LAUSR

The capability of detecting and tracking targets can play a significant role in mobile robot navigation systems. Visual tracking systems may control the direction and speed of motion of a robot to keep the target in its field of view either by moving the robot itself or the vision sensors. In this work, a compact size tracking and guiding system that could be mounted on weight-sensitive UAV platforms is presented. The system combines a motion detection technique that could be used with non-static cameras in addition to color filtering to detect and track objects in the field of view of a UAV. This hybrid system provides a reliable tracking system for low resolution images taken by a UAV camera. The proposed system implements keypoint detection algorithms including SIFT, SURF and FAST, a motion detection method using frame subtraction and object detection algorithms using color back projection in a hybrid approach that utilizes the best of each algorithm and avoids heavy usage of computing resources. Keypoint detectors SURF, SIFT and FAST are tested and implemented for the purpose of image alignment and frame subtractions. Experimental tests showed the system’s ability to detect and track low detailed targets. The system is tested on a UAV using a Raspberry Pi 2 mini-computer running OpenCV libraries and was able to process eleven frames per second implementing object detection and tracking. The test objects were mainly cars monitored from different altitudes through a UAV downward pointing camera.