Biomimetic Amphibious Spherical Robots And Its Object Detection And Recognition System

Recent years, the technology of robot is developing rapidly and becoming more and more widely applied. One of the most important development area is the intelligent mobile robot control including the target recognition and tracking. Therefore, the construction of mobile robot vision system is an important process whose function is to extract useful information from the outside world scenery image helping the moving robot implement target identification, tracking, and path planning.In order to achieve tasks like the detection of complex environment, target recognition, this paper proposed and designed a set of biomimetic robot system to meet the requirements of flexible movement stability and bearing capacity. At the same time we designed a realtime target detection and tracking process. Besides video fusion technology can be used for the establishment of multiple robot collaboration system.In this paper we focused on the structure design of the robot system and the construction of object detection system. Also we applied video fusion technique as the function extension of our vison system. The most important part of our paper is the algorithm optimization to further enhance the efficiency of the moving target detection and tracking, including reducing the candidate image patch numbers and lowering the computational complexity by utilizing the compressive projection. The proposed algorithm is based on the framework of TLD with the ability of long-time tracking, scale invariance in the tracking process, BING method was utilized to reduce the sample numbers, and compressive projection matrix was introduced to lower the complexity of feature computing. This paper also designed serval experiments to verify the efficiency of proposed algorithm, the optimization algorithm has the more obvious improvement in accuracy reached 58%, frame rate from the 16 FPS of TLD to 27 FPS, lower than CT algorithm 38 FPS, however, it could fully meet the requirements of the real-time target detection robot system. And we designed lots of experiments to verify the function of our system.At the last part of the paper, we present our contribution in the multi-robot system by utilizing the video fusion technique, which could resolve the problem of shade in single robot vision.