Moving Object Tracking Method Based On Curvelet Transform

Video target tracking technology is a cutting-edge field of computer vision research.The technology has broad application space in many aspects of traffic monitoring, medical diagnostics and safety monitoring. The nature of the video object tracking technology is by comparing continuous video sequence and the use of estimates methods and similarity comparison method to find the target, and then marking the position of the target. Most tracking algorithms are only carried out for a certain class of problems, such as fast-moving target, slow-moving target, and the target in occlusion. Wavelet transform has been widely used in target tracking, but the description of the characteristics of its image is not ideal.Curvelet transform is a new multi-resolution analysis method, which has strong directionality and anisotropy and can efficient extraction of image characteristics.Firstly, this paper describes a target tracking algorithm based on Curvelet transform.This algorithm extracts the target template as a target to be tracked, then selects the energy of Curvelet coefficients to establish feature vectors, and uses similarity function to determine the final destination. This algorithm only relies on Curvelet coefficient in addition to other parameters, so a small amount of calculation of the algorithm to implement. However, when the target was occluded by other object. the representations of target with the energy are no longer accurate and the position is not true. In response to this shortcoming, this paper proposed the particle filter tracking algorithm which combine Curvelet coefficient energy and color features. The algorithm takes full advantage of the energy invariance Curvelet coefficient between each frame of the image, and optimizes the particle filter tracking algorithm based on color characteristics when background interference or occlusion the target also can be tracked. Experimental results show that the particle filter tracking algorithm combined the color feature and Curvelet coefficient can effectively track the fast moving targets, slow moving targets and the targets which are blocked by other targets.