Image positioning and identification method and system for coal and gangue sorting robot

Abstract

ABSTRACT Accurately positioning and identifying coal and gangue under the heterogeneous and complex scenarios of coal production is a key and difficult problem for coal and gangue sorting robot. This paper proposes an image processing based method for the problem and further develops a positioning and identification system. The coal and gangue samples are selected from Hancheng Coal Mine, China’s Shaanxi province. Three automatic image threshold segmentation methods are used for image binarization and the comparative analysis shows that the clustering method outperforms the others. Particle analysis of coal and gangue samples is carried out by using the morphological corrosion and expansion methods and a complete, clean target sample is obtained; the center of mass is extracted by using the center of mass method. The features of the grayscale and texture of the samples are analyzed under the light of 300lux. It is found that coal and gangue have higher discrimination degrees in gray variance, skewness, texture contrast and entropy than other parameters. LS-SVM (Least squares support vector machines) is applied as the image classifier. Three classifiers are trained by using the feature of grayscale, texture, and the joint feature combining skewness with contrast respectively. The experimental results show that the classifier using the joint feature outperforms the others. Furthermore, we develop a positioning and identification system and further evaluate the system by using the images of goal and gangue which are randomly picked from the production environment. The results show that the average coordinate errors in x and y directions are 2.73% and 2.72%, the identification accuracy of coal and gangue samples is 88.3% and 90.0% respectively, and the sum of time of identification, positioning and opening the camera for a single sample in average is 0.130 s.