Target behavior recognition and prediction method in complex dynamic environment

Abstract

The invention provides a target behavior recognition and prediction method in a complex dynamic environment, which comprises the following steps: firstly, establishing a human behavior database and anIFLYTEK speech recognition database; then, carrying out real-time collection on the surrounding environment through a visual sensor and a voice sensor when a security inspection robot is in inspection; afterwards, merging the voice information and the visual video image information, judging the target from overall identification to area locking through target identification and then to individualidentification; finally, carrying out real-time matching on the locked target human motion data and the human behavior database data, performing identification and prediction on the target human behavior, and giving an early warning via the robot when the target human behavior is higher than a certain level of danger. The security inspection robot of the invention can extract and understand the behavior actions of the human body within the monitoring range automatically, efficiently and accurately, and can timely discover suspicious persons or suspicious behaviors and issue an early warning to notify the security personnel, thereby minimizing and preventing the occurrence of dangerous events.

Description

technical field [0001] The invention belongs to the field of robot intelligent security, and in particular relates to a target behavior recognition and prediction method in a complex dynamic environment based on a visual human motion capture technology and a human dangerous behavior intelligent recognition technology combining vision and hearing. Background technique [0002] In the work of intelligent security inspection robots, there is an urgent need to improve the basic ability to deal with emergencies, so higher and higher requirements are put forward for the intelligence of video surveillance systems. The current security inspection robot monitoring system mainly relies on manual identification and monitoring of video images. The disadvantage is that it is difficult for humans to monitor a large number of video images in real time for a long time, resulting in the inability to provide timely early warning and monitoring of sudden dangerous events. prevention. Moreover...

AI Technical Summary

Problems solved by technology

The current security inspection robot monitoring system mainly relies on manual identification and monitoring of video images. The disadvantage is that it is difficult for humans to monitor a large number of video images in real time for a long time, resulting in the inability to provide timely early warning and monitoring of sudden dangerous events. prevention

Moreover, in a complex dynamic environment, that is, in a patrol area with a large flow of people, it is difficult to make a comprehensive analysis and accurate identification and judgment of human behavior in the entire monitoring range even if manual monitoring is performed through video. In fact, We hope that robot vision can keep vigilance at all times like human eyes, and have the ability to identify and predict dangers in the surrounding environment under inspection

However, most of the current video surveillance is usually only used as a tool for deterrence and after-the-fact evidence collection

Images