|
|
CLASSIFICATION OF
COTTON AND SUGARCANE PLANTS ON THE BASIS OF THEIR SPECTRAL
BEHAVIOR
MUHAMMAD SHAHZAD SHIFA1, M. SHAHID NAWEED2,
MUHAMMAD OMAR2,
M. ZEESHAN JHANDIR2 AND TANVEER AHMED3
Abstract: The study
is about the classification of cotton (Gossypium hirsutum L.) and
sugarcane (Saccharum officinarum L.) crops based on spectral
behavior of their plants. The hand held ground based remote sensing
optical Multispectral Radiometer MSR5 has been used for this purpose.
MSR5 scans a scene and gives its digital representation in 5 separate
spectral bands compatible with Landsat satellite images, so the study is
also applicable to Landsat images. To judge the discrimination power of
five spectral bands, used as features to represent the scenes, K-means
algorithm is used for unsupervised clustering of reflectance sample data
set. Computational and visualization results of clustering through
K-means show that MSR5 scans are good candidates for classification
purpose. Supervised classification is achieved using K-nn algorithm, and
98% accurate results of classification are achieved. Choice of MSR5 for
crop classification is good for two reasons: the results are accurate
and the technique is an efficient way to represent an image with only
five real values covering a 1.5 square meter.
1Department of Physics, The Islamia University of Bahawalpur, Pakistan
2Department of Computer Science & IT, The Islamia University of
Bahawalpur, Pakistan
3Department of Computer Science COMSATS, Institute of Information
Technology, Abbottabad, Pakistan.
|
|
|