We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
DETECTION OF CIRSIUM ARVENSE L. IN CEREALS USING A MULTISPECTRAL IMAGING AND VEGETATION INDICES.
- Authors
Hamouz, Pavel; Hamouzová, Kateřina; Soukup, Josef
- Abstract
The objective of this study is to find algorithms for detection of Cirsium arvense in cereals using airborne high resolution multispectral imaging. The imaging of three winter wheat fields infested with C. arvense was realized in July 2007, just before harvest. The images were taken from helicopter using three-band (R,G,NIR) multispectral camera at spatial resolution of 0.1 m. Ground truth data were collected by ground imaging using SLR camera and separated into two classes according to coverage of C. arvense. NDVI and DVI vegetation indices were calculated. The classification accuracy was calculated by comparing each pixel with corresponding ground truth data and was defined as percentage of correctly classified pixels from all classified pixels. The best threshold value of vegetation indices, which provides the highest accuracy of classification was determined. Correlation coefficients were also calculated. Pearson`s correlation coefficient of r = 0.781 0.851 was achieved between NDVI and coverage of C. arvense. Optimal threshold value of NDVI varied depending on field from 0.41 to 0.47. Using of these threshold values for the classification resulted in 85.5% 90.5% overall classification accuracy. DVI provided the correlation of r = 0.779 0.870. For DVI, the classification accuracy ranged from 83.5% to 92.5 % with threshold value of 0.130 0.156.
- Subjects
ALGORITHMS; GRAIN; REMOTE sensing; MULTISPECTRAL imaging; VEGETATION management; STATISTICAL correlation; PRECISION farming; WEEDS
- Publication
Herbologia, 2009, Vol 10, Issue 2, p41
- ISSN
1840-0809
- Publication type
Article