Vehicle forward collision pre-warning method based on deep learning

Abstract

The invention discloses a vehicle forward collision pre-warning method based on deep learning. The vehicle forward collision pre-warning method comprises the steps that forward image information of avehicle is obtained in real time through a vehicle-mounted camera, and an image is pre-processed; the vehicle characteristics in the image are extracted through a multi-scale deep convolutional neuralnetwork, and a vehicle target is identified and positioned; according to the vehicle position in the image and based on the geometrical relationship projection and camera parameters, the distance between the current vehicle and a forward vehicle is calculated; the relative speed is calculated through the change of the real-time distance between the main vehicle and the forward vehicle; and the collision distance is calculated through the relative speed and the distance between the two vehicles, the danger level of the vehicles is judged according to the calculation result, and the corresponding collision pre-warning strategy is mobilized accordingly. According to the vehicle forward collision pre-warning method, the forward vehicle information can be obtained in real time, the distance between the current vehicle and the forward vehicle and the collision time of the current vehicle and the forward vehicle are estimated accurately, the proper pre-warning measure is made, safety drivingis guaranteed, and traffic accidents are reduced.

Description

technical field [0001] The invention relates to the field of pattern recognition technology and the field of automobile active safety technology, in particular to a vehicle forward collision warning method based on deep learning. Background technique [0002] With the gradual improvement of people's living standards, more and more people buy cars as means of transportation, and the number of cars is also increasing day by day. Traffic safety has become an increasingly serious problem, and people pay more and more attention to it. The frequent occurrence of traffic accidents is caused by the driver's negligence and the lack of vehicle safety performance. [0003] The assisted driving system is an active safety system that can monitor road information and the status of the driver in real time, judge the distance between the vehicle and the vehicle and pedestrians in front, and whether there is a possibility of collision. Brake to avoid accidents. Now, what people need is a ...

AI Technical Summary

Problems solved by technology

[0007] R-CNN (Regions with Convolutional Neural Network Features) obtains suggested regions of different sizes through a region of interest extraction step, and then scales these suggested regions to a fixed size and then enters the convolutional neural network, but this method is computationally intensive and Cannot meet real-time requirements

Faster-RCNN uses a region of interest generation network to achieve end-to-end detection. However, this method slides a set of convolution kernels on a fixed feature map to generate region proposals for multi-scale targets, resulting in variable targets. The contradiction between the size and the fixed feature map receptive field size cannot adapt to the diversity of target sizes in the real environment

Images