Research On Typical Target Image Recognition And Capture System Under Complex Background

As the tasks performed by robots become more and more complex,the requirements for their intelligence are also getting higher and higher.The robot visual serving refers to the use of image feedback information in robotic motion control to realize closed-loop control of robots so as to achieve the intelligent effect.The result of robot vision system's image processing will directly affect the stability of machine control system.In this paper,we study the general scheme of target image recognition and acquisition system under complex background,camera calibration technology,target feature extraction and recognition,target positioning and servo control system design.The robot arm developed independently by the laboratory is taken as the research object.In view of the shortcomings of the existing camera calibration methods,this paper proposes a camera calibration method combining Harris corner detection algorithm and Zhang two-step method.The industrial camera is calibrated by using calibration toolbox and the proposed approach.The results show that the error between the calibration result of the new method and the calibration result of the toolbox is small,which can calibrate the camera's internal and external parameters.And the calibration program is easy to expand and transplant.Aiming at the problem that the motion blur of the image during acquisition and the target difficult to identify in complex background,a suitable solution is proposed.Prior to the target recognition,motion blurred image restoration is performed,and then adaptive threshold target recognition and improved feature points matching are proposed to identify targets for different background environments and recognition targets.After the target is identified,the target is located in two ways,that is,no artificial marker and artificial marker.Experiments show that the two methods can accurately obtain the target location information to achieve the goal of rapid positioning to solve the problem of monocular vision is difficult to locate.Aiming at the shortcomings of visual controller designed by conventional methods,the visual controller is designed by using the adaptive fuzzy neural network(ANFIS).The hybrid learning algorithm is used to train the neural network.Finally,the network model is simulated.The experimental results show that using the adaptive fuzzy neural network can better fit the visual controller to achieve the goal of grasping.