Computational method for automatic recovery and extraction of weather information

Abstract

Some meteorological centers stored in their facilities a lot of paper sheets containing measurements of atmospheric pressure, temperature, humidity and precipitation. The meteorological measurements are shown as a continuous line drawn on a graph paper divided into periods of time. The paper sheets are in danger of being lost either by the deterioration suffered because of the passage of time, factors such as humidity or any calamity. Reading and recording of information is performed manually and therefore a considerable amount of time and effort is employed. The recovery of paper sheets will allow to have a database of images and the automatic extraction will allow to get more frequent and accurate measurement data. This paper proposes a computational method that involves digital processing and analysis of images scanned using a mathematical model for calculating the measurements. Binarization methods, morphological operators and filters were applied according to their performance metrics (accuracy, precision and error rate). The application of these procedures generated a representative curve that was introduced to the mathematical model and allowed to calculate the measurement values in the corresponding schedules. The results showed an average error of ±0,1676 hPa (hectopascal) of pressure and the average processing time per sheet is 1,03 seconds.