Surface Recognition for Cars: A Comprehensive Approach for Neural Networks

IF 1 Q4 AUTOMATION & CONTROL SYSTEMS

Mechatronic Systems and Control Pub Date : 2019-11-26 DOI:10.1115/dscc2019-9148

Ashkan Pourkand, Christopher White, Naghmeh Zamani, David I. Grow

引用次数: 1

Abstract

This paper explores the viability of neural-network-based classification of ground surface for vehicles. By classifying road surface in near real-time, improvements in vehicle performance (e.g. braking and cornering) may be possible. Classification performance for many combinations of feature encoding and neural network types are compared. The vehicle used here was an Audi “S3” with a magnetic suspension system on the Sport mode. An NI CompactRIO (or cDAQ) module was used to record from a lowing the cDAQ to communicate with the PCB 352C03 one-axis accelerometer. The accelerometer was firmly attached to the windshield of the car. This work focuses on the classification of four road surfaces (asphalt, dirt, concrete, and sand), though larger target sets were also considered. The most accurate method involved a MATLAB feature extraction package with a back-propagation neural network, yielding an overall accuracy of 97%. Lessons learned from this wide exploration of options may extend to other related classification problems.

查看原文本刊更多论文

汽车表面识别:神经网络的综合方法

本文探讨了基于神经网络的车辆地面分类的可行性。通过近乎实时的路面分类，车辆性能(例如制动和过弯)的改进可能成为可能。比较了多种特征编码和神经网络类型组合的分类性能。这里使用的车辆是奥迪“S3”与磁悬浮系统在运动模式。NI CompactRIO(或cDAQ)模块用于记录cDAQ与PCB 352C03单轴加速度计的通信。加速度计牢牢地固定在汽车的挡风玻璃上。这项工作的重点是四种路面(沥青、泥土、混凝土和沙子)的分类，尽管也考虑了更大的目标集。最准确的方法是使用MATLAB特征提取包和反向传播神经网络，总体准确率为97%。从这种广泛的选择探索中获得的经验教训可以扩展到其他相关的分类问题。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Mechatronic Systems and Control AUTOMATION & CONTROL SYSTEMS-

CiteScore

1.40

自引率

66.70%

发文量

期刊介绍： This international journal publishes both theoretical and application-oriented papers on various aspects of mechatronic systems, modelling, design, conventional and intelligent control, and intelligent systems. Application areas of mechatronics may include robotics, transportation, energy systems, manufacturing, sensors, actuators, and automation. Techniques of artificial intelligence may include soft computing (fuzzy logic, neural networks, genetic algorithms/evolutionary computing, probabilistic methods, etc.). Techniques may cover frequency and time domains, linear and nonlinear systems, and deterministic and stochastic processes. Hybrid techniques of mechatronics that combine conventional and intelligent methods are also included. First published in 1972, this journal originated with an emphasis on conventional control systems and computer-based applications. Subsequently, with rapid advances in the field and in view of the widespread interest and application of soft computing in control systems, this latter aspect was integrated into the journal. Now the area of mechatronics is included as the main focus. A unique feature of the journal is its pioneering role in bridging the gap between conventional systems and intelligent systems, with an equal emphasis on theory and practical applications, including system modelling, design and instrumentation. It appears four times per year.