Classification accuracy of a frequency analysis method: comparison between supervised SOM and KNN

Hannula, Manne; Laitinen, Jyrki; Alasaarela, Esko

doi:10.1109/itab.2003.1222525

Cited by 2 publications

(1 citation statement)

References 2 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The class vector has only one value with '1', while rest of the value of class vector is '0'. The location of value '1' in the class vector indicates the class of the particular input feature vector [9,[12][13][14]. During the training, both the feature vector and class vector are considered.…”

Section: Self-organized Maps (Som)mentioning

confidence: 99%

Self-organized maps based spectral prediction of Rotylenchulus reniformis numbers

Doshi

King

Lawrence

2007

2007 IEEE International Geoscience and Remote Sensing Symposium

View full text Add to dashboard Cite

Rotylenchulus reniformis nematodes present in the soil are one of the major nematode parasite species significantly affecting the growth and development of cotton plants. Recent studies have shown that the nematode numbers in the plant's rhizosphere has direct impact on the reflectance of the plants. In this paper, authors utilize this correlation in developing a field worthy methodology for predicting nematode population number extant in the plant's rhizosphere from variable plant's reflectance. To accomplish this task, a supervised Self-organized map (SOM) was trained using the hyperspectral data signatures of cotton plants affected by different known nematode numbers. The hyperspectral signatures used for training were collected from the cotton plants grown in controlled environment. Twelve field samples (uncontrolled environment) with known nematode numbers obtained from lab analysis of the soil were presented to the supervised trained Self-Organized Map. The location of the sample on the labeled supervised-SOM was used to determine the estimated nematode population of the field sample. In addition to the map grid, the locations of the samples were also visualized using U-matrix, to determine whether the samples were not corrupt or located in the junk part of the map. In addition to the primary goal, hyperspectral signatures of both training and testing data were divided into three sub-regions: Visible region, NIR region and Mid-IR region to observe whether any particular region was the most effective in predicting nematode population.

show abstract

Section: Self-organized Maps (Som)mentioning

confidence: 99%