Service robot visual picture privacy protection method based on generative adversarial network

Abstract

The invention discloses a service robot visual picture privacy protection method based on a generative adversarial network. The method is characterized in that, the privacy identification and protection comprises the functions of data preprocessing, privacy identification, picture conversion and the like. The method comprises the following steps: firstly, performing data preprocessing on data acquired by a visual data acquisition end, then judging whether the input preprocessed data has privacy or not by a privacy recognition module, if so, performing picture conversion, converting the pictureinto picture data which does not involve privacy, and storing the picture data, wherein the training data growth and feature learning are used for updating a training data set, and based on the training data set, a feature model is obtained through an improved Cycle-GAN algorithm and used for picture conversion. According to the method, the picture data does not involve the private content from the source, and the method has the characteristics of short training time and strong generalization ability of private picture conversion.

Description

technical field [0001] The invention relates to the field of privacy protection, in particular to a privacy protection method for visual pictures of service robots based on a generative confrontation network. Background technique [0002] The smart home system used to take care of the elderly has the risk of privacy leakage due to the extensive use of cameras and voice monitoring equipment, affecting people's psychological state and even causing psychological disorders. This is one of the biggest obstacles to the deployment and promotion of such systems. At present, the research mainly focuses on the data that already has potential safety hazards and contains private information, and does not make the data itself not involve private content from the source. [0003] In order to solve the problem of user privacy leakage caused by robot vision equipment from the source, the image conversion method can be used to make the original visual image itself not involve privacy content...

AI Technical Summary

Problems solved by technology

In the prior art, although the image conversion method adopts the Cycle-GAN derived from the Generative Adversarial Network (GAN) for conversion, this method has many advantages in image style transfer, and can convert the image from the X domain to the Y domain and then from the The Y domain is restored to the X domain. Structurally, one discriminator of Cycle-GAN shares two generators, each with a discriminator, but when the application scene only needs to convert the image from the X domain to the Y domain , there are redundant discriminators, which will generate losses, resulting in difficult convergence of the loss rate, increased training time, and low generalization ability

Images