Tulika Jain, Shah & Anchor Kutchhi Engineering College; Ashish Shethia, Shah and Anchor Kutchhi Engineering college; Siddhi Khanvilkar, Shah & Anchor Kutchhi Engineering College; Linesh Patil, Shah & Anchor Kutchhi Engineering College; Vidyullata Devmane, Shah & Anchor Kutchhi Engineering College; Srikanth Kodeboyina, Blue Eye Soft Corporation
Keywords: Asteroid, You Only Look Once (YOLO), Space, Threats, Convolutional Neural Network (CNN), Near-Earth Objects, Astrometric Parameters, Celestial Bodies, Risk Prediction, Asteroid Collisions.
Abstract:
Outer space surveys have obtained valuable information about celestial bodies which provides enough evidence to prove that the majority of asteroids flow inside the asteroid belt between Mars and Jupiter. Asteroids continue to move inside the same belt due to Jupiter’s gravitational force. However, on occasion due to the large planets gravitational force asteroids get defected and become an impediment for other planets. Asteroids are rocky bodies moving across the sun and are too small as compared to planets. These asteroids are formed due to the collision of two gas clouds or planets, so they have abnormal shapes. Whilst an asteroid passes very near to the sun, it becomes hot and starts burning to provide light, this kind of burning asteroid is referred to as comets. Sometimes a small part of an asteroid or a comet, called a meteoroid enters the Earths atmosphere and creates havoc on the earths floor that is dangerous for life on earth. Therefore, it is a necessity to detect an asteroid coming towards the Earth and analyze its behavior to avoid losses. In our work, we have worked on asteroid detection and predicted risk to the earth surface, considering the parameters such as distance from the Earth, detected asteroid velocity, collision time and size of the asteroid. We have used deep learning concepts and astrometric calculations to gain greater accurate consequences. The research outcome shows the estimated risk imposed by the asteroid on Earth using the fastest single neural network, YOLO.
The interest in the detection and tracking of Near-Earth Asteroids (NEAs) has grown dramatically over the previous few decades and NASA articles about the greatest threat to earth might come from. At the same time, technology has also been increasing at an alarming rate, enabling well-known surveys and amateur mini-surveys to decorate their solutions to attain better predictions. Based on this information presented in various surveys, we are making an attempt to use more reliable technology for detecting asteroids and predicting their collision threat to earth and its living being. Numerous attempts, studies, researches and software are made to locate near-earth objects. A number of the preceding software like Astrometrica, Astrometry.net, SCAMP packages were developed for the detection of asteroids. Astrometrica is a smart application for the logical assessment of astrometric data reduction of CCD pictures, focusing on the measurement of small celestial bodies. Astrometry.net software generates astrometric calibration meta-data and additionally recognizes other celestial bodies within the image. Many techniques used previously in Asteroid detection include Blink Technique, Moving Object Detection Technique, DRIFT-SCAN Technique, and step-stare Technique, few take a sequential scan of images of the same area and few uses multiple images, each being exposed while the telescope tracks along with the Earth rotation. The Lincoln Near-Earth Asteroid Research program funded by NASA conducted at the Massachusetts Institute of Technology Lincoln Laboratory Experimental Test Site with a goal of improving knowledge of Near-Earth Objects.
In this work, the input is given as a sequence of night sky images for the proposed system, if the input is a sequence of night sky photographs, a Python module called Pillow is used to eliminate background noise. These pre-processed images are now ready for the next step of the moving object detection procedure.
After getting scaled processed images in the previous stage, the detection part and also classification among unvarying and moving objects will be carried out using YOLO. The YOLO algorithm divides full images into regions called bounding boxes. YOLO gives high accuracy and is extremely fast. YOLO works on full images and gives bounding boxes on targeted objects as output. After successfully detecting the asteroid, astrometric calculations are done to figure out the parameters such as distance from Earth, the velocity of an asteroid, size of asteroid and time for collision. The further trajectory of the detected asteroid needs to be calculated so we can determine whether the asteroid falls on the earth’s orbit or not. If the asteroid doesnt fall on the earths trajectory so according to the proposed system, it poses a negligible risk to the earth. But if the trajectory of an asteroid falls on earth orbit it would be a threat to earth so we will predict the risk.
After knowing the trajectory and other astrometric parameters we would compare it with the catalogue of existing asteroids to verify whether the detected asteroid is known or not. If the asteroid is present in the catalogue, display all the known data and determine risk accordingly. Using astrometric calculations done in the previous stage, the risk is predicted as the final output of the system.
For risk prediction, we are considering the existing threshold values mentioned by the space agencies to calculate various factors. Risks do not depend on individual factors, for risk calculation, multiple factors are taken into consideration. Here, we are talking of four major factors, the distance of an asteroid from the Earth, the velocity of the asteroid, the size of the asteroid and the time taken to reach the Earth. An effort is made to predict risk from three factors out of four factors. This will help the user to understand the seriousness of the situation and according to the risk, planetary defence techniques can be implemented. This can save a significant amount of time for deciding on planetary defence measures by employing an automated approach based selection of techniques depending on the range of risk levels. We have successfully identified a solution for the quick detection of an asteroid and the risk considerations associated with it are discussed. The YOLOv4 algorithm is used in the suggested pipeline, which processes 45 frames per second, which is faster than in real-time. Estimation of risk is done after acknowledging the asteroid. Risk is the chance of the asteroid colliding with Earth. Risk estimation is done on the basis of defined parameters explained as per the astrometric calculations and these calculations are straightforward to understand.
Date of Conference: September 14-17, 2021
Track: Machine Learning for SSA Applications