Pedestrian Safety Distinguishment Technology For Vehicle

Pedestrian Safety Technology Distinguishment for VehicleWith the continuous development of China's economy and the increasing of cars,traffic safety problems have caused increasingly serious.As a pedestrian in the traffic environment,the security problem can't be underestimated,so the establishment of pedestrian security mechanism were brought into sharp focus in recent years.Although the degree of pedestrian passive security mechanism can protect pedestrians in traffic accidents,the study of effective pedestrian active safety technology has important significance to avoid the occurrence of traffic accidents.Study on pedestrian safety technology for vehicle is one of the most popular topics in the field of intelligent vehicle.In this paper,we research on the pedestrian safety technology,which avoid the collision between vehicles and pedestrians in the root.At present,most of the pedestrian detection system is still using single camera.To remedy the defects of single sensor,this paper carried out a series of information fusion by multi sensor to improve the reliability of the monitoring data.In addition,the depth learning algorithm is used to complete the accurate recognition of pedestrians.Based on the establishment of a three-dimensional model of the relationship between pedestrians and vehicles,the system determines whether the pedestrian is safety.In this paper,the research work and research results of vehicle pedestrian safety technology are summarized as follows:(1)Pedestrian candidate area acquisition based on multi source fusion.Using the hardware integration technology and data fusion technology,a multi-sensor fusion system is constructed,which makes up for the defects of the single sensor and enhances the robustness of the pedestrian detection system.This paper has carried out two information fusion.The first is the information fusion of radar data and speed sensor data.Through the establishment of the mapping model between the radar data and the world coordinate system,the mapping model between the world coordinate system and the pixel coordinate system,the radar data and image data are fused in the second.Finally,the candidate region is obtained by establishing the relationship model between depth information and pedestrian candidate regions.(2)Pedestrian detection model based on deep learning.This paper constructs a pedestrian detection model based on deep learning by considering the requirement of detection accuracy of pedestrian detection system in practical application,In order to speed up the feature abstraction and improve the detection efficiency of the model,this paper makes use of the advantage of the HOG feature to deal with the input of the depth model.In order to ensure the accuracy of detection in complex traffic environment,the detection efficiency of the model is improved.(3)A discriminant model for the conflict between pedestrians and vehicles.In this paper,the potential conflicts between pedestrians and vehicles are analyzed,and the randomness of pedestrian motion and the randomness of pedestrian motion velocity are considered.According to the motion characteristics of the pedestrian and the vehicle's own motion characteristics,a three-dimensional spatial temporal conflict model is established.According to the pedestrian detected by pedestrian detection system,the potential danger degree of pedestrians can be obtained.