An Approach to Rain Detection Using Sobel Image Pre-processing and Convolutional Neuronal Networks

Abstract

Rain fall detection has been an important factor under study in a multitude of applications: estimation offloods in order to minimize damage before an environmental risk situation, rain removal from images, agriculture field, etc. Actually, there are numerous methods implemented in order to try to solve this issue. For example, some of them are based on the traditional weather station or in the use of radar technology. In this work, we propose an approach to rain detection using image processing techniques and Convolutional Neuronal Networks (CNN). In order to improve the results of classification, images in rain and no rain conditions are pre-processed using the Sobel algorithm to detect edges. The architecture that defines the CNN is LeNet and it is carried out with three convolutional layers, three pooling layers and a soft max layer. With the proposed method, it is possible to detect the presence of rain in certain region of the image with an accuracy of 89%. The purpose of the proposed system is just to complete with a different added value, other traditional methods for detection of rain.