Foreground Objects Recognition in Video Based on Bag-of-Words Model

Abstract

Given a video sequence about the road scenarios, discovering the foreground objects and labeling them are crucial. This paper proposes to represent the foreground objects in videos by the bag of words model. The foreground object is treated as a positive document which is a set oi image patches, and the background region is a negative document. Initially the training image patches are sampled and represented by speeded up robust features (SURF) descriptor, and then the bag of words model is constructed by K-means clustering algorithm. Subsequently the document is represented as the histogram of the visual words which is the feature vector of the image. Finally, a naive Bayesian classifier is obtained by training these feature vectors. In the stage of foreground objects detection, the motion regions are detected firstly, and then classified by the naive Bayesian classifier. The experimental results demonstrate that the proposed algorithm is robust and efficient with the processing speed up to eighteen frames per second on a standard PC.