Design of Object-tracking Algorithm Based on Mean-Shift with Artificial Neural Network

How to make the distance learning process more interactive becomes one of the crucial issues in distance learning today. The use of object-tracking in the learning process to move the camera following the teacher in real-time during the learning process can improve the interactivity of the learning process. Mean-shift is one of the well known object-tracking algorithms because this algorithm is relatively easy to implement and has relatively low computational loads. It works iteratively to locate a predefined object in an image. To decrease the execution time of the mean-shift algorithm, the number of iterations required for the algorithm to reach convergence needs to be reduced. This study aims to develop a faster object-tracking algorithm for indoor distance learning cases where possible changes in lighting and other factors that may affect the color intensity of the image can be excluded. The proposed algorithm is modified from the original mean-shift algorithm, alterations from the original mean-shift algorithm are made to the determination of the starting point of the object search, the maximum number of iterations and threshold by using artificial neural networks as well as resampling the target object for each frame.