Workpiece Recognition And Localization Algorithm Based On Machine Vision

With the development of the robot industry,robots are now in widespread use in industrial production.Most of industrial robots are currently based on the teach-repeat paradigm and it can only handle particular and repeated tasks.These characteristics limit application of industrial robots in practice.In recent years,vision sensor has been widely used in robot perception systems,due to the advantages such as:low cost,high flexibility etc.and it can obtain more information.In order to achieve good performances in uncertain environments and perceive the external environment,this dissertation researches the workpiece recognition and localization algorithm based on machine vision.The main research contents of this dissertation are as follows:Firstly,the background and significance of the vision-based workpiece recognition and localization are introduced.The research status of the vision-based workpiece recognition and localization is reviewed.The problems and challenges of the subject are explained.Secondly,the pinhole model of camera and the binocular stereo vision model are elaborated,and the matching-constraints commonly used in binocular stereo vision are described.The commonly used camera calibration methods and binocular stereo vision system calibration methods are studied,and its experimental results are also given.Thirdly,the working principle and types of template matching are introduced,and a contrast experiment result of each method is shown.The computational efficiency and anti-interference of template matching algorithms are analyzed.The feasibility of using shape-based template matching algorithm for the recognition and localization of flat workpieces is discussed.The shape-based template matching algorithm is studied.The algorithm is optimized in calculation method and searching strategy.It is proved that efficiency of the algorithm has been improved.The experimental results are also given.Finally,a three-dimensional workpiece recognition and localization algorithm based on binocular stereo vision is designed to solve the current difficulties.The error of binocular vision is analyzed,and the influence of external parameter error on the reconstruction accuracy of the contour is proved.According to this analysis,an optimal projection plane-based feature contour measurement method using binocular stereo vision system is proposed,and the algorithm is described.Experimental results prove the effectiveness of the proposed algorithm.