DETECTION OF LANDSLIDES USING INTERNET OF THINGS AND A DEEP LEARNING APPROACH

Abstract

Abstract— Landslide may occur when the upper mass of the soil gets dislocated from the lower layers and travels down slope. It may be caused by a number of reasons like Heavy rainfall, steepy slopes, volcanic erruptions and man-made activities like deforestation, mining etc., increases the fragility of land and make it easier for landslides. In this system, a model is proposed to combine both the Internet of Things and Deep learning methodologies. Moisture sensor and accelerometer sensor are used to read the humidity values. Incase of any sensor failures, a camera module is fixed and is able to monitor the prone zone. To increase the packet reception ratio, the sensors are grouped without interfering each other and are uniformly distributed. Convolutional Neural Network is used to extract the features and be able differentiate the changes in images. As there was no available source for the dataset, a new dataset is built to test the model.

Keywords— Landslide, Sensor, Camera module, CNN, Sink node