Image-Based Concrete Crack Detection Method Using the Median Absolute Deviation.

Area of Science:

• Civil Engineering
• Materials Science
• Computer Vision

Background:

• Accurate concrete crack detection is crucial for structural health monitoring.
• Existing methods often struggle with varying crack widths and image conditions.
• Image processing techniques offer potential for automated crack analysis.

Purpose of the Study:

• To develop and evaluate an adaptive thresholding method for concrete crack detection and quantification.
• To compare the proposed method's performance against established techniques like Otsu and Niblack.
• To assess the method's capability in detecting and measuring narrow cracks.

Main Methods:

• An adaptive thresholding technique based on Median Absolute Deviation (MAD) was employed.
• Limited pre-processing followed by dynamic threshold determination for sub-images.
• Laplace edge detection for crack edge identification and centerline-based width measurement.

Main Results:

• The proposed method successfully detected cracks as narrow as 0.1 mm with high probability (94%).
• Detection probability reached 100% for wider cracks.
• Superior accuracy, precision, and F2 scores compared to Otsu and Niblack methods.

Conclusions:

• The adaptive MAD-based thresholding method provides a robust and accurate solution for concrete crack analysis.
• This technique demonstrates significant improvements over traditional image processing methods for crack detection.
• The approach holds promise for enhanced structural health monitoring and concrete infrastructure assessment.