Institute of Visual Informatics

Institute of Visual Informatics (IVI)

Leading Digital Technology Across Industrial Revolution

Occupant Analytics and Event Detection for Intelligent Building Application

Headed By Assoc. Prof. Dr. Rabiah Binti Abdul Kadir.

Geran Penyelidikan Khas Top Down UKM: Penyelidikan IR4.0



Impressive progress has been made in AI in recent years, driven by exponential increases in computing power and by the availability of vast amounts of data. AI in the fourth industrial revolution (4IR) is a continuum promising a convergence of new technologies. From software used to discover new behavior toward to algorithms used to predict problem interests. The world is moving towards automation in every field, which surveillance systems have also drawn the attention of researchers lately. Digital video surveillance systems will capture and store more video that can be analyzed for the purpose of monitoring suspicious activities in buildings area. However, there isn’t anyone with enough time to watch every hour of footage. This project will look at how building surveillance system using CCTV as a device to regulate the occupants inside the building. Using CCTV systems, data is collected from multiple sources with overlapping contents, which is mostly redundant and contains both informative and useless contents. The pre-processed of an effective data collection for which image is required to further occupant analysis and event detection process. The pre-processing involves several steps such as target annotation, segmentation, object extraction, scalling object, and eigenvalue conversion. Followed by identifying moving object candidates on the background identified image captured by the digital video surveillance systems using deep learning classifier. An appropriate classification algorithm will be chosen to analyzes a set of data points with one or more independent variables and finds the best fitting model to describe the data points. All features will be processed using a deep feed forward neural network with N hidden layers. Finally, the effectiveness and accuracy of deep learning classifier will be evaluated in detection of event, existence and movement of occupant.