A method of blasted rock image segmentation based on improved watershed algorithm.

It is of great theoretical significance and practical value to establish a fast and accurate detection method for particle size of rock fragmentation. This study introduces the Phansalkar binarization method, proposes the watershed seed point marking method based on the solidity of rock block contour, and forms an adaptive watershed segmentation algorithm for blasted rock piles images based on rock block shape, which is to better solve the problem of incorrect segmentation caused by adhesion, stacking and blurred edges in blasted rock images. The algorithm first obtains the binary image after image pre-processing and performs distance transformation; then by selecting the appropriate gray threshold, the adherent part of the distance transformation image, i.e., the adherent rock blocks in the blasted rock image, is segmented and the seed points are marked based on the solidity of the contour calculated by contour detection; finally, the watershed algorithm is used to segment. The area cumulative distribution curve of the segmentation result is highly consistent with the manual segmentation, and the segmentation accuracy was above 95.65% for both limestone and granite for rock blocks with area over 100 cm, indicating that the algorithm can accurately perform seed point marking and watershed segmentation for blasted rock image, and effectively reduce the possibility of incorrect segmentation. The method provides a new idea for particle segmentation in other fields, which has good application and promotion value.