Abnormal events detection method for surveillance video using an improved autoencoder with multi-modal input

摘要

This paper introduces an algorithm to solve the the anomaly behavior detection problem of surveillance videothrough an improved autoencoder with multimodal inputs. Using 3D convolution and 3D deconvolution, and the decoderadds a feature map corresponding to the encoder on a specific layer to enhance the image detail information. Taking theRGB frame and the optical flow as inputs, abnormality scores are calculated according to the reconstruction error forlocating the abnormal segment. Experiments conducted in the CUHK Avenue dataset, the UCSD Pedestrian dataset andthe Behave dataset, our approach works best compare to the original approach. While improving the AUC, due to the useof unsupervised learning, a lot of labeling time is saved, which is more in line with the diversity and contingency ofabnormal behavior in real life.