Video Object Tracking Research On Spatial Constrained Coding Algorithm

One of the biggest dream of entire human race is to let the robot have the same vision as human themselves. During the latest 100 years, the information technology have been developing very fasts. With the continuing development of computer vision field, the object tracking technology having been innovating persistently. The purpose of object tracking is designed for continuous image sequences which have different lengths of time, each of these images has the moving object which mast be locked down. A good video object tracking algorithm is generally required to meet two basic requirements:1. Performs on real-time, that means the algorithm’s processing speed should reach a certain degree; 2. Be outstanding robust, that means however composite the image scene is, or the target’s pose or position changes a lot, the algorithm can always locate the object accurately. But different from theoretical research, in practical applications, video tracking technology is yet have to dealing with easy or hard problems, such as complex scenes, object’s suddenly pose change, changes in illumination which happened in image scenes, etc.In this paper, common analysis of the situation that appear in real life such as a similar object interference, blur that happens in object’s motion, being short of contrast, partial occlusion, and last, the changes caused by illumination have been made, which exists in the tracked target video sequence a lot. And some major research results are also be received which shown as follows:1. A method of object tracking which lay the foundation of Spatial Neighborhood Constraints Coding is proposed. This method uses a new restrained policy, i.e. the constraints of spatial information of the encoding weighted code model. Different from other methods which only take the pixel’s color value into consider, this model is encoded by separately considering the gradation characteristics of the weighted pixel and adjacent pixels coded weighted distance between them. The model treating spatial information as an important part, the robust code of corresponding pixels of the frame in complex scenes is able to be received. It enhances the coding method’s stability, and the object tracker’s robustness is higher than before, during the whole act of object tracking, it has achieved tracking results of high efficiency and accuracy, compared with than other methods.2.Based on the space constraint-based video coding tracking method is proposed in (1),a target tracking method with Mean shift algorithm being combined is proposed. While using spatial neighborhood constrained coding model to obtain the target pixel’s accurate code, This method adds Mean shift algorithm, which advantages are:firstly, its computing cost can be quite cheap, so in the circumstances that target’s area is under known, its tracking speed is 24 frames per second; secondly, this algorithm is insensitive when dealing with the edge of the block, the rotation of the object, deformation caused by some aspects, background change. Thus while being sure the algorithm is highly rubust, its ability of doing real-time tracking is also improved, so that when faced with complex scenes the proposed method can also be accurate in location and tracking.