Research On Video Texture Recognition Based On LBP And KNN

In order to cope with more and more of the network is also more and more covert content such as reactionary, pornography, violence and other sensitive content, an effective method based on the content of technology to shield sensitive technology is needed. In order to retrieve the sensitive content and shield the main transmission data from the network, we can find a solution from many angles. An effective method is to design a reasonable description of the video texture, and to design and match the measurement method to achieve the recognition of video texture, and then realize the effective classification of sensitive video.We can use these descriptors to make the machine understand the content of the video and to handle a large amount of video in the network with the efficiency of its efficiency. A large amount of video in the network can be distinguished and perceived. These are distinguished from video and can be classified, and the video with sensitive content can be distinguished and shielded.The main work of this paper includes the following aspects:(1) We review the important concepts of the static image texture feature and highlight the operator, summarize the method of texture feature analysis and measurement, and compare the performance of some important texture features, including BRIEF, LAZY, ORB, RIEF, SIEF and SUEF. Through the comparative analysis, it is concluded that the LBP operator has a great advantage in describing the local texture features, so the follow-up work of this paper is based on the LBP operator.(2) The classical LBP operator based dynamic texture descriptor VLBP is deeply analyzed, including the definition of VLBP operator, rotation invariance and its measurement method, VBP is formed after the addition of the traditional LBP operator, which can effectively represent the video texture, which is simple and efficient. At the same time, it is pointed out that the characteristics of the operator in special cases, the characteristics of the higher dimensions of the shortcomings of the. On the basis of the above works, this paper designs a kind of description operator which can describe the video texture, which is the content of the video.(3) This paper proposes a video texture recognition algorithm based on LBP and KNN. This algorithm can be used to represent the similarity between two texture sets, so that we can calculate the similarity between two and two. If we can realize the similarity between these two video textures, we can realize the classification and matching of video texture.In order to realize the algorithm, the author consulted and consulted a lot of documents. Through the related experiments, this paper obtains the high experimental results in database Dyn Tex. By this algorithm, the data can be read from the video, the paper first in the VS2010+Open CV2.4.4 environment, such as JPG, BMP and other formats for grayscale, binarization and extract features, and from the characteristics of a part of the description operator. Then the feature of a group of images is extracted and the robustness is good. Finally, the feature extraction and verification of the extracted results are directly used to verify the contents of the video.Through this paper, we design and implement an effective recognition algorithm of video texture, which provides a solution for the video retrieval and video surveillance, especially in the present Internet environment. As an important means of communication, video is needed to create a green, healthy Internet environment.