Moving object detection for unseen videos via truncated weighted robust principal component analysis and salience convolution neural network

Moving object detection is a basic and important work in intelligent video analysis. Recently, a lot of methods have sprung up. Among them, the methods based on deep learning have achieved very amazing results. However, the methods based on deep learning rely on special annotated data to train the model. Thus they have weak generalization ability and can only deal with the data related to the training data. In order to handle this issue, this paper proposes a method based on Truncated Weighted Robust Principal Component Analysis and Salience Convolution Neural Network. Unlike other deep learning methods, the input of the proposed method does not contain the scene information. The proposed method uses the salient information obtained by the proposed Truncated Weighted Robust Principal Component Analysis as input. This improves the generalization ability of the proposed method. The experimental results show the superior performance of the proposed method for unseen videos on CDNET 2014 database.