People stream statistical method based on depth information

The invention relates to the technical field of people flow statistical methods, and discloses a people flow statistical method based on depth information, and the method comprises the steps: a feature extraction module: inputting a video sample into the feature extraction module, and generating a multi-scale feature; the patch selection module introduces the thought of local information, learns scales of different levels by adopting a quartile method strategy, divides pictures into four patches according to the distance to extract different feature information, and reflects crowd distribution information of dense and sparse degrees; according to the depth information module, depth information weight parameters are introduced to represent the distance degree of the distance between people and a camera, and crowds with different densities are divided according to scales according to the distance degree; and the feature fusion module is used for fusing the feature information of the four patches by adopting a m