Research On Moving Object Detection Based On Local Ternary Pattern And Hue

Computer vision is one of the important problems for artificial intelligence,computer science and technology.Because there is most information in the video scene,it is difficult to process these information using computers.The moving objects in the scene contain most of the interested contents,and therefore moving objects detection is very necessary in order to eliminate background interference.Usually moving objects detection is the first step in many applications for computer vision.After getting the moving objects region,the objects can be tracked,classified,identified and scene understanding on this basis.However,the complexity of video scene will seriously affect the accuracy of moving object detection,such as uneven illumination,illumination change and he widespread shadow.Aiming at widespread shadow in video scenes,a moving object detection method combining LTP texture and hue is proposed in order to reduce the influence of shadow.The main content of this thesis is as follows.Firstly,the algorithm idea of three kinds of moving object detection methods is summarized,and Gaussian Mixed Model is described because it is the most common method in moving object detection.Secondly,the related basic knowledge used in the thesis is described,including LBP operator,LTP operator,RGB and HSV color space and their mutual transformation,basic morphology processing and so on.Thirdly,the cause of shadow in the scene is analyzed.Using analysis and experiment,it is proved that the LTP operator and hue can resist shadow.Fourthly,LTP operator and hue are combined to describe the background model,and the comparison method of LTP texture and hue is provided according to the characteristics of them.A new method of moving object detection combined with LTP operator and hue is proposed using the idea of Gaussian Mixed Model.According to the experimental results,the proposed method has a better effect than the traditional Gaussian Mixed Model and the LBP method.This thesis can help to the related research of moving object detection.