Applying the latest developed video mosaicing technologies to practical video surveillance systems, the intelligent panoramic video surveillance technology attempts to solve the intrinsic contradictory between wide field of view (FOV) and high resolution existed in current surveillance systems. The intelligent panoramic video surveillance system we developed in this work has both theoretical significance and practical implication to real-world video surveillance system. We developed the concept, architecture, algorithm and software for building practical intelligent panoramic video surveillance system. Based on the high resolution and wide field of view image generated by the system, we can perform accurate and reliable target detection, tracking, classification and recognition algorithms. We also developed a new technique to calibrate the novel oval-fisheye image. The main contributions are summarized in the following. (1) We proposed the concept and architecture of the intelligent panoramic video surveillance system, studied the system hardware construction, the functions of the components, and the data flow as well as the control signal flow, analyzed the architecture of the software platform and flow of core functions, and tested the integrated system. (2) We investigated the theories and algorithms of surveillance video mosaicing. Based on some background knowledge, we analyzed their basic principles, implementation steps as well as the strong points and weak points when applied in surveillance environments. (3) According to the characteristics of the video stream, and based on manifold theory and graph theory, we proposed a novel constraints-based graph embedding optimal video surveillance algorithm and embedded it into a mature real surveillance system. (4) We studied the SIFT feature based video mosaicing method which includes feature extraction, homograph estimation and re-projection process and the cylinder model as well as sphere model based video mosaicing algorithms. We also tested these algorithms in real video frames. (5) According to the imaging characteristics of the latest designed oval-fisheye lens, we analyzed the theory and practice of the image correction algorithms and tested them with real video frames. KEY WORDS：Video Surveillance; Video Mosaicing; PTZ Camear; Panoramic Surveillance Systems; Fisheye Image Correction; Oval-fisheye Image
修改评论