AED-Net: An Abnormal Event Detection Network. (October 2019)

• Record Type: Journal Article
• Authors: Wang, Tian; Miao, Zichen; Chen, Yuxin; Zhou, Yi; Shan, Guangcun; Snoussi, Hichem
• Is Part Of: Engineering. Volume 5:Number 5(2019)
• Page Start: 930
• Page End: 939
• Publication Date: 2019-10
• Subjects: Abnormal events detection -- Abnormal event detection network -- Principal component analysis network -- Kernel principal component analysis; Engineering -- Periodicals; Engineering -- China -- Periodicals; 620.005
• DOI: 10.1016/j.eng.2019.02.008
• Languages: English
• ISSNs: 2095-8099

Abstract

Highlights: We propose a new self-supervised framework, AED-Net. This new network can be trained with unlabeled data and perform better by comparing with the state-of-the-art methods in an abnormal detection task, as tested on UMN dataset and UCSD dataset. We combine PCAnet, a network for feature extraction, and kPCA, an effective one-class classifier, delicately to form the AED-Net. We combine local response normalization layer (LRN layer), a trick used in CNN to aid generalization, with the AED-Net for improvement. It does improve the detection results. Abstract: It has long been a challenging task to detect an anomaly in a crowded scene. In this paper, a self-supervised framework called the abnormal event detection network (AED-Net), which is composed of a principal component analysis network (PCAnet) and kernel principal component analysis (kPCA), is proposed to address this problem. Using surveillance video sequences of different scenes as raw data, the PCAnet is trained to extract high-level semantics of the crowd's situation. Next, kPCA, a one-class classifier, is trained to identify anomalies within the scene. In contrast to some prevailing deep learning methods, this framework is completely self-supervised because it utilizes only video sequences of a normal situation. Experiments in global and local abnormal event detection are carried out on Monitoring Human Activity dataset from University of Minnesota (UMN dataset) and Anomaly Detection dataset from University of California, San Diego (UCSD dataset), and competitive results that yield a better equal error rate (EER) and area under curve (AUC) than other state-of-the-art methods are observed. Furthermore, by adding a local response normalization (LRN) layer, we propose an improvement to the original AED-Net. The results demonstrate that this proposed version performs better by promoting the framework's generalization capacity.