NEWS AI and hyperspectral imaging detect methane from orbit


A new system combining machine learning with hyperspectral imaging can automatically identify ‘super-emitters’ of methane from orbit, facilitating more effective measures to curb greenhouse gas emissions. Oxford University researchers


collaborated with social enterprise Trillium Technologies’ NIO.space (Networked Intelligence in Space) initiative to develop a system to detect methane plumes on Earth from orbit, according to details published in Nature Scientific Reports. “What makes this research particularly exciting and relevant is the fact that many more hyperspectral satellites are due to be deployed in the coming years, including from ESA, NASA, and the private sector,” said PhD student and lead researcher Vít Růžička. “In combination, these new sensors will provide global hyperspectral coverage, enabling worldwide automated detection of methane plumes.” The new machine-learning tool overcomes existing


challenges in mapping methane plumes from aerial imagery. This has traditionally been difficult as methane is transparent to both the human eye and the spectral ranges used in most satellite sensors. By detecting narrower bands


in hyperspectral data, noise can be filtered out more effectively than traditional multispectral satellites. Yet, because it collects more data, artificial intelligence is required for analysis. The model was trained using 167,825 hyperspectral tiles captured by NASA’s aerial sensor AVIRIS over the ‘Four Corners’ area of the US. It was also tested on other sensors in orbit, such as NASA’s new hyperspectral sensor, Earth Surface Mineral Dust Source Investigation mission (EMIT), which is attached to the International Space Station. Overall, the model was found


to deliver an impressive 81% accuracy in detecting large methane plumes, surpassing the previous most accurate approach by 21.5%. Růžička said: “Such on-


board processing could mean


The AI model was found to deliver an impressive 81% accuracy in detecting large methane plumes


that initially only priority alerts would need to be sent back to Earth, for instance a text alert signal with the coordinates of an identified methane source. This would allow for a swarm of satellites to collaborate autonomously: an initial weak detection could serve as a tip- off signal for the other satellites in the constellation to focus their imagers on the location of interest.” Prof Andrew Markham, supervisor for the research, added: “In the face of climate change, these kinds of techniques allow independent, global validation about the production and leakage


of greenhouse gases. This approach could easily be extended to other important pollutants. Our ambition is to run these approaches onboard the satellites themselves, making instant detection a reality.” The researchers will foster


further research by open- sourcing the annotated dataset and code on GitHub. They are also exploring the possibility of onboard satellite processing as part of the NIO.space initiative, enabling instant detection and collaboration among satellites. Read ‘Why hyperspectral Earth imaging has a bright future’ on Page 12


Seeing opportunity: highlights from PPMA


By Allan Anderson, UKIVA Chairman


Machine vision systems played a minor role at the recent PPMA show at Birmingham’s NEC, with most of the attention on packaging and processing machinery. However, visitors who sought them out will have found food for thought. A current trend is the use


of AI techniques such as deep learning (DL) to enhance machine vision. A good example was seen from Clearview Imaging, which helped a producer of fresh desserts to find defects – such as a single human hair – on its products. It devised a system – trained


using DL – to identify these kinds of items. Clearview captured many images of foreign objects – at various angles – and used these to train a convolutional neural network (CNN). The system was able to find defects with high accuracy. “One benefit is that you


can use this system even if you’re not competent at programming,” said Clearview’s Hashem Khan. It wasn’t just food under inspection. Observant Innovations has made a version of its Inspect 3D package – which analyses solid surfaces – available as a standalone application. Usually, it is supplied along with cameras and lighting to help clients reconstruct and measure very challenging surfaces – such as a scratch on a turbine drum. “Now, the output of a line


scan camera can be imported into the cut-down version of Inspect 3D,” said Gareth Edwards, the company’s technical director. “This will allow interactive lighting and measurement and analysis.” At Multipix Imaging, managing director Simon Hickman explained how a sophisticated 3D camera – from Photoneo of Slovakia – underpins a random bin-picking system. Here, the 3D camera took 17 fast scans of a container full of components, helping a robot arm to ‘decide’ which part to pick. This depended on factors such as accessibility, as some parts were impossible to pick (if they were against the container wall, for instance). As well as the camera, the system depended on a ‘black box’ – running on Linux – to process the data. Pharmaceutical inspection is


also critical – as nobody wants to ingest the wrong medicine. Nick Cox, area sales manager at Scanware, demonstrated a new system called Spectra Bulk QI that analyses tablets and capsules as they move along lanes prior to being bottled. These can be checked for size, for instance, or colour – which may indicate that the wrong tablet has made its way onto the line. Suspect products like these can be ejected in various ways. The ‘QI’ platform that underpins it will become standard on all Scanware products over the next 18 months, he said. More developments in vision


systems such as those seen at PPMA will be seen at next year’s Machine Vision Conference (www.machinevisionconference. co.uk), on 18-19 June 2024 in Coventry.


8 IMAGING AND MACHINE VISION EUROPE DECEMBER 2023 / JANUARY 2024


@imveurope | www.imveurope.com


NASA


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32