[Research on maize multispectral image accurate segmentation and chlorophyll index estimation].

In order to rapidly acquire maize growing information in the field, a non-destructive method of maize chlorophyll content index measurement was conducted based on multi-spectral imaging technique and imaging processing technology. The experiment was conducted at Yangling in Shaanxi province of China and the crop was Zheng-dan 958 planted in about 1 000 m X 600 m experiment field. Firstly, a 2-CCD multi-spectral image monitoring system was available to acquire the canopy images. The system was based on a dichroic prism, allowing precise separation of the visible (Blue (B), Green (G), Red (R): 400-700 nm) and near-infrared (NIR, 760-1 000 nm) band. The multispectral images were output as RGB and NIR images via the system vertically fixed to the ground with vertical distance of 2 m and angular field of 50°. SPAD index of each sample was'measured synchronously to show the chlorophyll content index. Secondly, after the image smoothing using adaptive smooth filtering algorithm, the NIR maize image was selected to segment the maize leaves from background, because there was a big difference showed in gray histogram between plant and soil background. The NIR image segmentation algorithm was conducted following steps of preliminary and accuracy segmentation: (1) The results of OTSU image segmentation method and the variable threshold algorithm were discussed. It was revealed that the latter was better one in corn plant and weed segmentation. As a result, the variable threshold algorithm based on local statistics was selected for the preliminary image segmentation. The expansion and corrosion were used to optimize the segmented image. (2) The region labeling algorithm was used to segment corn plants from soil and weed background with an accuracy of 95. 59 %. And then, the multi-spectral image of maize canopy was accurately segmented in R, G and B band separately. Thirdly, the image parameters were abstracted based on the segmented visible and NIR images. The average gray value of each channel was calculated including red (ARed), green (AGreen), blue (ABlue), and near-infrared (ANIR). Meanwhile, the vegetation indices (NDVI (normalized difference vegetation index), RVI (ratio vegetation index); and NDGI(normalized difference green index)) which are widely used in remote sensing were applied. The chlorophyll index detecting model based on partial least squares regression method (PLSR) was built with detecting R2=0. 5960 and predicting R2= 0. 568 5. It was feasible to diagnose chlorophyll index of maize based on multi-spectral images.

[Rapid Detection of Trace Dimethoate Pesticide Residues Based on Colorimetric Spectroscopy].

In order to detect dimethoate pesticide residues rapidly and safely, a feasible method based on colorimetric spectroscopy was developed. Because dimethoate is one of organophosphorus pesticides containing sulfur, its sulfenyl can react with Pd2+ to produce a yellow complex named palladium sulfide. PdCl2 was used as the color agent, which was dissolved in acetic acid instead of the common concentrated hydrochloric acid. The dimethoate solution was prepared by dissolving the commercial pesticides into distilled water at different concentrations. The pesticide samples were reacted with the same amount of PdC2 solution respectively. The absorbance spectra of the samples after coloring reaction were measured in the region of 300-900 nm by a spectrophotometer. The result showed that the effect of using acetic acid instead of concentrated hydrochloric acid was not only safe but also preferable, and 0.5 mg x kg(-1) was the minimum concentration of the pesticide that could be distinguished in the spectra. The result met the pesticide residue detecting requirements of part fruits and vegetables in the national standard GB2763-2012 regulations. Further studies on random 40 dimethoate samples from 0.5 to 88 mg x kg(-1) were carried out. Thirty samples were randomly selected to establish the training model and remaining 10 samples were used to test the model. The preprocessing methods were carried on the spectrum data such as normalization and smoothing to get a better effect through comparison their prediction results with the correlation coefficient (r) and the root mean square error of cross-validation (RMSEP). The principal component analysis (PCA) method and partial least squares (PLS) method were used to establish prediction models respectively in the different wave ranges. By calculating the correlation coefficient of dimethoate samples in 350-900 nm the maximum of 0.9572 was obtained at wavelength 458 nm, so 453-463 and 400-600 nm were selected as feather regions. Experiments showed that the effect of SG preprocessing on the absorbance spectra in the region of 350-900 and 400-600 nm was obvious, and PLS method were better than PCA method. The optimum model was obtained in the region of 400-600 nm, when principal component number was 4, the training set r=0.9941, RMSEP=2.7703 and the validation set r=0.9933, RMSEP = 2.2148. This method was safe in operation and the colorimetric reaction time was 2 min, which provided theoretical and technical support for further studying on development of rapid, safe organophosphorus pesticide detection instrument.

[Visible-NIR spectral feature of citrus greening disease].

Citrus greening (Huanglongbing, or HLB) is a devastating disease caused by Candidatus liberibacter which uses psyllids as vectors. It has no cure till now, and poses a huge threat to citrus industry around the world. In order to diagnose, assess and further control this disease, it is of great importance to first find a quick and effective way to detect it. Spectroscopy method, which was widely considered as a fast and nondestructive way, was adopted here to conduct a preliminary exploration of disease characteristics. In order to explore the spectral differences between the healthy and HLB infected leaves and canopies, this study measured the visible-NIR spectral reflectance of their leaves and canopies under lab and field conditions, respectively. The original spectral data were firstly preprocessed with smoothing (or moving average) and cluster average procedures, and then the first derivatives were also calculated to determine the red edge position (REP). In order to solve the multi-peak phenomenon problem, two interpolation methods (three-point Lagrangian interpolation and four-point linear extrapolation) were adopted to calculate the REP for each sample. The results showed that there were, obvious differences at the visible & NIR spectral reflectance between the healthy and HLB infected classes. Comparing with the healthy reflectance, the HLB reflectance was higher at the visible bands because of the yellowish symptoms on the infected leaves, and lower at NIR bands because the disease blocked water transportation to leaves. But the feature at NIR bands was easily affected by environmental factors such as light, background, etc. The REP was also a potential indicator to distinguish those two classes. The average REP was slowly moving toward red bands while the infection level was getting higher. The gap of the average REPs between the healthy and HLB classes reached to a maximum of 20 nm. Even in the dataset with relatively lower variation, the classification accuracy of threshold segmentation method by the REP could reach to more than 90%. The four-point linear extrapolation method had slightly better result than the three-point Lagrangian interpolation method. This study provided useful theoretical foundation to detect HLB by spectral reflectance.

[A novel vegetation index (MPRI) of corn canopy by vehicle-borne dynamic prediction].

Ground-based remote sensing system is a significant way to understand the growth of corn and provide accurate and scientific data for precision agriculture. The vehicle-borne system is one of the most important tools for corn canopy monitoring. However, the vehicle-borne growth monitoring system cannot maintain steady operations due to the row spacing of corn. The reflectance of corn canopy, which was used to construct the model for the chlorophyll content, was disturbed by the reflectance of soil background. The background interference with the reflectance could not be removed effectively, which would result in a deviation in the growth monitoring. In order to overcome this problem, a novel vegetation index named MPRI was developed in the present paper. The tests were carried out by the vehicle-borne system on the cornfield. The sensors which configured the vehicle-borne system had 4 bands, being respectively 550, 650, 766 and 850 nm. It would obtain the spectral data while the vehicle moved along the row direction. The sampling rate was about 1 point per second. The GPS receiver obtained the location information at the same rate. MPRI was made up by the reflectance ratio of 660 and 550 nm. It was very effective to analyze the information about the reflectance of the canopy. The results of experiments showed that the MPRI of soil was the positive value and the MPRI of canopy was the negative value. So it is easier to distinguish the spectral information about soil and corn canopy by MPRI. The results indicated that: it had satisfactory forecasting accuracy for the chlorophyll content by using the MPRI on the moving monitoring. The R2 of the prediction model was about 0.72. The R2 Of the model of NDVI, which was used to represent the chlorophyll content, was only 0.24. It indicates that MPRI had good measurement results for the dynamic measurement process. It provided the novel measurement way to get the canopy reflectance spectra and the better vegetation index to construct the prediction model of the contents of chlorophyll.

[Detection of chlorophyll content in water body based on two-dimensional correlation spectroscopy].

Twenty five samples were collected from 10 different ponds in Jiangsu Province of China. According to the different water status and surface area of each pond, different numbers of water samples were collected. The present paper aims to detect chlorophyll content in water body based on hyperspectrum. The visible and near infrared spectral transmittance of the water samples was measured by using a Shimadzu UV-2450 spectrograph. At the same time, the chlorophyll content of each sample was measured using hot-ethanol extraction method in the laboratory. Then the spectral characteristics were analyzed for the water samples and the results showed that with chlorophyll concentration increasing, spectral transmittance decreased gradually. There is an apparent transmission valley at 676 nm. And then two dimensional correlation spectrum technology was used to analyze the sensitive absorption band of chlorophyll in water body. Comprehensive observation of the spectral characteristics of water samples can be carried out much accurately by analyzing two-dimensional correlation spectra of synchronous and asynchronous spectrograms. And the effective spectral response bands of the chlorophyll content were found at 488 and 676 nm. Then the NDWCI (normalized difference water chlorophyll index) was established with the transmittance of red band and blue band. Two regression models were built to predict the chlorophyll concentration in water. One is a multiple linear regression model based on the original transmittances at 488 and 676 nm. The other is the linear regression model based on NDWCI. By comparison, the model based on NDWCI was better. The R2 of its training model reached to 0.7712, and the root mean square error of calibration was 45.5099 mg x L(-1). The R2 of prediction model reached to 0.7658, and the root mean square error of prediction was 39.5038 mg x L(-1). It reached to a practical level to predict the chlorophyll content in water body rapidly.

[Detection of onion soluble solids content based on the near-infrared reflectance spectra].

Onion soluble solids content (SSC) was detected using near-infrared (924-1720 nm) reflectance spectra. Three cultivars of onions, harvested at different period, were selected for experiment and the total number of samples is 268. SSC reference value of onion juice was determined using the temperature compensated refractometer. Some pre-processing methods, such as S-G smoothing, scatter correction, and derivation, were compared to establish a statistical model based on partial least squares regression (PLSR) method. The results show that the avitzky-Golay smoothing with window 32 and span 10 is more efficient. The determination correlation coefficient of prediction R2 is 0.87 and root mean square error (RMSEP) is 2.42 degrees Brix. Compared to the 2nd derivation, the 1st derivation got better prediction result, but the spectra scatter correction is the best (R2 = 0.88, RMSEP of = 2.31 degrees Brix). The optimal prediction (R2 = 0.90, RMSEP = 1.84 degrees Brix and RPD = 3) was built based on crossing validation modeling, which shows that infrared reflectance spectroscopy with scatter correction pre-processing is feasible for onions soluble solids detection.

[Effect of soil moisture on prediction of soil total nitrogen using NIR spectroscopy].

As one of the most important components of soil liutrient, it is necessary to obtain the soil total nitrogen(STN)content in precision agriculture. It is a feasible method to predict soil total nitrogen content based on NIRS. However, the effect of soil moisture content (SMC) on the prediction of STN is very serious. In the present research, the effect of SMC was discussed from qualitative analysis and quantitative analysis by the Fourier spectrum analyzer MATRIX_I. Firstly, sixty soil samples with different STN and SMC were scanned by the MATRIX_I. It was found that the reflectince of soil samples in near infrared region decreased with the increase in SMC. Subsequently, Moisture absorbance index (MAI) was proposed by the diffuse of absorbance at the wavelengths of 1 450 and 1 940 nm to classify soil properties and then correction factor was present Finally, the STN forecasting model with BP NN method was established by the revised absorbance data at the six wavelengths of 940, 1 050, 1,100, 1,200, 1,300 and 1,550 nm. The model was evaluated by correlation coefficient of Rc, correlation coefficient of Rv, root mean square error of calibration (RMSEC), root mean square error of validation (RMSEP) and residual prediction deviation (RPD). Compared with the model obtained from original spectral data, both the accuracy and the stability were improved. The new model was with Rc of 0.86, Rv of 0.81, RMSEC of 0.06, RMSEP of 0.05, and RPD of 2.75. With the first derivative of the revised absorbance, the RPD became 2.90. The experiments indicated that the method could eliminate the effect of SMC on the prediction of STN efficiently.

[Grey analysis of NIR spectra and prediction of nitrogen content in jujube leaves].

Jujube was chosen as the object in the present research. Spectra data of jujube leaves were collected during the period of budding, branch leaf, flowering and coloring. The nitrogen contents of jujube leaf samples were determined by Kjeldahl analysis method. Grey relation analysis between spectral reflectance and nitrogen content of jujube leaves was done based on Grey theory. It was found that the gray relation between spectral reflectance and nitrogen content of jujube leaves at 560, 678 and 786 nm was high. Nine kinds of vegetation index based on spectra data of NIR786, R678 and G570 were calculated. The gray relation of nine kinds of vegetation index was NDVI>GRVI>NDGI>GNDVI>CNDVI>RVI>GDVI>DVI>SAVI. NDVI, GRVI, NDGI, GNDVI and CNDVI were chosen to build prediction models of nitrogen content of jujube leaves. Spectra data of 560, 678 and 786 nm were also used to build prediction models of nitrogen content of jujube leaves. LS-SVM and GM(1, N) were used to build prediction module. The prediction R2 and verification R2 of LS-SVM module were 0.805 and 0.704 respectively when five kinds of vegetation index were used as input of prediction module. When when Spectra data of 560, 678 and 786 nm were used as input, the prediction R2 and verification R2 of LS-SVM prediction model were 0.772 and 0.685 respectively. The prediction R2 and verification R2 of GM(1, N) module were 0.927 7 and 0.895 8 respectively when spectra data of 560, 678 and 786 nm were used as input. The results of prediction GM(1, N) module which used five kinds of vegetation index as input were 0.547 6 and 0.489 7. From those results it was observed that grey theory only needed little information to build prediction module with high precision, so that it could be used in precision management of jujube plants.

[Prediction of chlorophyll content of greenhouse tomato using wavelet transform combined with NIR spectra].

In quantitative analysis of spectral data, noises and background interference always degrade the accuracy of spectral feature extraction. The wavelet transform is multi-scale decomposition used to reduce the noise and improve the analysis precision. On the other hand, the wavelet transform denoising is often followed by destroying the efficiency information. The present research introduced two indexes to control the scale of decomposition, the smoothness index (SI) and the time shift index (TSI). When the parameters satisfied TSI < 0.01 and SI > 0.100 4, the noise of spectral characteristic was reduced. In the meanwhile, the reflection peaks of biochemical components were reserved. Through analyzing the correlation between denoised spectrum and chlorophyll content, some spectral characteristics parameters reflecting the changing tendency of chlorophyll content were chosen. Finally, the partial least squares regression (PLSR) was used to develop the prediction model of the chlorophyll content of tomato leaf. The result showed that the predictiong model, which used the values of absorbance at 366, 405, 436, 554, 675 and 693 nm as input variables, had higher predictive ability (calibration coefficient was 0. 892 6, and validation coefficient was 0.829 7) and better potential to diagnose tomato growth in greenhouse.

[Detection of corn chlorophyll content using canopy spectral reflectance].

The canopy spectral reflectance and chlorophyll content of corn were measured and analyzed under different nitrogen treatments. The correlation between spectral reflectance and chlorophyll content was discussed based on different growth stages and different nitrogen levels. The results showed positive correlations under high and normal nitrogen treatment, while negative correlation under low nitrogen treatment. The relation between reflectance of normal fertilizer region and chlorophyll content was better than others, with r(Normal) > r(High) > r(Low). Normal fertilizer was the best condition to detect the corn chlorophyll content using spectral reflectance. Analysis of the relations at different growth stages showed that on the band of 400-1000 nm the absolute value of correlation coefficient increased and reached the maximum at shooting stage, it decreased until anthesis-silking stage, and then rebounded at milking stage. The positive correlations were found at shooting and milking stage, while negative correlations were found at seedling, trumpet and anthesis-silking stage. It was indicated that the sensitive stages to detect the chlorophyll content were shooting and trumpet stage with high absolute value of correlation coefficient above 0.6 around 550 nm. In order to detect the chlorophyll content of corn, 558, 667, 714 and 912 nm were selected to establish the MLR model and PLSR model. The results showed that PLSR was more capable of building chlorophyll content models reflecting correct relations among multi-variables compared with MLR. In the meanwhile, three wavelengths were selected (558, 667 and 714 nm) to build different vegetation indices such as GDVI, GRVI, GNDVI, DVI, RVI and NDVI. The correlation between DVI and chlorophyll con tent was better than others and DVI was used to establish binomial model and exponential model at shooting stage (R2 = 0.80) and trumpet stage (R2 = 0.66) respectively which was higher than PLSR It also provided a feasible method to detect chlorophyll content non-destructively.

[Monitoring of Cnaphalocrocis medinalis Guenee based on canopy reflectance].

The canopy reflectance of rice was measured in the filed in order to monitor the damaged region caused by Cnaphalocrocis medinalis Guenee. The characteristics of canopy spectral reflectance were analyzed in contrast region and damaged regions. When rice plant was damaged by Cnaphalocrocis medinalis Guenee, the chlorophyll absorption was decreased in the band of 600-700 nm. The canopy reflectance of moderate damage region was lower than that of the contrast region, while the reflectance of severe damage region rice was higher near 550 nm. The canopy reflectance of Cnaphalocrocis medinalis Guenee damaged rice was fluctuant and exhibited the significant peak in the NIR band of 750-770nm. Meanwhile, red edge inflection point as one of the most important spectral parameters was analyzed at different damage levels based on the first derivative of reflectance spectra. The analysis results indicated that red edge inflection position moved to direction of blue light (short wavelength) with the affection severity increasing. Then the modified reflectance of rice canopy was calculated based on zero-mean calculation and standard deviation. It was easy to find the degree of deviation from the average of samples and distinguish the damaged region from experiment plots. The canopy modified reflectance was gently in the contrast region, but changed violently in the affected regions in the band of 750-950 nm. The analysis of Cnaphalocrocis medinalis Guenee affected regions illustrated that the Cnaphalocrocis medinalis Guenee was increased with the increase in severity. The vegetation index was applied in detection of Cnaphalocrocis medinalis Guenee damaged regions because of the composition of multi-wavelength information. The wavelengths 762 and 774 nm were chosen to build detection parameters of Cnaphalocrocis medinalis Guenee such as NIR-RVI, NIR-DVI, NIR-NDVI and KI. The results indicated that the NIR-NDVI could be used to identify the damaged region with contrast region efficiently. The accurate rate of 25 verification samples selected randomly reached 70%. The preliminary studies on rice Cnaphalocrocis medinalis Guenee damaged regions provided a new method to detect the affected regions in the wide area.

[Development of a portable SOM detector based on NIR diffuse reflection].

A portable soil organic matter detector based on near infrared diffuse reflectance was developed. The detector uses a microprocessor 89S52 as the micro controller unit (MCU) and consists of an optical system and a control system. The optical system includes an 850 nm near-infrared LED lamp-house, a lamp-house driving-circuit, a Y type optical fiber, a probe, and a photoelectric sensor. The control system includes an amplifying circuit, an A/D circuit, a display circuit with LCD, and a storage circuit with USB interface. Firstly the single waveband optical signal from the near-infrared LED is transferred to the surface of the target soil via the incidence fibers. Then the reflected optical signal is collected and transferred to the photoelectric sensor, where the optical signal is converted to the electrical signal. Subsequently, the obtained electrical signal is processed by 89S52 MCU. Finally, the calculated soil organic matter content is displayed on the LCD and stored in the USB disk. The calibration experiments using the estimation model of the soil organic matter were conducted. Thirteen kinds of natural soil samples were prepared, each divided into five sub-samples. After measurement, the natural samples were dried under the condition of 105 degrees C for 24 h, and then the same measurements were performed. The analysis of the correlation between the detected SOM content and the measured reflectance was carried out. For the natural soil samples, R2 = 0.907, while R2 reached 0.963 for the dried soil samples. The average reflectance of the five sub-samples from the same kind soil was calculated for each kind of soil. And then the same correlation analysis was conducted, for the natural samples R2 = 0.950, and for the dried samples R2 = 0.982. The results showed that the developed detector is practical. And the soil moisture has an effect on the accuracy of the detector. It is necessary to correct the real time measurement result of the detector based on soil moisture.

[Spectral characteristics of corn under different nitrogen treatments].

The canopy spectral reflectance and chlorophyll content of corn were measured and analyzed under different nitrogen treatments. The trends of chlorophyll content were discussed based on different growth stages and different nitrogen levels. It was observed that the chlorophyll content increased with the increase in nitrogen, and could be affected by the environment changes including the temperature, rain, fertilizer treatment and so on. The characteristics of canopy spectral reflectance indicated that the canopy spectral reflectance changed significantly at different stages. In the visible region (400-750 nm), the reflectance increased and reached the maximum until the shooting stage, and decreased subsequently with the growth progress. In near-infrared region (750-1 000 nm), the spectral reflectance climbed sharply. It increased from tillering stage to shooting stage first, and then began to decline at trumpet stage and was raised again at anthesis-silking stage. At milking stage, the reflectance was decreased again. There were clear distinctions of visible reflectance in different nitrogen regions. At shooting stage, with the increase in nitrogen the reflectance decreased at chlorophyll absorption band (430-450 nm, 640-660 nm). Investigating the reflectance of the corn canopy under the different nitrogen treatment, it was found that the reflectance was higher in normal fertilizing region than others in 550 nm, with R(Normal) > R(Low) > R(High). At trumpet stage, the canopy reflectance in low fertilizing region was higher than others in the visible region. It was clear that the corn canopy reflectance of normal fertilizing region was the same as high fertilizing region. The results indicated over fertilizing could not help increase the corn nitrogen uptake. The study provided the basic information of chlorophyll measurement based on spectral technology and could help to guide the precision fertilizer in the field.

[The spectral characteristics and chlorophyll content at winter wheat growth stages].

The canopy spectral reflectance and chlorophyll content of winter wheat were measured and analyzed in the whole growing season. The characteristics of canopy spectral reflectance indicated that the canopy spectral reflectance changed signifi cantly at different stages. It decreased first and increased later with growth progress in the visible region (400-750 nm) after jointing stage, then the reflectance was lowest at booting stage. In near-infrared region (750-1000 nm) the spectral reflectance climbed sharply. The canopy reflectance was declined at booting stage and rose to the highest point at flowering stage. It dropped to minimum level subsequently at milk-ripe stage. However, the spectral reflectance characteristics at jointing stage and booting stage were used to detect the chlorophyll content. High correlation was observed between the canopy spectral reflectance and chlorophyll content. The positive correlation of canopy spectral reflectance with chlorophyll content was found at jointing stage and booting stage. Because the plant spectral reflectance was effected by chlorophyll greatly in visible region, the correlation coefficient reached 0.89 at 552 nm in booting stage. On the contrary, there was a negative correlation at flowering stage. Meanwhile, the red edge inflection point as one of the most important spectral parameters was analyzed at winter wheat growth stages based on the first derivative of reflectance spectra. The relation between the red edge inflection point and chlorophyll content was observed in each plot and the analysis results illustrated that the red edge inflection points could help detect the chlorophyll content at jointing stage and booting stage. The linear model of chlorophyll content with red edge inflection points was built at jointing stage (R2 = 0.92). High correlation was found between thered edge inflection point and chlorophyll content at booting stage. It was showed that the binomial model (R2 = 0.91) was better than linear model (R2 = 0.73). It was indicated that using spectral analysis to detect the winter wheat chlorophyll content non-destructively was feasible. The obtained conclusions also provided theoretical basis for further researches on measuring methods of chlorophyll content in the field.

[Nitrogen content testing and diagnosing of cucumber leaves based on multispectral imagines].

Using CCD camera and special filters, the growth parameters of cucumber plants, the nitrogen content and the area of leaves, were investigated in an experimental greenhouse. In order to make nutrient stress to the plants, different nitrogen levels were prepared. The basic nitrogen content was 0.067 kg x L(-1) and four different levels of nitrogen contents were made to be 2, 1, 0.5, and 0.2 N, respectively. The genetic and water-segment methods were used to separate IR and R2 images from the background. It was found that the result of water-segment is better. It has clearer boundary, less noise and closer result to the original image. After the reflectance information of cucumber leaves was obtained, the vegetation indexes, RVI, NDVI and GNDVI, were calculated and then the correlation coefficients between each vegetation index and nitrogen content or leave area were analyzed. The result shows that there is an obvious linear correlation between NDVI and nitrogen content of leaves or leave area and the R2 are 0.8209 and 0.7017, respectively. The high correlations were also observed between GNDVI and nitrogen content of leaves or leave area, and the R2 are 0.7625 and 0.6762, respectively. The reason is that the reflectance of IR is mainly affected by reflectivity and the canopy structure of cucumber leaves. As biomass and area of leaves increase with the nitrogen content, the reflectivity of leaves becomes stronger. And the gap among cells of high nitrogen content leaves is large. Cell wall has more water, which has a strong effect on the reflectivity of NIR At visible wavelength, the reflectance of cucumber leaves decreases as nitrogen content increases since the chlorophyll content increases as nitrogen content increases. The trend of correlation between RVI and nitrogen content disagreed with that of the correlation between RVI and leave area. There is an obvious linear correlation between RVI and leave area, and the R2 is 0.8577. However, the correlations between RVI and nitrogen content exhibit a nonlinear relationship, and R2 is only 0.5988. It is because as cucumbers grow older, the reflectance of canopy increases at visible wavelength but decreases at near infrared wavelength. The experimental result proves that CCD camera and special filters can be used as a good method for diagnosing nitrogen content of cucumber plants.

[The canopy and leaf spectral characteristics and nutrition diagnosis of tomato in greenhouse].

A green house experiment was conducted to research the characteristics of tomato canopy spectral reflectance and leaf spectral reflectance under different nutrition treatments, and the relationships between spectral reflectance and the water content, chlorophyll content, as well as nitrogen content were analyzed. Substrate cultivation method was used to grow the plants. The substrate was made from a mixture of peat and vermiculite. Test area was prepared for four levels of nutrition to form nutritional stress. There were 12 seedlings under each nutritional condition and a total of 48 seedlings were planted for the experiment. The canopy reflectance and leaf reflectance were measured by an ASD handheld spectroradiometer and a FT-NIR spectrometer respectively. It was observed that the trend of tomato canopy reflectance was similar to each others. There was a reflection peak at about 550 nm, and the reflectance in the visible light region was lower than that in near-infrared region. The results of analysis also indicated that under different nutrient conditions, canopy spectral reflectance characteristics of tomato took on disciplinary change. At near-infrared bands, the reflectance gradually increased with adding nutrition, while reduced at visible light bands. The leaf spectral reflectance characteristics at near-infrared bands had the similar change with the canopy reflectance. There were four sensitive wavelengths of water at near-infrared bands: about 980, 1450, 1 930, and 2 210 nm, and the results of single linear regression (SLR) and multi-linear regression (MLR) indicated that the reflectance at these sensitive wavelengths could be used to estimate the water content in tomato leaves. R2 were 0.5903 and 0.7437 respectively. NDCI as one of the most important spectral parameter was calculated by the spectral reflectance of 530 and 760 nm, and the result indicated that there existed a good correlation between NDCI and the nitrogen content, with R2 = 0.7511. Meanwhile, red edge inflection points were analyzed under four nutrition treatments based on the first derivative of canopy spectral reflectance. The analysis results illustrated that red edge inflection position moved to direction of red light (long wavelength) with the nutrition supply.

[Development of an analyzing system for soil parameters based on NIR spectroscopy].

A rapid estimation system for soil parameters based on spectral analysis was developed by using object-oriented (OO) technology. A class of SOIL was designed. The instance of the SOIL class is the object of the soil samples with the particular type, specific physical properties and spectral characteristics. Through extracting the effective information from the modeling spectral data of soil object, a map model was established between the soil parameters and its spectral data, while it was possible to save the mapping model parameters in the database of the model. When forecasting the content of any soil parameter, the corresponding prediction model of this parameter can be selected with the same soil type and the similar soil physical properties of objects. And after the object of target soil samples was carried into the prediction model and processed by the system, the accurate forecasting content of the target soil samples could be obtained. The system includes modules such as file operations, spectra pretreatment, sample analysis, calibrating and validating, and samples content forecasting. The system was designed to run out of equipment. The parameters and spectral data files (*.xls) of the known soil samples can be input into the system. Due to various data pretreatment being selected according to the concrete conditions, the results of predicting content will appear in the terminal and the forecasting model can be stored in the model database. The system reads the predicting models and their parameters are saved in the model database from the module interface, and then the data of the tested samples are transferred into the selected model. Finally the content of soil parameters can be predicted by the developed system. The system was programmed with Visual C++6.0 and Matlab 7.0. And the Access XP was used to create and manage the model database.

[Application of wavelet packet analysis in estimating soil parameters based on NIR spectra].

Using the method of wavelet analysis, the NIR spectra of soil samples were decomposed and reconstructed, and higher precision PLS models were established to estimate soil parameter (TN, SOM). One hundred fifty soil samples were collected from a winter wheat field and the NIR spectra of all samples were measured. Firstly, experiment statistic features were analyzed aiming at all soil samples, and the system clustering was carried out for TN and SOM respectively. Then 50 new TN samples and their corresponding spectra, and 50 new SOM samples and their corresponding spectra were obtained. Secondly, the PLS models were established with these new samples based on their corresponding spectra. The models showed a certain amount of accuracy, but it was still not practical. Therefore, wavelet analysis of NIR spectra was tried. The wavelet packet decomposing by eight-level biorthogonal algorithm was carried out, and 256 nodes were gotten. The lowest approximation signal is corresponding to soil moisture and soil texture spectrum trend. The maximal detail signal is corresponding to the high-frequency turbulence caused by the soil particle size, precision of spectrometer, and other uncertainties. After reconstructing these two nodes and then removed from the original spectra, the characteristic spectra corresponding to each soil parameter were acquired. Finally, the PLS models were established for TN and SOM content respectively: for TN content, the calibration coefficient of the PLS model is 0.960, the validation coefficient is 0.920; and for SOM content, the calibration coefficient of the PLS model is 0.922, and the validation coefficient is 0.883. It was showed that the accuracy of each model was highly improved and the models were able to meet the needs of actual production. The research results conclude that wavelet analysis can eliminate or substantially reduce the factors outside the parameters. It can also remove the obstacles in establishing linear models of soil parameters, and it is feasible and potential method for the real-time estimation of soil parameters.

[Characteristics of soil spectral reflectance and estimation of soil parameters in Fuxin opencast coal mine].

The surface soil samples of Fuxin opencast coal overburden dumps were collected in the field and the spectral reflectance and characteristic parameters of the soil samples, such as moisture content, pH, electrical conductivity (EC), available potassium (K+) content, and soil organic matter content (SOM), were measured. The Analysis results indicated that the curves of the soil spectral reflectance decreased with increasing the laid years. The possible reasons were the influences of soil texture and color. Although the valleys of the spectral reflectance appeared at 1 420, 1 910 and 2 210 nm, they were not conspicuous on the lime soil and mixture soil reflectance curves at 1 420 and 2 210 nm. With discussing the spectral reflectance of different types of soil texture, it's easy to find that the reflectance of fine grained soil was higher than the rough grained soil. Correlations between soil spectral reflectance and soil parameters were analyzed. The results showed that there was a positive relation between reflectance and pH, and correlation coefficient decreased with the wavelength increasing. There was no relationship between spectral reflectance and EC, and negative relations were observed between spectral reflectance and soil parameters, K+ and SOM, respectively. A high correlation coefficient was found between spectral reflectance and SOM, and the highest correlation coefficient reached -0. 76. The exponential correlation was found between sol spectral reflectance and soil moisture content to analyze all samples. According to different years and textures, more detail was described about the correlation between spectral reflectance of characteristic wavelength (1 910 and 1 943 nm) and soil moisture content. Meanwhile, the linear correlations were found under different conditions and higher correlation coefficients were obtained. In order to estimate SOM, five wavelengths (1 350, 1 602, 1 862, 2 160 and 2 227 nm) were selected based on principal component analysis to build a multiple linear regression model. The multiple correlation coefficient of calibration model (R2(C)) was 0.737 4, and the multiple correlation coefficient of validation (R2(V)) was 0.682 4. It indicated that the model was able to meet the needs of monitoring SOM in Fuxin opencast coal mine.

[Estimation of soil organic matter and soil total nitrogen based on NIR spectroscopy and BP neural network].

Estimation models of soil organic matter (SOM) and soil total nitrogen (TN) were established based on NIR spectroscopy and BP neural network A total of 150 soil samples were collected from the tested farm, and the NIR spectra of all soil samples were measured. First, data pretreatment was performed for each sample with the method of locally weighted scatter plot smooth filtering. Then the box plot analysis for the measured SOM data and TN data were conducted separately and the information about the shape, location, and distribution of the target data was obtained. The variance between the SOM data was very small, and most of them were concentrated on the median. This was also observed from TN data. Thus clustering analysis was carried out for the target parameters of the soil samples so that the original dataset with 150 spectra was clustered to 50 groups. For each group, the average of spectral data was calculated at every wavelength to obtain a new spectrum. The new spectrum was calculated with natural logarithm and normalized, which was taken as a new sample. Principal component analysis (PCA) was executed for 50 new samples and the principal components with over 99.98% of cumulative proportion of correlation matrix were extracted to establish BP neural network. According to the analysis result of SOM content, the calibration accuracy of the model was 0.999, and the validation accuracy reached to 0.854. According to the analysis result of the soil TN content, the calibration accuracy of the model was close to 1, and the validation accuracy reached 0.808. The result shows that the smooth filter can weaken the noise in the data, expose the data features, provide a reasonable starting approach for parametric fitting; and improve the prediction accuracy; It is feasible and practical to estimate soil parameters by using BP neural network with the prediction accuracy of 0.854 (SOM) and 0.808 (TN); Compared to the other prediction modeling method, the BP neural network model has higher robustness and better fault tolerance, and the model accuracy would not be affected by the several outline samples when the number of samples is large enough.