MACHINE LEARNING ALGORITHMS FOR LANDSLIDE DETECTION AND FORECASTING USING SATELLITE IMAGERY

Landslides in steep regions may be caused by man-made elements such unforeseen structures or natural ones like excessive rainfall, earthquakes, and soil moisture. Automatic prediction may help prevent landslides, which can be catastrophic and result in significant property and human loss. Recently, landslides have been automatically identified using machine learning techniques. For the semiautomatic identification and prediction of landslides, a variety of feature extraction and classification-based techniques have been used to satellite pictures. Nevertheless, there hasn't been much study on totally automated detection with respectable accuracy. Finding a suitable training database and producing very accurate test results is the most difficult job in the categorisation and prediction of landslides using satellite photos. Finding the research gap is the main goal of a thorough analysis of the many methods utilised for landslip detection and classification using satellite imagery. Proposing a unique strategy prototype for the same job is the secondary goal. For analysis, fifty publications from reputable journals that use machine learning and deep learning algorithms are taken into consideration. The performance of several categorisation methods from current research is compiled in this article, which then compares and discusses their accuracy. An efficient landslip categorisation technique prototype is suggested based on the gap found. Tested using an expanded Beijing dataset of 770 satellite photos, a slightly altered version of the deep learning network ResNet101 produces an accuracy of 96.88%. Additionally, the study gives the researchers an overview, current state, and future directions of machine and deep learning algorithms for landslip detection. The methods covered will be a useful tool for spotting research gaps, mentoring aspiring scholars, and encouraging creative inquiry in the area of satellite image-based landslip categorisation.