Estimation Method Of The Crowd Under Security Surveillance Video

China is the most populous country in the world, with the "urbanization" process continues to accelerate, the rapid development of urban construction has led to a lack of overall construction process planning, caused the population of cities and industry too aggregation, which brought many unsafe factors, safety accident death rate to rise. How can we live in a safer place, how to build a strong security network to ensure the safety of the city, it has become a pressing need to resolve the matter. Through intelligent video surveillance monitoring for each region to provide real-time crowd number estimation can not only help the relevant departments for effective public safety maintenance of key areas, but also can improve the efficiency of their work, to the timely processing feedback; finally realizes safety accidents from the "post control" to "preventive" transformation. Therefore, the use of video surveillance carried out accurately and efficiently estimate the number of people become very meaningful.In this paper, the number of people at this stage for the estimation algorithm is studied through comparative analysis shows that the number of people directly estimation methods need to be very accurately detect pedestrians in the video can be accurately counted, and in the relatively large number of pedestrians or the crowd there are a lot of occlusion when the crowd hardly accurate single pedestrian detection, segmentation; estimation method for number of pixel statistic characteristics of the population based on the algorithm is simple and easy to use, fast operation speed can meet the real-time demand, in small number and no effect under the condition of occlusion is good, but affected by the perspective effect and occlusion factors; the number of population estimation method based on texture features suitable for the detection of high-density population under the influence unobstructed and perspective effects and other factors, but its very operation complex; the population based on the number of feature points similar to the estimation method based on the statistical characteristics of the pixel method, but also to establish a function relationship between the number of features and the number, and is used to solve the problem, but feature selection and extraction is more complicated.In view of the above problems, this paper presents a method to estimate the number of people taking into account a priori knowledge. In real life, when the population density, the relative space is limited, the movement of people will be smaller. Consider this a priori knowledge defines a description of the degree of change of population movement characteristics, named sports strength characteristics. In foreground block units, using background subtraction and frame difference method to get exercise intensity characteristics, combined with the statistical characteristics of the pixels characterize by the SVR. Experimental results show that this method can increase the number of people without completely blocked in case of estimation accuracy. In addition, for appearing in the video surveillance perspective effect, this paper proposes a perspective correction method based on statistical regression.