Image anomaly detection method based on variational auto-encoder

Abstract

The invention discloses an image anomaly detection method based on a variational auto-encoder, in particular to an anomaly detection method integrating a variational auto-encoder and support vector data description, solving the problems of separation of two stages of anomaly detection and feature extraction, limitation of anomaly detection performance and incapability of coping with high-dimensional and large-scale anomaly detection tasks in traditional anomaly detection in the prior art. The image anomaly detection method comprises the steps of collecting image data; performing data set division and data preprocessing; constructing an anomaly detection model based on the variational auto-encoder; training an anomaly detection model; calculating a threshold value for distinguishing normaland abnormal image data according to the trained model; and judging whether the to-be-detected image is an abnormal image or not by using the trained model. According to the image anomaly detection method, support vector data description is adopted to perform distance constraint on the features extracted by the variational auto-encoder, and the extracted features are more suitable for anomaly detection, and the memory complexity is low, and the image anomaly detection method can be applied to high-dimensional and large-scale anomaly detection tasks.

Description

technical field [0001] The invention belongs to the technical field of deep learning, and mainly relates to image anomaly detection, in particular to an image anomaly detection method based on a variational autoencoder, which is used for image anomaly detection in combination with support data vector description. Background technique [0002] The task of anomaly detection is to identify data that is inconsistent with expectations, and such inconsistent data is usually defined as an anomaly. Anomaly detection has important responsibilities in many domains. Among the anomaly detection technologies that have been researched and developed, they involve applications such as credit card fraud detection, network intrusion detection, and medical diagnosis. The present invention addresses the problem of anomaly detection on images. [0003] The methods applied to image anomaly detection are divided into traditional anomaly detection technology and deep anomaly detection technology....

AI Technical Summary

Problems solved by technology

proposed a support vector machine method (OC-SVM) in the paper "Estimating the Support of a High Dimensional Distribution", which can construct a hyperplane model of positive data and divide the data on the other side of the hyperplane into Anomaly classes are used to obtain ideal anomaly detection results. The disadvantage of this method is that the result depends largely on the selection of regularization parameters and kernel functions. When the amount of data is huge, the memory complexity is very high, which limits this method. Applications in Large-Scale Image Anomaly Detection Tasks

The disadvantage of this type of method is that because the general loss function is used to replace the customized anomaly detection target, the target of anomaly detection cannot affect the hidden layer features extracted by the network at all, making this method often suboptimal

[0005] In summary, the traditional anomaly detection method applied in the image field will not have a good anomaly detection effect due to the large amount of data in the image dataset. The deep anomaly detection method represented by the encoder and the variational autoencoder is used in the anomaly detection task. When , due to the lack of customized anomaly detection targets or the lack of tolerance for the diversity of normal samples, the performance of the deep model is often not fully utilized

Images