Research On Image Analysis And Location Of High-voltage Copper Finger Burr Based On Machine Vision

The high-voltage circuit breaker is one of the important electrical components in the highvoltage circuit,which is used widely in power generation,transmission,substation and other systems.The high-voltage copper finger is an important contact element on the high-voltage circuit breaker,which is responsible for the task of turning on and off the high-voltage circuit and the load current.Since lots of burrs are generated on the high-voltage copper figure in machining,it is easy to produce a sharp discharge phenomenon during utilization,causing arc ablation,which greatly shortens its service life,significantly reduces the dielectric insulation strength and affects the impact breaking capacity,causing great safety hazards to the power system equipment.Therefore,the finished high-voltage copper fingers are required to be free of burrs,smooth in surface and good in quality.In addition,due to the material properties of the high-voltage copper fingers,no other substances or residues are required on the processed surface.Therefore,the physical contact method can only be selected for processing.At the same time,due to the characteristics of small size,small batch,and many varieties,most factories currently still use manual deburring,resulting in low efficiency and poor surface consistency,thus affecting the product performance.To sum up,this dissertation proposes an image analysis and location method for high-voltage copper finger burr based on machine vision.Through image analysis and processing,the accurate position of workpiece feature points can be obtained,thereby providing position information for automatic deburring of industrial robots,which lays a foundation for hand-eye cooperation of robots to complete similar work.In this dissertation,based on industrial Ethernet,FANUC industrial robot and Hikvision industrial camera are connected,and the communication between the industrial robot and camera is realized by installing ROS system on the upper computer,and the platform construction of hand-eye cooperative robot deburring based on machine vision is completed.On this platform,the following work has been completed:Burr image denoising method based on block cosparse over-complete learning transform is studied.Created a high-voltage copper finger burr image data set,classified the image data set according to the image content and image size.Refers to the real noise and additive Gaussian noise,carried out image denoising processing of the high-voltage copper finger images.The reference and full-reference image quality evaluation methods evaluate the image denoising quality of the proposed algorithm,the traditional algorithm and the deep learning model.The results show that the proposed algorithm reaches the advanced level of image denoising at present.The block cosparse over-complete learning transform burr image denoising algorithm completes the sharpening of high-voltage copper finger burr image.However,because single or multiple images with little similarity are difficult to be used for subsequent feature point extraction and matching,a burr video denoising algorithm based on online sparse transform is proposed.According to the different rotation directions of the adjustable fixture,the video sets of high-voltage copper finger burr are divided into 4 different types.The algorithm uses a temporary sliding window strategy to extract image block information in noisy video frames at any time,and uses a block-based online 3D denoising mechanism to generate denoising estimates of these frames with controllable delay,so as to finish the clarification processing of the captured video.Burr image segmentation is carried out on the obtained clear images and video frames,and the position information of the edge pixels of the high-voltage copper finger workpiece is obtained.This paper proposes an algorithm for segmentation of burr images based on block sparse over-complete clustering transformation.The algorithm consists of a set of transformations.The k-means clustering method is used to initialize the clustering results,and the discrete cosine transform is used to initialize the sparse over-complete transformation matrix,image segmentation algorithm is affect by the burr parameters of sparse constants and regularization coefficient,the experimental results show that the algorithm preserves the weak edge information of the workpiece in the segmentation results of high-voltage copper finger burr image and avoids the influence of three-dimensional structure on the segmentation results greatly.Based on the above theoretical analysis,the pixel position of the workpiece edge of the highvoltage copper finger obtained by image segmentation is intersected with the pixel position obtained by feature point extraction and matching to obtain the pixel position information of the feature point of the workpiece edge.Using epipolar geometry and triangulation methods,the three-dimensional coordinates of the workpiece edge of the high-voltage copper finger are calculated,and the error analysis is carried out through the average Euclidean distance similarity.The results show that the edge positioning result of the high-voltage copper finger workpiece based on machine vision can meet the processing accuracy requirements of industrial robots to realize three-dimensional location of the edge of the workpiece.