Advancing Crack Segmentation Detection:Introducing AAMC-Net Algorithm for Image Crack Analysis

WANG Xiaofang1, LIU Chenfang2, Hou Junliang1 and Zhou Liang1

  1. Geely University of China
    Chengdu Sichuan, 641423
    939549393@qq.com; houjunliang@guc.edu.cn; zhouliang@guc.edu.cn
  2. Chengdu College of University of Electronic Science and Technology of China
    Chengdu Sichuan, 611731
    2975431869@qq.com

Abstract

This research paper presents an innovative solution to address the challenges of poor detail detection effectiveness and prolonged training time in image segmentation. The proposed approach leverages the Adaptive Attention Multiscale Convolution Network (AAMC-Net), incorporating a multi-scale dilated convolution VGG L network for feature extraction and a deconvolution method for image segmentation. Extensive experiments demonstrate the superior performance of the proposed algorithm concerning intersection over Union (IOU), accuracy, precision, recall, F1, average training efficiency, and segmentation efficiency when compared to several traditional algorithms. On average, the proposed algorithm achieves remarkable improvements of 3.9%, 3.1%, 1.7%, 4.9%, 17.9%, 14.8% ,and 20.2% in these metrics. Moreover, the enhanced algorithm exhibits notable advantages in detail processing and real-time image segmentation detection.

Key words

Image crack segmentation, Convolutional neural network,VGG L, Attention mechanism

Digital Object Identifier (DOI)

https://doi.org/10.2298/CSIS230725042W

Publication information

Volume 21, Issue 4 (September 2024)
Year of Publication: 2024
ISSN: 2406-1018 (Online)
Publisher: ComSIS Consortium

Full text

DownloadAvailable in PDF
Portable Document Format

How to cite

Xiaofang, W., Chenfang, L., Junliang, H., Liang, Z.: Advancing Crack Segmentation Detection:Introducing AAMC-Net Algorithm for Image Crack Analysis. Computer Science and Information Systems, Vol. 21, No. 4, 1435–1455. (2024), https://doi.org/10.2298/CSIS230725042W