360°Full Viewing Angle High-resolution Network Video Surveillance System

In recent years, with the network technology, transmission technology and video compressiontechnology and other related technology continues to evolve, network video monitoring system hasbeen has penetrated into all areas of our life. With the continuous expansion of its applications,more and more new demands have been proposed, such as a wide range of public places’ real-timemonitor with no blind spot, that is to achieve continuous monitoring360-degree range goals at thesame time and the same region. As the view of traditional video surveillance system’s lens is lessthan60degrees, to achieve large-scale non-blind control, we need to stitch multiple video cameras’images or use a fast mechanical rotation camera. However, using these planes, the control systemwill become complex and expensive, and it is difficult to achieve seamless between the screens andthere are blind points on time.In order to achieve a single non-rotating lens panoramic wide range of monitoring, this paperuse the fish-eye lens which have more than180degree view as surveillance cameras, to build a360degree high-resolution network-based video surveillance system. This so-called360-degreeviewing angle control means, when the fish-eye lens shooting straight down, we can get all the360degrees scenes under the lens, without turning on a camera to monitor the blind spot monitortargets.In this paper, first analysis the existing fish-eye lens imaging principle, domestic and foreignfish-eye correction algorithm. Then in order to get a set of low distortion and real-time fish-eyealgorithm for the surveillance video, we improve the algorithm according the need for videosurveillance, including fish-eye surfaces projection model to improve investment, all roamingmodeling perspective, select the appropriate difference algorithm, look-up table design. Systemmonitoring using the existing camera with a fish-eye lens to capture and compress the video data.The compression format is MPEG-4and we use the SDK comes with a video camera to get videocompressed data. We send data to the client through the P2P technology based on UDP networkprotocol. The client receives and extracts network data in real time base on DirectShow framework.According to the selected fish-eye calibration mode to correct the decoded image data of eachframe, and roaming with the mouse for interactive display. System test results show that single frame size1600*1200fish-eye image correction time isno more than40ms, meeting the real-time requirement of video surveillance. Fish-eye correctionworks well, the corrected image is very close to the real scene. Moreover, we realize two modes ofinteractive panoramic roaming view. All these results achieve the desired goals.