Spatial heterogeneity-based Voronoi matching method for GlobeLand30 data inconsistency detection: a case study of Linqu County, Shandong, China

Remotely sensed data have become a valuable source of data for mapping regional or global land cover. Referable to the spectral complexity and the quality of remotely sensed data selected, classification errors usually occurred in land cover mapping, where the errors occurred can result in an issue in which the spatial heterogeneity in remotely sensed data is inconsistent with that in land cover data. To ensure data quality, land cover data checking mainly depends on manual labor, which is time-consuming and labor-intensive. To tackle this issue, a spatial heterogeneity-based Voronoi matching method for automated detection of land cover data inconsistency is proposed. First, Harris corner-based Voronoi diagrams (HCVDs) are constructed utilizing the image noise filter and Harris corner detector. Second, the regional statistics of range on land cover data using HCVDs is calculated. Ultimately, by matching the heterogeneity in HCVDs and using HCVDs-based first-order neighbor filter, where inconsistent spatial heterogeneity occurs in land cover data can be found. A GlobeLand30 case study from Linqu County, Shandong, for the year 2010 in China is carried out. After detection and validation, this method is shown to be workable for detecting spatial regions where inconsistent spatial heterogeneity occurs in land cover data, and the data quality is improved by 6.3% in this study area after manual verification.