Anomaly Detection Based on Vision Transformer Model and Texture Features

Abstract

Anomaly detection is one of the video surveillance applications,which aims to detect and analyze abnormal behaviors and risky situations inorder to prevent accidents. Various deep learning models have beenpreviously developed for this purpose, such as CNN, RNN, and VisionTransformer, each one has its strengths and weaknesses based on the qualityof input data. This paper proposes a novel approach based on the texturecharacteristics of input frames. In order to enrich the input data of the visiontransformer model, and enhance feature extraction for the detection ofanomaly, we combine the original image with its texture extracted usingLocal Binary Pattern (LBP), and fed them into a fine-tuned pre-trainedVision Transformer, enabling the automatic classification of video framesinto abnormal and normal categories. The results demonstrate theeffectiveness of our approach in identifying risky situations in videosequences.