Evaluación y Comparación de Algoritmos de Detección de Bordes aplicados a coordinómetros digitales: Caso de Estudio Presa de Chanlud

Abstract

This research project aims to evaluate and compare various edge detection algorithms applied to images obtained through measurement instruments such as digital coordinometers, with the goal of improving the accuracy of the deformation-reading digitization process, reducing the dependence on manual calibration in structural monitoring systems of the Chanlud Dam. Digital coordinometers are key tools in civil engineering for monitoring deformations and displacements in hydraulic infrastructure. However, it is important to emphasize that the quality of data acquisition largely depends on image processing, particularly the accuracy of edge detection. The selection and analysis of different traditional and modern algorithms such as Sobel, Canny, Laplacian, and DOG along with Deep Learning models like CNN and MLP, were carried out to assess performance in terms of accuracy, computational efficiency, noise tolerance, and applicability under real field conditions. The images used come from a database provided by the Austro Seismic Network, which contains recent images, enabling automated analysis and validation within a contextualized results system. The design and implementation of the system propose a hybrid methodology that combines the strengths and benefits of the algorithms and models used, verifying the most efficient technique using environments such as PC and Raspberry Pi, thereby determining computational performance. This project contributes to the development of more effective tools for data analysis and optimization in infrastructure monitoring, enhancing the response capacity to potential civil failures and thus strengthening risk management for the Chanlud Dam.