Study On Target Tracking System Of Mobile Robot Based On Visual Guidance

War on terrorism has become a hot issue in today's world; bomb explosion with high lethality and strong destructive power is used as a primary mean for terrorist activities. The main problem we are facing is that: the traditional way for explosive ordnance disposal (EOD) is operated manually and very dangerous; the EOD field should be screened wirelessly in order to make the telerobot unable to work so that the terrorists cannot detonate explosive through remote control system. Therefore, in this paper, we developed a visual navigation system based on color image for the wheel-pedrail EOD robot. This system can solve the problem by effectively identifying and tracking the target, and inducing the robot to achieve the EOD independently.In this study, we adopted the method of target recognition based on color feature. Hue of singal-channel color image in hue saturation value (HSV) color space was used to establish color histogram for the target area. Using back projection algorithm, the histogram was projected to the original image and we finally got the gray image which reflected the main color distribution of the target. In this image, the value of grayscale is high in area containing the target color; while in other areas the value of grayscale is low. Using this method, it is easy to separate the target from the background. Besides, this method has good robustness on recognizing the target deformation, rotation, and light intensity changes in environment validated by experiment.We used the Mean Shift search algorithm for target tracking. This method using the convergence of Mean Shift vector, could recursively find the center of the target image. For continuous color image, the target position of each picture is supposed to be the starting point for the next Mean Shift searching. Then target tracking can be completed circularly in this way. Kalman filter was adopted to improve the effect of fast-moving target tracking. We proposed a selective strategy that Mean Shift was used for searching window radius in order to avoid the loss of tracking targets. The effectiveness of this selective strategy was validated by imitating the typical moving frequencies which was in accordance with the actual use of the system.In conclusion, a mobile robot visual tracking system was established by using the target tracking algorism described above. This research achieved visual servo control in both vertical and horizontal direction during the process of robot moving. According to the practical application, we did the tracking experiment for typical target, guiding the robot to the specific location to complete EOD. The results showed that the target tracking algorithm has good real-time and accuracy, and this target tracking system has certain flexibility.