Automatic recognition system for concrete cracks with support vector machine based on crack features.

Area of Science:

• Civil Engineering
• Computer Vision
• Materials Science

Background:

• Concrete cracks are a prevalent issue in infrastructure, necessitating reliable automated inspection methods.
• Machine vision systems offer advanced solutions for concrete crack detection and recognition.
• Effective crack identification is crucial for structural health assessment and maintenance.

Purpose of the Study:

• To develop and validate an automated system for concrete crack detection and recognition.
• To enhance the accuracy and efficiency of crack identification in concrete surfaces.
• To establish a robust framework for distinguishing cracks from other surface anomalies.

Main Methods:

• Image segmentation using a combined Prewitt operator and Otsu threshold algorithm.
• Application of mathematical morphology for refining crack zone identification.
• Feature extraction including compactness, occupancy rate, and length-width ratio.
• Support Vector Machine (SVM) with a radial basis function for crack classification.

Main Results:

• The proposed segmentation algorithm effectively produces complete crack zones with reduced interference.
• Preliminary discrimination successfully removed 99% of impurity areas, ensuring sample quality.
• The SVM model achieved a high recognition accuracy of 97.14% for concrete cracks.

Conclusions:

• The developed machine vision system demonstrates high effectiveness and practical applicability for concrete crack detection.
• The integration of advanced image processing techniques and SVM classification provides a reliable solution for structural monitoring.
• The proposed method satisfies practical requirements for automated inspection of concrete surfaces.