Abstract
For the actual demand of videos filtering on the internet, an objectionable internet videos filtering technology base on Probabilistic Latent Semantic Analysis (PLSA) model is proposed. Optical flow is first used to extract key frames of the video. And feature points are found out in key frames by Scale Invariant Feature Transform (SIFT). Then the visual words corresponded with the feature points are generated by k-means clustering method. Finally PLSA model is used to judge whether each of these visual words just mentioned is similar with the visual words in the training library, and corresponding videos with similar words is filtered out. The experimental results show that the detect rate of the filtering technology in the study is up to 86%, and it is fit for the video detection environment.