Abstract
An improved concrete structure health monitoring method based on G-S-G is proposed, which fully combines an optimized Gray-Level Co-occurrence Matrix (GLCM) with an improved Self-Organizing Map (SOM) neural network to achieve accurate and real-time concrete structure health monitoring. First of all, in order to obtain a dynamic image of the crack damage region of interest (ROI) with clear contrast and obvious target, the image acquisition system and image optimization method are used to process the damaged image. Moreover, in order to realize the accurate location of crack damage, crack damage identification research based on GLCM-SOM effectively eliminates the interference of honeycomb and pothole damage on crack damage. In order to obtain the indicators for monitoring the health status of the structure, the damage characteristics and probability distribution characteristics of the concrete structure in the gray level co-occurrence matrix are combined to extract the probability range index (PRI). On the basis of extracting the crack damage index, in order to verify the reliability of the sensitive feature index, starting from the two dimensions of damage texture feature and data expansion, through reverse research of the damage model, the damage index of accurately locating the crack damage was selected. It follows that the final sensitive indicators: entropy (ENT) and PRI can be used for structural health monitoring because of their strong damage characterization ability and sensitivity to damage characteristics. This research shows that is helpful to realize the high precision intelligent concrete structure health monitoring of modern concrete structure crack damage.