{"title":"A Kalman filter-Hungarian algorithm with a postprocessor for tracking aeolian saltating particle in high-speed video","authors":"Fanmin Mei, Hongji Zhou, Jin Su, Jinguang Chen","doi":"10.1002/esp.6014","DOIUrl":null,"url":null,"abstract":"<p>Saltating particle tracking (SPT) is an essential visualized channel to understand the dynamics of aeolian saltation at sand particle size scale, while the published SPTs could have low recall or accuracy rate and misestimate further saltation intensity. Hence, a Kalman filter-Hungarian algorithm with a postprocessor (KF-H-<i>k</i>) was proposed, where the Kalman filter was employed for predicting particle motion, and the Hungarian algorithm for optimizing global assignment, as well as the postprocessor with <i>k</i>-means cluster for correcting the erroneous recovered tracks by Kalman filter-Hungarian algorithm. The new SPT was validated in a digital high-speed video with various particle concentrations from a wind tunnel experiment. It demonstrated that compared with the previous SPTs, KF-H-<i>k</i> kept the highest and most stable accuracy (85% ~ 93%), the best spatial resolution, the moderate recall rate (50% ~ 70%) and time cost. The present work offers a new hybrid scheme for tracking sand particles accurately but alsodatasets for automatically identifying saltating tracks via machine learning models, very critical for insight into new hypothesis on sand ripple formation.</p>","PeriodicalId":11408,"journal":{"name":"Earth Surface Processes and Landforms","volume":"49 15","pages":"5086-5097"},"PeriodicalIF":2.8000,"publicationDate":"2024-11-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Earth Surface Processes and Landforms","FirstCategoryId":"89","ListUrlMain":"https://onlinelibrary.wiley.com/doi/10.1002/esp.6014","RegionNum":3,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"GEOGRAPHY, PHYSICAL","Score":null,"Total":0}
引用次数: 0
Abstract
Saltating particle tracking (SPT) is an essential visualized channel to understand the dynamics of aeolian saltation at sand particle size scale, while the published SPTs could have low recall or accuracy rate and misestimate further saltation intensity. Hence, a Kalman filter-Hungarian algorithm with a postprocessor (KF-H-k) was proposed, where the Kalman filter was employed for predicting particle motion, and the Hungarian algorithm for optimizing global assignment, as well as the postprocessor with k-means cluster for correcting the erroneous recovered tracks by Kalman filter-Hungarian algorithm. The new SPT was validated in a digital high-speed video with various particle concentrations from a wind tunnel experiment. It demonstrated that compared with the previous SPTs, KF-H-k kept the highest and most stable accuracy (85% ~ 93%), the best spatial resolution, the moderate recall rate (50% ~ 70%) and time cost. The present work offers a new hybrid scheme for tracking sand particles accurately but alsodatasets for automatically identifying saltating tracks via machine learning models, very critical for insight into new hypothesis on sand ripple formation.
期刊介绍:
Earth Surface Processes and Landforms is an interdisciplinary international journal concerned with:
the interactions between surface processes and landforms and landscapes;
that lead to physical, chemical and biological changes; and which in turn create;
current landscapes and the geological record of past landscapes.
Its focus is core to both physical geographical and geological communities, and also the wider geosciences