Early detection of plant diseases and weeds with Support Vector Machines.
geosensor, data mining, support vector machine, pattern recognition, feature selection, curve fitting
The aim of this project is the identification of spatio-temporal patterns in geosensor data, which appear in the evolution of crops or plant diseases. This study examines and transfers the research into methodical problems which occur during the process of generating weed distribution maps in geoinformation systems (GIS) of. M. Backes (Backes 2005: Methodische Probleme bei der Erstellung von Unkrautverteilungskarten mit Geoinformationssystemen (GIS)).
This project is a cooperation inside the Research Training Group with Weed Science at Hohenheim University and Phytomedicine at Bonn University. Based on hyperspectral information for the identification of plant diseases or shape parameters for crop identification different methods of pattern recognition in the field of machine learning and curve fitting methods are used. First of all with data mining patterns in geosensor data are recognise and parameterise. Here especially data preprocessing, feature selection, curve fitting and methods of parameter estimation are used. Afterwards the detected patterns are upgraded through machine learning, where discrimination from classes is learned. Therefore training data is required. At least a classification based on Support Vector Machines, which is a part of learning, is carried out.
Publication - Rumpf, T.; Mahlein, A-K.; Dörschlag, D.; Plümer, L. (2009): Identification of combined vegetation indices for the early detection of plant dieases. Proceedings of the SPIE Europe Conference on Remote Sensing, 31.August 2009 - 3.September 2009, Berlin, Germany
Publication - Weis, M.; Rumpf, T.; Gerhards, R.; Plümer, L. (2009): Comparison of different classification algorithms for weed detection from images based on shape parameters. Proceedings of the 1st International Workshop on Computer Image Analysis in Agriculture, 27. 28. August 2009, Potsdam, Germany
Publication - Römer, C.; Bürling, K.; Rumpf, T.; Hunsche, M.; Noga, G.; Plümer, L. (2010): Early identification of leaf rust on wheat leaves with robust fitting of hyperspectral signatures. Proceedings of 10th International Conference Precision Agriculture, 18. - 21.July 2010, Denver, CO, USA
Publication - Rumpf, T.; Mahlein, A.; Römer, C.; Plümer, L. (2010): Optimal wavelengths for an early identification of Cercospora beticola with Support Vector Machines based on hyperspectral reflection data. In: IEEE (Hg.): 2010 IEEE International Geoscience and Remote Sensing Symposium, 25. - 30.August 2010, Honolulu, Hawaii, USA
Publication - Rumpf, T.; Mahlein, A.; Steiner, U.; Oerke, E-C.; Dehne, H-W.; Plümer, L. (2010): Early detection and classification of plant diseases with Support Vector Machines based on hyperspectral reflectance. Computers and Electronics in Agriculture (74), 91-99
Oral presentation at the SPIE Remote Sensing for Agriculture, Ecosystems, and Hydrology XI, 31.August - 03.September 2009, Berlin.
'Identification of combined vegetation indices for the early detection of plant diseases'
Participation at the 10th International Conference Precision Agriculture, 18. - 21.July 2010, Denver, CO, USA.
Poster presentation at the IEEE International Geoscience and Remote Sensing Symposium, 25. - 30.July 2010, Honolulu, Hawaii, USA.
'Optimal wavelengths for an early identification of Cercospora beticola with Support Vector Machines based on hyperspectral reflection data'