Automatic recognition of suspicious activity for camera based security systems

We present a digital processing based method of intruder detection. CCTV (Closed Circuit Television) camera images are processed and analysed. An object detection algorithm is used to determine whether a camera image contains a possible intruder, while further feature extraction processing can isolate the image containing any possible intruder for further analysis. An artificial intelligence based approach is used for analysing processed images. A technique for the detection and identification of postures is performed. The methods used are described along with the results obtained. Conclusions are drawn and proposals for commercial exploitation are made.