Aprimoramento da robustez de sistema de medição de nivel de corpos d’agua com base em imagens digitais

Abstract

It is safe to say that the majority of cities built near rivers, lakes, or basins experience flooding. Not every one of them is capable of investing in sensor-based solutions to monitor the behavior of these bodies of water in order to prevent or reduce the loss of material goods or of life. There exist quite a few systems to handle the task, but the few image-based solutions lack flexibility and reliability. This work is an improvement of the Thiago Nagaoka undergraduate thesis’ system that solves the orientation issue by using the SIFT and RANSAC pipeline. The system is composed of a specifically designed ruler, a camera, and a computer. Over the original work, in which the ruler and its extraction from the image captured by the camera here used, a new process that bases its measurements on the ruler pattern’s frequency was developed to infer the water level measure. During testing, it was also subjected to a more adverse set of conditions. The proposed system computes measurements with an error concentrated in the range of 0 to 10% when the process runs successfully. Adjustments to increase the success rate and precision are also proposed.