Research on Electricity Personnel Apparel Monitoring Model Based on Auxiliary Categorical-Generative Adversarial Network

Abstract

This article aims at the problems that exist in the monitoring process of employees’ standard clothing. Adopting the reverse way of thinking to carry out the normative identification work, that is, the classification ability depends on the feature fitting ability. First of all, building an employee work card template library, which is used to realize personnel authentication in combination with template matching method. The recognition result is given as a label to the real captured image. Then, the construction of the auxiliary categorical-generative adversarial network was done with a few real images. The training of generator G and discriminator D is completed through iterative interactive update, so as to realize the fitting of the real clothing image. The trained ACat-GAN is used as a standard clothing monitoring model. Respectively, based on the discriminant image and parameter adjustment data, discriminator D and generator G push out the list of well-dressed people and Image of non-standard. Based on a power supply business hall as the background platform, this article collected personnel clothing images for experiments. The experimental results show the feasibility and practicability of the method which described in this paper.