Information such as cracks and deflections is the important basis for structural safety. Existing methods have not achieved simultaneous detection. In most existing computer vision measurement systems, the view is fixed due to the fixed position of the camera. Thus, it is difficult to obtain the structures’ overall crack and deflection information. An automatic response measurement method is proposed in this study including () continuous image acquisition and signal transmission system based on self-walking bracket and Internet of Things (IoT), () an image splicing method based on feature matching, and () a crack and deflection measurement method. The self-walking bracket allows the industrial camera to move at a fixed distance to obtain the continuous image of the beam. Next, the spliced image is obtained through the PCA-SIFT method with a screening mechanism. The cracks’ information is acquired by the dual network model. The simplified AKAZE feature detection algorithm and the modified RANSAC are used to track the natural features of the structures. The curve fitting is performed to obtain the deflection curve of the beam under different loads. Experimental results show that the method can directly reflect the crack and deflection information of the beam. The average deviation of width is 11.76%, average deviation of length is 8.18%, and the average deformation deviation is 4.50%, which verifies the practicability of the method and shows great potential to apply it in actual structures.

中文翻译：

---

一种基于拼接图像的测试光束响应自动测量方法

裂缝、挠度等信息是结构安全的重要依据。现有方法尚未实现同时检测。在大多数现有的计算机视觉测量系统中，由于相机的位置固定，视图是固定的。因此，很难获得结构的整体裂纹和挠度信息。本研究提出了一种自动响应测量方法，包括（）基于自行走支架和物联网（IoT）的连续图像采集和信号传输系统，（）基于特征匹配的图像拼接方法，以及（）裂纹和挠度测量方法。自走式支架允许工业相机以固定距离移动以获得光束的连续图像。接下来通过PCA-SIFT方法得到拼接图像，s筛分机制。裂缝信息由双网络模型获取。使用简化的 AKAZE 特征检测算法和改进的 RANSAC 来跟踪结构的自然特征。进行曲线拟合，得到梁在不同载荷下的挠度曲线。实验结果表明，该方法可以直接反映梁的裂纹和挠度信息。宽度平均偏差为11.76%，长度平均偏差为8.18%，变形平均偏差为4.50%，验证了该方法的实用性，在实际结构中具有很大的应用潜力。