Abstract
Conference Title: 2014 13th ACM/IEEE International Conference on Information Processing in Sensor Networks (IPSN) Conference Start Date: 2014, April 15 Conference End Date: 2014, April 17 Conference Location: Berlin, Germany This article describes a machine learning approach to water level estimation in a dual ultrasonic/passive infrared urban flood sensor system. We first show that an ultrasonic rangefinder alone is unable to accurately measure the level of water on a road due to thermal effects. Using additional passive infrared sensors, we show that ground temperature and local sensor temperature measurements are sufficient to correct the rangefinder readings and improve the flood detection performance. Since floods occur very rarely, we use a supervised learning approach to estimate the correction to the ultrasonic rangefinder caused by temperature fluctuations. Preliminary data shows that water level can be estimated with an absolute error of less than 2 cm. [PUBLICATION ABSTRACT]