The Research Of Scale And Orientation Adaptive Research On The Dense Feature-Weighted

The object tracking technology is an important research direction in the field of computer vision, how to calculation the location of target accurately, and find a way to estimate the scale and orientation of the target are two difficult problems needs to solve. At the same time, The real-time is also an important factor in evaluating the object tracking algorithms. Mean-shift tracking algorithms has the characteristics of simple and efficient, therefore, Mean-shift tracking algorithms is widely used in object tracking. Most of Mean-shift based object tracking algorithms have neglected the dense feature spatial distribution information.Objective: Most of Mean-shift based object tracking algorithms have neglected the dense feature spatial distribution information. in this paper, making effective use of the dense feature to enhance the reliability of tracking.Method: There are some color features gather together on tracking objects, and each of those features forms a region of certain size. These dense feature regions play an important role in human vision. The spatial structure information of these feature dense regions can be used in object tracking. An effective and efficient tracking object model is presented. The intensive features are found, and then the areas and distances between the dense region centroids and the target object center are calculated to obtain the features? weight, which is applied to describe the tracked object. The intensive features in the target model have bigger weights, at the same time the weights of the discrete features are smaller relatively. At the same time, zeroth order moment and the similarity coefficient between the target model and candidate models are used to estimate the area of the target. Then an area compensation method is used to compensate the object areas which are weakened due to background weighting processing. Finally, the estimated area and the second order center moment are used to adaptively estimate the object scale and direction. The object models are updated when there are great changes in background.Result: Experimental results show that the proposed method can adapt well to the object scale changes, and the average accuracy of tracking is above 94.6%, has higher accuracy and efficie ncy and more robust in comparison with some state-of-the-art methods.Conclusion: The proposed method increases the weight difference between different features in the target model and has a great advantage to distinguish the target from background; The area compensation method solve the problem that estimate the target area is less than the actual area which cased by the target feature-weighted was weakened.