Unmarked robust monitoring of structural strain based on computer vision

The accuracy of traditional visual monitoring methods for the monitoring of field structures is often controlled by factors such as artificial targets and lighting conditions. In order to overcome the influence of illumination conditions on the accuracy of visual measurement, a method combining the phase-based dense optical flow algorithm and the support vector regression (SVR) algorithm was proposed to achieve robust monitoring of structural strain in the field. The method consists of image pre-processing based on 2D Gabor filter, phase-based dense optical flow sub-pixel displacement field matching calculation and displacement field smoothing based on SVR algorithm. After that, a strain conversion method based on the principle of strain sensors was used to realize the calculation of continuous strain fields on the surface of the structure, and the feasibility of the proposed method was verified by simulation test experiments and field test experiments. In the simulation test, the proposed method has a computation speed 50% faster than the conventional DIC algorithm with comparable measurement accuracy, and a clearer and more complete strain cloud map can be obtained. In the field test, the proposed method shows better environmental immunity, and the strain measurement error can be controlled within 2. 0% compared with the conventional test method. Compared with the traditional visual monitoring method, the proposed method improves the computational speed and robustness while ensuring the accuracy requirements, and it is applicable to the field monitoring of strains on the surface of specific large engineering structures without the need for artificial targets. © 2023 Science Press. All rights reserved.