Research on Crack Width Measurement Based on Binocular Vision and Improved DeeplabV3+

Crack width is the main manifestation of concrete material deterioration. To measure crack information quickly and conveniently, a non-contact measurement method planar structure based on binocular vision proposed. Firstly, an improved DeeplabV3+ semantic segmentation model proposed, which uses L-MobileNetV2 as backbone feature extraction network, adopts IDAM to extract high-level information, introduces ECA attention mechanism, optimizes loss function achieve high-precision areas. Secondly, plane space coordinate equation was constructed principle SIFT point matching, calculated by combining segmented image. Finally, verify performance above method, test platform built. The experimental results show that RMSE using algorithm less than 0.2 mm, error rate 4%, has stable accuracy in different angles. It solves problem fast convenient structures outdoor environment.