ArticlePDF Available

Abstract and Figures

The paper presents several technical issues and software implementations proposed in order to decode, process, and render images received from NOAA satellites. NOAA satellites are seeking extreme weather events on Earth. The proposed solutions offer the possibility of decoding and processing the NOAA signals, by using the PC and its audio-in port, without any extra signal processor-based boards. Also, system architecture is open, according to the GNU requirements and hardware independent, due to its capability of working offline, by decoding a pre-recorded .wav file. Finally, the results are converted to a standard image file that lends itself easily to processing for relevant feature extraction. In spite of its hardware independency, the algorithm is capable to operate also online with SDR receivers. These receivers must follow the Universal Software Radio Peripheral specifications on USB ports.
Content may be subject to copyright.
Volume 49, Number 4, 2012 ACTA TECHNICA NAPOCENSIS
Electronics and Telecommunications
________________________________________________________________________________
Manuscript received February 29, 2010; revised April 17, 2010
1
NOAA SIGNAL DECODING AND IMAGE PROCESSING
USING GNU-RADIO
Nicolae Crisan, Ligia Cremene
Technical University of Cluj-Napoca
E-mail: nicolae.crisan@com.utcluj.ro
Phone: 0264401241
Abstract: The paper presents several technical issues and software implementations proposed in order to decode, process, and
render images received from NOAA satellites. NOAA satellites are seeking extreme weather events on Earth. The proposed
solutions offer the possibility of decoding and processing the NOAA signals, by using the PC and its audio-in port, without any
extra signal processor-based boards. Also, system architecture is open, according to the GNU requirements and hardware
independent, due to its capability of working offline, by decoding a pre-recorded .wav file. Finally, the results are converted to
a standard image file that lends itself easily to processing for relevant feature extraction. In spite of its hardware independency,
the algorithm is capable to operate also online with SDR receivers. These receivers must follow the Universal Software Radio
Peripheral specifications on USB ports.
Keywords: Software Defined Radio (SDR), Universal Software Radio Peripheral (USRP), GNU radio.
I. INTRODUCTION
GNU radio [1] is an open development toolkit that
provides free processing components in order to implement
software radio interfaces. GNU radio toolkits are primarily
written in Python language based on Linux scripts and
tools. Some performance critical signal processing routines
are written in C++ using floating point processor
extensions. Thus, a programmer can imagine simple and
rapid applications still compatible with SDR receivers that
follow the Universal Software Radio Peripheral (USRP)
specifications.
USRP from Ettus research [2] enables users worldwide
to address a wide range of research topics, opening the way
through the software radio approach with the help of the
GNU radio toolkit [3] which it is compatible with.
The Automatic Picture Transmission (APT) system [4]
uses data streams to transmit a visible image during
daylight using channel A and an infra-red image on
channel B (all day and night long). The APT analog signal
is transmitted continuously via NOAA-8 through NOAA-
14 satellites. This signal was designed especially for
inexpensive receivers and ground stations.
National Oceanic and Atmospheric Administration
(NOAA) [5] is a US department that administrates an
environmental satellite system that sends data from space
in order to monitor and track weather change and extreme
weather events. NOAA operates two types of satellites:
geostationary satellites, for US monitoring, and polar-
orbiting satellites that circle the planet and transmit
pictures continuously from about 540 miles above the
Earth. Transmitted data may also be used to calculate ocean
temperature and for a better understanding of the Earth
climate change.
This paper focuses on the algorithm and software
implementation in order to decode and process the APT
signal using original routines. The hardware platform used
for receiving and recording the APT signal in a .wav file is
outside the topic of this paper. The main objective of this
work is to reduce complexity by intensively using offline
signal processing techniques and image processing in order
to counteract the Doppler effects.
The paper is organized as follows. Section II overviews
the main challenges in the reception of NOAA satellite
images and presents the core ideas of the proposed
solutions. Section III discuses the block-diagram for
NOAA satellite reception and the D-APT-AM signal
recording. Section IV describes the use of GNU radio
companion to decode the APT-wav files. Section V
presents the adjustments (processing) performed on the
received image. Section VI presents the conclusions.
II. PROBLEM FORMULATION
The present paper aims not to highlight the hardware
platforms or routines involved in converting the APT
analog signal into a .wav file. This is a common issue
within the ham radio community, together with the open
distribution of the APT-wav files [6].
For instance, a low-cost ground station uses a SDR
receiver (SDR1000-6000) with a satellite track device
entirely controlled by the computer [7]. This tracking
device follows the NOAA-X satellites according to the
telemetry data, assisted by software (see Fig. 1) [8].
Recorded .wav files are also available from other
communities who share it freely via servers. The APT-wav
files are available for download in 8 or 16 bits/sample, the
Volume 51, Number 1, 2010 ACTA TECHNICA NAPOCENSIS
Electronics and Telecommunications
________________________________________________________________________________
2
last ones having higher definition and offering a better
quality. Fig. 1 captures an example of how Keplerian data
together with software and internet are tracking the satellite
24 hours a day. The trajectory of the NOAA satellites is
elliptic, so the distance varies with time.
Figure 1. Online Keplerian data via Internet using
tracking satellite software NOAA-18
The question that emerges is whether the .wav file can
offer at least as much information as the decoded image,
from Fig 2. The answer is no. The image in Fig. 2 is
acquired with a professional hardware platform. This
platform performs Doppler shift, frequency correction and
line synchronization based on the Keplerian data extracted
directly from the APT signal [4]. Most of the wav-APT
signals are stored as they are making it possible to do these
corrections offline, by software, providing the data without
any change at the receiver level in terms of frequency auto-
tuning or Doppler shift compensation.
Figure 2. NOAA-18 image with each line
synchronization and Doppler shift compensation [6]
The wav-APT file becomes very interesting from this
point of view for any programmer who is interested in
intensively using software radio. The main goal of image
and signal processing is in reducing the complexity of the
necessary hardware platform. This means that the use of a
very low-cost ground station becomes affordable for
anyone who uses a SDR receiver and an omni-antenna. The
tracking device is no longer necessary, at the cost of SNR
loss. Of course, a noisier image requires more processing
and the need for processing increasing.
Another challenge emerges from the wide use of the
cheap narrow band transceivers with a narrower IF
bandwidth (only 25 KHz instead of 34 KHz) as it is
required for the APT signal. This leads to the loss in image
contrast. Fortunately, the reduced range of luminance
along the black and white grayscale lends itself very well
to histogram equalization techniques. The main concerns
in the next sections are whether image processing
techniques can cope with these problems.
III. APT DATA FRAME FORMAT
The APT analog signal [4] is FM modulated using an
RF carrier between 137.1 and 137.9125 MHz, depending
on the index of the NOAA satellite. For example, NOAA-
18 transmitter is tuned on 137.9125 MHz.
A low-cost ground station demodulates the FM signal
and applies the FM demodulated signal to the PC-audio-in
board. The APT-AM demodulated analog signal (Fig. 3) is
a 2.4 KHz subcarrier in the range of the PC sound card. The
D-APT-AM signal is sampled by the sound board and
recorded as is on the hard-disk without any receiver
adjustments in a .wav format.
Figure 3. The most simple block-diagram for NOAA
satellite reception and the D-APT-AM signal recording
The figure 4 shows the basic APT format. The
transmitted carrier is FM modulated on 137.5 MHz and
approaches 137.9125 MHz depending on the satellite
index. An AM subcarrier on 2400 Hz modulates image
data, as amplitude variation along a gray scale. Each word
is sampled using 8 bits/pixel along one row, which lasts
exactly 0.5 seconds (2 lines/second). The equivalent data
rate is 4160 word/second with a Digital/Analog accuracy
of 8 MSB’s of each 10 bit word. For each image, only 909
words/row are useful data; the remaining ones are for
synchronization and telemetry.
Figure 4. APT data frame format [4]
Sync A is a 1040 Hz square wave of seven cycles. Sync
B is an 832 pulse train having also seven cycles [4]. The
total number of pixels along a row is 2080, taking into
account A and B images, Sync A and B and telemetry
frames. There are 128 telemetry frames with 8 lines each.
All those 128 frames put together a total of 1024 lines. The
image resolution is 909/1024 pixels.
FM-receiver
Omni-antenna
audio-in
WAV file
I/O
PC
APT-AM
D-APT-AM
y
n
y
n
S
p
a
c
e
A
Minute
marker
(four lines
two white
two black)
S
p
a
c
e
B
Minute
marker
(four lines
two white
two black)
Image A
Visible/IR
spectrum
Image B
IR
spectrum
Telemetry data frame A
Telemetry data frame B
128 frames
0.5 seconds (APT video line time)
Volume 51, Number 1, 2010 ACTA TECHNICA NAPOCENSIS
Electronics and Telecommunications
________________________________________________________________________________
3
Additional explanations on Telemetry and Sync frames
are not necessary at this point because we are not using any
online adjustments based on these frames. These
adjustments are performed offline, after the signal
acquisition. The signal acquisition is running continuously,
and the PC is saving data in a .wav file, according to the
block diagram from Fig. 3.
Offline Doppler shift compensation and others
adjustments will be the topic of Section V and are entirely
software defined.
IV. USING GNU RADIO COMPANION TO
DECODE THE APT-WAV FILES
The sampling rate of the recorded .wav file is 11.025
KHz. This sampling frequency depends especially on the
technical performance of the sound board and may be even
higher. Nevertheless the most notable thing is that the D-
APT-AM is an AM 2.4 KHz signal.
After the reading of the .wav file (see Fig. 5) with the Wav
File Source block, the signal is filtered with a low pass
filter. The higher cut-off frequency of the filter upper
bounds, the higher are the components at 2.4 KHz. The
filtered signal is re-sampled to 9.6 KHz, which is four times
higher than 2.4 KHz.
The Rational Resampler must interpolate the 9.6 KHz
signal because it is not an integer multiplier of the input
frequency (11.025 KHz). After the interpolation, the 9.6
KHz signal presents four samples for each word of the AM
2.4 KHz signal. So we have four samples/period and the
samples come with 90 degrees phase shift (figure 5).
Using two consecutive samples we can detect the
instantaneous amplitude of the AM 2.4 KHz signal at a rate
of 4.8 KHz. The amplitude will be 𝐴 =
𝑥
2
+ 𝑦
2
, if every
pair of two consecutive samples is converted to a complex
number 𝑉 = 𝑥 + 𝑗𝑦, where x and y are float variables and
the block Float to Complex does this conversion. The use
of the complex number V offers some advantages making
it easier to calculate the value of amplitude A.
The block Complex to Magnitude acts like an AM
demodulator and thus the signal caring the useful data is
present at the output of the block. However, the
demodulated signal is not yet ready for saving because the
true symbol rate in APT format is 4.16 KHz. The next
block interpolates the signal again to 4.16 KHz. The
interpolated signal is now ready for saving, after a range
expansion between 0 and 255. This range happens to be
equal to the number we can render on using eight bits.
Taking into account that the CAN’s resolution is 8
bits/word this processing follows a logic choice. As a rule
of thumb, it is mandatory to use a minimum number of bits
for a word in a file. Pursuing this idea the next block
converts float numbers to unsigned char. Next, one block
detects the first Sync A synchronization burst of 1040 Hz
in order to skip the first row of the image which is
incomplete. Finally, the signal is saved as is, in a .dat
file, by the last block.
All Radio Companion blocks are written in Pyton
under the GNU Radio packages [3]. The block-diagram
presented in Fig. 5 is responsible only for reading,
decoding and saving data in a file. The main challenge here
is the .dat file which is not a compatible image format.
Figure 5. Block diagram of the second down-converter
This problem is avoided with the help of the Image
Magic program installed also with the Linux packages. The
conversion from .dat to .png image file is done by the
following command from the Linux terminal:
convert -size 2080x1024 -depth 8 gray: output.dat
output1.png
The converter puts together every 2080 words into a
row, generating 1024 rows, according to the number of
APT frames (see Fig. 4). The gray depth is 8 (0-255 levels)
and the results are saved in a .png image format.
V. THE IMAGE ADJUSTMENTS
When the satellite approaches the ground station [9],
travelling at about 7.5-7.6 Km/s, at an altitude of 700 Km
above the Earth, the Doppler effect shifts the frequency
with ±3 KHz. At this point, the carrier frequency is higher
than 137.9125 MHz with maximum 3 KHz, just when the
relative velocity reaches a maximum pick.
a.
b.
Figure 6. Row shift effect over the decoded image
The frequency drops suddenly with 6 KHz after the
satellite passes over the view point, from front to back. This
shift affects the data rate and also the geometry of the
image. The effect is captured in Fig. 6. In Fig. 6.a the image
is shifted to the right when the satellite comes from the
front. The same image is shifted to the left when the
satellite is moving away from the reference view point
(Fig. 6.b).
1-WAV File
So
urce
2- LPF
3-Resample
5-Float to
Complex
𝑥
2
+ 𝑦
2
y
x
V
7-Resample
8-Multiply
X255
9-Float to
UChar
10-Skip
incomplete row
11-Save in
DAT File
Volume 51, Number 1, 2010 ACTA TECHNICA NAPOCENSIS
Electronics and Telecommunications
________________________________________________________________________________
4
The correction process (image geometry
compensation) shown in Fig. 7.a shifts each row in the
opposite direction using a vertical line as reference in order
to counteract the Doppler effect. The black space A region
(see Fig. 4) is used to define this line and the transitions
from black to gray at the beginning of each row of the
image. In this way, the Doppler shift effect is compensated
for each row entirely by software.
Another effect that needs correction at the level of the
satellite could avoid the geometry distortions by
maintaining almost the same resolution of 4 Km along the
scan line [10][11][12]. This correction takes into account
the instantaneous distance between the satellite and the
earth [12] and its rotation speed. Fortunately, NOAA
satellites make this correction online. Nevertheless all
these undesired factors like earth rotation, earth curvature
and Doppler shift effects are cumulative and all distort
geometry of the image. This is shifted to the left or to the
right and could be corrected by software following the rule
of shifting the image lines in the opposite direction. From
these factors, only the Doppler shift alters both the
geometry and the luminance level of the image [6].
a.
b.
Figure 7. Image adjustments a. Geometric corrections
b. Histogram equalization
The 25 KHz-channel transceiver limits the range of
the image luminance along the gray scale. The use of a
common transceiver with a 25 KHz FM band instead of the
34 KHz band decreases the image contrast. This loss
increases with the increasing of the satellite velocity and
with the carrier frequency. A 34 KHz receiver using a
wider band could counteract better this effect. The best way
to counteract the Doppler shift is to tune the transceiver
using the telemetry data transmitted together with the
image. Here the proposed method is compensating the
luminance loss depicted in Fig. 7.b by using a histogram
equalization technique [8].
Histogram equalization is a common technique,
broadly used in image processing every time an image
suffers from a loss of contrast. For a dedicated transceiver
with a 34 KHz FM band, 26% of its the power spectrum is
outside the range of a commonly used transceiver. The loss
in terms of the contrast is not 26% because the signal
energy is not uniformly distributed along the spectrum. The
loss affects only higher frequencies that have not as much
occurrence probability like the lower frequencies. This
leads to the conclusion that the depreciation of the contrast
with 26% represents the worst case. However, this depends
on the image content and affects more the brighter images.
After the histogram equalization, the image becomes
clearly visible with a significant gain in quality (Fig. 7.b).
VI. CONCLUSIONS
A new method has been proposed for NOAA satellite
image reception based on GNU radio technique. The
procedure is based on a sequence of image processing
procedures for restoring the distorted image, instead of
using the more complicated frequency correction and line
synchronization based on the Keplerian data extracted
from the APT signal. The advantages of the proposed
method are the accuracy and affordability, no professional
hardware being needed.
REFERENCES
[1] Alexander M. Wyglinski, Maziar Nekovee, Y. Thomas Hou,
Cognitive Radio Communications - Principles and Practice,
Elsevier Academic Press - British Library Catalog, pp 621-631,
2010, ISBN 978-0-12-374715-0
[2] Documentation available about USRP Universal Software
Radio Peripheral is available at http://www.ettus.com/
[3] Documentation about GNU radio is available at
http://gnuradio.org/redmine/projects/gnuradio/wiki/BuildGuide
[4] Documentation about APT Automatic Picture Transmission
is available at
http://www2.ncdc.noaa.gov/docs/klm/html/c4/sec4-2.htm
[5] Information about NOAA The National Oceanic and
Atmospheric Administration is available at
http://www.noaa.gov/satellites.html
[6] The .wav file used in the paper is available at
http://www.fredvandenbosch.nl/satellites_WAV.html
[7] A list of low cost SDRs capable of receiving and decoding
APT signals is available at http://www.flex-radio.com/
[8] A software tool for real time satellites tracking is available at
http://www.n2yo.com/satellites/?c=4
[9] T. Langdon, A close up on Doppler Shift, The AMSAT Journal,
2010
[10] Bate, Roger; Mueller, Donald; White, Jerry , Fundamentals
of Astrodynamics, Dover Publications, Inc., New York, 1971,
ISBN 0-486-60061-0.
[11] Acharya and Ray, Image Processing: Principles and
Applications, Wiley-Interscience 2005 ISBN 0-471-71998-6
[12] John, A. Richard, Remote Sensing Digital Image Analysis,
Springer New-York Dordrecht London, pp 56-62, 2013, ISBN
978-3-642-30062-2
Volume 51, Number 1, 2010 ACTA TECHNICA NAPOCENSIS
Electronics and Telecommunications
________________________________________________________________________________
5

Supplementary resource (1)

... For transmission of pictures in APT format, the data stream is modulated in amplitude (AM) at a frequency of 2400 Hz, resulting in a signal that varies in amplitude depending on the information coming from the AVHRR/3 sensor, which is capable of measuring radiation in different areas of the electromagnetic spectrum. Then the signal is modulated in frequency (FM) between 137 and 138 MHz with a maximum peak deviation of 17 KHz, and finally the signal is amplified and transmitted to earth [5]. The frame of a picture in APT format can be seen in Fig. 2. A complete line of video of this frame has a duration of 0.5 seconds; so too, a complete video line of an APT frame is 2080 pixels long, of which 1818 are used to form the image, the rest are used for telemetry and synchronization. ...
Conference Paper
Full-text available
From the specific view of space, sophisticated Low Earth Orbit (LEO) satellites provide us with information on cloud movements, precipitation, land and sea surface temperatures. These satellites are fascinating as they continuously transmit an image signal that is easy to decipher of what they are actually looking at. This data is useful to predict weather and rainfall. The main purpose of this work is to develop an inexpensive ground station for receiving the images from these satellites and give a rough prediction of weather and rainfall. Software Defined Radio (SDR) was used for the reception of these signals along with V-Dipole, QFH Antenna and several open source software's. So far this work is successful in receiving signal from 40+ satellite passes of the National Oceanic and Atmospheric Administration's (NOAA) and METEOR-M2 satellites over the location Bangalore, India (Grid: MK83SC) and decoded the same to obtain real time images as seen by the satellite. The images obtained were further analyzed for rough weather predictions over the above mentioned location.
Chapter
This chapter presents methods that allow us to analyse or modify the geometric properties of an image. Our attention, first, is on techniques for filtering images to remove noise or to enhance geometric detail. We will then look at means by which we can characterise geometric properties like texture, and processes that allow us to analyse objects and shapes that appear in imagery.
Chapter
Image analysis by photointerpretation is often made easier when the radiometric nature of an image is enhanced to improve its visual characteristics. Specific differences in vegetation and soil type, for example, may be brought out by increasing the contrast of an image. Highlighting subtle differences in brightness value by applying contrast modification, or by assigning different colours to different brightnesses in the method known as colour density slicing, will often reveal features not otherwise easily seen.
Article
This book gives comprehensive and balanced coverage of the principles of cognitive radio communications, cognitive networks, and details of their implementation, including the latest developments in the standards and spectrum policy. Case studies, end-of-chapter questions, and descriptions of various platforms and test beds, together with sample code, give hands-on knowledge of how cognitive radio systems can be implemented in practice. Extensive treatment is given to several standards, including IEEE 802.22 for TV White Spaces and IEEE SCC41. Written by leading people in the field, both at universities and major industrial research laboratories, this tutorial text gives communications engineers, R&D engineers, researchers, undergraduate and post graduate students a complete reference on the application of wireless communications and network theory for the design and implementation of cognitive radio systems and networks.
  • T Langdon
T. Langdon, A close up on Doppler Shift, The AMSAT Journal, 2010