Methods for accelerating geospatial data processing using quantum computers

Abstract

Quantum computing is a transformative technology with the potential to enhance operations in the space industry through the acceleration of optimization and machine learning processes. Machine learning processes enable automated image classification in geospatial data. Quantum algorithms provide novel approaches for solving these problems and a potential future advantage over current classical techniques. Universal Quantum Computers, developed by Rigetti Computing and other providers, enable fully general quantum algorithms to be executed, with theoretically proven speed-up over classical algorithms in certain cases. This paper describes an approach to satellite image classification using a universal quantum enhancement to convolutional neural networks: the quanvolutional neural network. Using a refined method, we found a performance improvement over previous quantum efforts in this domain and identified potential refinements that could lead to an eventual quantum advantage. We benchmark these networks using the SAT-4 satellite imagery dataset in order to demonstrate the utility of machine learning techniques in the space industry and the potential benefits that quantum machine learning can offer.