Error Analysis Of Binocular Vision System Based On Field Robots

Robot navigation in field is becoming more and more important in modern agriculture. Binocular vision system is seen as a simple and effective way which has been concerned by many scholars. Binocular vision system has many problems in the navigation of the field robot, which leads to excessive error during the navigation. The purpose of this article is to study the law of the errors and find a way to reduce them.At first, this paper introduces the research status and research significance of binocular vision system, then it introduces the basic principle of binocular vision system. The world coordinate, image coordinate and pixel coordinate system and the transformations among them are also introduced. And then we focus on the calibration of the binocular vision system. This paper also introduces the way of using Jean Yves camera kit in detail for the calibration. Later, in this paper, we studied the error factors which affect the system accuracy of the binocular vision. The error factors include the numbers of images used on calibrating the system, the distance between the camera and the object, the angle of the camera, the base length and the focal length. The paper used the fitting method to acquire the error laws and put forward the corresponding methods to reduce the errors. Experiments show that the error on the edge of the image is greater than the error in the middle of the image. Moreover, the errors appear the shape of "V" from the left side of the image to the right side. In the depth direction (Z direction), the errors in the same round of the image are same. The error in the middle of the image is bigger than that of the edge of the image. Based on these error curves, it is easy to get a fitting error curve. Then the absolute errors are easily obtained that can be reduced to 1/8～1/4. At the same time, this paper used a matching technology based on SURF (Speeded Up Robust Features) avoid the situation that the objects appear in the edge of a image which can increase the error of the system.