Low-Cost Optical Displacement Measurement for SHM Applications Supported by CNN Object Detection

This paper outlines the considerations involved in designing an algorithm that can accurately and cost-effectively determine displacements for structural health monitoring, using standard image processing techniques and a convolutional neural network (CNN)-based object detection system. The algorithm can effectively identify and track geometric measurement motives across image sequences, enabling the precise determination of position and displacement. High levels of precision can be achieved through careful optimisation of geometric shape selection and pre-processing methods. A minimal example implementation using open-source, Python-based libraries demonstrates good agreement with reference measurements and minimal noise, highlighting the algorithm's potential. This flexible, robust approach offers substantial opportunities for further optimisation and application-specific scalability, making it a promising solution for structural health monitoring across various infrastructures.