Research Of Pedestrian Distribution Model Based On Analysis Technology Of Multiple Intelligence Surveillance Video

With the rapid development of society and vigorous construction of infrastructure, the proportion of building energy in the total energy consumption is growing quickly. Due to the current building intelligent level is not high enough, there is serious waste phenomenon in the building energy. The energy consumption of air conditioning system and lighting system occupies relatively larger proportion in the building energy consumption. Currently these systems are usually controlled by detecting ambient temperature, humidity, air quality and other factors. But these systems does not consider pedestrian distribution in different areas and different time periods. In this thesis, the pedestrian distribution is estimated by the technology of intelligent monitor video in the view of building energy-saving. The pedestrian tracking results are combined with building plan in order to determine the real-time distribution of pedestrian. The number of pedestrian is counted in a certain period of time and region and the statistical results of pedestrian is used to model the distribution of pedestrian. According to the distribution model, we can set up the operating parameters beforehand for the whole building system and formulate the corresponding energy saving strategies. This thesis mainly includes the following aspects:1. The algorithm of moving target detection based on the classification of background is proposed under the frame of non-parametric background model and the segmentation results is compared with other classical moving object detection algorithms. The hollow number of target object of our algorithm are lesser and the target objects are more complete in real time detection.2. The advantages and disadvantages of the classic target tracking algorithm are analyzed such as Meanshift algorithm, Kalman Filter algorithm and Particle Filter algorithm and the principle of Meanshift algorithm is expounded. In order to track the target more accurately, the color and texture information are combined. The joint information is used as the characteristic of continuous adaptive Meanshift algorithm. In order to improve the resistance to occlusion, the continuous adaptive Meanshift algorithm is combined with Kalman Filter algorithm and then the improved target tracking algorithm based on Camshift is proposed.3. A model of real-time pedestrian distribution is discussed based on multi-source information. By establishing the mapping model, the target objects in the image can be mapped to a building floor plan. We can use the mapping results to complete the continuous tracking of the target objects in the monitoring area of adjacent cameras.4. The algorithm of two-way flow of pedestrian based on moving target classification and machine learning is presented. This algorithm is used to achieve the pedestrian count for a period of time at certain region and then the statistics result is adopted to build a mathematical model. The pedestrian distribution model based on pedestrian flow rate is obtained.5. Our algorithms that involved in this paper are integrated into a unified frame system with the aid of OpenCV and MFC in order to form a complete analysis process about the pedestrian distribution. The integrated system can achieve foreground detection, target tracking, target mapping and human traffic statistics. The system provides a good interface for the further development.