Study On Embedded Monitoring System Used For Crowd Density Estimation

Crowd density estimation is the important content of intelligent crowd surveillance. By estimating the crowd density under different scenarios the relevant departments can safeguard the public security and facilitate the traffic. We try to study an efficient estimation method of crowd density level and apply it to embedded system for real-time monitoring.This paper consists of two parts: the first part is the research on feature extraction and recognition of crowd density image. The gray-level co-occurrence matrix(GLCM) is used to get the texture information of digitized images and how to choose the information is studied, then the optimum crowd density feature vectors were found from GLCM. The feature vectors are used by a support vector machine(SVM) based on statistical learning theory for automatic classification. For better classification precision, the multi-class SVM model is build and it's classification results were compared with artificial neural network to testify the feasibility and advantage of this method. The second part is to build a low cost and high performance embedded crowd density surveillance system and design it's component modules reasonably to accomplish the system function. The system with web server applies the theories above and can satisfy real-time monitoring requirement basically and runs steadily, having strong practical application value.