Semi-automated crack detection using computer vision

Abstract

The characteristics of surface cracks reflect the health state and functional degradation level of structures. Inspection and measurement of cracks on concrete surfaces are therefore of crucial importance and considered to be a fundamental component of safety and health monitoring of concrete structures. The conventional manual tracing of crack detection is time-consuming and depended largely on the inspector’s experience and knowledge. This paper presents a novel semiautomated approach for crack detection using computer vision and image processing techniques. The proposed method requires a pre-defined length to determine the calibration factor which gives the relation between the pixels of the image and the actual scale. High-resolution images of surface cracks are captured and subsequently processed according to the proposed method to obtain accurate crack measurements. The acquired digital images are subjected to a few image pre-processing steps for noise reduction and accurate determination of crack boundaries. The study also includes a comparison of the efficiency of edge detection and segmentation techniques in crack detection. Furthermore, this approach can be easily accommodated on mobile phone platforms that simplify and accelerate the process of crack detection. Experimental results demonstrate that the established method is capable of precisely perceiving the surface crack measurements.