Based On The Space-time Characteristics Of The Movement Of The Vehicle Across The Camera Tracking Technology Research And Implementation

In recent years, people is more and more care about the security of surveillence.Intelligent video surveillance system attracts people’s attention increasingly.Intelligent video surveillance system can track the target and analyze its behavior, These information can be helpful for decision.At present,the technology of moving target tracking is mainly in the single camera view.There are some problems due to the scope of single camera.Therefore,the implemention of tracking is limited. Do research and realize the camera surveillence system under the cross-regional to track moving target is very necessary.Multiple camera surveillence system has a wide field of view and can track target for a long time.Therefore,this system has a wide application in the modern life.In the field of camera surveillance, the technology of data fusion is a obstacle. The camera data fusion technology is also can be regard as target handover. It is the key point of multiple cameras surveillance, Only under the condition of the problem being solved,We can achieve the goal of cross-regional tracking.In this paper, we design and implement a mutiple camera surveillance system for overlap cross-regions,the system can track moving targets under the complex envirorment.The main contributions in this paper are as follows:At first,we introduce the related key techniques and algorithms, such as objects detection, objects tracking and objects handoff in multi-camera surveillance in detail. We make analysis and summary some key technologies so as to lay the foundation for the subsequent research work.Second,according to the spatial information of background images,we do reseach and analysis the image matching technology.Then,we put forward a approach to recover camera field of view lines based on SURF feature.This method can draw the public horizon area between two cameras accurately and fastly. Through a large number of experiments,We can verify the accuracy and stability of the algorithm.Third,based on field of view which is generated from spational information and the time characteristics of moving target,we design and implement a method to handover target in multiple cameras. When the target enter the public area of cameras,We can predict the target’s postion in the next camera.A large number of experimental results show that using this method can handover target quickly and accurately.Fourth,under the completion of two works,We design and implement a prototype system based on two cameras which can track vehicle through cross-regional areas. through a large number of experiments, we can verify the system can achieve the goal of cross-regional continuous tracking with high robustness and practicality.