Research On Monocular Vision-based Realtime Distance Measurement Algorithm

Visual can help people measure the distance between the thing in the external environment and themselves, thus for they can avoid to hit them. But for the constriction of software and hardware, mobile robots are difficult to navigate efficiently like humans just by the machine vision. But compared to the traditional way of using sensors for navigation, the visual navigation(especially monocular vision) had showed great advantages in low-cost, real-time, high precision and other aspects, which increased its research value. In this paper, we have made a detailed analysis of the edge detection and ranging monocular vision algorithm which were used in the autonomous navigation system of mobile robot based on visual conducting.In view of the theory of machine vision systems, we set up a vision rang measuring system of the mobile robot firstly, then we studied and improved the algorithm to measure range and the algorithm which be used by the monocular vision to detect edges of an image. So a simple and effective visual inspection method is to detect the line of intersection which between the detected obstacle and the ground within its visual range and the use the line to be its local movable area boundary. And the method is also the main idea through this paper.In order to obtain the intersection between the object and ground, we need to detect the edge of the image. While external environmental factors will bring greater influence on the edge detection of obstacles, especially the influence of illumination factor is more prominent, HSI color model can effectively reduce the influence of light on the edge detection as it has the properties of invariant chroma component and separation among the three-component of color, intensity and saturation. In this paper, we have made a measuring method of color better and applicated it in the edge detection algorithm, in this way we could increase the accuracy of edge detection effectively. We improved Prittew operator and the edge information of the pixel poin. Finally, we obtained sequence points of the edge of the target by the template scanning. This article verified the method of obstacle edge detection based on HSI color space by the simulation. The experimental results show that the algorithm of this article applies to the edge detection of unknown indoor environments obstacles, and the algorithm can improve the detection accuracy and reduce the impact of light on the detection res ults effectively.After acquiring the intersection line between the obstacle and the ground, we need to select apt quantity points on the line, then measure the distance between these points to acquire the distance between the mobile robot and the obstacle, furthermore plan the localized movable area. By the means of analyzing the imaging model of the camera and changing the focal length of the camera to calibration many times used Zhang Zhengyou camera calibration method, we can gat the parameters of the camera both inside and outside. This paper has presented a monocular vision ranging algorithm based on spatial geometric constraints, which can bring about the measurement of the distances between the mobile robot and obstacles. The algorithm has many adva ntages such as small amount of computation, good real-time performance and high precision etc.