Research On Obstacles Detection Method For Mobile Robots Based On Active Panoramic Vision

With the development of economy and society, mobile robots will participate in more and more social fields and play an increasing role. In many mobile robot applications, the obstacle detection is an indispensable task. It is extremely important to detect obstacles quickly and effectively for some tasks of mobile robots such as obstacles avoidance and navigation.In recent years, due to the rapid development of digital image processing technology as well as a large number of digital image processing equipment performance improves, the application of vision sensors used in mobile robots is more and more widely. But the existing mobile robot vision systems have faced a lot of problems, such as high cost, difficult to maintain, large amount of calculation, limited detection range, difficult to detect dynamic obstacles and other issues. To solve these problems this thesis designs and realizes a mobile robot vision system based on Active Omnidirectional Vision Sensor(AODVS) for omnidirectional and fast detection.The main research work is as follows:1. An Active Omnidirectional Vision Sensor(AODVS) is proposed based on the single-view panoramic imaging technology. AODVS is integrated ODVS with single view point and a planar laser generator composed by four red liner lasers that installed on the same plane. All obstacles around the mobile robot can be detected in real time by AODVS.2. To solve the problems of excessive consumption of computing resource, poor real-time performance and limited detection range of current mobile robot vision system, a fast and efficient obstacle detection method is proposed in this thesis based on AODVS. According to laser information on the surrounding obstacles projected by planar laser generator, the obstacles distance and direction information are obtained with robot vision method.3. According to the characteristic of the active panoramic vision obstacle detection method, an omnidirectional obstacle avoidance strategy is proposed. The robot moving direction is divided into 360 candidate orientation, using the obstacle information and direction while the evaluation function calculates the optimal traffic direction to help realize the local path planning of mobile robot.4. Moving obstacle detection method was discussed under the dynamic background based on AODVS. AODVS can acquire a wide view of 360 degrees, it will cause inconsistencies in the background motion image due to distortion of the omnidirectional image. So it could not extract moving obstacle through compensating frame difference directly. In this thesis, the image is divided into grid windows and the area of moving objects can be obtained from the background transformation-compensated frame difference using every affine transformation for each window. Finally, according to the imaging characteristics of AODVS, we can obtain the distance and orientation information of moving obstacle by visual method.