A recognition method of apples based on texture features and EM algorithm

This paper presents an apple recognition method based on texture features and Maximum Expectation (EM) algorithm for Gaussian Mixture Model (GMM). The images were converted to HSV space from RGB space and the H channel images were selected as interested images to be processed. The images of H channel were divided into blocks of 8*8 pixels and the texture features of the blocks were calculated. Angular second moment was selected for clustering using EM algorithm. Their prior probability, mean and variance were computed. And then the images were segmented according to these parameters obtained. Results showed that the proposed segmentation method could recognize the apples effectively, and 85.33% of apples were successfully recognized.