Research On Pedestrian Detection Algorithm Based On Background Modeling And The Deformable Part Model

Pedestrian detection is a research hotspot in the fields of pattern recognition,computer vision,and artificial intelligence.The complex detection environment and pedestrians have large non-rigid body characteristics,which brings great challenges to the detection process.Based on this,this paper proposes a pedestrian detection framework based on background modeling and Deformable Part Model(DPM),and implements the pedestrian detection system programmatically.The deformable part model is one of the pedestrian detection algorithms that has received much attention in recent years and has achieved better pedestrian detection results.It relies on the calculation of image features and sliding window determination,which has the disadvantage of large computational complexity.When the background is more complex,there will be some misdetection and leakage.Check the situation,Pedestrians often have the attributes of motion in the video.Therefore,this article puts forward the idea of combining the moving target extraction and the pedestrian detection system,and enhances the real-time and accuracy of the pedestrian detection system.The main work and results of this article are as follows:1)Moving target detection method.After analyzing the advantages and disadvantages of inter-frame difference,optical flow method,background difference,and Vibe background modeling,Gaussian Mixture Model(GMM)is used to model the background,which can suppress most of the interference.Under the premise of the effective extraction of the sports goal.At the same time,the target shadow is removed by the hue judgment,and the influence of the light mutation is suppressed by changing the global gray scale.2)Deformable part model design.Based on the analysis of pedestrian gesture diversity and target occlusion,an improved Histogram of Oriented Gradient(HOG)feature and extended Support Vector Machines(SVM)deformable components are designed and implemented.Pedestrian detection system model,and gives the choice of the sample and the model training process.3)Fast pedestrian detection system design.For the shortcomings of large amount of calculation and poor real-time performance of the deformable part model,firstly the moving target is extracted from the video frame,only the foreground area is pedestrian detected,the real-time performance of the system is effectively improved,and the complex background is further suppressed as the detection band.The interference that comes from improving the accuracy of detection and reducing the false detection rate of the system.4)Design and implementation of a pedestrian detection system based on a network camera.A rapid pedestrian detection system based on C++ is designed and implemented.This system is used to test in the actual scene and the test results are analyzed.