智能公路系统下的路面病害图像检测技术综述

Development of road distress image detection based on intelligent road system

本文梳理了国内外路面病害图像检测技术的研究进展。在智能公路系统的框架下,尝试提出路面病害检测的发展趋势。研究表明:路面病害检测已经历3个阶段,人工检测阶段、半自动化检测阶段、无损自动检测阶段。基于2D数字图像的路面检测具有数据采集便利、识别算法成熟的特点,尤其是近年深度神经网络在基于2D数据的病害检测上取得了高实时性和高鲁棒性的成果,适用于海量道路图像的自动化、实时化病害筛查;基于3D数字图像的路面检测具有识别精度高但采集效率有限的特点,适用于局部范围获取病害多维度信息。本文结合不同方法的现状与优势,尝试展望路面病害检测技术阶段4——多源分级检测阶段。该阶段依托高速物联网技术和人工智能技术,打破传统道路巡检车的局限,以大批量民用车辆为主要数据采集载体,分级利用深度神经网络的高实时性以及3D图像识别的高精度性,有针对性地实施道路监测与养护。

This paper summarizes the road disease detection and proposes its trend based on intelligent highway system.The road disease detection has gone through three stages:manual detection,semiautomatic detection,and non-destructive automatic detection.2D digital image is convenient and has mature algorithms.In recent years,deep neural network has achieved high real-time and high robustness,therefore 2D image is very suitable for automatic and real-time detection of massive road images.3D digital image has high accuracy but limited efficiency.It is available for medium range detection to obtain multi-dimensional information.This paper combines the advantages of above stages and proposes the road disease detection stage 4——a multi-source hierarchical road detection system.This stage integrates the High speed Internet of Things(IOT)and artificial intelligence,breaking the limitations of traditional road inspection vehicles,taking civil vehicles as the main data collection carrier,making full use of the real-time disease recognition of deep neural network and the high precision of 3D image recognition,targeting at implementation of road monitoring and maintenance.