US 12,423,970 B1
Method and system for crop mapping across large regions with low sample dependence
Hongwei Zhao, Beijing (CN); Jia Liu, Beijing (CN); Wenbin Wu, Beijing (CN); Lei Wang, Beijing (CN); and Ruoyun Xing, Beijing (CN)
Assigned to INSTITUTE OF AGRICULTURAL RESOURCES AND REGIONAL PLANNING, CHINESE ACADEMY OF AGRICULTURAL SCIENCES, Beijing (CN)
Filed by Institute Of Agricultural Resources And Regional Planning, Chinese Academy Of Agricultural Sciences, Beijing (CN)
Filed on Jan. 14, 2025, as Appl. No. 19/020,448.
Claims priority of application No. 202410711366.3 (CN), filed on Jun. 4, 2024.
Int. Cl. G06V 10/771 (2022.01); G06V 10/26 (2022.01); G06V 10/72 (2022.01); G06V 10/75 (2022.01); G06V 10/764 (2022.01); G06V 10/77 (2022.01); G06V 10/776 (2022.01); G06V 20/10 (2022.01)
CPC G06V 20/188 (2022.01) [G06V 10/267 (2022.01); G06V 10/72 (2022.01); G06V 10/751 (2022.01); G06V 10/764 (2022.01); G06V 10/7715 (2022.01); G06V 10/776 (2022.01)] 9 Claims
OG exemplary drawing
 
1. A method for crop mapping across large regions with low sample dependence, comprising:
acquiring remote sensing data, ground sample data, meteorological data, soil data, basic geographic data and disaster data, the ground sample data comprising crop type and geographic crop information;
establishing geographically divided crop planting regions according to crop growth periods, the meteorological data, the soil data, and basic geographic data;
establishing key growth period model libraries corresponding to individual crop regions, comprising: calculating a normalized difference vegetation index and an enhanced vegetation index, and performing declouding using a declouding algorithm by means of a cloud mask produced by cloud projection of dark pixels; removing outliers using a smoothing method, and extracting, based on sample points, pixels points as a crop supervision dataset; and assigning the ground sample data to the geographically divided crop planting regions according to geographical coordinates, supplementing sample data based on a network platform according to sizes of the geographically divided crop planting regions, and cleaning and correcting offset ground data;
constructing machine learning models based on a plurality of machine learning algorithms in combination with the remote sensing data in the individual geographically divided crop planting regions and during the individual crop growth periods, respectively, to obtain machine learning crop extraction models corresponding to the individual machine learning algorithms;
calculating overall classification accuracy of the individual machine learning crop extraction models, selecting an optimal machine learning crop extraction model among the regions, and determining model parameters;
acquiring a spatial crop distribution base map using the optimal machine learning crop extraction model;
performing product correction based on the disaster information, comprising: in a case where a disaster occurs, determining an abnormal crop recognition region by comparing a current crop map with the spatial crop distribution base map; increasing crop extraction samples in the abnormal crop recognition region, establishing crop extraction models adapted for a disaster response based on a plurality of machine learning algorithms, respectively, and evaluating the individual crop extraction models adapted for the disaster response based on the overall classification accuracy, to obtain a target crop extraction model adapted for the disaster response; and supplementing sample information for a current growth season to correct a target crop extraction model adapted for the disaster response for a last growth period; and
acquiring a regional crop map using the target crop extraction model adapted for the disaster response.