Research On Crowd Density Estimation In Video Surveillance Of Airport Terminals

With the sharply increase of the air passenger volume in recent years, various emergent accidentsare more and more frequent, causing heavy casualties and property damage. It is of great importancefor society stability to study how to prevent the crowd density from being too large. Due to the highpassenger volume and wide surveillance region in the airport terminals, research on crowd density es-timation using video surveillance systems is of profound significance and scientific practicability.The traditional method of crowd density estimation based on video surveillance systems is stu-died in this paper. A new effective background generating and updating algorithm for color video isproposed first and detects crowd by subtraction from the background using the HIS color space. Thismethod improves the accuracy of crowd density estimation based on pixel statistics. In the estimationof crowd density based on texture analysis, gray level co-occurrence matrix is used to extract the tex-ture features of crowd images. Then the parameters of SVM with RBF kernel are rapidly selected onthe basis of segmented dichotomy and the SVM obtained has a high generalization performance toclassify the degrees of crowd density.According to the environmental characteristics and applied requirements in the airport terminals,the methods for crowd density estimation are optimized. This paper proposes a moving object seg-mentation algorithm based on on-line Gaussian mixture model and texture information to solve thecolor similarity problem between foreground and background. And then the crowd is extracted com-bined with subtraction. A new technique for calibration to enhance the previously-developed methodof crowd density estimation using a reference image is presented aiming at the case that the number ofcameras is large and there are many similar surveillance scenes. The experiment results show that thismethod is fast and effective in the airport terminals and it can assist in the warning system of airport.