Cotton Diseases And Pests Identification And Remote Sensing Monitoring

Cotton is the major cash crop in Xinjiang. The rapid and accurate monitoring over planting area, growth vigor and distribution of cotton in Xinjiang is of great importance for acquiring cotton pests and diseases, assessment of cotton yield and quality. HJ-1A/1B developed by China is rendered with comprehensive advantages and will expect huge application prospect in information extraction of crop planting areas owing to its large coverage area, short revisiting period, high spatial resolution, multispectral dimension, free data access and simple data processing. In this paper, Xinjiang cotton growing areas in China is taken as the research object. The HJ-1A/1B satellite data combined with the canopy level spectral data were extracted for the cotton area, and pests and diseases early identification and extraction. The main contents are as follows:(1) Study on identification of early cotton pests and diseases at the canopy level. Based on the spectral response characteristics of differences of different diseases and insects at different levels, the study selects vegetation index and wavelet features to identify the verticilliums(hwb) and cotton spider mites(mym). Then the discriminant models were constructed. The results show that the discriminant model based on vegetation index and wavelet feature can distinguish established pests type well, in which the precision of the model established by vegetation index is 97%,while and the precision of the model established by wavelet feature is 99%. Relatively speaking, the precision of the wavelet feature is better than vegetation index. The study chooses 6 kinds of vegetation index and 9 wavelet features from 10 level, which have the maximum correlation coefficient with sample category. The chosen vegetation index and wavelet features are taken as discriminant model input, which have high sensitivity to distinguish the two diseases and pests.(2) Study on multi-temporal HJ Satellite image-based cotton planting area and location information Method.According to HJ-CCD data image during cotton bud stage and boll opening stage in the 125 th Regiment of the 7th Agricultural Shi in Kuytun City, Xinjiang and based on The phenological and spectral difference between cotton and other crops in the study region and variation difference in NDVI values, SVM-based stratified supervised classification method and CART algorithm classification method are adopted respectively to extract cotton planting area and their extraction precision is compared from the perspective of total area and space position, etc. The results show that: The efficiency of area extraction can be greatly improved by making use of HJ images in critical stage. SVM-based stratified supervised classification method and CART algorithm classification method can both be used to effectively extract cotton planting area, in which the precision in extraction of total area and position precision of the former are 96.34% and 82.98% respectively, while those of the latter are 96.51% and 87.23%. By comparison and analysis, the complicated SVM-based stratified supervised classification method has high classification precision but poor position precision, while the simple CART algorithm classification method has both high classification precision and sound position classification precision.(3) Study on HJ satellite image-based cotton pests and diseases information extraction. The study use binary logistic regression to establish the relational model between measured spectrum of the canopy vegetation index(TVI) and health state of cotton, then the model were applied to HJ images. Based on an assumption of neighborhood consistency among3?3 pixels, the each pixel probability value is calculated to give a range of pests and diseases plots information. The results shows that 6 plots are classified correctly in the 10 plots pest verification points. Cotton diseases distribute in remote sensing images spotty, which is consistent with the actual situation. The extraction of cotton pests and diseases abnormal information with the satellite data has potential.