Research And Implementation On Video Surveillance Intelligent Detection System Based On Ant Colony Optimization Algorithm

Currently, video surveillance system has been widely used in all kinds of industries, and the difficulty of artificial maintenance have been increased due to its extending scale, which will affect the efficiency of real-time monitoring. The video surveillance intelligent detection system mentioned in this dissertation is one of the best solution to solve the problem above. However, in practical application, the system has met several kinds of performance issues such as high memory consumption due to high concurrency level and detection thread overload, which lead to the high concurrency bottleneck of the system. Therefore, the load balancing mechanism is applied to optimize its performance.A variety of relatively mature load balancing algorithm have been invented. Among them, ant colony algorithm performed better in load balancing scheduling efficiency and system throughput because of its solving and information maintenance mechanism. General assignment problem is the main research direction in the field of load balancing. However, the concurrent detection model of the video surveillance intelligent detection system can not be described as this problem due to its sequence-dependent scheduling.According to the problem above, the parallel detection module is described as sequence-dependent load balancing model in parallel machines in basis of the general assignment problem and the characteristics of the parallel detection module of the system. Meanwhile, the local search algorithm of intelligent equalization algorithm is improved, and the improved algorithm is applied to the genetic algorithm and ant colony optimization algorithm. And then the optimization algorithm mentioned above is simulated with the model in order to conclude the most optimal load balancing algorithm applicable to it. Finally, the simulation results are verified in the practical svstem.The innovation of this dissertation mainly includes the following three points:1. Development and optimization of the video surveillance intelligent detection system are completed. The overall architecture of the system is stratificated, and the key technology used for each layer is analyzed. The communication between the modules is implemented using IIS technology and Socket technology. The concurrency performance bottlenecks of database has been solved with data cache technology. The detection algorithm is encapsulated with the help of the dynamic link library technology,.2. Ant colony algorithm and its application in load balancing mechanism is studied. The local search algorithm of the algorithm is improved on the basis of the research. The new transfer price and the concept of neighbor nodes is introduced in local search algorithm to reduce the transfer cost and improve the algorithm performance. The improved algorithm is simulated in sequence-dependent load balancing model for parallel machines, which concluded that the improved local search algorithm can lead to the improvement of execution time and relative load imbalance rate when applied to the ant colony optimization algorithm.3. Load balancing parallel detection framework suitable for the parrelled detection is designed according to the current concurrent detection module. The improved ant colony optimization algorithm proved to be the most optimal in the simulation above is applied to the detection module. Through several rounds of testing, the results concluded in the simulation are further verified, and the concurrent performance of the system is improved.