Font Size: a A A

Monocular Vision For Autonomous Mobile Robot Tracking And Error Analysis

Posted on:2015-01-09Degree:MasterType:Thesis
Country:ChinaCandidate:Q X ZhangFull Text:PDF
GTID:2268330425487604Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
This article mainly researches on mobile robot which is based on vision and error analysis, including design of the robot system, camera calibration, the stability of video, target tracking and distance measurement and so on.By choosing appropriate algorithms,the whole system to achieve the desired results. In this article, the research in the followings:According to a demand report, a robot platform is designed. Its turning way of four-wheel vehicle are rear-wheel steering. And it is controlled by differential steering model and navigated by a camera. We drive the robot to the target by calculating each motor speed and sending it to the motor after getting the distance and angle between target and robot by processing the image come from the navigation camera.Camera calibration is the foundation of computer vision system. The result of calibration directly affects the quality of the image. So selecting an appropriate calibration algorithm for the robot navigation camera has a crucial effect. In this paper mainly discuss the system model, image noise, quantization error and the data of training quantity to affect to the calibration result by compared the calibration algorithm based on radial with calibration algorithm based on plane. And choose the calibration algorithm according to the results of the comparison.It is unavoidable appear jitter when the robot is moving. Different monocular vision measuring methods for the sensitivity of the jitter is different. After compare all the model, the least sensitive model is selected. And in order to weaken the influence of jitter on video come from the navigation camera, this paper adopted a method of stable video base on the navigation position reference system.In object recognition, this article uses the Kalman filter for target tracking. And then extract the edge of the target by using Canny operator. After getting the edge of the target, a fast ellipse fitting method is adopted. Combination with the monocular vision measuring model, we get distance and angle between target and robot. Testing the whole system by using the hardware, Experiments show that the article algorithm has high feasibility and real...
Keywords/Search Tags:error analysis, Monocular vision measurement, video steady, targetrecognition, Kalman filter, ellipse fitting
PDF Full Text Request
Related items