Abstract:In this paper, a monitoring system grabs the fragment images at the end of the conveyor belt and does the image processing in time, and the key task and the problem for the system is to deal with the touching fragments quickly and accurately. To resolve the problem, an adaptive image segmentation algorithm is proposed based on the analyses of fragment size, shape and gradient etc. Firstly a fragment image is roughly segmented by a recursive thresholding algorithm, and then the treatment is repeated for the regions of touching fragments. The judgment for touching fragments is made based on fuzzy comprehensive evaluation on fragment size, shape and gradient etc. In one Mine, we took 300 fragment images to do testing, the experimental results show that as compared to Clustering analysis and Graph based algorithms, the new algorithm can successfully make fragment image segmentation online, the accuracy is about 95%. The new algorithm can also be expanded into other complicated particle online monitoring systems.