Research On Workpiece Recognition Method Based On Kinect

In many industrial production lines,random stacking of workpieces often results in occlusion between multiple workpieces.For the identification and location of stacked workpieces,this thesis mainly studies the following aspects:(1)The structure of Kinect v2 camera and principle of acquiring depth images were analyzed;the coordinate conversion relationship between coordinate systems in the camera's geometric model was studied;the calibration experiment was performed using the Zhang Zhengyou calibration algorithm;Calibration principle of the hand-eye system was studied.(2)The preprocessing method of images collected by Kinect v2 camera was studied.The alignment methods of depth and color images were analyzed.In this thesis,median filtering algorithms,Laplace enhancement algorithms,and LBP operators were used to process images,and compare these with other algorithms to verify their effectiveness.(3)The method of identifying artifacts was studied.First,the principle and process of template production were introduced.Then,for the case where the upper workpiece was tilted due to the stacking of the workpiece,a multi-template matching method was adopted so that multiple templates cover multiple forms of the workpiece.According to the actual situation,In the recognition process,the method of rotating template images was used to generate templates with different orientations to match to identify the workpiece.Finally,study the image template making method and the workpieces image recognition algorithm.According to the algorithm,the image was recognized,and the experimental results were analyzed to verify the effectiveness and stability of the algorithm.(4)The methods of workpiece positioning were studied.The Huff gradient method was used for workpiece positioning.First,Used the weighting method to grayscale the color image of the workpiece.Second the edge detection was performed on the gray image using the Canny operator.Then the principle of the Hough gradient method was described,and the algorithm was used to locate the target workpiece.Finally,analyzed the error results of the detected coordinate to verify the accuracy of the method.