Geospatial Data Analytics lab excels in global competition
For the second consecutive year, researchers from Ohio State were recognized internationally for their research in remote sensing. Rongjun Qin, assistant professor of geodetic engineering and director of Ohio State's Geospatial Data Analytics Lab, and team members Huijun Chen, Changlin Xiao and Wei Liu were named winners in two separate tracks of the 2020 Institute of Electrical and Electronics Engineers (IEEE) Geoscience and Remote Sensing Society (GRSS) Data Fusion Contest.
Remote sensing involves scanning the earth by satellite, unmanned aerial vehicles (UAVs) or aircraft flying at very high altitudes, in order to obtain data about the planet and its features. The annual competition, the largest of its kind in the geosience and remote sensing community, challenged 159 teams from around the globe to engage machine learning problems related to data aquisition of earth's terrain, oceans, atmosphere and space. Using a standard data set supplied to all participants, Ohio State's team created global land cover maps of terrain such as forests, wetlands, and urban environments with low- and high-resolution labels.
Qin recalled that the "remote" element of the competition took on new meaning during this year's competition. "We had team members in Columbus but also in China," he said. Furthermore, team members were working from home, without the high-performance computing facilities in their lab space to rely on. Sharing large files of computational results (30+ GB) and analysis for every single algorithmic trial, coupled with members working in vastly different time zones on machines with vastly different computational power, required innovation on the part of all involved.
"We mirrored the environments and sent the codes directly to each team member to allow them to run computations and generate results and analysis locally," Qin stated. The group utilized live chats and whatever other tools that were readily available to share data, results and ideas. For its efforts, the Ohio State team was awarded a first place finish in the Land Cover Classification with Low- and High-Resolution Labels track and a fourth place finish in the Land Cover Classification with Low-Resolution Labels track. Qin likened the contest's empasis on accuracy to auto racers' quest for still-faster times on the track. "Whichever team achieves better accuracy, pushes the capability boundary a bit more," he stated.
The win, further cemented the stature of Ohio State's Geodetic Engineering program as a leader in not only photogrammetry but in geosience and remote sensing as well. Noting that many teams in the field were full-time operations with dedicated staff, Allison MacKay, Professor and Chair of the CEGE applauded Professor Qin's team of student researchers. "In addition to the talents of Rongjun's team at algorithm development, this also highlights his skills as a leader by coordinating a team remotely," she remarked.
Qin, whose work is supported by the Office of Naval Research, noted that winning the competition winners reflects well on Ohio State's machine learning practice in remote sensing. "I'm very proud that all went well," he said of his team's efforts during the contest.
IEEE, an association dedicated to using data from satellites to understand the Earth and to advance innovation and technological excellence for the benefit of humanity, is the world's largest technical professional society.