Video human body attribute identification method based on deep adversarial network

Abstract

The invention provides a deep architecture for detecting attributes (such as gender, race and clothing) of a human body in a monitoring video. The method is characterized by, through capability of a hybrid neural network and a part-based method, carrying out decomposition forecasting on an object in an image, and giving robustness; carrying out training based on weighted loss, obtaining attributeprediction scores, re-constructing a network to enable the architecture to have robustness to occlusion, eliminating obstacles, and carrying out classification on occlusion images through a discriminator network; and finally, improving resolution of the images through a super-resolution network, and obtaining attribute identification result of the video human body. The method can improve the resolution of the low-resolution images, process occlusion problems, and can effectively extract attributes even under conditions of poor resolution and strong occlusion, so that identification efficiencyis improved greatly; and the method is suitable for a plurality of application fields.

Description

Technical field [0001] The invention relates to the field of human body attribute recognition, in particular to a video human body attribute recognition method based on a deep confrontation network. Background technique [0002] Surveillance cameras are spreading rapidly in most cities around the world. In surveillance tasks, the focus of computer vision is to track and detect targets described by their visual appearance, and capture as many features as possible for users to better geographic conditions. Especially important, attribute recognition uses security cameras to detect people’s attributes (such as age, gender, etc.) and objects (backpacks, bags, etc.), which are often used for video retrieval, dangerous behavior warning, traffic flow video monitoring, industrial automation monitoring, security, and automatic sales. The behavior of suspicious characters in surveillance videos in public places such as cargo planes, ATM machines, shopping malls, and stations, etc. Although...

AI Technical Summary

Problems solved by technology

[0002] Surveillance cameras are rapidly spreading in most cities around the world. In surveillance tasks, computer vision focuses on tracking and detecting objects described by their visual appearance, and capturing as many human characteristics as possible is better for users. Geographical situation Especially important, attribute recognition detects people's attributes (such as age, gender, etc.) and items (backpacks, bags, etc.) through security cameras. In areas such as cargo planes, ATM machines, shopping malls, stations and other public places, surveillance videos of suspicious people's behaviors, etc., although the attribute recognition of video people has made great progress, most of them are based on the face of the video person. , age, and race) are rarely identified based on the entire body of the person, and most of them think that the surveillance video is full-resolution and the person is not closed, but in fact, surveillance cameras with far-field vision are usually subject to The resolution problem and the serious impact of people occlusion, so the commonly used methods still have certain limitations in the attribute recognition of people in surveillance video

Images