Computer Vision Research Group

APSRA : Automatic Person identification from CCTV video feeds installed at Security Risk Areas

Recently the immense cost of successful terrorist attacks on soft targets such as mass transport systems has indicated that forensic analysis of video after the event is simply not an adequate response. Indeed, in the case of suicide bombings there is simply no possibility of prosecution after the event and thus no deterrent effect. A pressing need is emerging to monitor all surveillance cameras in an attempt to detect events and persons-of-interest. One important issue is the fact that human monitoring requires a large number of people, resulting in high ongoing costs. Furthermore, such a personnel intensive system has questionable reliability due to the attention span of humans decreasing rapidly when performing such tedious tasks. A solution may be found in advanced surveillance systems employing computer monitoring of all video feeds, delivering the alerts to human responders for triage. Indeed such systems may assist in maintaining the high level of vigilance required over many years to detect the rare events associated with terrorism. Because of this, there has been a significant need in both the industry and the research community to develop advanced surveillance systems, sometimes dubbed as Intelligent CCTV (ICCTV). In particular, developing total solutions for protecting critical infrastructure has been on the forefront of R&D activities in this field.

Amongst the various biometric techniques for person identification, recognition via faces appears to be the most useful in the context of CCTV. Our starting point is the robust identification of persons of interest. While automatic face recognition of cooperative subjects has achieved good results in controlled applications such as passport control, CCTV conditions are considerably more challenging. Factors such as varying illumination, expression, and pose can greatly affect recognition performance. According to the very recent Face Recognition Vendor Test 'FRVT' report issued by NIST (National institute of standards and technology, USA), head pose is the hardest factor to model. In mass transport systems, surveillance cameras are often mounted in the ceiling in places such as railway platforms or other public areas. Since the subjects are generally not posing for the camera, it is rare to obtain a true frontal face image. A further complication is that we generally only have one frontal gallery image of each person-of-interest (e.g. a passport photograph or a police mugshot). In addition to robustness and accuracy, scalability and fast performance are also of prime importance for surveillance. A face recognition system should be able to handle large volumes of people (e.g. peak hour at a railway station), possibly processing hundreds of video streams. While it is possible to setup elaborate parallel computation machines, there are always cost considerations limiting the number of CPUs available for processing. In this context, a face recognition algorithm should be able to run in real-time or better, which necessarily limits complexity.

The research is being carried out with the following main aims: (i) To develop and evaluate the effectiveness of a novel method, where we explicitly remove the effect of pose from the face model creating pose-robust features. (ii) Develop a fast illumination compensation method to deal with poor imaging conditions (iii) To evaluate the extent of speedup possible in the developed methods and whether such approximation affects recognition accuracy. (iv) To compare the performance, robustness and speed of new methods with contemporary approaches in the context of face classification under pose and illumination variations. (v) To compare the developed methods with existing state-of-the art approaches in terms of scalability and computational advantage. (vi) To deliver a fully functional and automatic prototype system to be used at public security risk areas.(see related publications)



Recent News
© 2010 Department of Electrical Engineering, CIIT Lahore  |  Terms and Conditions