Publication Type:

Conference Paper


First International Conference on Networks Soft Computing (ICNSC), 2014 (2014)



Approximation methods, Artificial neural networks, auto face detection, BLOB(Binary Large Object), CCTV videos, closed circuit television, eigenface histograms, Face detection, Face recognition, fisherface histograms, FLANN(Fast Accurate Nearest Neighbours), Haar Cascade Classifier, image database, LBP histograms, learning (artificial intelligence), Machine learning techniques, moving body identification, SIFT(Scale Invariant Feature Transform), speeded up robust features, SURF, SURF(Speeded Up Robust Features), Surveillance cameras, Video signal processing, Video surveillance


The idea of auto face detection from surveillance cameras and CCTVs is very relevant today. More and more CCTVs and surveillance cameras are being installed everyday. If there is a database of facial data present then the task of recognition boils down to comparison of each and every face detected from the video with every face saved in the database. Now this process involves capturing the faces before hand. This is actually a very tedious job. So the database of images is created (/updated) as and when new faces come into the camera view. The labeling of the faces can be done at leisure (by a human) or not be done at all. The current system once deployed does not need a database of images to start with. It creates its own collection of images, and then tracks the future occurrences of those images. Eigenface, fisherface, LBP histograms and SURF are different algorithms used for face recognition. We have tried all these algorithms.but among these surf shows better result. So the paper uses SURF for comparing image descriptors.

Cite this Research Publication

Shiju Sathyadevan, Balakrishnan, A. K., S, A., and S, A. R., “Identifying moving bodies from CCTV videos using machine learning techniques”, in First International Conference on Networks Soft Computing (ICNSC), 2014, 2014.