Research On Binocular Stereo Vision Of Unmanned Vehicles

The traditional ground unmanned devices usually use sensors like single line laser,four lines laser and CCD to perceive complex environmental information. Since each sensor has it’s own unique limitations,and some error factors exist when multiple sensors have their data fusion, the system accuracy is low when the system is used in the natural environment. For mobile unmanned vehicle which used in the actual battlefield environment and complex road environment, the context-aware sensing precision subsystems, environmental adaptability and rapid real-time detection determines whether an unmanned vehicle can truly reflect its autonomy and manipulative. Binocular stereo vision technology has incomparable advantages like fast sampling rate,image-rich information,strong anti-interference ability and low maintenance costs compared to other sensors and you don’t have to install other equipment, just to do the CCD camera calibration and improve the algorithm for binocular stereo vision. In this way you can enhance the environmental awareness of unmanned vehicles and meanwhile save costs. The issue takes an unmanned vehicle of No.58 Research Institute of China Ordnance Industry as a platform, using Binocular stereo vision technology to solve problems which the unmanned vehicle could meet in environment perception process. Such as other sensors like laser radar has low precision and real time,and easy to be interfered. This paper describes the form of environment perception system,system principle, hardware platform design, and system work flow of an unmanned vehicle platform.Doing the following jobs such as calibrating and fusing data between the CCD and other sensors,researching on binocular stereo matching technology of CCD,testing of matching algorithms on Middlebury platform,the practical application of binocular stereo matching technology in the environment perception system on an unmanned vehicle.To enhance the system performance of the unmanned vehicle’s environment perception system, this paper made the following research:1. High precision calibration technology for the unmanned vehicle. Improving the measurement accuracy by calibrating CCD camera with the one line laser radar, four lines laser radar, and millimeter wave radar.2. Efficient binocular stereo matching technology for the unmanned vehicle. Choosing the Yoon algorithm to realize by comparing the advantages and disadvantages of many kinds of binocular stereo matching algorithms. Writing the test procedures and getting the disparity maps to prove the superiority of Yoon algorithm compared to the SSDA algorithm.The application results shows that the binocular stereo vision technology has greatly improved the unmanned vehicle platform environment perception system precision, speed and real-time performance.