Research On Small Moving Target Detection Based On Background Mode Ling

Small moving target detection has always been a difficult problem in moving target detection methods.A small moving target has fewer pixels in the image and a smaller proportion of the entire image.When small moving target blend with a complex environment,it is difficult to extract its edge or color features and segment it in a single image accurately.The problem of detecting small moving targets in video are explored in this thesis,which is divided into the following three parts.(1)For the traditional Gaussian mixture model,only the correlation of pixels in the time domain is considered.In this thesis,the spatial information of images is introduced into the background modeling process and an algorithm of small moving target detection based on improved Gaussian mixture model is proposed.In the model matching stage,the current pixel is represented by the neighborhood feature value of the current pixel to perform matching determination.The experimental results show that the improved method can obtain better detection results for the small moving targets in the video,and obtain higher recall and precision.(2)In view of the actual situation,the historical gray value sequence of video pixels does not strictly obey the Gaussian distribution problem,and a small moving target detection algorithm based on the skew normal mixed model is proposed.A skew normal mixture model is established at each pixel position,and after a new image arrives,parameter update and foreground target detection are performed.The experimental results show that the proposed algorithm has more complete contour of the target,more adaptability to the dynamic background,and a certain robustness to camera shake.(3)Aiming at the background of motion in the video,an improved Wronskian skew normal mixture model is proposed.The current pixel is represented by a support domain composed of each pixel and its neighboring pixels,and the Wronskian function is used to classify the foreground and background pixels.The experimental results show that the proposed method can effectively cope with the noise generated by the motion background and obtain better detection results of small moving targets.