Remote brain-like three-dimensional gait recognition system for complex visual scene and implementation method

Abstract

The invention provides a remote brain-like three-dimensional gait recognition system for a complex visual scene and an implementation method. According to the invention, intelligent detection for the object motion is realized through a low-power-consumption ARM embedded system, a somatosensory camera, an intelligent light sensing and night LED light supplementing system by integrating voice, infrared and visual sensors, and human body motion data is acquired through a gait sensing system formed by a fixed-focus camera, a laser ranging sensor, a high-definition aerial photographing unmanned aerial vehicle and a wireless charging system and transmitted to a cloud brain-like gait recognition system by using a mobile network for analysis and recognition. On a cloud brain-like computing platform, gait data under various complex scenes is generated according to a virtual pose synthesis method, and the moving body detection accuracy in a complex scene is improved by using a method of training and learning a moving body detection model by the data. The remote brain-like three-dimensional gait recognition system for a complex visual scene has broad application prospects in various fields and has excellent economic and social benefits.

Description

technical field [0001] The invention relates to a long-distance brain-inspired three-dimensional gait recognition system and an implementation method for complex visual scenes. Background technique [0002] The current detection of human gait behavior can be divided into image-based 2D gait recognition, 3D gait recognition and other gait recognition methods according to the data source. The 2D method uses ordinary cameras for data collection, training and recognition. Due to the wide distribution of 2D cameras and the fast processing speed of 2D images, it is relatively widely used. The 3D gait recognition method mainly realizes the human gait recognition through the method of 3D human body modeling, and the 3D human body modeling usually requires the cooperation of cameras from multiple angles to complete data collection and 3D model synthesis. [0003] Other gait recognition methods mainly include: wearable gait recognition method, which collects human gait data through m...

AI Technical Summary

Problems solved by technology

However, the shooting distance of ordinary two-dimensional cameras is limited. Common focal length cameras generally have an effective distance of 5-100 meters, and long-focus cameras can reach up to 1-5 kilometers. The worse the effect

At the same time, 2D image data has limited processing methods for external contour interference such as occlusion, clothing, wearing and holding objects, as well as changes in the horizontal angle and pitch angle of the camera installation.

[0005] In the process of 3D gait recognition, a 3D human body model is used, and the acquisition of 3D video and the processing of the model are usually more complicated than 2D recognition

At the same time, since the walking and movement posture of the human body are affected by various factors, such as the shooting distance of the camera, complex scene changes, changes in clothing conditions, occlusion interference, and different viewing angles, the versatility and recognition rate of the gait behavior recognition algorithm are still far away. inferior to the human brain

Especially in the 3D modeling of gait behavior human body, there are few technical achievements, which also limits the application of gait recognition technology in reality

Images