RGB-D Based Abnormal Event Detection

Depth camera or depth camera network can also be applied in abnormal event detection, especially in some scenarios where traditional cameras cannot work properly, e.g., crimes such as people fighting always occur in the dark, or in some situations where privacy is a big concern. One typical application is to detect human getting-up from the bed in a hospital ward, as the automation of this detection task can help to minimize the risk of patient falling to the ground. For this task, we have developed extensive motion and shape features from both color and depth image sequences. We also developed a multiple kernel based multiple modality feature fusion scheme for training the event detector/classifier. In our own work on recognizing the activities of daily life, we obtained a 98% success rate in recognizing when someone is getting up from bed, under laboratory conditions.

[Publication]
RGB-D Camera Based Get-up Event Detection for Hospital Fall Prevention, Bingbing Ni, Chi Dat Nguyen, Pierre Moulin. In ICASSP, 2012 [pdf].

RGB-D Based Abnormal Event Detection