基于卷积神经网络的自动行驶小车研究与设计

Research and Design of Automatic Traveling Car Based on Convolution Neural Network

针对车辆的自动驾驶问题,提出了一种卷积神经网络模型,并设计制作了一辆小车实现了自动驾驶。使用开源神经网络构建工具Tensorflow进行神经网络模型的构建和训练,用车载笔记本电脑实现神经网络的训练和计算;使用CAD工具设计小车车体,用一枚前置摄像头采集环境信息;在实际道路上收集数据并训练神经网络模型,在未收集过训练数据的道路上进行测试。测试结果表明,小车具有一定程度的自动驾驶能力。

Aiming at the problem of vehicle autopilot,a convolution neural network model is proposed and a car is designed to realize automatic driving.Using the open source neural network construction tool Tensorflow to construct and train the neural network model,carrying out the training and calculation of the neural network with the vehicle laptop computer;using the CAD tool to design the car body and using a front camera to collect the environmental information;on the actual road collect the data and train the neural network model for testing on the road where no training data has been collected.Test results show that the car has a certain degree of autopilot ability.