High-precision Monocular Vision Positioning Technology And Its Application In PCBA

With the popularization of 5g technology,the size of components of PCBA(printed circuit board assembly)of smart phone gradually tends to be miniaturized and precise.High precision visual positioning technology is one of the keys to promote the development of precision,flexibility,lean and high precision of smart phone automatic production line.Aiming at improving the visual positioning accuracy of the assembly robot in the PCBA automatic production line of smart phone,this paper carries out research from four aspects: the construction of high-precision monocular vision system,target area detection,edge contour extraction,sub-pixel edge contour thinning and high-precision center positioning.The main research contents include:(1)research on the construction and calibration of high-precision monocular vision positioning system.According to the requirements of PCBA visual positioning in the actual production line,the light source,lens and camera are selected,and a high-precision monocular visual positioning system is built.It is calibrated by Zhang Zhengyou plane calibration method,and the average calibration error is solved by re projection error.The calibration accuracy of the system is verified by experiments;(2)PCBA parts target area detection In order to accurately and quickly match the target area of PCBA parts,firstly,histogram equalization and adaptive guided filtering based on gradient information are used to complete the image preprocessing,which can adaptively identify edges and other textures,and protect the image edge information while removing noise.The experimental results show that the algorithm has good denoising effect Then,using the template matching method based on the similarity of the best partner,the similarity between the matched image and the target image is determined by the way of point set,and the best matching region is obtained.The non repeated sub-pixel region is matched by segmenting the image to realize fast traversal.The experimental results show that this algorithm has higher accuracy and timeliness than other comparison algorithms;(3)In order to solve the problem that the common threshold segmentation Otsu algorithm and its improved algorithm are difficult to deal with PCBA images with complex background and texture,this paper proposes an adaptive Otsu algorithm,which describes the segmentation state by using three states: foreground,background and intermediate state,and determines the wave crest for the intermediate state region,so that the target gradually deviates to the bottom of the target valley,The experimental results show that the proposed algorithm has better segmentation effect than the comparison algorithm.In order to select the target contour from the complex contour of different features,the cross star datum point of rectangular contour is equivalent replaced by taking its minimum circumscribed circle;For the circular contour,using the adaptive filtering algorithm proposed in this paper,the non-circular contour is eliminated by calculating the change rate of image edge gradient,and then the contour filtering operation is realized.Experimental results show that this algorithm has higher accuracy and timeliness than other comparison algorithms.(4)In order to improve the positioning accuracy,this method uses an improved Gaussian curve fitting method to remove the existing false edges on the basis of pixel-level contour extraction.For problems such as burrs and edge synapses in the actual edge contour,random sampling is used first.The consensus method is used to eliminate the point set with mutation,and then the least square method is used to realize the final positioning of the center position of the sub-pixel contour.Comparative experiments show that the method can achieve ±0.03 mm for the center positioning accuracy of the target area of PCBA parts.