Research And Application Of Ore Particle Size Detection Method Based On Image Segmentation

In mining enterprises,the particle size information of crushed ore is an important parameter to evaluate the blasting effect,and also the main index to control the operating parameters of crushers at all levels.Therefore,the particle size information of crushed ore is of great significance to the efficient production of mines.In view of the traditional algorithm in dealing with complex environment of ore image segmentation problem such as low accuracy,owe segmentation serious,based on the further study of artificial intelligence application in the field of image segmentation,on the basis of the deep learning as ore image segmentation technology,artificial intelligence and ore particle size test method of research.Firstly,the preprocessing method of ore image is selected.Then,the features of ore images are acquired automatically through neural network,and the contour detection model and contour optimization model are trained to improve the segmentation accuracy of ore images.Finally,through the study of ore particle size information,relevant parameters that can quantitatively describe ore particle size are selected,and the software development of ore particle size detection system is completed.The main research work and technical innovation points of this paper are as follows:(1)Make data sets suitable for deep neural network training.First of all,the operation images of explosive heap and conveyor belt under various environments were collected from the mine as the image set,and the images of non-overlapping areas,different lighting conditions and different color of ore surface were selected from the image set as the sample set.Then,draw the edge line of the image of the sample set as the label set,and combine the sample set and its corresponding label set into the training set.Finally,the training set is greatly expanded by image clipping,and the training set is taken as the input data set of the training model.(2)A three-step data processing method was established.After graying the RGB images in the data set,the image noise is reduced by means of mean filtering,and the image is enhanced by adaptive histogram equalization.This three-step data processing method of"grayscale+mean filter+adaptive histogram equalization processing" not only reduces image noise and enhances ore edge information,but also improves model training speed and reduces data processing workload of model training.(3)The ore image segmentation algorithm based on double model is established.After graying the ore image into a binary image,and then denoising and enhancing the image,the pre-trained contour detection model is used to extract the edge features of the blast and conveyor belt images respectively.Then,the pre-trained contour optimization model is used to optimize the image edges of blast reactor and conveyor belt respectively.Finally,the connected region is marked to obtain the segmentation result.(4)Development of ore size detection system software.Based on the in-depth study of the shape and texture characteristics of the ore image,the expression method of the particle size parameters was determined,and the particle size parameters such as the perimeter diameter,area diameter,minimum outer ellipse length diameter and short diameter of the ore were obtained.Then,combined with the ore image segmentation algorithm in this paper,a particle size detection system suitable for ore image in complex environment is developed.The system was applied to Anqian iron mine,and through the particle size detection and analysis of the operation images of explosive heap and conveyor belt under the actual operation environment,the results showed that the system had a good detection effect and could meet the actual operation needs of mining enterprises.