Overlapping Pellet Size Detection Method Based on Marker Watershed and GMM Image Segmentation

The particle size of pellets is an important parameter in steel big data, and the high density overlap rate bring a great challenge to detection. To address this problem, intelligent detection algorithm with improved watershed Gaussian mixture model (GMM) proposed. First, initial segmentation background achieved by using adaptive binary segmentation, then secondary fine combining morphological operations such as skeleton extraction marked segmentation; then, contour connected domain calculated, non-overlapping foreground overlapping are filtered according roundness their contours. Finally, number predicted reconstruction grayscale image pellets, granularity pellets. experimental results showed that 91.98% accuracy images. Compared other algorithms, can also effectively suppress over-segmentation under-segmentation problems, it realize pellet dense, those on pelletizing disk, which provides effective technical means for metallurgical performance analysis ore pellet-making driven data.