| With the development of society, human activities are increasingly diverse and complex. Wide-area video monitoring is one of the most important means of maintaining social security and stability and criminal prevention. In order to meet the needs of monitoring environment and obtain the moving object’s comprehensive information, we need multi-cameras’collaboration. Duo to the bad factors of economic costs and environment, multi-cameras monitoring technology with non-overlapping field of view (FOV) is more and more popular. Because there are more and more cameras distributed in our life, the data needed to be dealt with is increasing. In order to improve its efficiency, using distributed system is an important method.But how to make multi-cameras distributed on different computers collaborate with each other is the key point. So, in this paper, we combine Agent with the optimal search theory to build a theoretical model, and apply it to the non-overlapping cameras system. The main research work is as follows:Firstly, we propose an optimal search control model based on Markov. Considering the process of object moving among cameras being similar to the transition of states of Markov, we apply the Markov to the cooperation among cameras. Besides, we compare this model with other models to prove its feasibility.Secondly, we use Socket to realize the real-time two-way communication between Agents based on ACL. With Socket, Agents will improve its stability and security of communication. What’s more, the system will realize distributed processing.Thirdly, the framework of system based on Agent and optimal search theory is put forward. In this framework, the intelligent Agents stand for cameras. With the new model, the system will search as the optimal search plan and know the next camera. Then it will inform the next camera to do object tracking.Finally, we use JADE platform to build an intelligent Multi-Agent system, and realize the Non-overlapping cameras system based on multi-agent and optimal search theory initially. Then we use real video data to conduct a series of experiments to testify the effectiveness of the framework. |
