An Algorithm for Concrete Crack Extraction and Identification Based on Machine Vision

This paper proposes solutions to the large extraction error, the difficulty of identification, and other problems existing in crack processing. The first solution entails enlarging the grayscale difference between the crack and background via adaptive grayscale linear transformation using the OTSU algorithm for segmentation and combining the extending direction of the skeleton line and the grayscale feature of the crack edge to fill the broken part of the binary image to obtain a complete image of the crack. The second solution is to improve several major characteristic parameters of the crack image to be more suitable for the characteristic description of the crack. Finally, a comparison of different types of input features and different accuracies performed using the training support vector machine verifies the accuracy and practicability of the proposed algorithm for extracting and recognizing cracks.