Research On The Wooden Workpiece Positioning Technology Based On Binocular Vision

Wooden furniture is a kind of industrial products,consumer's personalized demand is growing.Custom wooden furniture manufacturing industry is transforming towards digital,intelligent direction.The introduction of the industrial robot has solved the wooden furniture production in manufacturing process workpiece stacking,sorting,handling,grinding problem.Binocular vision system is able to rapidly locate the industrial robot wooden workpiece location information,intelligent implementation of the corresponding process,however,traditional binocular stereo matching algorithm to calculate the wooden artifacts binocular disparity map in the weak object edge area and texture area prone to failure and matching error rate high,and then affect the high precision positioning of the wooden workpiece.Therefore,the positioning technology based on binocular vision for wooden workpiece is studied in this paper.The main work contents and research results of this paper are as follows:(1)According to the characteristics and requirements of wooden workpiece,the system framework of wooden workpiece positioning system based on binocular vision was designed.First of all,the research object and system requirements were analyzed,then according to the system requirements,the wooden workpiece positioning system for the overall was designed and the algorithm process of the wood workpiece positioning system was designed,finally according to the design of the wood workpiece positioning system the required hardwares were selected.(2)The coordinate transformation and calibration method of wooden workpiece positioning system based on binocular vision were studied and the calibration of binocular camera was realized.First according to the principle of binocular camera imaging,the camera coordinate system and world coordinate transformation was realized,and lens distortion of the common forms and the lens distortion model in the camera coordinate transformation were realized,then using Zhang Zhengyou calibration method completed the ZED binocular camera calibration,and solved the internal and external parameters of binocular cameras and distortion parameters,finally,a hand-eye calibration of camera coordinate system and the mechanical arm were obtained between the base coordinate system transformation matrix(3)A stereo matching algorithm based on binocular vision for wooden workpiece positioning system was studied,and a deep learning stereo matching algorithm based on adaptive cost aggregation and edge preservation loss function was proposed.First the difficulties in the current stereo matching algorithm was analyzed,and then using a traditional widely used in stereo matching algorithm of BM local stereo matching algorithm calculated the wooden artifacts disparity map,it is pointed out that the algorithm is not effective in solving the parallax between the object edge area and the weak texture area in the binocular image of wooden workpiece,a set of end-to-end stereo matching algorithm based on depth learning,keep the edge adaptive cost aggregation module and loss function is designed,the high precision solution of the disparity between the edge area and the weak texture area in the binocular image of wooden workpiece is realized,and the better accuracy and real-time performance of the disparity estimation were obtained.(4)The positioning of the center point of the wooden workpiece was realized through experiments,and the positioning accuracy of the positioning system was verified to meet the system requirements.Firstly,an experimental platform for the positioning system of wooden workpieces was built,and the parallax diagram of wooden workpieces was obtained by using the stereo matching algorithm proposed in this paper,and the 3D point cloud reconstruction of wooden workpieces was realized.Then observed hoff circle detection and connected domain were used to detect the cylindrical type and plate type wooden workpiece center pixel coordinates,using random consistency algorithm for segmentation of point cloud surface on the wooden parts,pursuing the wooden workpiece surface cloud data fitting equation of a plane,according to the fitting of the equation of a plane with the center through coordinate transformation to obtain the coordinates of the wooden workpiece center coordinates of mechanical arm relative to the base coordinate system,realized the mechanical arm grab to a wooden workpiece is realized,Finally,the positioning accuracy of the wooden workpiece positioning system along the Z axis direction was analyzed by calculating the error between the fitting distance between the wooden workpiece's upper surface and the desktop point cloud and the actual distance.This thesis has 50 figures,12 tables,115 references.