Multi-target Intelligent Detection Method of Support Structure Defects Based on Digital Image Processing Technology

Support structure defects seriously affects the sustainability and safety of the engineering field. Currently, the detection of support structure defect mainly relies on manual inspection, which has the disadvantages of poor accuracy, low efficiency and insufficient automation. This paper takes cracks, concrete scaling and concrete weathering of three kinds of common defects as the research object, a multi-target intelligent detection method of support structure defects based on digital image processing technology is designed. Firstly, a typical defect image recognition library is constructed by image datasets of three kinds of defects. YOLOv5 neural network is built to accurately and classify identify defects. Then, the improved Gaussian filter algorithm combined with histogram equalization is used for image preprocessing. An improved adaptive Canny algorithm combined with Otsu is proposed for image segmentation. According to the different characteristics of defects, the crack algorithm based on morphology and the defect area algorithm based on the connected component labelling are designed for feature extraction. Finally, the experimental platform of defects intelligent detection is constructed to complete the multi-target intelligent detection. The experimental results show that the multi-target intelligent detection method proposed in this paper can effectively achieve high-precision and high-accuracy detection of cracks, concrete scaling and concrete weathering. The proposed method is of certain significance for the digital and intelligent development of support structure detection.