Research And Implementation Of Surveillance Video Target Retrieval Technology Based On PCA-SIFT Algorithm

In recent years,with people's attention to the field of security,more and more monitoring equipment has been applied to public places such as universities,road traffic,residential quarters,etc.However,in the face of massive surveillance video libraries,how can we find interest efficiently and quickly Content is a problem.At present,the search for specific targets usually uses manual methods,but the accuracy and efficiency of this search method are relatively low.In order to solve the shortcomings of manual browsing and annotation,and to meet people's urgent need to find targets in surveillance video more quickly and accurately,this paper combines the characteristics of surveillance video data to research and implement surveillance video target retrieval technology based on PCA-SIFT algorithm.First,preprocess the read surveillance video to remove the noise that may exist in the surveillance video.Then,in view of the feature that the surveillance video contains a lot of useless background frames,the key frame of the video is extracted using the background difference method.Through in-depth study of various background modeling algorithms,it is found that Vi Be + background modeling algorithm has the highest comprehensive index.Based on Vi Be + background modeling,differential operation is performed,and key frames are extracted according to the set threshold.In order to reduce the amount of calculation during feature extraction,the ROI in the key frame is extracted as the source data for later feature matching.Experiments show that the key frame and key frame ROI extraction algorithm based on background difference used in this paper has a good effect.Then,this paper deeply studies the local feature extraction and matching algorithm of the image.In view of the shortcomings of the SIFT extraction algorithm,this paper adopts the PCA-SIFT algorithm after reducing the dimension of the SIFT algorithm,and then uses the FLANN algorithm to increase the calculation speed,while using the RANSAC algorithm to optimize Matching pairs to improve retrieval efficiency.In view of the shortcomings of the lack of color information in the algorithm,the HSV color histogram is used to initially screen the images in the database,and then the HSV color histogram features and PCA-SIFT features are used to extract the similarity in a weighted fusion method in the remaining images.High image.Experiments show that the algorithm has high calculation efficiency and accuracy.Finally,on the basis of the above algorithm research,a detailed demand analysis of the surveillance video target retrieval system was carried out,and the software design of surveillance video target retrieval was completed using the open source computer library Open CV and the Microsoft basic class library MFC.The system is simple and easy to use.The test results prove that its operation is stable and stable,which can improve the real-time and accuracy of surveillance video target retrieval and meet the needs of surveillance video target retrieval.