Improved Hyperspectral Vegetation Detection Using Neural Networks with Spectral Angle Mapper

Date

2017-04-13

Author

Özdemir, Okan Bilge
Çetin, Yasemin

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

32
views

0
downloads

Cite This

Hyperspectral images have been used in many areas including city planning, mining and military decision support systems. Hyperspectral image analysis techniques have a great potential for vegetation detection and classification with their capability to identify the spectral differences across the electromagnetic spectrum due to their ability to provide information about the chemical compositions of materials. This study introduces a vegetation detection method employing Artificial Neural Network (ANN) over hyperspectral imaging. The algorithm employed backpropagation MLP algorithm for training neural networks. The performance of ANN is improved by the joint use with Spectral Angle Mapper (SAM). The algorithm first obtains the certainty measure from ANN, following the completion of this process, every pixels' angular distance is computed by SAM. The certainty measure is divided by angular distance. Results from ANN, SAM and Support Vector Machine (SVM) algorithms are compared and evaluated with the result of the algorithm. Limited number of training samples are used for training. The results demonstrate that joint use of ANN and SAM significantly improves classification accuracy for smaller training samples

Subject Keywords

Hyperspectral, Classification, SVM, Neural Networks, Spectral Angle Mapper

URI

https://hdl.handle.net/11511/51834

Collections

Graduate School of Informatics, Conference / Seminar