Development of anomaly detection using vision system

Harizan, Ahmad Hafiy (2019) Development of anomaly detection using vision system. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

Preview

Text (Full Text) Development of anomaly detection using vision system.pdf - Submitted Version Download (1MB) | Preview

Abstract

As crime rate greatly increases over the years along with technological advances on vision systems, surveillance systems are being installed in almost every possible location such as banks, offices, and residential areas. With continuous recording of videos, humans are facing difficulty to monitor them manually for the detection of anomalous activity. In attention to this problem, the proposed work designs a semi-supervised anomaly detection technique in stationary backgrounds to reduce the need of human intervention in detecting anomalies. Segmentation, detection of objects and noise removal in foreground are the main phases of the proposed work. Identification of important target features such as centroid and the bounding box assists in object detection and localization of objects in frames. The performance evaluation of the simulation results based on localization error and rate of accuracy validated that the proposed work could deal with various anomaly activity in a static background scene and successfully detect anomaly without much human intervention.

Actions (login required)

View Item