Crack and Noncrack Damage Automatic Classification from Concrete Surface Images using Broad Network Architecture

In concrete structures (bridge, road and building), surface cracks are important indicators of structural durability and serviceability. Manual visual inspection is usually adopted but often considered ineffective in terms of cost, safety, and reliability. In this paper, an automatic damage classification method using broad network architecture is proposed to apply on concrete surface images. Such method intelligently learns cracking and noncracking structures from raw input images by a flatted broad network, and gives superiority in effectiveness and efficiency compared with many popular mainstream deep structure methods.