Pedestrian fast detection method based on videos

Abstract

The invention pertains to the technical field of image processing and pattern recognition, and the invention discloses a pedestrian fast detection method based on videos. Human positive samples are collected in an actual scene in two modes including the automatic mode and the manual mode to train a human star detection template in a particular scene; video monitoring images are acquired and image pretreatment is performed so as to perform appropriate compensation on illumination mainly; treatment is performed on time domain information of an image moving object to extract the foreground region of the moving object; the pyramid layering range is determined according to the size of the foreground region to judge the foreground region whether needs to be fused, and the pyramid layers are redivided; matching is performed between the trained pedestrian template and the processed image to determine the pedestrian position; and fusion optimization is performed on the pedestrian position to select the optimal pedestrian position and mark the pedestrian. The method of the invention is suitable for the field of tracking of a plurality of human objects in different scenes to satisfy the detection robustness and real-time performance requirements at the same time.

Description

technical field [0001] The invention belongs to the technical field of image processing and pattern recognition, and relates to a video-based rapid detection method for pedestrians, which is a real-time detection method for multiple pedestrians under different lighting and background environments, and is suitable for detecting dense crowds in various scenarios detection. Background technique [0002] The purpose of pedestrian detection is to enable the computer to intelligently and automatically identify pedestrian targets from image or video data. Human detection and recognition is also an important research field in computer vision and pattern recognition, which can be widely used in video surveillance, robot vision and virtual reality technology. Conditional disturbances such as pedestrian movement occlusion, illumination changes, and complex background environments. [0003] Existing pedestrian detection systems generally include two parts: ROI segmentation and object ...

AI Technical Summary

Problems solved by technology

However, the calculation of the detection process is huge, and it cannot adapt to lighting changes and complex background environments.

[0006] Title of paper: Cascade object detection with deformable part models, Journal: IEEE conference on. Computer vision and pattern recognition, Year: 2010, Page 2241 to 2248, proposed a cascade detection scheme to improve detection without reducing detection accuracy speed, but still cannot achieve real-time detection

Images