Image Processing-based Flood Detection for Online Flood Early Warning System

Abstract

This paper will discuss the design of an online flood early warning system. This system will use a single board computer Raspberry-PI as the main controller, and a webcam to capture image. This system is integrated to Twitter. In hardware section, Raspberry-PI has main tasks as an image processor and do an update request to Twitter. In software section, OpenCV will be used as Image Processing software. Some method which used in this system is: 1) Region of Interest: this method is to create a portion of an image that you want to filter or perform some other operation on. Brightness and contrast: these methos is used in order to get brighter and better image before next process. 3) Grayscale and threshold: this method is to create an object segmentation. Otsu-thresholding is used on this step. 4) Edge detection: edge detection algorithm to find edge points on a (relatively) horizontal water line and point of dam’s height. By using these methods, the system can read and monitor the water level in the dam. If the water level exceeds the specified threshold, this system will generate an early warning of impending floods by doing update time line (text and image) of water level conditions to Twitter. The public will get the information if they following early warning system’s Twitter. Simulation test results show the system can read water level with an accuracy nearing 96%.