Identify Method Of Blasting Pile Fragmentation Based On Computer Vision

The fragmentation of blasting pile is a key data index of mine blasting,which is crucial to the evaluation of blasting effect and optimization of blasting parameters.The non digital the fragmentation statistics method has many problems,such as low efficiency and many limitations.It has become a hot research direction in the industry to use computer vision methods to quickly segment ore and rock images and then calculate the distribution of the fragmentation.At present,the traditional ore and rock image segmentation algorithm is difficult to deploy,and the degree of intelligence is not enough.How to improve the self perception ability of ore and rock image segmentation and achieve the statistics of fragmentation distribution is very necessary.In order to improve the self perception ability of ore and rock image segmentation,and quickly and efficiently calculate the fragmentation distribution.The method of computer vision is used to carry out experiments and research on the segmentation and fragmentation statistics of ore and rock images.The following conclusions are drawn:(1)In the second chapter,an improved Mask B-R-CNN model based on Mask RCNN is proposed for ore and rock segmentation,which realizes self perception segmentation of ore and rock images and ensures the accuracy of ore and rock image segmentation.It provides the research basis for the experiment of the fragmentation distribution statistics.(2)In the third chapter,it is found that Mask B-R-CNN converges faster than the original Mask R-CNN through the laboratory experiment of the fragmentation distribution statistics.Furthermore,a morphological optimization fragmentation statistics method combining Mask B-R-CNN and HSV transform is proposed.By comparing the results of the fragmentation statistics method and screening experiment,theoretically,the fragmentation statistics error of each level is within 3%,which verifies the accuracy and effectiveness of the fragmentation statistics method combining Mask B-R-CNN.(3)Through experimental research on block size statistics at the blasting site,a block size mapping relationship between the blasting site image and the physical world was established.The mining and rock image segmentation experiments and block size statistics experiments at the blasting site were completed.Compared with traditional mining and rock area extraction methods,the morphological optimization method combining Mask B-R-CNN and HSV transformation was more accurate in extracting mining and rock areas.The block size statistical method in this paper has been applied to the mine site of Anqian Mine,Anshan Iron and Steel Group Co.,Ltd.The cumulative probability curve of the Degree distribution of three blasting pile blocks is similar,and the block rate is 4.21%,3.37% and 3.12% respectively.