| At present,the sluice video monitoring system lacks intelligent analysis function,and the sluice management relies on the subjective judgment of managers.In order to improve the level of sluice intelligent management,this paper designs an intelligent video image processing system suitable for sluice scenes,which has the functions of detecting and tracking moving targets.The work completed in this paper is as follow:Firstly,according to the actual situation of sluice management,the demands of intelligent video processing system are analyzed.On this basis,four basic functions are designed: video acquisition function,moving target detection function,moving target tracking function,intrusion warning area detection function,and the overall architecture of the system is designed with layered concept.Secondly,the Gaussian mixture model(GMM)used in the target detection part of the system is introduced,and the experimental results show that the algorithm meets the prospect extraction requirements in the sluice scene.Then the morphological opening operation is carried out on the detected moving target image frame to make the detected moving target contour more complete,so as to meet the requirements of target detection in the design system.In addition,in order to improve the information quality of transmitted video and achieve the desired effect of foreground detection,the video image is filtered and the color space is converted before the target detection.Thirdly,the Kernel Correlation Filter(KCF)algorithm used in the target tracking part of the system is introduced.Aiming at the tracking problem,that it is easy to lose the tracking target and the tracking effect is poor using KCF algorithm,when the scale of the target changes,the scale pyramid idea is integrated into the KCF algorithm to construct pyramids of different scales for the detected target area.By adjusting the scale space search area,the tracking box size can automatically adapt to the tracking target,so as to reduce the calculation and improve the tracking accuracy.At the same time,aiming at the occlusion problem in the process of target tracking,the Histogram of Oriented Gradient(HOG)feature in KCF algorithm and the SIFT feature in Scale-Invariant Feature Transform(SIFT)algorithm are fused to obtain a new feature model,which realizes the function of re-positioning after target occlusion.Finally,the user interface in the intelligent video processing system is designed by using the Gaussian mixture model and the improved KCF algorithm.And the functions of moving target detection,moving target tracking and alert area intrusion detection are tested in the user interface.The experimental results show that the purpose of timely reminding managers is realized when the tracking target in the monitoring screen is illegal,and the intelligent management level of sluice video monitoring is improved. |
