基于小波变换的棉花黄萎病严重度高光谱识别研究  

Hyperspectral identification of cotton verticillium disease severity based on wavelet transform

作　　者：靳宁[1] 王大勇[1] 相栋[1] 赵永强[1] 

机构地区：[1]山西省气候中心,山西太原030002

出　　处：《山西气象》2013年第1期31-36,共6页Shanxi Meteorological Quarterly

摘　　要：小波变换能有效提取光谱信息并降低数据维数，本研究探讨了基于小波变换的棉花黄萎病病情严重度高光谱识别的可行性。通过测量不同病情严重度叶片的反射光谱数据，采用小波变换对高光谱数据进行处理，并以提取的小波能量系数建立了基于判别分析、BP神经网络、GA—BP神经网络及SVM支持向量机的病情严重度识别模型。识别结果表明，GA—BP神经网络识别效果最好，而支持向量机SVM识别结果较稳定。利用小波变换对高光谱数据进行小波能量系数提取，有效降低了数据维数，且建立的识别模型相对稳定，能有效识别棉花黄萎病不同严重度，为棉花黄萎病遥感监测提供理论依据和方法。Wavelet transform can effectively extract spectral information and reduce data dimension. The hyperspectral identification feasibility of cotton vertieillium disease severity that based on wavelet transform was studied in this paper. The hyperspectral reflectance of different disease severity blades was measured and then processed by wavelet transform to extract wavelet energy coefficients. Furthermore, the identification models of discriminant analysis, back propagation （BP） neural network, genetic back propagation （GA-BP） neural network and support vector machine （SVM） were built. The results indicate that the identification effect of GA-BP neural network was the best while SVM was more stable. The data dimension can be effectively reduced after the wavelet energy coefficients being extracted from the hyperspectral data using wavelet transform. The identification models are stable and disease severity can be identified effectively. Finally, the practical methods are provided in monitoring the cotton verticillium by remote sensing.

关 键 词：小波变换 棉花黄萎病 高光谱 严重度识别 

分 类 号：P458.121[天文地球—大气科学及气象学]