基于高光谱图像的新疆红富士苹果颜色分级研究

[目的]以新疆红富士苹果为研究对象,探讨应用高光谱图像技术对其着色面积进行的研究方法.[方法]对852/713双波段比图像作阈值分割,以及形态学开运算去除果梗区域,提取色调H灰度图像对应去除果梗的二值图像像素值为1的累计频度值,依据AdaBoost算法将15个BP神经网络弱分类器训练组成强分类器,对苹果的着色面积进行分类.[结果]采用AdaBoost_NN对苹果着色面积的分级与人工分级一致率达到97.7;.其中45个优等果有2个被错分为一等果,27个等外果有1个被错分为二等果.[结论]利用高光谱图像技术提取的特征波长图像能够很好的对苹果着色面积进行分级,为今后多光谱成像技术在线分析苹果品质奠定研究基础.

Study on Color Sorting for Xinjiang Fuji Apples by Using Hyperspectral Imaging Technology

[Objective]The study aimed to investigate the prediction performance of the coloration area of Xinjiang Fuji apple by using hyperspectral imaging technology.[Method]Firstly 852/713 band ratio image was segmented using threshold and removed stem's region of binary image by morphological opening operation, and then the accumulating of the gray value was abstracted from the hue gray image that corresponded to the value 1 in the binary image that was removed from stems region.15 BP weak classifiers were trained and then the strong classifier was got by AdaBoost algorithm,and the strong classifier was used to sort the coloration area of apple.[Result]It has the result with a veracity of 97.1%.2 premium apples of the total 45 apples were misclassified as first class apple,1 substandard apple of the total 27 apples was misclassified as second class apple.[Conclusion]The image of feature wavelengths that was abstracted from the hyperspectral image can sort well the coloration area of apple,which provides research basis for apple quality online analysis by using multispectral imaging technology.