Abnormal gait behavior recognition method based on virtual posture sample synthesis

Abstract

The invention discloses an abnormal gait behavior recognition method based on virtual posture sample synthesis. The virtual posture sample synthesis model is trained by using a three-dimensional parameterized abnormal gait real training sample; and then a virtual sample is generated by using the trained model, a SoftMax abnormal gait behavior classifier is trained by using the real training sampleand the virtual sample, and the collected abnormal gait behavior can be accurately recognized through the classifier after the training is completed. According to the invention, a virtual attitude sample synthesis method and a characteristic modulation mechanism are adopted; the problem of few human body abnormal gait real training samples can be effectively solved; the abnormal gait behavior detection and recognition model has wide application prospects, including dangerous area abnormal gait behavior recognition, elder abnormal gait behavior detection, pedestrian gait behavior analysis in intelligent driving and security and protection and the like, and has good economic and social benefits.

Description

technical field [0001] The invention relates to an abnormal gait behavior recognition method based on virtual posture sample synthesis. Background technique [0002] Abnormal gait usually refers to various abnormal movements and postures when the human body walks. As an important biological feature, it can be used to detect abnormal gait behavior of the elderly, monitor abnormal behavior in the security field, analyze pedestrian abnormal behavior in an autonomous driving environment, evaluate and analyze gait symptoms in medicine, etc. It has broad application prospects and great practical value. [0003] At present, there are two main methods for detecting and identifying abnormal gait behavior by visual means: one is based on traditional two-dimensional image data, and the other is based on three-dimensional point cloud data collected by structured light sensors. It is more intuitive and convenient to use the two-dimensional vision method. The two-dimensional color camer...

AI Technical Summary

Problems solved by technology

Due to the inability to collect training samples on a large scale, the small sample problem in pattern recognition is formed

The problem of small samples will make abnormal gait detection and recognition directly affect the recognition accuracy of the recognition model in the face of external factors such as viewing angle and shape change, occlusion and clothing change.

Due to the lack of depth information, the 2D gait image cannot construct a 3D space model, and cannot fully utilize the characteristics of the 3D human body model for virtual sample expansion

Mirroring virtual samples is a completely symmetrical operation, but even for normal and healthy people, their human bodies are not completely symmetrical. Mirroring symmetry is an ideal assumption, and there are deviations from reality, which will affect the real recognition to a certain extent Effect

The method of adding Gaussian noise can solve the problem of ordinary white noise interference in image recognition, but it cannot solve the problem of individual differences in the same movement of different people, that is, there are differences in the changes of posture joints of different people in the same movement. difference

At the same time, it is difficult to automatically select the mean and variance of Gaussian noise. If the noise is too small, the effect will not be obvious, and if the noise is too large, it may affect the classification results. It is usually set based on personal experience, so the versatility is not strong.

Images