Crop and weed species recognition based on hyperspectral sensing and active learning
In this paper, a new active learning method is proposed, which distinguishes between crop and weed based on their different spectral reflectances. A method comprising novel detection and incremental class augmentation was implemented. Novel detection was based on one class classifiers. Under field conditions, crop and weed species can be recognised successfully. Best results for the active learning were obtained for the one-class Mixture of Gaussians (MOG) and Self Organising Map (SOM) classifiers. The MOG and SOM performance in crop recognition was 100% and 96% respectively while the recognition for different weed species ranged from 31-98% (MOG) and 43-94% (SOM).
Keywordsspectrograph self-organizing map mixture of Gaussians one-class classifier
- Duda, R. and Hart, P. 1973. Pattern Classification and Scene Analysis. John Wiley & Sons, New York, 512 pp.Google Scholar
- Herrala, E., Okkonen, J., Hyvarinen, T., Aikio, M. and Lammasniemi, J. 1994. Imaging spectrometer for process industry applications SPIE 2248 33–40.Google Scholar
- Saunders, R. and Gero, J. S. 2001. A Curious Design Agent: A Computational Model of Novelty-Seeking Behaviour in Design. In: (Eds.) Gero, J. S., Chase, S. and Rosenman, S., Proceedings of the Sixth Conference on Computer Aided Architectural Design Research in Asia (CAADRIA 2001). Key Centre of Design Computing and Cognition, University of Sydney, Sydney, pp 345–50.Google Scholar