Φ-lab visiting professor advances the application of quantum computing in EO
Prof. Mihai Datcu, Visiting Professor at Φ-lab, has recently published a paper on the use of Quantum Computing in hyperspectral imaging, a compelling and topical area which is very much in line with the QC4EO initiative. The paper is included in the prestigious IEEE J-STARS journal and describes the first application of a quantum annealer in EO data analysis.
Quantum technologies are constantly evolving, with Quantum Computing (QC) enjoying fast-track development and considerable investment within the ICT industry. The Quantum Computing for Earth Observation (QC4EO) initiative is one of Φ-lab’s flagship programmes, aimed at exploring new computing paradigms to solve previously intractable data-analysis problems in EO.
Initially created as a collaborative venture between ESA and CERN, QC4EO now has a number of partners including DLR, ECMWF, ELLIS, LRZ and TUM. The initiative seeks to leverage the increasing power of quantum computers to explore new EO applications. Some of the potential of QC4EO lies in the intersection of EO problems with QC’s capabilities for optimising large-scale systems and Quantum Machine Learning (QML).
One of the most interesting applications is the processing of images from hyperspectral satellite sensors. The substantial volumes of data from the HYP image sensors are organised in a number of spectral bands, but not all the bands have the same information value for the classification of features such as land cover types. Prof. Datcu explains: “Selecting the most informative and trustworthy bands for each land-cover class is an important analysis step for high-accuracy classification tasks and saving internal storage space. Although there are conventional computational tools for this selection process, known as annealers, we knew from theoretical studies that quantum technologies have the potential to carry out the optimisation more efficiently. Until now though, there hasn’t been any real-world research in this area.”
Prof. Datcu’s work involved comparing the performance of quantum and traditional computational techniques in processing the hyperspectral data. First, a D-Wave Quantum Annealer (D-Wave QA) was benchmarked against a conventional annealer for the task of band selection. Next, the D-Wave QA-selected bands were split into the respective land-cover classes with both quantum and conventional classifiers.
The results were encouraging: “In the first part of the study, we proved that the D-Wave QA selected the appropriate bands and that the quantum methodology reduced both storage space and computational load,” Prof. Datcu comments. “In the case of the classifiers, the quantum versions even outperformed their conventional counterparts in most instances.”
Leading on from the success of the study, Prof. Datcu’s future work at QC4EO will develop the concept of a hybrid quantum-classical network for EO datasets, creating a tool that can operate across a range of dataset types and sizes.
When asked about this and other collaborative research he is working on at ESA, Prof. Datcu is decidedly enthusiastic: “Φ-lab is a fantastic opportunity to study interdisciplinary, highly complex EO activities such as mission design, remote sensing, environmental parameter retrieval and the interpretation of calibrated results. I see my job as putting together a new vision of AI4EO [Artificial Intelligence for Earth Observation], and inevitably the promising resources of quantum computing will play a major part in that.”
The full paper, A Quantum Annealer for Subset Feature Selection and the Classification of Hyperspectral Images by Soronzonbold Otgonbaatar and Mihai Datcu, can be found here. Other recent papers co-authored by Prof. Datcu include:
- S. Otgonbaatar and M. Datcu, Natural Embedding of the Stokes Parameters of Polarimetric Synthetic Aperture Radar Images in a Gate-Based Quantum Computer, in IEEE Transactions on Geoscience and Remote Sensing, DOI: 10.1109/TGRS.2021.3110056, downloadable here
- S. Otgonbaatar and M. Datcu, Classification of Remote Sensing Images With Parameterised Quantum Gates, in IEEE Geoscience and Remote Sensing Letters, DOI: 10.1109/LGRS.2021.3108014
To know more: QC4EO, Φ-lab Flagship Programmes