An Improved Object Tracking Algorithm Combining LCT Kernel Correlation Filtering and Kalman Prediction

In the process of long-time object tracking based on LCT, aiming at the problems of re detection after tracking failure in the case of scale change, background similar interference and severe occlusion, the paper presents a better algorithm based on the combination of LCT kernel correlation filter and Kalman prediction. The Kalman filter is introduced into the re detection module of LCT. When the object tracking fails, the Kalman filter is used to predict the position of the object in the current frame before the re detection. Compared with the original algorithm, it avoids the whole frame image traversal search, reduces the search range of the object, and lower the interference of similar objects in the background. The experimental results show that the improved LCT algorithm has better accuracy and rapidity, and has better tracking performance in the case of object occlusion.