A novel evaluation method of degree of order for passengers on trains

Existing detection methods of abnormal movements focus on model construction and similarity identification of objects in the images. The result of detection is not satisfactory when the number of individuals in the detection area is very large. A new method is proposed to solve such difficulties by analyzing the vehicle environment and passenger groups. Firstly, human head is segmented and extracted based on hair color and contour characteristic, and then, the degree of dispersion of head centroid is calculated precisely. At last, an evaluation model is built to evaluate the degree of order. Experiments show that the results of the evaluation can well reflect the actual situation and the method is able to carry out real-time monitoring of degree of order.