Research On Moving Target Detection Technology

Moving target detection technology plays an important role in the military field,intelligent transportation,intelligent video surveillance,industry,advanced human-computer interaction and image compression.For the field of computer vision,moving target detection technology is its research focus.The key of motion target detection lies in how to get interesting motion targets from video images.Accurate extraction of moving target plays an important role in the later of target tracking and target recognition.However,the dynamic changes of background scenes,such as the change of light,the shadow of objects and the obstructions made the practicability of moving target detection technology influences.The practicality,real-time,robustness and accuracy of the target detection became the focus for researchers.The main purpose of this paper is to improve the shadow removal algorithm and remove the shadow components in the image better.In order to realize target recognition and behavior comprehension for different scenes,the traditional contour extraction algorithm is improved and the contour of motion target is extracted.The algorithm in this paper is feasible and can remove the shadow components of the image well and achieve the extraction of the target contour.The paper is summarized as follows:(1)The basic principles and performance of commonly used target detection methods,such as frame difference method,optical flow method and background subtraction method.Analyzed its principle and compared and summarize it,could better grasp its application field.On this basis,the image features,thresholds and morphological processing related knowledge were correspondingly studied,in order to improve the accuracy of moving target detection.(2)The basic idea and performance of classic background modeling algorithms included averaging background method,background modeling based on mixed Gaussian model,background model method based on nonparametric kernel density estimation,ViBe algorithm and background model based on codebook law.Analyzed each background modeling method,compared its performance and applicable scenarios,and give some suggestions for improvement.(3)Based on the removal of the shadow of the moving target detection algorithm,the shading detection algorithm is analyzed and the shading problem caused by the change of illumination is analyzed.The improved algorithm proposed in this paper is a shading removal algorithm based on Gaussian mixture model.From the above analysis of the classical background method,Gaussian mixture model can well extract the moving target.In this paper,the Gaussian mixture model is firstly used to detect and extract the moving target.Then,the shading removal algorithm is combined with the shadow removal algorithm,and the subsequent processing is performed to obtain the moving target.Compared with other shading removal methods,the detection result of this algorithm is more accurate,which can better remove the shadow components in the image.(4)The target contour extraction based on the moving target detection algorithm,the edge detection operator and contour extraction algorithm are analyzed firstly,and then the improvement is made.The improved algorithm is an iterative contour extraction algorithm.The algorithm in this paper is still based on Gaussian mixture model.The initial detection is the same as above.The foreground image detected by Gaussian mixture background model is preprocessed.Then an outline model based on iterative algorithm is input to complete the contour extraction.Experimental results show that the iterative algorithm can extract the target contour well in both indoor and outdoor scenes.