Computer vision inspired real-time autonomous moving target detection, tracking and locking

Moving object analysis is one of the most important and challenging research subjects in computer vision, video processing and robotics. This analysis is three folds: object detection followed by its tracking and locking. Today's applications demand object analysis in real time, which further complicates the problem. Autonomy is another desired feature of such a system since manual handling of moving targets with a camera platform requires constant attention of humans thus making the tracking task cumbersome, time consuming and erroneous. To addressing these challenges, this research paper presents a complete system for handling moving objects i.e. from their detection to locking. Detection and tracking has been achieved using frame differencing and mean-shift algorithms whereas locking is accomplished using the developed control system. The presented strategy has been validated in simulation environment as well as using a custom-developed hardware platform. The prototype consists of an on-board camera, pan-tilt system and a wooden assembly. The performance of the algorithms has been tested based on recorded video as well as on real time live streaming. Experimental results demonstrated that the system has ability to track moving objects efficiently. The proposed system finds application for simple object detection to complex multiple tracking in various industries like transport, military, sports etc