ass日本风韵熟妇pics男人扒开女人屁屁桶到爽|扒开胸露出奶头亲吻视频|邻居少妇的诱惑|人人妻在线播放|日日摸夜夜摸狠狠摸婷婷|制服 丝袜 人妻|激情熟妇中文字幕|看黄色欧美特一级|日本av人妻系列|高潮对白av,丰满岳妇乱熟妇之荡,日本丰满熟妇乱又伦,日韩欧美一区二区三区在线

In order to fast, accurately and nondestructively detect pesticide residues in lettuce, combining discrete wavelet transform (DWT) algorithm with the approximate position of frequency doubling center of organic compounds in near infrared spectra, a method of feature extraction algorithm of piece-wise discrete wavelet transform (PDWT) was proposed. PDWT was used to extract the feature of four different concentrations of pesticide residues on the lettuce leaves. Transmission electron microscope (TEM) was carried out to detect the microstructure of lettuce. Hyperspectral image acquisition system was used to get information of near infrared hyperspectral image of lettuce, and the region of interest (ROI) was selected to get the near infrared spectrum data of the lettuce samples, which was ranged from 870 nm to 1800 nm. According to the approximate position of organic compounds in near infrared spectral region, appropriate piecewise paragraphs were selected. Each section of the spectral data was divided into seven layers by PDWT in turn, using sym5 as the basis function. Then, based on the analysis of the singular value of the high frequency wavelet coefficient curve, the characteristic band of lettuce was extracted by the optimal decomposition layer, which was the largest corresponding to the characteristic difference of the singular value. In order to evaluate the value of the feature extracted by singular value, a parameter of fit degree (FD) was proposed. Combined with the SVM classification accuracy, the feature extracted by PDWT was further evaluated. The results showed that under different concentrations of pesticide residues, the arrangement and structure of internal cells of lettuce leaves were different. The spectra of different concentrations of pesticide residues were different. PDWT had a higher classification accuracy of predictive classification compared with that of SVM. The classification accuracy of FD, calibration, cross validation and predictive classification accuracy of SVM were 75%, 95%, 92.86% and 90.63%, respectively, under the N value of 4 with PDWT. PDWT combined with FD was suitable for the feature extraction of spectrum, it provided a novel method for fast and nondestructive identification of lettuce pesticide concentration.