Real-Time Multi-object Detection and Tracking for Autonomous Robots in Uncontrolled Environments

in this paper a new system is developed for autonomous robots to detect and track multi-objects in uncontrolled environments and in real time for the purpose of decreasing the processing time needed and obtaining better error rates than current systems. To achieve this, a novel multi object tracking algorithm is introduced, implemented and enhanced using multithreading where every thread corresponds to a detected motion area. The implementation considered not only objects tracking but also the object motion estimation to speed up the overall tracking process. In addition, a modified color tracking algorithm is also introduced and integrated in the built system. It is based on the HSV color space. This avoids the problems of overlapping detected motion areas. Experimental results on the built system demonstrate that the proposed system reduced the computational time to approximately 54% compared to published results and reduced the position error to less than 1%.