Moving Object Detection Based On Super-pixels Classification And Codebook Model

Along with the rapid development of computer and related technology,intelligent video surveillance system based on video image sequenceshad been widely used in the military,transportation,security and other fields.The accurate detection of moving object had become a hot issue in the field of intelligent video surveillance.In recent years,domestic and foreign scholars had improved many moving object detection algorithms in order to improve the robustness,real-time and accuracy of moving object detection in complex environment.In this thesis,the main research was the moving object detection technology.The videos with simple background,complex background,slow change of light and light mutation were selected as test video.This thesis deeply studied the Codebook algorithm and summarized the advantages and disadvantages of the algorithm.In order to solve the shortcomings of this algorithm,this thesis further studied and inproved the algorthm.The performance and feasibility of the improved algorithm were verified by simulation experiments,and the moving object detection system in video image was designed and implemented.The main research work was as follows:1.Through reading a large number of domestic and foreign literature,this thesis summarized the current research situation of moving object detection in intelligent video surveillance system,and deeply studied the motion object detection algorithm based on the background subtraction method in the video sequence image,and analyzed the difficulties existing in the process of moving object detection.2.Aiming at the problem that the traditional Codebook need to set more experience parameters and low the detection rate of illumination changes.An adaptive Codebook background model of illumination change prediction was proposed to improve the accuracy of detection based on the research of Codebook background model.In this thesis,the background model was updated by adding the learning rate.The color distortion and the luminance range formulas in the model were improved and the setting of the empirical parameters was reduced.By increasing the light forecasting mechanism,the background model was more accurate,and detection accuracy was improved.3.In order to improve the integrity of the moving object detection,this thesis studied the super-pixel segmentation,and proposed a Codebook model based on simple linear iterative clustering.The background model was more close to the real background by using the segmented similar pixel region block instead of the single pixel for background modeling.In the process of updating the background model,different updating strategies were adopted for dynamic background and mutation background in complex environments.Experiments showed that the algorithm proposed in the thesis improved the performance and feasibility.4.On the basic of the research and improvement of the moving object detection algorithm in the video sequence images,this thesis designed and implemented a system which could accurately detect the moving target in the complex environment.At the end of this thesis,the content of the research was summarized,and the future research direction and ideas were pointed out.