Traffic police gesture recognition method based on double-branch space-time diagram convolutional network

Abstract

The invention relates to a traffic police gesture recognition method based on a double-branch space-time diagram convolutional network. The method comprises the following steps: 1) extracting trafficpolice joint points and a skeleton from a traffic police gesture video by adopting a deep convolutional network; (2) representing input information of a time-space diagram convolutional network in a two-way mode through an information representation method, fully utilizing and uniformly expressing the traffic police joint time-space characteristics and skeleton physical characteristics, and completing the traffic police action analysis from the two levels of traffic police joints and traffic police skeletons; 3) constructing a double-branch space-time diagram convolutional network according tothe natural skeleton structure of the human body and the time sequence, and respectively inputting the traffic police joint information and the traffic police skeleton information into the double-branch space-time diagram convolutional network to realize traffic police gesture recognition. The method can overcome the influence of factors such as the height, clothes, traffic scene illumination andcomplexity of the traffic police, effectively improves the gesture detection precision of the traffic police, and guarantees the real-time performance of a recognition algorithm, so as to meet the demands of actual application.

Description

technical field [0001] The invention relates to the field of environment perception of smart cars, in particular to a traffic police gesture recognition method based on a dual-branch spatio-temporal graph convolution network applied in a traffic scene based on artificial intelligence technology. Background technique [0002] For autonomous driving, traffic police gesture recognition is a key part of its environmental perception and environmental cognition tasks. Traffic police gestures are a method for directing traffic and ensuring safe and smooth road traffic. Familiarity with traffic police gesture signals is helpful for us to achieve safe and civilized driving. Traffic driving regulations require drivers to be able to accurately recognize eight traffic police gestures, and traffic police command signals have priority over traffic lights in traffic scenes, and vehicles must obey traffic police commands when there is traffic police intervention. Therefore, self-driving ca...

AI Technical Summary

Problems solved by technology

However, due to the influence of the traffic police's height, clothing, traffic scene lighting and complexity, there are large errors in the feature extraction of the traffic police gesture space; at the same time, the traditional feature classification method can only be used for simple specific scenes, and the generalization ability of the model is poor. Unable to meet real-time and precision requirements in practical applications

Images