High-Density Crowd Counting Method Based on SURF Feature

This paper presents a SURF-based method for high-density crowd counting,focusing on overlaying the low counting accuracy in a high-density crowd or open environment.The traditional density-based clustering algorithm(DBSCAN) by adopting minimum spanning tree(MST) is improved,making its minimal search domain adaptive to the distribution of clustering data.Then the SURF features of moving crowd through the improved DBSCAN algorithm is classified.An eigenvector which can represent the moving crowd is built on the clustering results.Finally,the number of crowd through a support vector regressor(ε-SVR) is got.The experimental results confirm that the proposed method have a high accuracy and robustness to the high-density crowd counting.