Research On Real-time Pedestrian Vehicle Detection And Early Warning Algorithm For Driver Assistance

Increasing car ownership not only brings convenience to people’s life,but also makes driving safety a huge challenge.Driver assistance technology avoids accidents by detecting pedestrians and vehicles ahead on the road and providing early warning,and the use of platforms such as android systems can easily and conveniently achieve the function of driver assistance.Aiming at the resource-constrained hardware such as mobile devices,this study proposes a detection and early-warning algorithm that can detect pedestrians and vehicle objects in real time,and design a safety warning system for driver assistance on mobile devices.The research content and innovations are as follows:First,this study systematically analyzes existing object detection algorithms based on comparative experiment.The object detection model based on deep learning consists of two parts: feature extraction network and detection network.At first,the convolutional neural network for feature extraction is introduced in detail,and the advantages and disadvantages of the network structure are analyzed specifically for the application scenarios.Then,it introduces and compares the widely used detection models,analyzes the characteristics of the models and the problems existing in the assisted driving scenarios,and proposes corresponding solutions to lay the foundation for the subsequent algorithms.Second,a real-time pedestrian detection algorithm for assisted driving scenarios is proposed.Aiming at the real-time demand of driver assistance technology and the limited computing resources of android systems,a lightweight convolutional neural network based on ratio consistency is designed,and pedestrian detection algorithm is designed based on the statistical characteristics of pedestrian objects.Research consists of such as the design of network structure,the design of pre-selection rectangle and multi-stage training method.For the assisted driving scenario on android systems,the channel compression method is used to improve the running speed of the algorithm,and a real-time pedestrian safety warning system is effectively designed based on the proposed algorithms.Third,a real-time vehicle detection and early warning algorithm based on hierarchical regions of interest is proposed.Aiming at the problem that the computing resources are limited on android systems and the detection algorithm is not sensitive to small objects,a single-scale detection algorithm is designed to improve the speed,and the vehicle objects in the near,middle and long distance are detected in the multi-level regions of interest.In turn,multi-level warnings are given to objects of different distances.Based on the proposed algorithm,the early warning system for vehicle safety is designed on android systems to detect vehicle objects from near and far in real time,and carry out risk assessment and vehicle safety warning for assisted driving.