Anomalousness Detection for Surgery Videos Using CHLAC Feature

We propose a chapter mark addition method for surgery video application that adopts cubic higher-order local auto-correlation (CHLAC) feature. In our method normal motions, which frequently occur in front of video cameras, are learnt statistically with CHLAC in combination with the subspace method. An anomalous motion is detected as a motion that exists far from the learnt subspace for the motions frequently-observed, and a chapter mark is placed just before the position the anomalous motion is recorded on the video. We conducted preliminary experiments using surgery video data to confirm effectiveness of the method we propose. The results show that the proposed method can detect the motions not frequently-observed in a surgery operation.