Research On A Feedback Control System Based On Video Tracking

Machine vision is one of the hot topics in scientific research today.It is widely used in driving,drone,robot navigation,industrial robot and robot arm control,virtual reality,image measurement and so on.It uses a camera,computer,linear motor and other devices to form a visual feedback system to replace the human eye to measure,monitor,track,control and other complex tasks,while showing more excellent precision,speed,range and other properties.Therefore,the study of machine vision has important theoretical research value and a wide range of applications.This paper systematically studies the use of tracking control system based on visual feedback to study machine vision and three-axis linear motor precision as a control system.First of all,several successive calibration principles and methods used at present are studied,and the initial calibration results are obtained through experiments.After fully studying several lens distortion models,the camera internal parameters and external parameters are then optimized to control the calibration error and improve the accuracy of successive calibration.The visual distance model is used to solve the localization problem of the target point in this experiment platform.Second,after the camera calibration is complete,the image for the video sequence is processed.Firstly,the characteristics of the centralized filter are studied,and the simulation experiments of their filtering effects and image processing results are analyzed.The median filter and mean filter are validated.Binarization of target detection.Using Otsc algorithm to improve the binary processing,to facilitate the next experiment.Then,according to the introduction of target detection technology,according to platform characteristics and experimental environment,we select the appropriate target detection method and verify the experiment.Analysis of the experimental environment background selection method to extract the target.Then the trajectory of the target point planning,in this experiment we plan to run a straight line of the target point,and then draw 20 points at equal intervals for experimental demonstration.The tracking problem in the moving target is researched and explored.The common tracking methods are compared first,and then the Kalman filter algorithm and the centroid algorithm are compared in detail.According to the experimental accuracy and the algorithm complexity,we choose the centroid algorithm To pinpoint the location without environmental impact.The last article through visual system and three-dimensional linear motor motion control platform system in the open-loop and closed-loop case study,set the test evaluation index,the tracking error quantization,control experimental error.It verifies that the theoretical performance of the tracking control experiment method and control strategy designed in this paper has been greatly improved.