Concrete Surface Crack Detection Using Image Enhancement, Segmentation and Deep Learning: A Computer Vision Approach

Research Proposal: Concrete Surface Crack Detection Using Image Enhancement, Segmentation and Deep Learning This research proposal presents a computer vision-based approach for detecting and segmenting cracks on concrete surfaces using digital image processing and deep learning techniques. The proposed methodology combines image preprocessing, contrast enhancement, conventional threshold-based segmentation, and a U-Net-based convolutional neural network (CNN). Publicly available datasets, including the Crack Forest Dataset (CFD), Crack500, and DeepCrack, are considered for the proposed investigation, subject to their availability and licensing conditions. The study aims to investigate whether image enhancement techniques such as Contrast Limited Adaptive Histogram Equalization (CLAHE) can improve crack visibility and segmentation performance. A conventional threshold-based method will be used as a baseline and compared with a U-Net-style deep-learning segmentation model. The proposed system will be evaluated using precision, recall, F1-score, Intersection over Union (IoU), and inference time. Special attention will be given to challenges such as uneven illumination, shadows, stains, low contrast, and complex concrete surface textures. ResearchObjectives: 1. To study existing image-processing and deep-learning techniques for concrete crack detection. 2. To investigate image enhancement and preprocessing methods. 3. To implement conventional threshold-based segmentation. 4. To investigate U-Net-based crack segmentation. 5. To compare the proposed methods using standard evaluation metrics. 6. To identify limitations and possible future improvements. Expected Contribution: The proposed work aims to develop a documented image-processing and deep-learning workflow for concrete surface crack detection. It is intended to support preliminary infrastructure inspection and maintenance planning. Research Status:Research proposal. The methodology and expected outcomes are proposed, and experimental performance results have not yet been established. Author: Harsh Patel Department: Mechatronics Engineering Institution: LSP Government College of Engineering, Ratnagiri, Maharashtra, India Affiliated: Dr. Babasaheb Ambedkar Technological University, Lonere

Authors

Institutions

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-19
DOI
https://doi.org/10.5281/zenodo.24370748
Primary Topic
Infrastructure Maintenance and Monitoring
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Concrete Surface Crack Detection Using Image Enhancement, Segmentation and Deep Learning: A Computer Vision Approach

Harsh Patel
Zenodo (CERN European Organization for Nuclear Research)
Infrastructure Maintenance and Monitoring
article

Concrete Surface Crack Detection Using Image Enhancement, Segmentation and Deep Learning: A Computer Vision Approach

Harsh Patel
article en

Abstract

Research Proposal: Concrete Surface Crack Detection Using Image Enhancement, Segmentation and Deep Learning This research proposal presents a computer vision-based approach for detecting and segmenting cracks on concrete surfaces using digital image processing and deep learning techniques. The proposed methodology combines image preprocessing, contrast enhancement, conventional threshold-based segmentation, and a U-Net-based convolutional neural network (CNN). Publicly available datasets, including the Crack Forest Dataset (CFD), Crack500, and DeepCrack, are considered for the proposed investigation, subject to their availability and licensing conditions. The study aims to investigate whether image enhancement techniques such as Contrast Limited Adaptive Histogram Equalization (CLAHE) can improve crack visibility and segmentation performance. A conventional threshold-based method will be used as a baseline and compared with a U-Net-style deep-learning segmentation model. The proposed system will be evaluated using precision, recall, F1-score, Intersection over Union (IoU), and inference time. Special attention will be given to challenges such as uneven illumination, shadows, stains, low contrast, and complex concrete surface textures. ResearchObjectives: 1. To study existing image-processing and deep-learning techniques for concrete crack detection. 2. To investigate image enhancement and preprocessing methods. 3. To implement conventional threshold-based segmentation. 4. To investigate U-Net-based crack segmentation. 5. To compare the proposed methods using standard evaluation metrics. 6. To identify limitations and possible future improvements. Expected Contribution: The proposed work aims to develop a documented image-processing and deep-learning workflow for concrete surface crack detection. It is intended to support preliminary infrastructure inspection and maintenance planning. Research Status:Research proposal. The methodology and expected outcomes are proposed, and experimental performance results have not yet been established. Author: Harsh Patel Department: Mechatronics Engineering Institution: LSP Government College of Engineering, Ratnagiri, Maharashtra, India Affiliated: Dr. Babasaheb Ambedkar Technological University, Lonere

Zenodo (CERN European Organization for Nuclear Research)
Dr. Babasaheb Ambedkar Technological University (IN)
Infrastructure Maintenance and Monitoring
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon