| Belt conveyor is an important tool for transporting coal and ore.Belt longitudinal tear accidents can cause huge economic losses.At present,the research on belt longitudinal tear detection technology only focuses on a single image for detection.In the actual operation of the enterprise,wear,scratches and patches on the belt surface will bring interference to the detection and lead to misjudgment,and the detection reliability is not high.In addition,belt deviation and vibration also bring difficulties to the detection.In view of these phenomena,this paper proposes a new machine vision method based on dynamic template matching to detect longitudinal tearing of the belt.The raw data of the belt surface is collected in advance as a template.The data collected during the operation is matched with the template to determine the new feature data and then improved accuracy of test results.Specific work includes:Firstly,in order to collect wide belts,this design uses two CCD cameras for acquisition and is assisted by a line laser.The introduction of line laser simplifies the problem of tear feature extraction in two-dimensional images and improves the detection efficiency to meet the real-time requirements of detection.The high-contrast feature of the line laser fringe reduces the interference caused by factors such as dust and light,and improves the accuracy of the detection results.Secondly,in the image preprocessing stage,several commonly used preprocessing methods are used,and the advantages and disadvantages of several algorithms are analyzed for the longitudinal tear detection of the belt.Finally,median filtering,piecewise linear gray transform image enhancement and the Otus method image segmentation algorithm is prepared for the projection algorithm extraction and the Hough detection algorithm of the surface features of the back belt.In order to solve the impact of the belt deviation and vibration during the detection process,the edge of the belt is determined based on the distortion characteristics of the "one" line laser image at the edge of the belt.Belt data on both sides are stitched.Thirdly,in order to solve the belt surface wear scratches and patches that interfere with the detection during the inspection process,a template matching method is used to treat the belt as a whole,and periodically collect the raw data of the belt surface that runs for one cycle as a template.The data collected during the operation is used to match the image with the template,and the template is matched using the improved normalized product correlation algorithm to eliminateinterference and determine new cracks,and improve detection efficiency.Finally,the experimental algorithm is verified,and the experimental results verify the effectiveness of the algorithm.The actual measurement data shows that under the requirements of ensuring the real-time performance of the system,the accuracy and stability of the detection system are greatly improved,which is of great significance to the safety of industrial and mining enterprises. |
PDF Full Text Request