Camera active sensing method and system

Abstract

The invention designs an active sensing method and system for video monitoring. The invention provides a high-efficiency cooperative control method for master and slave cameras. A heuristic algorithm is adopted to automatically adjust camera parameters to establish a master-slave camera coordinate mapping model, and a geometric model between a slave camera and a scene is constructed, so that the slave camera can quickly position a target. The invention provides a pedestrian attribute identification method based on target state perception. Pedestrian state judgment is carried out according to the human body key points to guide validity judgment of a pedestrian target attribute recognition result. On the basis, a pedestrian target attribute recognition result is updated based on the multi-frame image sequence, and then a complete pedestrian attribute recognition result is obtained. The invention provides an automatic data labeling method based on master-slave collaboration. The slave camera is utilized to confirm the target detected by the master camera, the master camera periodically generates the background image, the confirmed target and the generated background image are fused to form new annotation data, then the target detection model of the master camera is adjusted and optimized, and the adaptive capacity of the target detection method to a new scene is improved.

Description

technical field [0001] The invention relates to a camera active sensing method and system. Background technique [0002] With the continuous development of society, video surveillance has become an important infrastructure in the fields of urban security, road safety and military. The sharp increase in the amount of surveillance video data makes it more difficult to analyze and use surveillance video. Actively obtaining effective information of targets from surveillance videos is a research hotspot in the field of video surveillance. [0003] The current video surveillance system is becoming more and more intelligent, but there are still the following problems: [0004] (1) The current monitoring system is still based on "passive" perception, and the active perception ability is insufficient [0005] Most of the existing video surveillance systems are mainly based on "passive" information acquisition. They only perform target recognition based on the data obtained by imag...

AI Technical Summary

Problems solved by technology

[0011] Aiming at the problem of slow registration speed and low precision of the master-slave camera, the present invention adopts a two-stage camera parameter automatic adjustment method based on heuristic information, using the positional relationship between the feature point pairs of the master-slave camera picture as the heuristic information, by first adjusting The two-stage method of azimuth and pitch angle and then adjusting the field of view angle realizes the alignment of the master and slave camera images, and then extracts the feature point pairs of the images to establish the coordinate mapping between the master and slave cameras

Images