Smart automation system human operator monitoring and anomaly detection using camera and node-red

Ong, Wei Ying (2021) Smart automation system human operator monitoring and anomaly detection using camera and node-red. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

	Text (24 Pages) Smart automation system human operator monitoring and anomaly detection using camera and node-red.pdf - Submitted Version Download (741kB)
	Text (Full text) Smart automation system human operator monitoring and anomaly detection using camera and node-red.pdf - Submitted Version Restricted to Repository staff only Download (2MB)

Abstract

Smart automation system is defined as the combination of automatic control of electronic devices and internet connection which allow them to be controlled remotely. It is becoming necessary in many fields and one of it is monitoring system. In the field of monitoring system, it is important to inspect and detect the action of human in order to detect and diagnose the unwanted behaviour deviation of a human operator from the pre-set sequence of actions, taking into account their possible response to the current state of the system. Security cameras such as closed-circuit television is commonly used by citizens but there is still a lot of features can be added in to the system to provide intelligence and automation. In this project, the focus is on human operator monitoring and anomaly detection with the aids of camera and Node-Red. Artificial technology is added to the monitoring system to determine the abnormal action of human. Hardware components such as laptop and laptop VGA camera are used in this project. Tensorflow.js. pre-trained model is used to perform pose estimation. Pose classifier is trained using ML5.js neural network. P5.js is used as the code editor. Node-Red is used to wire together the flows by using the nodes and make them communicate with each other. Three experiments are carried out and the results obtained will be analysed in terms of accuracy in pose classification. This system will detect the normal and abnormal poses of human.

Actions (login required)

View Item