基于深度學(xué)習(xí)的裂隙智能提取研究

打開文本圖片集

中圖分類號(hào)：TP18 文獻(xiàn)標(biāo)志碼：A 文章編號(hào)：2095-2945（2025）13-0009-05

Abstract：WiththerapiddevelopmentofChina’seconomicconstruction，engineeringconstructionismovinginahigher， deperandbroaderdirection.Inordertoensurethesafetyof projectconstruction，itisnecessarytoidentifytheenginering geologicalconditionsofthesitebeforeprojectconstruction.Asanimportantpartofrockmassstructure，fractureshavean importantimpactonengineeringgeologicalconditions.Therefore，extractingthestructureofcracksisveryimportantfor engineeringconstruction.Traditionalcrack structureextractionmethodsaretime-consumingandlabor-intensive，andhavepoor operability;theacuracyofcrackextractionmethodsusingtraditionalcomputertechologycannotmeetengineeringneedsandare poorinpracticality;however，therearefewresearchonusingdeeplearningtechnologytoextractfractures.Bystudyingthedeep learningnetworkoftheencoder-decoderarchitecture，thispaperdeterminesthattheimagesegmentationmodeltrainedbythe "U-Net++"frameworknetworkandthe"Resnet5O"encoder-decodernetworkcanefectivelyextractthefisurestructureoffield outcrop photos.

Keywords： crack extraction; image segmentation； deep learning; crack structure; extraction method

隨著我國(guó)經(jīng)濟(jì)建設(shè)的高速發(fā)展，房屋建筑越來(lái)越高，地基基礎(chǔ)越來(lái)越深，穿山隧道越來(lái)越長(zhǎng)，水利水電工程越來(lái)越龐大，相應(yīng)地，對(duì)于工程建設(shè)過(guò)程中的工程地質(zhì)條件要求也越來(lái)越高，對(duì)工程地質(zhì)信息的掌握也要越來(lái)越精確。（剩余6462字）