Intelligent Coal Transport Detection System Based On Machine Vision

Coal industry is the backbone of the energy industry in China,but at this stage of the coal industry caused by a lack of technology,tend to cause the secondary waste of energy,low production efficiency,high accident rate,especially in the process of transportation of the coal,transportation belt idler cause waste of electricity and transportation belt running deviation,safety accidents caused by coal mining enterprises are still facing problems.Based on the electronic monitoring system in coal mining enterprises,this paper designs a set of complete intelligent coal transportation detection system based on machine vision.In this paper,transportation belt area detection,coal volume detection and tear detection are taken as the main research contents.For each part,this paper will discuss and study from the perspective of traditional machine vision method and deep learning method.The research contents are as follows:For the detection task of transport belt area,this paper starts from the significant straight edge of transport belt,USES Hough transform to detect the straight line in the screen,and then designs a series of rules to select the belt edge.This method is simple and fast,but has poor robustness.Therefore,this paper improves the semantic segmentation network based on RefineNet to realize the region detection of transport belt.This method is relatively slow,but has high recognition rate.Secondly,for the coal volume detection,considering the actual demand,the realtime requirement of coal’ width is higher than that of coal’ height.This paper divides the coal volume detection into coal’ width detection and coal’ height detection.Coal’ width detection.In this paper,the image is segmented into a single area,and each area is identified separately to achieve the identification of coal.This method has a low computational load and meets the actual requirements.Aiming at the detection of coal height,this paper implements and compares the monocular depth estimation based on markov random field and the depth estimation method based on convolutional network.The traditional markov random airfield-based method has some disadvantages,such as poor robustness,fuzzy depth boundary and large computation amount.The experiment shows that the depth estimation method based on convolutional network can completely replace the traditional method to realize the detection of coal’ height.In the final task of tear detection of transport belt,this paper uses the collection characteristics of line array camera to realize tear detection by dividing the captured images by using the statistical information of row and row pixels.Experiments show that this method is fast,stable and has high recognition rate.