Research On Vision-based Caliper Glitch Detection And Positioning Method

At present,automobile calipers mainly rely on manual grinding,and there are problems such as low grinding efficiency,poor working environment of workers,and the quality of polishing relying on the experience of workers.The use of robots to polish the burrs of automobile calipers can solve these problems well.In order to be able to automatically plan the robot grinding path of the car caliper,it is necessary to know the grinding area of the burr and its position information.The traditional burr measurement method is manual measurement,and there are problems such as low measurement efficiency,wear on the burr and the measuring tool.The use of vision to detect and locate the burr processing area of the car caliper is a non-contact measurement that improves efficiency without causing damage to the burr area.Therefore,this paper studies the visual algorithm for detecting and locating the burr processing area,which mainly includes the following aspects.1)By comparing the effects of different image smoothing algorithms,it is found that adaptive filtering based on image gradient information can sharpen the edges of the caliper image and have a good smoothing effect on the inner region of the contour.Different threshold segmentation algorithms are studied.It is found that the local adaptive threshold segmentation algorithm and K-means cluster segmentation algorithm can deal with the segmentation of the caliper illumination uneven image well.For the caliper threshold segmentation,there are a large number of reflective regions,and a reflective region search algorithm based on contour tracking is proposed.The reflective area is filled,and the experiment proves that the method can obtain a complete binary image of the caliper.2)subtracting the binary image 1 of the removed glazing template from the binary image 1 of the unpolished workpiece and the binary image 2 of the unpolished workpiece to obtain a binary map of the burr processing area,and then performing contour search on the binary image and calculating the centroid position of the contour The effect of the whole pixel edge detection algorithm and the improved algorithm on the detection of the rectangular edge of the caliper is studied.It is found that the improved Canny algorithm can detect a single edge well.The sub-pixel edge subdivision algorithm is studied,and the image sub-pixel positioning for Zernike moment is studied.The selection of the threshold K is improved,and the image subpixel positioning combined with Canny’s Zernike moment is selected.Experiments show that the method can accurately obtain a single sub-pixel edge of the caliper image;study the hough transform line detection and least squares line fitting algorithm,first use The cumulative probability Hough transform detects the edge of the rectangle,then performs sub-pixel positioning on the detected line,and finally performs a least squares fit on the obtained sub-pixel edge points to obtain the sub-pixel edge of the center rectangle.3)According to the experimental needs,select the appropriate camera,light source and lens to capture the caliper picture;explain the imaging principle of the camera and the cause of the distortion,use the Zhang Zhengyou calibration method to correct the distortion of the camera,and then linearly calibrate the corrected image;Finally,the caliper center rectangle detection and burr processing area detection experiments are carried out in the VisualStudio environment.The experimental results show that the algorithm can detect the caliper burr processing area and the sub-pixel edge of the center rectangle captured by the robot.In this paper,we use visual technology to detect the caliper burr grinding area and the caliper center rectangular area,and obtain the burr grinding area information and the burr relative central rectangular position information,and then guide the robot to realize the caliper grinding offline programming.