Huirong Xu’s research while affiliated with Zhejiang University and other places

By using imaging techniques, plant physiological parameters can be assessed without contact with the plant and in a non-destructive way. During plant-pathogen infection, the physiological state of the infected tissue is altered, such as changes in photosynthesis, transpiration, stomatal conductance, accumulation of Salicylic acid (SA) and even cell death. In this study, the different temperature distribution between the leaves infected by tobacco mosaic virus strain-TMV-U1 and the noninfected leaves was visualized by digital infrared thermal imaging with the microscopic observations of the different structure within different species tomatoes. Results show a presymptomatic decrease in leaf temperature about 0.5-1.3 °C lower than the healthy leaves. The temperature difference allowed the discrimination between the infected and healthy leaves before the appearance of visible necrosis on leaves.

Development of nondestructive measurements of soluble solids and firmness, which are two important ripeness and quality attributes of fruits, benefits the producers, processors and packers. The objective of this research was to evaluate the use of near-infrared (NIR) spectroscopy in detecting soluble solid content (SSC) and firmness for pears of three cultivars 'Cuiguan', 'Xueqing' and 'Xizilv' (n=160 of each cultivar). Relationships between nondestructive NIR spectral measurements and firmness and SSC of pear fruits were established by partial least square regression (PLSR) method. Models were developed for each cultivar, every two cultivars, and for all three cultivars in the spectral range of 800-2500 nm. The results of the models for all three cultivars turned out the best. For SSC assessment: correlation coefficients of calibration (rcal), root mean standard errors of calibration (RMSEC) and root mean standard errors of prediction (RMSEP) were 0.93, 0.35 °Brix and 0.50 °Brix for all three cultivars, respectively. For firmness assessment: rcal, RMSEC and RMSEP were0.92, 2.29 N, 2.95 N for all three cultivars, respectively. The results indicate that NIR spectroscopy can be used for predicting SSC and firmness of pear fruit and are the basis for the development of NIR analyzer suitable for on line application.

The use of Fourier transform near infrared spectrometry (FT-NIR) was explored as a tool to discriminate the pear samples of three different varieties (‘Xueqing’, ‘Cuiguan’ and ‘Xizilv’) which were mainly planted in Zhejiang province, China. Discriminant models were developed using discriminant principal component regression (PCR) and partial least squares (PLS) regression, and discriminant analysis (DA) in three NIR spectral regions, respectively. 240 samples (80 samples of each variety) were separated into two sets: half for calibration and the remaining half for validation. Both PLS and DA calibration models developed in the region of 800-2500 nm gave high accuracy of classification for the calibration set, 100% samples were correctly classified. For validation, ‘Cuiguan’ pears were classified correctly in 100% of cases, ‘Xueqing’ and ‘Xizilv’ were both 97.5% correctly classified (one sample of each variety misclassified) based on PLS discriminant mode. The validation result based on DA model was still 100% correctly classified for each variety. The results of this study indicated that FT-NIR had great potential for discriminating different varieties of fruits without detection depending on chemical components.

The objectives of this study were to characterize leaf reflectance spectra of tomato leaves damaged by leaf miner and to determine those leaf reflectance wavelengths that were most responsive to plant damage caused by the pest. Near infrared (NIR) Spectral characteristics of single tomato leaves at various levels of infestation by the leaf miner, were measured and analyzed using a spectrometer. Tomato leaf damage was classified into five scales, i.e., 0 (no damage), 1 (light damaged), 2 (10-25% damaged), 3 (more than 25% damaged), and 4 (severe damaged), based on the scale of infestation displayed on the surfaces of plant parts. Spectral parameter such as reflectance sensitivity was used to find the optimal wavelengths to determining and evaluating the damage level. Results showed that there were significant differences in reflectance among infestations at wavelengths of 1450nm and 1900 nm particularly. The determining coefficients (R2) for a linear relationship were 0.98 and 0.91 for the spectral-infestation levels relations. Thus, both of these wavelengths were good indicators of leaf senescence caused by the leaf miner.

Genetic algorithms (GAs) are used to implement an automated wavelength selection procedure for use in building multivariate calibration models based on partial least squares regression. The GAs also allows the number of latent variables used in constructing the calibration models to be optimized along with the selection of the wavelengths. This method was applied to fundamental study of non-destructive measurement of intact fruit quality with Fourier transform near infrared spectroscopy (FT-NIR). The experiments tested in this method are sugar content, titratable acidity and valid acidity. The optimal configurations for the GAs were investigated for each data set through experimental design techniques. Despite the complexity of the spectral data, the GA procedure was found to perform well (RMSEP=0.395, 0.0195, 0.0087 for SC, TA and pH respectively), leading to calibration models that significantly outperform those based on full spectrum analyses (RMSEP=0.512, 0.0198, 0.0111for SC, TA and pH respectively). In addition, a significant reduction in the number of spectral points required to build the models is realized and all of the numbers of wavelengths for building the models can reduce by 84.4%. It is instructive for the further study of the theory of non-destructive measurement of the fruit internal quality with FT-NIR spectroscopy.

Nondestructive method of measuring soluble solids content (SSC) of kiwifruit was developed by Fourier transform near infrared (FT-NIR) reflectance and fiber optics. Also, the models describing the relationship between SSC and the NIR spectra of the fruit were developed and evaluated. To develop the models several different NIR reflectance spectra were acquired for each fruit from a commercial supermarket. Different spectra correction algorithms (standard normal variate (SNV), multiplicative signal correction (MSC)) were used in this work. The relationship between laboratory SSC and FT-NIR spectra of kiwifruits were analyzed via principle component regression (PCR) and partial least squares (PLS) regression method using TQ 6.2.1 quantitative software (Thermo Nicolet Co., USA). Models based on the different spectral ranges were compared in this research. The first derivative and second derivative were applied to all measured spectra to reduce the effects of sample size, light scattering, noise of instrument, etc. Different baseline correction methods were applied to improve the spectral data quality. Among them the second derivative method after baseline correction produced best noise removing capability and to obtain optimal calibration models. Total 480 NIR spectra were acquired from 120 kiwifruits and 90 samples were used to develop the calibration model, the rest samples were used to validate the model. Developed PLS model, which describes the relationship between SSC and NIR spectra, could predict SSC of 84 unknown samples with correlation coefficient of 0.9828 and SEP of 0.679 Brix.

Water and chlorophyll content were analyzed in tomato leaves by near infrared (NIR) spectroscopy without any previous sample treatment. A total of 120 leaves were collected as experimental materials, a set of 80 samples was used to calibrate the instrument by partial leastsquares regression. In order to get a best model, four different mathematical treatments were used in spectrums processing: offset correction, smoothing, first and second derivative. Different preprocessing of spectra gives different performance of the prediction model; the original spectrums with smoothing treated spectra give the lowest RMSEP value and highest correlation coefficients value ,the best model of chlorophyll content has a root mean square error of prediction (RMSEP) of 8.92 and a calibration correlation coefficient value of 0.96297 and the best model of water content has a root mean square error of prediction (RMSEP) of 2.36 and a calibration correlation coefficient value of 0.99264.

An algorithm for the automatic recognition of citrus fruit on the tree was developed. Citrus fruits have different color with leaves and branches portions. Fifty-three color images with natural citrus-grove scenes were digitized and analyzed for red, green, and blue (RGB) color content. The color characteristics of target surfaces (fruits, leaves, or branches) were extracted using the range of interest (ROI) tool. Several types of contrast color indices were designed and tested. In this study, the fruit image was enhanced using the (R-B) contrast color index because results show that the fruit have the highest color difference among the objects in the image. A dynamic threshold function was derived from this color model and used to distinguish citrus fruit from background. The results show that the algorithm worked well under frontlighting or backlighting condition. However, there are misclassifications when the fruit or the background is under a brighter sunlight.

To identify fruits on the tree and determine their locations are the key to harvest fruits by robots. The main features and applications of infrared thermal imaging were reviewed, and main methods to locate fruits on trees were compared. As the low identification rate of common machine vision system, a new method to identify the citrus in a tree canopy by means of infrared thermal imaging was put forward. About 45 infrared thermal images of citrus on trees were acquired from the citrus orchard. It was found that the different thermal distribution among citrus, leaves and branches was about 1°C and these differences clearly appeared in the gray-level image, which could be easily used to segment the citrus from other parts in the image by using binary image at T=190. A multilayer-masks edge operator was used to extract edge of the image. The results indicated that it was possible to identify citrus on trees using infrared thermal imaging, and it was much easier than the methods presently used.