Research On Automatic Localization Technology Of Robot ARM Based On Machine Vision

Machine vision technology has always been an important research direction in the computer field,and it plays a key role in the fields of human-computer interaction,aerospace and factory automation.With the continuous improvement of production requirements,the robot arm has become a part of production and life,and can replace some humans to do some work efficiently,but the robot arm can only work according to the preset process.If the working environment changes,it needs reset the workflow.Therefore,to improve the flexibility and intelligence level of the robot arm and realize the automatic localization of the robot arm to the target,the machine vision technology can be combined with the robot arm.The existing binocular vision localization method has achieved certain research results,but the binocular vision localization method in the complex environment still has the following two problems:(1)The binocular visual positioning method has low accuracy;(2)The real-time nature of binocular vision positioning methods is not well guaranteed.In view of this,this paper studies the problem of target localization based on binocular vision.Based on the traditional localization method,two improved binocular vision localization methods are proposed.The methods are validated under the same experimental conditions.The improved methods have improved the localization efficiency.The specific research work of this topic is as follows:Firstly,the imaging model and calibration method of the camera are introduced.The camera calibration experiment is carried out by using Zhang Zhengyou calibration method,and the experimental result is analyzed.Secondly,the measurement principle and scheme of the vision system are given.The visual positioning system is built based on the theoretical calculation of system hardware selection.The hardware composition and software platform of the vision system are introduced in detail.Again,the image pre-processing operation is performed on the acquired target image,so that the feature information of the localization target in the image can be better extracted,and the localization accuracy is improved.Then,the traditional localization method is realized,and the shortcomings of the method in localization are pointed out.On this basis,according to the research objectives of this paper,two improved binocular vision localization methods are proposed:(1)The localization method based on region of interest matching;(2)The localization method based on the pixel coordinate offset.Finally,under the same conditions,the traditional binocular vision localization method and two improved localization methods are used to locate the target respectively,and the experimental results are compared with the two performance indexes of localization accuracy and localization time efficiency.The experimental results show that,to a certain extent,the improved two localization methods can improve the localization efficiency of the vision system,and have a positive effect on improving the working performance and flexibility of the robot arm.However,the localization method based on the pixel coordinate offset is better,the localization accuracy is improved by 56.8% compared with the traditional localization method,and the localization time efficiency is improved by about 32.6%,so it can be applied to the binocular visual localization system of the robot arm.