proposed
approach in terms of ensuring an even crop growth.
According to CIE S 008/E-2001 [
36
], the uniformity measure
u1
is defined as the
ratio of the minimum to the average value of the illuminance measured horizontally on
a reference surface. Because of the intended use of photon-based instead of photometric
quantities as part of the present work,
u1
is re-formulated accordingly to represent the ratio
between the minimum Ep,min and the average Ep,mean PPFD values:
u1=Ep,min
Ep,mean . (5)
Similarly,
u2
can be re-defined as the ratio between
Ep,min
and the maximum PPFD
value
Ep,max
, representing a measure for the span of the corresponding PPFD distribution.
The corresponding equation reads
u2=Ep,min
Ep,max . (6)
When applying Equations
(5)
and
(6)
, the lowest PPFD values are expected to be found
close to the edges between individual UAs. These specific boundary regions, however, are
less relevant for an even plant growth since, by definition, the UAs are typically larger than
the plant’s canopy surface area at each growth stage (only negligibly small portions of the
plant might temporarily extend a UA’s boundary before the next LED segment is turned
on by the algorithm discussed in Section 2.5). Hence, in order to provide a more accurate
indication of potential non-uniformities within the relevant areas, a more suitable metric
should be introduced that evaluates homogeneity by relating the fractional area in which
the PPFD lies within a specific range around the desired PPFD value for an optimal lettuce
growth (i.e.,
200 µmol m−2s−1
) to the total area
Atotal
of the respective UA. The so-defined
uniformity measures u10% and u20% are thus given by
u10% =Abet.180..220(EP(x,y))
Atotal
, where 180 µmol
m2s
<Ep(x,y)<220 µmol
m2s, (7)
u20% =Abet.160..240(EP(x,y))
Atotal
, where 160 µmol
m2s
<Ep(x,y)<240 µmol
m2s, (8)
where
Abet.180..220
and
Abet.160..240
are the areas for which the corresponding PPFD distribu-
tion Ep(x,y)deviates from the desired value by less than 10 % and 20 %, respectively.
2.5. Automated Crop Size Detection and Lighting Adjustment
Each LED segment can be dimmed separately to adapt the luminaire’s overall light
distribution to the crops’ current growth stage. In order for the lighting system to do
this automatically and without the need for additional human intervention, the diameter
of the canopy surface area must be determined first by using the Raspberry Pi camera
module which is mounted to the system’s reference luminaire and programmed to take a
corresponding image every hour. Based on this periodic image acquisition, an appropriate
color detection algorithm can be used to determine the crop diameter in a time-resolved
manner. If the observed crop diameter is larger than the area illuminated by an inner LED
segment, the next outer LED segment will be switch on, too. The Raspberry Pi is not only
used to capture and analyze the crop images but also to control the corresponding LED
driver electronics and, thus, the PPFD on the plant canopy by using dedicated pulse-width
modulated (PWM) signals to drive the individual LED segments.

Agronomy 2021,11, 2570 10 of 17
To determine the actual crop size, a real-time green-pixel analysis is performed using
Python and the OpenCV library. Figure 5illustrates the corresponding workflow. First,
the original picture as obtained from the Raspberry Pi camera module is converted from
RGB to HSV color space since the latter is most suitable for pixel selection based on hue
values [
37
,
38
]. A green hue filter is then applied to determine the number of pixels of leaf
content within the range of each of the different UAs. If, for example, more than
10 %
of
the pixels assigned to the second UA are found to be green, i.e., the corresponding crop
leaf area is expected to have a diameter larger than the area that can be illuminated by
the inner LED segment alone, the middle LED segment of the luminaire will additionally
be switched on. The same holds true if a continuing crop growth results in more than
10 %
of the pixels belonging to the third UA to be registered as green. In this case, all
three LED segments will be energized, complying with a full-coverage lighting. Here,
the
10 %
threshold was determined empirically in a series of pre-tests during system
development. It basically represents a good compromise between system sensitivity and
the need, for energy saving reasons, to avoid switching on the next LED segment too
early as long as only a small portion of the plant leaves extends the current UA. Moreover,
it should be noted that the PWM signals that control the intensity of the different LED
segments via the LED driver electronics are defined by corresponding code values stored
in a dedicated look-up table and vary with the number of illuminated UAs. The code
values were initially determined during the calibration process of the luminaire intended
to provide a homogeneous illumination at a target PPFD of
200 µmol m−2s−1
for each of
the different UA settings, see Section 3.1.
software workow
new picture
LED setting
lter green pixels
activate next area keep LED setting
> 10 %
pixel
outside?
yes
no
Figure 5.
Workflow to determine the crop diameter for automated LED control. If more than 10% of
the green pixels are outside the current UA which is illuminated by a certain LED segment, the next
LED segment will additionally be switched on to increase the size of the illumination spot accordingly.
Obviously, the proposed lighting system uses only a single reference luminaire to
extract the required plant growth parameters from image acquisition. The assumption that
underlies this system design aspect is that crops of the same species sowed at the same time
and grown under the same conditions can be expected to show a similar growth behavior
and, in particular, yield similar diameters. Thus, using a single reference/master luminaire
appears to be sufficient for controlling the whole network of slave luminaires. Nonetheless,
extending the lighting system by individually controlled luminaires is straightforward and
might yield some further energy savings in case of an unequal plant growth.

Agronomy 2021,11, 2570 11 of 17
3. Results
3.1. Uniformity Analysis
Luminance image captures of the different light settings are taken to analyze the
uniformity and size of the illuminated areas as provided by the proposed horticultural
luminaire. As each of the three LED segments can be dimmed individually, the dimming
values are varied until the highest uniformity is reached. A sheet of white paper with
printed diameters of the different relevant UAs is placed below the reference luminaire
at a distance of
40 cm
and aligned accordingly. Measurements were performed using a
TechnoTeam LMK-5 luminance camera (TechnoTeam Bildverarbeitung GmbH, Ilmenau,
Germany) positioned at a
0°
/
45°
measurement geometry. The sheet of paper is assumed to
have a Lambertian reflection characteristics [
39
,
40
]. Additional reference measurements
of the resulting PPFD are performed using a spectral irradiance sensor CSS-45 (Gigahertz
Optik GmbH, Türkenfeld, Germany), which is placed onto the reference sheet right in
center of the inner UA. Because of the spectral homogeneity of the illuminated areas,
luminance and irradiance are proportional to each other so that the pixelwisely measured
luminance values can directly be mapped to corresponding PPFD values. This leads to the
resulting PPFD distributions shown in Figure 6.
0 200 400 600
450
400
350
300
250
200
150
100
50
0
0 200 400 600 0 200 400 600
0
50
100
150
200
250
PPFD
in µmol m-2 s-1
Pixel Pixel Pixel
Pixel
UA 1 UA 2 UA 3
Figure 6.
Measured PPFD distributions of the different UAs. Note that
x
and
y
coordinates are given
in pixels as the optical distortion of the measurement geometry of the used setup prevented a proper
assignment of pixel coordinates and their absolute position in metric units. As can be seen, the
proposed luminaire prototype configuration shows the desired adaptive UA coverage capabilities.
Based on these results, minimum, maximum, and averaged PPFD values were deter-
mined for each of the specific UAs and are summarized in Table 2. As can be seen, the
largest measured PPFD value is
252.45 µmol m−2s−1
located within UA 3. It is approx-
imately
26 %
larger than the target PPFD value of
200 µmol m−2s−1
. The lowest PPFD
value, on the other hand, is
88.68 µmol m−2s−1
and can be identified to be located close to
the edge of UA 3, which is expected from the discussions of Section 2.4. The corresponding
average PPFD values (i.e.,
191.79 µmol m−2s−1
for UA 1,
180.7 µmol m−2s−1
for UA 2,
and
215.69 µmol m−2s−1
for UA 3) are all within a
10 %
range deviating from the target
PPFD value.
Table 2. Uniformity comparison of the different UAs. Epvalues in µmol m−2s−1.
u10% u20% u1u2Ep,min Ep,max Ep,mean
UA 1 0.58 0.9 0.69 0.57 131.58 231.76 191.79
UA 2 0.38 0.7 0.49 0.36 88.68 243.57 180.7
UA 3 0.23 0.92 0.61 0.52 131.16 252.45 215.69
Regarding the considerations of uniformity, it must be stated that UA 2 shows the
smallest metric values with 0.49 and 0.36 for
u1
and
u2
, respectively. These findings can
be explained by the low PPFD values found near the boundary to UA 3 when only the
two inner LED segments are lit, see Figure 6. For both UA 1 and UA 3, similar and slightly

Agronomy 2021,11, 2570 12 of 17
larger metric values are observed. While UA 1 shows values of 0.69 for
u1
and of 0.57 for
u2, values of 0.61 and 0.52, respectively, must be reported for UA 3.
Because of the strong impact of the lower PPFD values close to the boundaries between
adjacent UAs, a more reliable estimate of uniformity that is less susceptible to these kind of
“outliers” was needed. In particular, since potential effects of resulting inhomogeneities can
easily be avoided in practice by increasing the system sensitivity in such a way that the next
LED segments are already switch on at an earlier growth stage for a smoother transition but
at the costs of reduced energy savings, see Section 2.5. Hence, for providing more suitable
homogeneity metrics that better reflect the relevant UA contributions, the ratios
u10 %
and
u20 %
have been introduced in Section 2.4.
u10 %
shows values of 0.58 for UA 1, 0.38 for UA 2
and 0.23 for UA 3, indicating that in the worst case of UA 3, only about a fifth to a fourth of
the illuminated area shows PPFD values deviating less than
10 %
from the target value of
200 µmol m−2s−1
. However, as can be seen when looking at the
u20 %
metric calculations,
considerably larger fractions of the UAs show PPFD values that deviate less than
20 %
from
this target value. Compared to UA 2, which shows a value of
u20% =
0.7, both UA 1 and
UA 3 yield a slightly higher uniformity of 0.9 and 0.92, respectively. In order to reach an
approximately equal uniformity
u20 %
of 0.9 in all three UAs, the radius of UA 2 needs to
be reduced to
100 mm
. However, changing the diameter of UA 2 accordingly also leads to
a decrease in the potential power savings from
79 %
to
73 %
, when applying
Equation (4)
.
Thus, the current selection of UA diameters seems to provide a good balance between
energy savings and homogeneity of the illumination, in particular since the proposed
luminaire prototype configuration shows the desired adaptive UA coverage capabilities as
emphasized by the corresponding PPFD distributions of Figure 6.
3.2. Calculation of Utilance
The determination of utilance was performed by using the light simulation software
LightTools (Synopsys, Inc., Mountain View, CA, USA). Luminous intensity distribution
curves provided by the lens manufacturer for the LED-lens combination using the stan-
dardized photometric format of the Illuminating Engineering Society (IES format) were
used for an accurate modelling of the light emission characteristics of the luminaire system.
The assumed mounting height above the cultivation area was
40 cm
, where again the same
three annular evaluation areas (diameters of
62 mm
,
116 mm
and
150 mm
) as in Section 3.1
were used to define representative UAs. Corresponding utilance values were then calcu-
lated by adopting Equation (1). Results are summarized in the first column of Table 3. As
can be seen, the lowest utilance is obtained for UA 1 with only
59 %
of the emitted photons
being allocatable to this area. UA 2 and UA 3, on the other hand, show considerably higher
utilance values of 81 % and 83.05 %, respectively.
Table 3. Calculated utilance.
ηUtarget ηUtotal
UA 1 0.59 0.22
UA 2 0.81 0.63
UA 3 0.83 0.83
For a better comparison, utilance values as obtained for a full-coverage lighting
strategy using the same luminaire (i.e., all LED segments constantly turned on) are also
provided and given in the second column of Table 3. Again utilance is calculated for
the three different UAs. It should be noted that the values for UA 3 are the same for
both targeted and full-coverage lighting. This basically represents the fact that the former
converges against the latter as further LED segments are energized due to the progressing
plant growth. Significantly enhanced utilance, on the other hand, can be reported for the
targeted lighting strategy when considering the earlier growth stages, i.e., for UA 1 and
UA 2
. In addition, less fluctuation between the different UAs is observed in this case, which

Agronomy 2021,11, 2570 13 of 17
indicates that the utilance is less dependent on the growth stage for targeted compared to
full-coverage lighting.
3.3. Calculation of Electrical Power Consumption for One Head of Lettuce (kWh)
Using the growth data
rlettuce(t)
reported by Li et al. [
28
], the electrical energy con-
sumption for the production of a single head of lettuce is calculated when applying targeted
versus full-coverage lighting using the test luminaire. For this purpose, the power con-
sumption per UA is calculated from the measured voltage and current of the different LED
segments. The corresponding LED driver’s efficiency depends on the duty cycle. Hence,
for each LED segment the duty cycle is adjusted such that the driver efficiency can be
constantly set to
90 %
. The resulting formula for calculation the respective electrical energy
consumption (Pel.) then reads
Pel. =1.11 ·
24
∑
i=1
U(i)·I(i)·16 h, (9)
where the factor 1.11 represents the driver’s efficiency and
U(i)
and
I(i)
are the voltage
and current of the active LED segments. Note that the summation index
i
denotes the
i
th
day of the assumed growth experiment with each active LED segment being constantly
turned on for
16 h
per day. Using Equation (9) to calculate the total energy consumption
for growing a single lettuce head over the period of 24 days eventually gives a value of
3.43 kW h
and
7.03 kW h
for the targeted and full-coverage lighting strategy, respectively.
Thus, targeted lighting consumes less than half of the electrical energy required by the
full-coverage approach.
Based on the maximally achievable efficiencies of current and future LED genera-
tions, see Kusuma et al. [
23
], the electrical energy consumption for a targeted lighting
strategy applied to grow a single lettuce head is further calculated for an idealized test
luminaire with ideal optics and optimal LED driver efficiency. For this purpose, each UA
is multiplied with the assumed PPFD of
200 µmol m−2s−1
and divided by the estimated
LED conversion efficiency to create the required input data for re-evaluating Equation
(9)
.
Table 4summarizes the results for various LEDs and LED combinations.
Table 4.
Theoretical comparison of electrical energy consumption for different LED efficiencies under
the assumption that 100 % of the generated photons reach the UAs.
LED Efficiency Power Consumption per
in µmol W−1Head of Lettuce in kWh
Horticulture white 2.28 1.33
blue/red current 3 0.92
white/red current 2.78 0.99
blue/red prognosed 4.1 0.67
white/red prognosed 3.4 0.81
450 nm physical maximum 3.76 0.81
660 nm physical maximum 5.52 0.55
3.4. Visualization of the Automated Lighting Adjustment
In order to visualize the system behavior of the proposed test luminaire prototype,
a dedicated growth experiment using pre-grown lettuce heads (14 days old) has been
performed over a period of 30 days, where the size of the illuminated area has been
adjusted automatically depending on the registered leaf surface area. Figure 7illustrates
the respective changes in illumination according to the different growth stages. As can be
seen, a good coverage of the leaf surface area can be achieved at all stages while, at the
same time, minimizing the amount of wasted photons.

Agronomy 2021,11, 2570 14 of 17
Figure 7.
Illuminated UAs in dependence of the size of the crop leaf area.
Upper row
: Transition
in illumination from UA 1 to UA 2 due to progressing plant growth.
Lower row
: Transition in
illumination from UA 2 to UA 3 due to progressing plant growth.
4. Discussion
In this work, a novel, fully-automated lighting strategy has been proposed to adapt
the light distribution of the horticultural luminaire to match the effective leaf surface area at
each stage of plant growth in order to minimize the wasting of optical radiation throughout
the growing process. A corresponding system prototype has been developed and tested
accordingly. It has been shown that the proposed lighting system is capable of properly
adjusting its light emission to comply with the canopy surface area by automated crop size
detection, which considerably increases the utilance compared to (standard) full-coverage
lighting, in particular for the earlier growth stages. In addition, it was found that the
targeted approach shows a higher utilance constancy across the different growth phases
than its full-coverage counterpart, emphasizing its suitability for great energy savings in
horticultural lighting applications. For example, it was found that the proposed targeted
lighting strategy is likely to consume less than half of the electrical energy required by
full-coverage lighting.
Future work thus addresses the large-scale integration of this promising lighting
approach to be able to collect and analyze real data based on comprehensive field studies.
The long-term goal is to properly quantify the energy savings in relation to crop yield for a
variety of different plants and horticultural scenarios. In this context, it may be expedient to
not only consider in a future system iteration the crop diameter but also the plant height in
order to keep the PPFD at a constantly optimal level throughout all growth phases. Usually,
as the plant continues to grow, its distance to the luminaire gets reduced, which increases
the resulting PPFD on the leaf surface area. A corresponding density regulation can for

Agronomy 2021,11, 2570 15 of 17
example be realized by dimming the individual LED segments based on the distance
obtained from an image-based plant height estimation or by applying an adequate leaf
area–plant height model.
With the developed luminaire prototype, it is further possible to customize light
distributions in such a way that different intensities are applied to different parts of the
plant’s canopy surface area. For example, the PPFD in the central regions can be set to a
higher value than in the outer parts so that potential effects of these spatial inhomogeneities
on the plant growth can be investigated in a systematic manner. In addition, the proposed
lighting system shall be further extended by integrating spectrally tunable LEDs that, in
combination with the luminaire’s spatial flexibility, allow for using different light spectra to
illuminate different crop parts to potentially trigger and examine certain grow mechanism
affected by changes in the spectral light composition.
Finally, it is of interest to further investigate, and conceivably compensate, the effects
of a reduced absolute crop yield for targeted compared to full-coverage lighting due to
a reduction of the amount of diffusely reflected lighting components, as discussed in the
literature [
41
]. So far, corresponding experiments to confirm this effect are limited to
greenhouses only. Future experiments should therefore focus on indoor cultivation, where
lighting conditions are not comparable to those observed for summer greenhouses. In
particular, no compensation strategies have been proposed yet.
Author Contributions:
Conceptualization, J.B. (Jens Balasus), S.B. and T.H.; Data curation,
J.B. (Jens Balasus)
and J.B. (Janis Blank); Formal analysis, J.B. (Jens Balasus); Methodology,
J.B. (Jens Balasus),
S.B. and T.Q.K.; Software, J.B. (Jens Balasus) and J.B. (Janis Blank); Supervision,
T.Q.K.; Validation, J.B. (Jens Balasus), J.B. (Janis Blank) and T.H.; Visualization, J.B. (Jens Balasus);
Writing—original draft, J.B. (Jens Balasus); Writing—review & editing, J.B. (Jens Balasus), S.B.,
T.H., J.B. (Janis Blank) and T.Q.K.; All authors have read and agreed to the published version of
the manuscript.
Funding:
We acknowledge support for the publication of this work by the Deutsche Forschungsge-
meinschaft (DFG—German Research Foundation) and the Open Access Publishing Fund of Technical
University of Darmstadt.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Not applicable.
Conflicts of Interest: The authors declare no conflict of interest.
References
1.
Kozai, T. Sustainable Plant Factory: Closed Plant production systems with artificial light for high resource use efficiencies and
quality produce. Acta Hortic. 2013, 1004, 27–40. [CrossRef]
2.
Despommier, D. The vertical farm: controlled environment agriculture carried out in tall buildings would create greater food
safety and security for large urban populations. J. Verbraucherschutz Leb. 2011,6, 233–236. [CrossRef]
3.
Kozai, T. Resource use efficiency of closed plant production system with artificial light: Concept, estimation and application to
plant factory. Proc. Jpn. Acad. Ser. B 2013,89, 447–461. [CrossRef] [PubMed]
4.
Graamans, L.; van den Dobbelsteen, A.; Meinen, E.; Stanghellini, C. Plant factories; crop transpiration and energy balance. Agric.
Syst. 2017,153, 138–147. [CrossRef]
5.
SharathKumar, M.; Heuvelink, E.; Marcelis, L.F. Vertical Farming: Moving from Genetic to Environmental Modification. Trends
Plant Sci. 2020,25, 724–727. [CrossRef] [PubMed]
6.
Murase, H. Development of micro-precision agriculture by plant factory. J. Soc. High Technol. Agric.
2000
,12, 99–104. [CrossRef]
7.
Watanabe, H. Light-controlled plant cultivation system in Japan—Development of a vegetable factory using LEDs as a light
source for plants. Acta Hortic. 2011,907, 37–44. [CrossRef]
8.
Shimizu, H.; Saito, Y.; Nakashima, H.; Miyasaka, J.; Ohdoi, K. Light environment optimization for lettuce growth in plant factory.
In Proceedings of the 18th IFAC World Congress; Sergio, B., Angelo, C., Sandro, Z., Eds.; International Federation of Automatic
Control (IFAC): Kyoto, Japan, 2011; pp. 605–609. [CrossRef]
9.
Dutta Gupta, S.; Agarwal, A. Artificial lighting system for plant growth and development: Chronological advancement, working
principles, and comparative assessment. In Light Emitting Diodes for Agriculture; Dutta Gupta, S., Ed.; Springer: Singapore, 2017;
pp. 1–25. [CrossRef]

Agronomy 2021,11, 2570 16 of 17
10.
Kozai, T.; Niu, G.; Takagaki, M. (Eds.) Chapter 3—PFAL business and R&D in Asia and North America: Status and perspectives.
In Plant Factory (Second Edition), 2nd ed.; Academic Press: Cambridge, MA, USA, 2020; pp. 35–76. [CrossRef]
11.
Yokoyama, R. Chapter 4.2 —Energy Consumption and Heat Sources in Plant Factories. In Plant Factory Using Artificial Light;
Anpo, M., Fukuda, H., Wada, T., Eds.; Elsevier: Amsterdam, The Netherlands, 2019; pp. 177–184. [CrossRef]
12.
Nakabo, Y. Design and control of smart plant factory. In Smart Plant Factory; Springer: Berlin/Heidelberg, Germany, 2018; pp.
51–55.
13.
Graamans, L.; Baeza, E.; van den Dobbelsteen, A.; Tsafaras, I.; Stanghellini, C. Plant factories versus greenhouses: Comparison of
resource use efficiency. Agric. Syst. 2018,160, 31–43. [CrossRef]
14.
Lin, K.H.; Huang, M.Y.; Huang, M.Y.; Huang, W.D.; Hsu, M.H.; Yang, Z.W.; Yang, C.M. The effects of red, blue, and white
light-emitting diodes on the growth, development, and edible quality of hydroponically grown lettuce (Lactuca sativa L. var.
capitata). Sci. Hortic. 2013,150, 86–91. [CrossRef]
15.
Yorio, N.C.; Goins, G.D.; Kagie, H.R.; Wheeler, R.M.; Sager, J.C. Improving spinach, radish, and lettuce growth under red
light-emitting diodes (LEDs) with blue light supplementation. Hortscience 2001,36, 380–383. [CrossRef]
16.
Son, K.H.; Oh, M.M. Leaf Shape, Growth, and Antioxidant Phenolic Compounds of Two Lettuce Cultivars Grown under Various
Combinations of Blue and Red Light-emitting Diodes. Hortscience 2013,48, 8. [CrossRef]
17.
Pennisi, G.; Orsini, F.; Blasioli, S.; Cellini, A.; Crepaldi, A.; Braschi, I.; Spinelli, F.; Nicola, S.; Fernandez, J.A.; Stanghellini, C.; et al.
Resource use efficiency of indoor lettuce (Lactuca sativa L.) cultivation as affected by red:blue ratio provided by LED lighting.
Sci. Rep. 2019,9, 14127. [CrossRef] [PubMed]
18.
Weiguo, F.; Pingping, L.; Yanyou, W.; Jianjian, T. Effects of different light intensities on anti-oxidative enzyme activity, quality and
biomass in lettuce. Hortic. Sci. 2018,39, 129–134. [CrossRef]
19.
Urrestarazu, M.; Nájera, C.; Gea, M.D.M. Effect of the Spectral Quality and Intensity of Light-emitting Diodes on Several
Horticultural Crops. Hortscience 2016,51, 3. [CrossRef]
20.
Pennisi, G.; Distal, A.; Landolfo, M.; Orsini, F.; Landolfo, M.; Pistillo, A.; Pistillo, A.; Crepaldi, A.; Crepaldi, A.; Nicola, S.; et al.
Optimal photoperiod for indoor cultivation of leafy vegetables and herbs. Eur. J. Hortic. Sci. 2020,85, 329–338. [CrossRef]
21.
Carotti, L.; Potente, G.; Pennisi, G.; Ruiz, K.B.; Biondi, S.; Crepaldi, A.; Orsini, F.; Gianquinto, G.; Antognoni, F. Pulsed LED Light:
Exploring the Balance between Energy Use and Nutraceutical Properties in Indoor-Grown Lettuce. Agronomy
2021
,11, 1106.
[CrossRef]
22.
Kanechi, M.; Maekawa, A.; Nishida, Y.; Miyashita, E. Effects of pulsed lighting based light-emitting diodes on the growth and
photosynthesis of lettuce leaves. Acta Hortic. 2016,1134, 207–214. [CrossRef]
23.
Kusuma, P.; Pattison, P.M.; Bugbee, B. From physics to fixtures to food: current and potential LED efficacy. Hortic. Res.
2020
,7, 56.
[CrossRef] [PubMed]
24.
Lee, X.H.; Chang, Y.Y.; Sun, C.C. Highly energy-efficient agricultural lighting by B+R LEDs with beam shaping using micro-lens
diffuser. Opt. Commun. 2013,291, 7–14. [CrossRef]
25.
Nelson, J.A.; Bugbee, B. Economic Analysis of Greenhouse Lighting: Light Emitting Diodes vs. High Intensity Discharge Fixtures.
PLoS ONE 2014,9, e99010. [CrossRef] [PubMed]
26.
Seginer, I.; Ioslovich, I. Optimal spacing and cultivation intensity for an industrialized crop production system. Agric. Syst.
1999
,
62, 143–157. [CrossRef]
27.
Poulet, L.; Massa, G.; Morrow, R.; Bourget, C.; Wheeler, R.; Mitchell, C. Significant reduction in energy for plant-growth lighting
in space using targeted LED lighting and spectral manipulation. Life Sci. Space Res. 2014,2, 43–53. [CrossRef]
28.
Li, K.; Li, Z.; Yang, Q. Improving Light Distribution by Zoom Lens for Electricity Savings in a Plant Factory with Light-Emitting
Diodes. Front. Plant Sci. 2016,7, 92. [CrossRef] [PubMed]
29.
ILV: International Lighting Vocabulary, 2nd ed.; CIE S 017/E:2020; Technical Report; Commission International de l’Eclairage (CIE):
Vienna, Austria, 2020. [CrossRef]
30.
Berry, J.; Bjorkman, O. Photosynthetic response and adaptation to temperature in higher plants. Annu. Rev. Plant Physiol.
1980
,
31, 491–543. [CrossRef]
31. Gosselin, A.; Trudel, M.J. Root-zone temperature effects on pepper. J. Am. Soc. Hortic. Sci. 1986,111, 220–224.
32.
He, J.; Lee, S.K.; Dodd, I.C. Limitations to photosynthesis of lettuce grown under tropical conditions: Alleviation by root-zone
cooling. J. Exp. Bot. 2001,52, 1323–1330. [CrossRef] [PubMed]
33.
Wolfe, D.W. Low temperature effects on early vegetative growth, leaf gas exchange and water potential of chilling-sensitive and
chilling-tolerant crop species. Ann. Bot. 1991,67, 205–212. [CrossRef]
34. Gates, D.M. Leaf temperature and transpiration. Agron. J. 1964,56, 273–277. [CrossRef]
35.
Pallas, J.E.; Michel, B.E.; Harris, D.G. Photosynthesis, transpiration, leaf temperature, and stomatal activity of cotton plants under
varying water potentials. Plant Physiol. 1967,42, 76–88. [CrossRef]
36.
Lighting of Work Places—Part 1: Indoor, CIE S 008:2001; Technical Report; Commission International de l’Eclairage (CIE): Vienna,
Austria, 2001.
37.
Narkhede, P.R.; Gokhale, A.V. Color image segmentation using edge detection and seeded region growing approach for CIELab
and HSV color spaces. In Proceedings of the 2015 International Conference on Industrial Instrumentation and Control (ICIC),
Pune, India, 28–30 May 2015; pp. 1214–1218. [CrossRef]

Agronomy 2021,11, 2570 17 of 17
38.
Narkhede, P.R.; Gokhale, A.V. Color particle filter based object tracking using frame segmentation in CIELab and HSV color
spaces. In Proceedings of the 2015 International Conference on Communications and Signal Processing (ICCSP), Melmaruvathur,
India, 2–4 April 2015; pp. 804–808. [CrossRef]
39.
Akao, Y.; Tsumura, N.; Miyake, Y. Modeling gonio-spectral reflection properties of paper sheets for efficient gonio imaging.
In Proceedings of the 2004 Conference on Colour in Graphics, Imaging, and Vision (CGIV). Society for Imaging Science and
Technology, Aachen, Germany, 5–8 April 2004; pp. 414–417.
40.
Akao, Y.; Tsumura, N.; Nakaguchi, T.; Miyake, Y. Characterization of white paper sheets by BRDF model parameters estimated in
the specular reflection plane. J. Imaging Sci. Technol. 2010,54, 60503. [CrossRef]
41.
Tani, A.; Shiina, S.; Nakashima, K.; Hayashi, M. Improvement in lettuce growth by light diffusion under solar panels. J. Agric.
Meteorol. 2014,70, 139–149. [CrossRef]