基于Faster RCNN的轻量化车辆测距模型
Lightweight Vehicle Ranging Model Based on Faster RCNN
桑振 1屠晓涵1
作者信息
- 1. 郑州警察学院,河南 郑州 450000
- 折叠
摘要
道路上车辆安全事故的发生常见于车辆之间不同程度的碰撞,多是由于车辆没有保持安全的行驶距离,因此在实际道路行驶中,对于车辆距离感知至关重要.本文基于Faster RCNN深度神经网络对目标车辆进行识别,利用Inception v2 模型对原有网络结构进行调整,在保持目标特征量的同时减少计算量,提升模型收敛速度.同时基于数据回归原理搭建图像像素与实际距离映射模型,隐性解决了单目相机成像过程中存在的畸变问题.实验结果表明,搭建的模型对车辆识别的精度达到 82.83%,在前方 40 m范围内车辆测距误差小于 4%,可以实现前方目标车辆的距离判断,为安全驾驶决策提供理论依据.
Abstract
The vehicle accidents on the road refers to the different degrees of collision between vehicles,mostly because the vehicle does not maintain the safe driving distance,so in the actual road driving,the perception of vehicle distance is very important.In this paper,the target vehicle is identified based on Faster RCNN deep neural network.The Inception v2 model is used to adjust the original network structure,which can reduce the computational load while maintaining the target feature quantity and improve the model convergence speed.At the same time,based on the principle of data regression,the mapping model between image pixels and the actual distance is built,which implicitly solves the distortion problem in the imaging process of monocular camera.The experimental results show that the accuracy of vehicle identification by the established model reaches 82.83%,and the vehicle ranging error is less than 4%within the range of 40 meters in front,which can realize the distance judgment of the target vehicle in front,and provide theoretical basis for safe driving decision-making.
关键词
单目视觉/车辆安全距离/Inception/v2/Faster/RCNNKey words
monocular vision/vehicle safety distance/Inception v2/Faster RCNN引用本文复制引用
出版年
2024
NETL
NSTL
万方数据