Research And Implementation Of Embedded Target Detection Algorithm

Along with the progress of the society, the demand is growing for intelligent video surveillance. The common intelligent video surveillance system will send the video data to the PC, and realize the target detection in the PC. But such a program has many problems in the video transfer process, such as large data transfers, time-consuming and Long-distance. And the general target detection algorithm robustness to external environmental changes, such as illumination changes and camera jitter, is relatively poor. Once the light changes or the camera has a slight jitter in target detection, the target detection result will be great fluctuations.In response to these issues, On the basis of in-depth study of the general target detection and image processing algorithms, this paper proposes a target detection algorithm for the embedded environment. The algorithm realizes the moving target detection in embedded environments, and uploads the statistical target detection results. This avoids the cumbersome process of video data transmission. On the foundation of the traditional background subtraction, the target detection algorithm integrates the image matching and edge detection algorithms, so that its light changes and a slight jitter robustness is enhanced.The embedded target algorithm include:On the basis of retaining the traditional background subtraction’s background model creating and updating, the SSDA image matching process is added to the algorithm, In order to achieve the detection of frame around the image dislocation caused by jitter; The algorithm uses the characteristics that edge feature is not sensitive to the change of the illumination, and changes image segmentation process of background subtraction into the refined edge’s mutual fuzzy comparison process between the background model and the foreground frame; Finally, the fuzzy comparison results are statistical processed to obtain the final target motion information.In order to design and verify the embedded target detection algorithm. Firstly, Visual Basic.NET-based software platform is built up, this platform realize the validation and comparison about the relevant target detection algorithm, the image matching algorithm, the edge detection algorithm and the proposed embedded target detection algorithm. The experimental results show that the algorithm has good robustness to illumination variations and the slight shaking of the camera, and can accurately detect the emergence and movement changes of the target. Secondly, STM32-Based embedded target detection system is built up, and the proposed embedded target detection algorithm is transplanted to this system for testing. The test results show that the algorithm is simple, easy to transplant to the embedded environment. The disadvantage is that the algorithm has lots of calculation, and the system real-time is poor, but some of the higher-frequency micro-controller can be used to improve the real-time.