University of Twente Student Theses


Vegetable mapping using fuzzy Dynamic Time Warping

Moola, Wisdom Simataa (2020) Vegetable mapping using fuzzy Dynamic Time Warping.

[img] PDF
Abstract:Fuzzy classification is an approach that is used in image classification when pixels are assigned membership to classes there are closely related. This research focuses on the concept of fuzzy classification to classify Time Weighed Dynamic Time Warping with Spring (TWDTWS) distances generated by TWDTWS algorithm on Sentinel 1A images. Determination of class separability was done using Transformed Divergency (TD) on the Vertically transmitted and Horizontally received (VH), Vertically transmitted and Vertically received (VV) and VH minus VV. The TWDTWS distances in each of the class queries for Chili, Tomato, Cucumber, Rice, Maize, Trees and Others were normalised. TWDTWS distances in each class query were first normalised and then used to compute fuzzy memberships for each class query using the Gaussian membership function. The computed fuzzy memberships for the best class and runner-up class in each pixel were then normalised to a sum of 1. Measure of similarity for Confusion Index (CI), Ambiguity Index (AI) fuzziness and fuzzy membership were used and different thresholds on each of the measures where applied in defuzzification. The result for TD shows that the classes are more separable when using the VH polarisation as compared to VV and VH minus VV. The results for accuracy assessment of the defuzzification output had an overall accuracy of 0.86 and kappa coefficient of 0.83 were obtained for thresholds of 0.6, 0.3, 1.2 and 0.7 for CI, AI, fuzziness and fuzzy membership respectively. The improved overall accuracy and kappa coefficient are due to pixels with less fuzzy memberships to the best class that remained unclassified. Therefore, further work should be undertaken using the approach used in this study when determining better sites for training and validation samples in remote sensing.
Item Type:Essay (Master)
Faculty:ITC: Faculty of Geo-information Science and Earth Observation
Programme:Geoinformation Science and Earth Observation MSc (75014)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page