Research On Algorithm Of Counting High-density Crowd Based On SURF Feature

With economic development, populous cities have sprung up all over the world one by one, and the people’s living standards have improved physically and mentally. The people’s flocking to the cities frequently, not only bring new labor and new atmosphere to the city, but also make the city more and more crowded, especially on weekends or during the holidays. In the public places, unfortunate accidents caused by high-density crowd occur frequently. At the same time, the video surveillance systems are ubiquitous. If we make use of the existing resources, so these intelligent systems can effectively forewarn and avoid disaster events. Compared with the traditional approach, the intelligent system of counting and density estimation can also improve the utilization rate of public facilities, and arrange the allocation of manpower and material resources.These fellows are main innovation of this paper:1) To improve the accuracy, the shadow removal was added in the process of feature extraction. The traditional statistical algorithms groups considered excessive feature extraction, but the shadows noise processing was ignored invariably. In order to improve the efficiency of feature detection, a shadow removal approach base on HSV color space was used for processing the image in this paper.2) To take full account of the distortion generated in the video, a perspective transformation method of correction was introduced in the process. Because of the angle of the camera and other issues, image distortion exist mostly, the distortion will influence the accuracy of feature extraction, so, a perspective correction method based on normalized was used of. Then the SURF feature was extracted, and feature vector was constructed, so machine-learning approach was used for learning and testing video. The experimental results show that the correlation of SURF points and the people’s number is increased by correcting.Finally, several sets of data’s comparison with experimental results also proved that adding shadow removal and perspective transformation in process is good for experimental results.