A New Water Level Measurement Technique Using Artificial Intelligent

Abstract

Flash floods are a growing concern worldwide, causing economic and social losses, increased death rates, and damage to infrastructure. The rapid nature of these disasters has led to delayed and inaccurate flood event information, causing public confusion and delays in response. This study aims to use AI to measure flood levels in real-time to improve flood information during flash floods. In this study, an Axia automobile as a model will be tested in an open space area. Then, box and manilla card is used as a level to mark the height of flood water, which is 15cm, 30cm, and up to 105cm. Data was collected by taking pictures of the vehicle from a distance of 620cm, 720cm, and 820cm. Teachable Machine applications will be used in this experiment to train the model for the data analysis. Image processing methods from the data will be used to identify flood elevation.  Key findings show the true percentages and false percentages accuracy of AI measurements on water level and distances measurement. Accuracy of AI measurements for distance represent 80% accuracy for correct value and 20% for the wrong values. Other than that, for accuracy of AI measurements on water level shows 90.5% indicates the accurate percentage and 9.5% indicates the inaccurate percentages. Additionally, the comparison in measuring water level between two devices, which is camera and Iphone show that the camera achieves 87% is accurate meanwhile the Iphone reached 62% of accurate values. Good agreement shows based on findings. However, some areas need to be improved especially for Iphone devices.