Reserch On Intelligent Vehicle Pedestrian Testing Technology

With the progress of science and technology,intelligent vehicle research has been developed rapidly,and pedestrian detection is an important part of smart cars.Therefore,pedestrian detection research has become very popular.Pedestrian detection has become very complicated in the real-time driving environment due to pedestrian walking posture,pedestrian dress,pedestrian complicated background,and external conditions such as light.Therefore,pedestrian detection can not have a "universal" system can detect pedestrians in any scene can only be a specific problem-specific analysis.In this paper,the HOG+SVM pedestrian detection method proposed by previous researchers is improved.HOG+SVM detects the pedestrian’s speed is relatively slow,and the detection accuracy is not very good.This paper proposes a PCA(principal component analysis)dimension reduction for HOG and also interpolates it.Reduces the dimensionality of individual HOG features and increases their accuracy,and fuses them with LBP features.The features of the fusion of HOG features and LBP features can not only express pedestrian profile information but also obtain pedestrian texture information,which can improve the speed of pedestrian detection and improve the accuracy of detection,which is beneficial to reduce false detection and missed detection in detection.Although some researchers have combined the two features of HOG and LBP,after simple fusion of these two features,the experimental results show that the detection effect is not much improved.In this paper,we first improve the HOG features,and use a simplified three-line interpolation method,that is,the two-dimensional convolution interpolation method,which reduces the dimensionality of the original three-line interpolation HOG and reduces the dimension of its PCA.The fusion of improved HOG features and LBP features does not lose the main features while its dimensions are reduced.This makes the detection speed faster and the detection accuracy better.This article uses an improved method to detect pedestrians in front of traffic.Secondly,for the existing detection method sliding detection method,in many studies it has been a forgotten,basically little research,this article compares the principle of the two most popular sliding window detection methods in detail,one is a change.The detection window detection does not become an image,and the other is a fixed detection window that detects a changed image.At the same time,aiming at multi-scale fusion problem,this paper proposes a non-maximal suppression algorithm,which solves the fusion of multi-sliding windows.At the same time,scored pedestrians were scored to solve the problem of pedestrian occlusion.In this paper,different formats of video detection material,written in the MFC platform,an application that makes the pedestrian detection of material can be quickly verified,so that the data is conducive to pedestrian detection results record.