Single Object Tracking Using Estimation Algorithms

The application of Kalman Filter in the process of state estimation and thereby tracking a single object in motion is explored in this paper. A collection of images consisting of 200 different instances of the single object's position has been taken into consideration, whose location has been found with the help of background subtraction technique. The actual trajectory has been obtained by connecting the centroid locations of the obtained images of moving object. This paper incorporates the use of traditional Kalman filter to estimate the position and the trajectory of the single object in motion. The performance of the traditional Kalman filter has also been compared with a proposed modified version of Kalman filter for this challenging job. An exponential function has been multiplied with the Kalman gain in the modified Kalman filter. The performance evaluation shows that the modified Kalman filter generates improved results with high convergence rate and low tracking error compared to Kalman filter. The work presented here has enormous potential in the field of object tracking and navigation for different practical applications.