A Framework of Filtering Rules over Ground Truth Samples to Achieve Higher Accuracy in Land Cover Maps

Abstract

Remote Sensing (RS) digital classification techniques require sufficient, accurate and ubiquitously distributed ground truth (GT) samples. GT is usually considered “true” per se; however, human errors, or differences in criteria when defining classes, among other reasons, often undermine this veracity. Trusting the GT is so crucial that protocols should be defined for making additional quality checks before passing to the classification stage. Fortunately, the nature of RS imagery allows setting a framework of quality controls to improve the confidence in the GT areas by proposing a set of filtering rules based on data from the images themselves. In our experiment, two pre-existing reference datasets (rDS) were used to obtain GT candidate pixels, over which inconsistencies were identified. This served as a basis for inferring five key filtering rules based on NDVI data, a product available from almost all RS instruments. We evaluated the performance of the rules in four temporal study cases (under backdating and updating scenarios) and two study areas. In each case, a set of GT samples was extracted from the rDS and the set was used both unfiltered (original) and filtered according to the rules. Our proposal shows that the filtered GT samples made it possible to solve usual problems in wilderness and agricultural categories. Indeed, the confusion matrices revealed, on average, an increase in the overall accuracy of 10.9, a decrease in the omission error of 16.8, and a decrease in the commission error of 14.0, all values in percent points. Filtering rules corrected inconsistencies in the GT samples extracted from the rDS by considering inter-annual and intra-annual differences, scale issues, multiple behaviours over time and labelling misassignments. Therefore, although some intrinsic limitations have been detected (as in mixed forests), the protocol allows a much better Land Cover mapping thanks to using more robust GT samples, something particularly important in a multitemporal context in which accounting for phenology is essential.

Automatic Detection of Aquatic Weeds: A Case Study in the Guadiana River, Spain

“…Several research efforts have considered digital aerial orthophotographs collected by PNOA for photointerpretation tasks, due to their high spatial resolution [44], [45], [46]. Some projects managed by the Ministry of Development (Spain) follow the INSPIRE Directive [47], which establishes a spatial data infrastructure for geographic data in Europe to provide geometric and temporal coherence of cartographic and geographic databases.…”

Driving Forces of Forest Expansion Dynamics across the Iberian Peninsula (1987–2017): A Spatio-Temporal Transect

“…The k-nearest neighbor (kNN) classifier was implemented and parallelized in MiraMon GIS and remote sensing software [53]. The classification process followed the methodology detailed in [54][55][56]. In Padial-Iglesias et…”

“…The k-nearest neighbor (kNN) classifier was implemented and parallelized in MiraMon GIS and remote sensing software [53]. The classification process followed the methodology detailed in [54][55][56]. In Padial-Iglesias et al [54], the authors generated accurate ground truth areas using the SIOSE database and filtering rules based on the inner imagery NDVI data to correct inconsistencies in the initial samples by considering inter-annual and intra-annual differences, scale issues, multiple behaviors over time, and labelling misassignments.…”

“…The classification process followed the methodology detailed in [54][55][56]. In Padial-Iglesias et al [54], the authors generated accurate ground truth areas using the SIOSE database and filtering rules based on the inner imagery NDVI data to correct inconsistencies in the initial samples by considering inter-annual and intra-annual differences, scale issues, multiple behaviors over time, and labelling misassignments. Furthermore, the authors considered phenologically representative imagery, image preprocessing lines, and auxiliary variables estimation for the classification process.…”

Characterization of land use cover datasets from a global to an Andalusian level: an “obstacle course” for users

“…Even though this variety, the large amount of categorical information is also a valuable source of reference data (training areas), which properly managed and filtered, can serve to generate new and more reliable LUC maps series. Examples of the use of different datasets for classification purposes can be found (Vidal-Macua et al, 2017;Gonzalez-Guerrero et al, 2020;Padial Iglesias et al, 2021).…”