一体化摄像机的数字图像自动聚焦技术分析-analysis of digital image autofocus technology for integrated camera.docxVIP

AbstractZoom camera is applied widely in the security surveillance field, its rapid expansion is largely due to the development of auto-focus technology. In an intelligent way instead of the traditional manual focus, auto-focus technology not only simplifies the procedure and saves adjusting time , but also greatly improves the focus accuracy. The digital image auto-focus technology analyses the collected digital images, obtains the clarity evaluation, then the result is fed to the optical lens to control the movement of the focus motor util it’s at the clearest position. In this paper, a study of key technologies in auto-focus process is implemented, the problems in the current auto-focus algorithms are analysed and improved, and an auto-focus prototype system is achieved.The influence for focus curve of different auto-focus area selecting methods is analyzed based on the study of the field depth and focus depth principle of the optical imaging system. In order to sovle the problem of poor real-time and self-adaptive in present auto-focus area selecting algorithms, a new algorithm based on particle swarm optimization method is improved, which can greatly improve the speed of selecting focus area and meet the self-adaptive requirements.This paper analyzes the rationality of existing image clarity evaluation algorithms, and summarizes the judging standards of the performance of the clarity evaluation algorithm based on the results of previous studies. In order to solve the problem of poor single-peak and robustness of the corrent evaluation algorithms when the image has greater impact with noise, an weighted algorithm based on the wavelet transform in the first level is proposed. The result of the experiment shows that the algorithm has better performance and can meet the requirements of the fine focusing of the auto-focus system.A focus curve peak search algorithm based on fuzzy reasoning is proposed based on the studies of the typical algorithms. The algorithm com