Research And Application Of Underground Drill Pipe Counting Based On Visual Object Tracking

Under the background of the popularization of underground video surveillance technology and the continuous maturity of downhole network infrastructure,the application of computer vision technology in the coal industry has continued to increase.Aiming at the issue of downhole drill pipe counting verification,the paper uses field research,theoretical analysis,experimental testing and engineering practice.Research and application of drill pipe counting in drilling construction are carried out by visual object tracking.First,this article analyzes the downhole drilling construction process,and determines the plan of calculating the number of drill pipes driven by extracting the motion trajectory of the impact power head.Secondly,the Retinex algorithm,which is more advantageous when dealing with images with insufficient light,is used to preprocess the image data collected downhole.The deficiencies exposed by the Retinex algorithm in the face of the special underground environment are analyzed.Gamma correction and Contrast Limited Adaptive Histogram Equalization(CLAHE)are introduced for power enhancement and CLAHE transformation of the initial image and enhancement results.Through experimental tests,it is found that the improved Retinex algorithm has a better enhancement effect on underground video image data.Then,the Kernel Correlation Filter Algorithm(KCF)is used to track the impact power head and extract the motion trajectory of the impact power head during the drilling construction process.Aiming at the single-scale and target loss problems of the KCF algorithm,a multi-scale and re-detection strategy is introduced to improve the algorithm.Through experimental tests,it is found that the target tracking accuracy of the algorithm is significantly improved,and the accuracy of the extracted drill pipe motion trajectory is greatly improved.Finally,data processing is performed on the motion trajectory extracted by the target tracking algorithm,and wavelet packet analysis is introduced to denoise the displacement of the impact power head,so that the motion trajectory tends to an ideal state without losing key information.Set the threshold to effectively count the back and forth movement frequency of the impact power head,and get the number of drills in the drill video.The drill pipe counting method is packaged and the field data is used to verify the drill pipe counting system proposed in this paper,and the drill pipe counting system is successfully used for the drilling construction of Beiyu Mine.There are 32 figures,14 tables,and 80 references in the thesis.