Will It Blend? Visualization and Accuracy Evaluation of High-Resolution Fuzzy Vegetation Maps

Abstract

ABSTRACT:Instead of assigning every map pixel to a single class, fuzzy classification includes information on the class assigned to each pixel but also the certainty of this class and the alternative possible classes based on fuzzy set theory. The advantages of fuzzy classification for vegetation mapping are well recognized, but the accuracy and uncertainty of fuzzy maps cannot be directly quantified with indices developed for hard-boundary categorizations. The rich information in such a map is impossible to convey with a single map product or accuracy figure. Here we introduce a suite of evaluation indices and visualization products for fuzzy maps generated with ensemble classifiers. We also propose a way of evaluating classwise prediction certainty with "dominance profiles" visualizing the number of pixels in bins according to the probability of the dominant class, also showing the probability of all the other classes. Together, these data products allow a quantitative understanding of the rich information in a fuzzy raster map both for individual classes and in terms of variability in space, and also establish the connection between spatially explicit class certainty and traditional accuracy metrics. These map products are directly comparable to widely used hard boundary evaluation procedures, support active learning-based iterative classification and can be applied for operational use.

Mapping grassland plant communities using a fuzzy approach to address floristic and spectral uncertainty

“…To improve visualization of the fuzzy classification, class membership maps for each plant community were featured using hue‐preserving colour blending (Chuang, Weiskopf, & Moller, ). The main advantages are that each vegetation cluster has a specific colour and that the colour's saturation level indicates the certainty of class assignment (Zlinszky & Kania, ). The spatial distribution of grassland habitats along wetlands is consistent with their ecological preferendum : for example, Cluster F2 is clearly distributed in topographical depressions and long‐term flood areas, while Clusters F7 and F8 occur at higher elevations that are rarely flooded.…”

“…Subset of classified vegetation maps derived from the fuzzy noise clustering: the fuzzy blended image (left) using the hue‐preserving algorithm (see Zlinszky & Kania, ) shows gradual transitions between classes; the uncertainty image (right) shows areas of high uncertainty in black and more certain areas in white [Colour figure can be viewed at wileyonlinelibrary.com]…”

Detecting shrub encroachment in seminatural grasslands using UAS LiDAR

“…To increase the amount of training data used in the final model for extracting biomass metrics, we applied a model validated by the 90/10 split (Appendix ). From the predictions, we obtained the class probabilities allowing fuzzy classifications, that is, the membership of a class is represented by a probability value between 0 and 1 rather than a Boolean value (true or false) as with traditional hard‐boundary classification(Foody, 1996; Zlinszky & Kania, 2016). Finally, a classification was performed with 100% of the reference data and printed into the full point cloud for visualization purposes.…”

Biodiversity Mapping via Natura 2000 Conservation Status and Ebv Assessment Using Airborne Laser Scanning in Alkali Grasslands

“…All in all this resulted in a multi-band pseudo-image of 70 LIDAR data products, which was the basis of vegetation classification and feature detection. For this, we developed and used the Vegetation Classification Studio (VCS) software tool (Zlinszky and Kania, 2016) to train a random forest machine learning classifier (Breiman, 2001;Pedregosa et al, 2011) that supports fuzzy class prediction output, visualization and accuracy evaluation. In fuzzy classification, each pixel is assigned not only a single class but a vector representing the probabilities of the pixel belonging to each of the individual classes.…”