Crowd density Test paper notes __ pedestrian density detection

Forecast: Video surveillance applications, the camera often observed a constant background, and this is where we are not interested in, therefore, we need to extract from each scene the need to study the moving objects or targets, commonly used in the following three methods: A. Frame difference method to obtain the target by the difference operation between two adjacent frame images Contour, which is the simplest method for fast detection of motion foreground in computer vision, but because of the complexity of the actual scene, this method has poor ability of suppressing noise and limited use situation.
B. Background subtraction
C. Optical Flow method
Four methods based on gradient, matching, energy based and phase based
Shadow elimination
Shadow detection and elimination algorithm: a. Attribute-based methods The shadow detection based on the properties of shadows in brightness, saturation, chroma and texture is robust to different scenes and illumination conditions. B. model-based approach
Based on the model, a shadow model is built by using the prior knowledge of moving object and scene, which is usually used in a specific scene, such as aerial image understanding.
Feature extraction in pedestrian statistics (2)