Concrete Crack Detection and Recognition Based on an Improved Canny Algorithm

To address the issues of low efficiency, strong subjectivity, and inadequate adaptability to complex backgrounds and fine cracks in traditional manual inspection and conventional image processing methods for concrete crack detection, this study conducts research on concrete crack detection and recognition technology based on an improved Canny algorithm. A database of positive and negative crack samples is constructed using a public crack image dataset. Non-local means filtering is introduced to denoise crack images, effectively suppressing noise while maximally preserving crack edges and texture details. On this basis, an improved Canny edge detection algorithm is employed to extract crack edge features, and morphological operations combined with connected component analysis are applied to optimize the edge results. Experimental results demonstrate that, compared with the traditional Canny, Sobel, and LoG algorithms, the proposed method achieves superior edge extraction integrity and segmentation accuracy under uneven illumination and noise interference. The SSIM reaches 0.8939, representing improvements of approximately 0.77%, 1.41%, and 0.40% over the Canny, Sobel, and LoG methods, respectively. These results indicate that the proposed method better preserves structural information and is well suited for engineering applications involving low-contrast and fine crack detection.