Research On Static Background Modeling And Moving Object Detection Technology

With the development of science and technology,moving object detection in visible light images and infrared light images has been applied to various fields in daily life and plays an important role.There is a lot of interference information in visible light images,and the complexity of the scene,changes in the environment,light intensity,and light uniformity can all affect the accuracy of moving object detection.The existing visible light image moving object detection algorithms have the characteristics of high algorithm complexity or poor anti-interference ability.Infrared light can to some extent eliminate the influence of environmental factors,but infrared images have the characteristics of weak signal strength and low signal-to-noise ratio,which makes it difficult to use image intensity based algorithms for moving object detection.In addition,small target detection in infrared images is also a major challenge.In order to improve the accuracy and efficiency of moving object detection,a background reconstruction algorithm suitable for both visible and infrared images was studied.This algorithm can quickly and accurately reconstruct the background in video images.Moving object detection based on the reconstructed background image can improve the accuracy and efficiency of moving object detection.The moving object detection algorithm studied in this article is divided into two parts:background reconstruction and object detection.Through experiments,it was found that the changes in background pixels in image sequences have certain patterns.In continuous video images,the pixel changes in the background and moving target areas exhibit different patterns.That is,in several consecutive video images,at the same pixel position,the difference between the pixel values of the static background,fluctuating background,and moving target shows the characteristics of small difference and large difference.A background reconstruction algorithm was designed using this rule.Firstly,N consecutive frames of images are read from the video,and for any pixel position,each frame of the image is sequentially subtracted from the corresponding positions of the other images to obtain N sets of difference sequences;Then,based on the rectangular radial basis function,count the number of differences within the rectangular width in each difference sequence;Finally,use the pixel values corresponding to the difference sequence of the maximum frequency as the background.The background reconstruction algorithm studied in this article can quickly and accurately extract the background of visible light images,as well as accurately extract the motionless parts of infrared images.In order to detect small moving targets in infrared images,the original image is magnified before background reconstruction.Reduce the background image after obtaining it.Afterwards,the moving target is extracted through background subtraction.The experimental results show that for moving object detection in visible light images,under certain data conditions,the SSIM values of the background and real background constructed by our method are 0.162 higher than those of the Vi Be algorithm.The precision,recall,F1 mean,and FPR indicators of the moving object detection results are better than those of the Vi Be algorithm and GMM algorithm.For infrared images,this algorithm can also quickly and accurately detect moving targets.