Methods Of Image Object Detection For Regional Security

In recent years,computer vision technology has developed rapidly,especially in image processing,target detection and other fields.In the application of regional security,sudden occurrence of moving targets and pedestrian intrusion may pose a threat to regional security.Therefore,the research on the detection of moving targets and pedestrians in the field of regional security has important research value and practical significance.In this paper,the moving target and pedestrian detection algorithms in current regional security applications are studied in depth,and corresponding improved algorithms are proposed.This paper mainly includes image segmentation methods in image preprocessing,analyses the advantages and disadvantages of existing common methods in moving object detection and pedestrian detection,and puts forward corresponding improvement methods combined with practical application requirements.The principal work and research production are as follows:1.Aiming at the problem of poor segmentation accuracy and loss of details in existing image segmentation methods,an image segmentation method based on improved genetic algorithm is proposed.This method utilizes the relevant information of the pixels of the image itself and combines genetic algorithm to redesign the fitness function and other modules.The algorithm can clearly segment the target,retain enough feature details,and effectively improve the segmentation effect of the target in the regional security scene.2.In order to overcome the shortcoming of "ghost" in Vibe moving object detection algorithm,a Vibe moving object detection algorithm based on image segmentation is proposed.This algorithm uses image segmentation algorithm,and uses the threshold information to filter the foreground points in Vibe algorithm again,so as to get the foreground points that meet the judgment conditions.The experimental results show that the algorithm can effectively improve the "ghost" can not be quickly eliminated.3.A pedestrian detection algorithm based on S-HOG feature is proposed to solve the problem that the detection accuracy of pedestrian detection using a single feature is low and the real-time performance is reduced by using feature fusion method.This algorithm uses feature fusion method to process HOG features and SIFT features,and then uses PCA to reduce the dimension of features,and then reduces the dimension of the two fused high-dimensional features.Finally,a pedestrian detection model based on S-HOG features is trained by using SVM classifier.Experiments verify the performance of the algorithm.