Research On Corn Borer Detection Method Based On Hyperspectral Imaging Technology

Agricultural pest detection on the quality and quantity of agricultural products plays a crucial impact. Agricultural Pest information is relatively small, more subtle, sometimes the human eye can not recognize discovered therefore, Thus, consistent with agricultural pests rapid non-destructive testing is a difficult and hot, but less at home and abroad to study agricultural pest detection. The purpose of this study is the use of hyperspectral imaging technology accurately, quickly detect whether corn stalks corn borer pest. Through this study, it can provide a strong basis for the future of the corn borer pest Detection.In this study, experimental study is corn stalks, using hyperspectral imaging technology to study corn stalk corn borer pest situation. First, collecting the corn stalk samples hyperspectral images. the correction formula based on black and white board hyperspectral images were in black and white calibration plate, by comparing the image preprocessing methods, method of using a mask image noise pretreatment, in addition to a large number of unwanted non-target information interference. For full-band spectral reflectance corn borer pest of corn stalks and normal corn stalks statistics, in determining the effective region of the spectrum 750～1000nm, based on the effective spectral region for data processing and analysis. By comparison the principal component analysis method can effectively achieve high spectral image data dimensionality reduction. Getting the main component of the image corn stalks samples using principal component analysis, Experiments were obtained PC-1, PC-2, PC-3, PC-4, PC-5 image. By comparison the first principal component shows a clear image pest site, and corn stalks distinguished from the epidermis, is conducive to pest site extraction and recognition. Using threshold method to split corn stalks first principal component image. Corn stalks establish pest detection algorithm model for testing, the test set overall detection accuracy of 92.9%, the overall validation set detection accuracy is 90%. Since the main component analysis is based on the analysis of 313 bands, data analysis capacity, wasting time and energy. In order to improve test accuracy and efficiency, the introduction of a new test segmentation algorithm mixing distance method, this method can help us find the most suited band split during the experiment. Several experiments to determine the best single-band and dual-band best combination, and ultimately determine the best features of the image is based on 754.8nm wavelength, threshold segmentation, segmented insect parts. Overall then subjected to detection detected, the test set accuracy was 97.1%, the overall validation set detection accuracy is 93.3%.In this study, the two methods mentioned detection and identification of corn borer pests, namely principal component analysis corn of corn borer detection method and optimal band of corn borer detection method. However, by comparing the two methods, based on sub-mixing distance calculation corn borer achieve higher accuracy detection research, but also save time, greatly improving the efficiency and accuracy. Detection of the ultimate realization of the corn stalk corn borer pest...