Research On LED Filaments Location Technology Based On Stereo Vision

With the wide use of LED lamps,people have higher requirements for the performance of LED lighting products.The production technology has made great progress,since The technical means of processing such products with high efficiency and high precision are closely related to the enterprises.A highly automated assembly line has become an important means to achieve this technology,As the research of machine vision deepened,and Compared to traditional technology,visual location and detection technology,the machine vision plays an important role in the positioning of the LED chip and the recognition and positioning of the filament.This paper has proposed a positioning scheme based on stereo vision to solve the problem which is the limitations of the existing semi automated assembly of LED lamps and lanterns,and In-depth research and exploration of the scheme.This paper has proposed a Image processing algorithm scheme with Features of LED filament image collected,and this paper give a key research on image stereo matching algorithm.The specific research contents are as follows:(1)with the Requirements for the shape of LED filament and the design of pending assembly,this paper Build a stereo vision work platform,which is an important platform for image acquisition.Studying The transformation relationship between four large coordinate systems.the system is calibrated by using the calibration algorithm of the Zhang's camera calibration,to obtain The internal and external parameters of the camera in the stereo stereo vision system and using the parameters to correct the image distortion and rectify the epipolar line.(2)Using the image processing algorithm(filtering algorithm,image binaryzation,morphological processing,Roberts algorithm)to obtain filament edge contour by processing the images.Study on the points detection algorithm.(3)By using Image point feature matching of stereo vision,Polar line constraint of reference image feature points and search method of estimated points in pending graph,limited the range of image search for stereo matching,reduce the time consuming of thematching algorithm.(4)3D reconstruction of the feature points of the LED end filament.Firstly,Finding the pixel coordinates of the camera projection matrix and the matching feature point pair,Secondly,get the parallax value of the same point in the left and right image.Finally,the actual coordinates of the space point on the filament are solved.In order to verify the real time and feasibility of the scheme,the verification and comparison of the stereo matching arts of the key steps of the scheme.The experimental results show that in the same environment,the result of the depth information reduction error of a target object,the algorithm proposed in this paper improves the accuracy of the traditional feature point matching algorithm.The experimental results show that the scheme is feasible and effective.