Tracking System Of Moving Object In Video Based On Kernerlized Correlation Filter

Detecting and tracking moving objects in the moving background has been a hot point in intelligent video domain for a long time. In this paper, the single moving object with the moving background in videos was detected and tracked by the image pre-processing technique and moving objects detecting and tracking technique.Due to the research involving the simultaneous movement of object and background, the background of video is complex specially, the pre-processing of captured video images was performed firstly in the study. The grey level transformation and gray equalization were used to improve the visual quality of images. The median filter and mathematical morphology dilatation were used in the image denoising.Prior to designing the moving target detecting and tracking system, the local filtering method was employed to remove the irrelevant part of video images for reducing the detecting and tracking function area. Then the moving objects from the processed image were detected based on the combining method of background subtraction and three-frame differencing. The experimental results confirmed that the design has a good detection results and it is not restricted to the abrupt change of nature environment such as light mutation.Finally, in order to reduce the amount of images sequences and computation in tracking part, the KCF algorithm with the small computation and fast operation was introduced in the design of moving target object tracking algorithm. The experiment results showed that this approach can reduces computation in the object tracking.