Detecting Space Debris using Deep Learning Algorithms: A Survey

A collection of natural micrometeoroids and artificial items orbiting the earth is known as space debris. Artificial space debris consists of non-operational satellites, parts of rocket, and junk ejected by the spacecrafts etc. Kessler Syndrome dictates that the danger of space debris could be accelerated by collisions among them which will cause a chain reaction that will result in uninhabited earth orbit and leading to the growth of a belt of debris around the earth. The substantial rise in the total quantity of space debris in Earth's orbit poses a severe risk to active spacecraft and satellites. Impacts of orbital space debris on satellites, spacecrafts and huge space assets are serious risk that can seriously affect mission success. Therefore, space debris detection is crucial for satellite protection and space disaster prevention. Furthermore, space debris identification is complex due to its rapid relative velocity and distortions caused by solar radiation in orbit-based monitoring systems. Moreover, space based known objects can be tracked with a telescope from here on earth, while unknown objects cannot. Vision sensors such as monocular photographic cameras are considered as a major job for the detection of space debris, however, these tasks are difficult due to a semantic division between observable characteristics of space debris, as well as a lack of measurable and credible visual information in space environment. Various strategies are available for detecting space debris. However, image-based space debris detection algorithms inspired by powerful pattern recognition ability of deep convolutional neural networks (DCNN) have been examined in this survey and a comprehensive analysis is carried out to define the research issues and progress in this domain.