Target Detection And Optimal Coverage Method Based On Embedded Smart Camera

With the increasing application of security and defense system, intelligent video surveillance system is becoming more and more important for preventing crime or terrorism. Because of the high cost and the limitation in processing and storage ability, the current center-based intelligent video surveillance system cannot be adopted in large scale systems. On the other hand, the embedded smart camera has characters of low cost, low power consumption and easy to expand. Therefore, embedded smart camera networks are becoming the trend of future large-scale video surveillance application. In this thesis, we mainly study the system building, detecting algorithm and coverage in ARM-based embedded smart camera systems. It is organized as follows:1. The platform construction of the high performance embedded smart camera system.Based on the analysis of the existing smart camera system, a design scheme of high-performance embedded smart camera is proposed. Our platform adopts ARM processer and PXA270 as hardware system. In addition, we develop a number of software modules for surveillance. 2. The algorithm design for target detection based on embedded smart camera.Considering limited computing ability in smart camera system, we propose a lightweight object detection method based on contour decision and classification. The method consists of contour detection algorithm and regional classification algorithm. Herein, Additive Increase-Multiple Decrease (AIMD) algorithm is developed to detect the contour and then identifies the foreground and background through the classification algorithm. Experiment results show it has high processing speed and robust performance.3. The multi-cameras'optimized coverage strategy.Because of the single camera's limited field of view, it always needs several cameras to cover an interesting scene. Unlike the existing research on coverage, we argue that the coverage algorithms can exploit the dynamic field of view nature of PTZ cameras to reduce the configuration cost by taking into account the heterogeneous coverage quality demands for different targets. Based on the observation, we propose a new coverage definition, termed probability coverage (p-coverage) for PTZ camera networks. To solve the above probability-based optimal coverage problems, a greedy-based heuristic algorithm is proposed. The effectiveness of the method is verified by simulation.