Design And Implementation Of Motion Objects Detection And Tracking In Campus Environment

With the rapid development of the car, smart driving or automatic driving was needed by people more and more urgently, and automatic driving based on computer vision has get wide attention among the world. In recent years, automatic vehicle research has been processed in lots of universities, research institutions and car companies, and many gratifying achievements have been brought up. Our lab has engaged in automotive electronics for more than one decade and accumulated lots about smart car. From years ago, we started to study a type of pure electric smart vehicle, which had a tiny body. The smart car focused on the more efficient use of energy and automatic driving, and now the tiny smart car could be able to cope with the complex campus environment.Firstly, this thesis described the history of intelligent car. Then it studied the automatic driving in campus environment based on computer vision. Through this solution, we were able to detect motion objects and pedestrian, track all the objects, which providing a visual perception for automatic driving and be able to meet the requirements of real-time automatic driving in a low speed.The main contributions of this thesis were as follows:1) By analyzing the particularity condition of the campus environment, we proposed a solution for motion objects detection and multi-feature detection for pedestrians, which could detect about 80 percent of motion objects and pedestrians, also the solution could meet correct rate and real-time performance.2) For possible problems in multi-feature detection, such as overlapped windows of one object, we fused the windows to make sure there was only one detection window for each pedestrian.3) To deal with motion target tracking, a CamShift algorithm combined with Kalman filter was presented, which can predict the state of the motion objects and improve the tracking accuracy rate.4) Finally, we verified the accuracy and real-time performance of the system by some experiments, and raise up our opinions on automatic driving in campus.