AdaBoost, HOG, LBP, Pedestrian Detection, Urban Navigation.
This article proposes the design and implementation of a low-cost vision based navigation mobile robot that tracks pedestrians in real time using an IP camera onboard. The purpose of this prototype is the navigation based on people tracking keeping a safe distance by PID and on-off controllers. For the implementation we evaluate two pedestrian detection algorithms: HOG cascade classifier and LBP cascade classifier off-line and onboard the robot. In addition, we implement a communication system between the robot and the ground station. The metrics of evaluation for the pedestrian detection proposals were precision and sensibility, obtaining better results with HOG. Finally, we evaluate the communication system, computing the delay of the controller response; the results show that the system works properly with a transmission rate of 115200 bauds.