A novel approach for automatic avalanche detection System Mr Ashish Sharma¹, Dr Farzil Kidwai¹, Mr Yogesh Sharma¹, Dr Sandeep Tayal¹, Prakhar Sharma¹, Harsh Rajput¹, Divleen Kaur¹, Tripti Jain¹, Mridul Gulati¹, Ishaan Sangwan¹ prakharsharma22@gmail.com ashish@mait.ac.in sandeeptayal@mait.ac.in yogeshsharma@mait.ac.in farzilkidwai@mait.ac.in ¹Computer Science and Engineering Department, Maharaja Agrasen Institute of Technology, Affiliated to G.G.S.I.P. University. dchugh121@gmail.com jaintripti59@gmail.com b.birla167@gmail.com vkharshrajput@gmail.com mridulgulati18@gmail.com ishaansangwan@gmail.com Abstract -Cities worldwide are reaching operational excellence and growing smarter with the introduction of innovative technologies. These technologies, if utilized efficiently, can become assets of tomorrow. Many weather-related natural disasters have increased in frequency and severity in recent times. These natural disasters can devastate the environment, the economy, and life as we know it. Technology, if used as an instrumental asset can enhance our preparedness for emergencies. Utilizing vision-based strategies to prevent natural disasters like avalanches have been one of the significant advancements. A valuable tool for avalanche forecasting and evaluating the efficacy of avalanche prevention measures in low visibility conditions is automated snow avalanche prediction. Automation offers an easy escape from the hassle of continuous human monitoring of avalanche activity and increases detection speed and accuracy. Our solution offers an economic and precise alternative to traditional prediction systems. It takes into account different climatological factors related to snow stability to predict avalanche danger. The system analyses past and present activity to predict if avalanches can be developed. It consists of a thermal imaging camera that collects infrared images of cracked snow. These images are then subsequently examined with the help of a deep-learning image processing model to investigate the likelihood of an avalanche occurring in the crack. The prediction, if positive the authorities can be alerted and further necessary actions can be taken. Keywords: avalanche 1 , automated avalanche detection 2 , cracks 3 , thermal camera 4 , image processing 5 , Convolutional Neural Networks (CNN) 6 , Geotagging 7 , OTSU algorithm 8 . INTRODUCTION Natural calamities have always existed. However, many weather-related natural disasters have increased in frequency and severity due to the modernization of several cultures around the world and the impact human industrial activities have made on the environment. This results in a greater overall worldwide impact of natural disasters. The number of lives lost due to these mishaps is appalling. Graph 1: It shows the number of deaths due to avalanches over the decades These unfortunate instances, though cannot be avoided, can be less destructive if one is better prepared. Technology has empowered us to do this. Multiple approaches to help predict the arrival and effect of natural calamities have been proposed using different technological stacks. The motive of this study is to put forward one such strategy for avalanches. When a layer of snow moves downhill and collapses, it triggers an avalanche. The release of snow is attributed to a number of mechanisms, including snow distortion, damage accumulation, fracture initiating, and crack propagation. This results in either dry slab avalanches or loose avalanches. The release of snow is also a result of complex interactions between the terrain, the snowpack, and the environment. Our prediction system consists of a thermal camera and an image-processing deep-learning model. The DL model is based on OTSU Image Processing Algorithm that quantifies cracks and checks the images for damage severity. As part of Electronic copy available at: https://ssrn.com/abstract=4392679