一种基于YOL0的低对比度道路裂缝检测方法

打开文本图片集

中图分类号：TP391.4 文献标识码：A 文章编号：2096-4706（2025）19-0145-08

Abstract： Low-contrast crack detection is particularly important in road crack detection.Especially in complex environments，existing methodsarediffculttoefectivelydealwithlow-contrastcracks，resultingininsuficientdetection accuracandrobustness.Inordertosolvethisproblem，thispaperproposesalow-contrastroadrackdetectionalgorithmbased onYOLOv8.Byintroducing SimConv，Transposeoperation，RepBlock module，andSlideLossandShapeIoUlossfunctions，the performance of the model inlow-contrastcrack detectionissignificantlyimproved： SimConv enhances featureextractionability Transposeoperationoptimizes featurefusioneffect，andRepBlock module improves feature representationability.Inadition， the SlideLossandShapeloUloss functionscanenhancetheleaingabilityofthemodelfordiffcultsamplesandaccelerate the convergence process.Through the experimentalverificationontheUAV-PDD2023 dataset，the proposed model issuperiorto othermainstreamdetectionmethodsintesofaccuracyecall，F1soreadtheridcators，specallinthelowotrast crack detection task， showing superior performance and low computational complexity.

Keywords： YOLO; crack detection; loss function; feature fusion; UAV image

0 引言

道路裂缝作为常见的路面缺陷，通常表现为细长的裂纹或网状结构，主要由多种因素引起，如水分渗透、交通荷载、地基沉降等。（剩余14001字）