In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B 
340 
Sugar beets (I) 
■* 
Peas (II) 
rn 
Radish (III) 
m 
Peas (after barley, 
IV) 
■■ 
Winter rape (V) 
Sunflower (VI) 
CD 
Spring barley (after 
sugar beets, VII) 
□ 
Winter wheat (after 
peas, VIII) 
■ 
Spring barley (after 
cereals, IX) 
hb 
Winter wheat (after 
radish) 
In the case of normalized distance procedure application 
the classifier calculated standard deviation for reflectance 
values around the mean value and made contours of standard 
deviations. The pixel is assigned to the closest category in 
the form of standard deviations. 
The maximum likelihood procedure is one of the most 
sophisticated, and the most widely used classifier. The 
classification assumes that the statistics for each class in each 
band are normally distributed and calculates the probability 
that a given pixel belongs to a specific class. Each pixel is 
assigned to the class that has the highest probability. 
Multiple schemes of image classification used in 
environmental and agricultural mapping are either 
traditionally statistical or heuristic (region growing, fuzzy 
classification etc.) (Kovalevskaya and Pavlov, 2002). 
Agricultural crops within the fields are not homogeneous 
objects because of the spatial variability of soil quality and 
soil moisture. In this research project we used the general 
schemes of supervised image classification. Training sites 
included only homogeneous areas within fields. 
The final stage of the classification process usually 
involves an accuracy assessment. There are several types of 
accuracy assessments. Usually it is done by generating a 
random set of locations in the field conditions to verify the 
true land cover type. A simple value file is then made to 
record the true land cover class for each of locations. This 
values file is then used with the vector file of point locations 
to create a raster image of the true classes found at the 
locations examined. This raster image is then compared to 
the classified map (Eastman, 2006). 
The Kappa coefficient is another measure of the accuracy 
of the classification. The coefficient is calculated by 
multiplying the total number of pixels in the ground truth 
classes by the sum of the confusion matrix diagonals, 
subtracting the sum of the ground truth pixels in a class times 
the sum of the classified pixels in that class summed over all 
classes, and dividing by the total number of pixels squared 
minus the sum of the ground truth pixels in that class times 
the sum of the classified pixels in that class summed over all 
classes. We used the Kappa coefficient to estimate the 
accuracy of the classification and to evaluate the results of 
applied methods of classification for each crop. 
The error matrix produced was used to identify particular 
crop types for which errors are in excess of that desired. The 
information in the matrix about which crops are being 
mistakenly included in a particular class (errors of 
commission) and those that are being mistakenly excluded 
(errors of omission) from that class can be used to refine the 
classification approach. 
Results showed that in May winter wheat and winter rape 
had the highest value of Kappa coefficient compared to 
another crops in the crop rotation (0.727 and 0.835). For the 
other crops the very low to low level of aboveground 
biomass is characterized. Radish can be recognized only 
during this period because of the early harvesting. The 
Kappa coefficient is 0.647. 
Some fields in May are still bare and some crops have very low 
aboveground biomass. Therefore, accuracy of the classified 
fallow land is above 0.760. 
Because of the weak biomass development for sugar beet, peas, 
sunflower and spring barley in the second decade of June the 
classification results shown that accuracy varied from 0.483 (sugar 
beet) to 0.614 (sunflower) (Figure 4). Winter rape and winter 
wheat had the higher value of Kappa coefficient (0.702 to 0.712) 
compared to the other crops. 
Spring barley is one of the main agricultural crops in Ukraine. 
Results showed that accuracy assessment for this crop in June 
varied from 0.445 to 0.633. The higher value has been obtained for 
crop after the previous winter wheat. 
—♦— Sugar beets I —■— Peas II —6— Radish III 
Peas IV —*—Winter rape V —•—Sunflower VI 
—i— Spring barley VII —-— Winter wheat VIII —™— Spring barley IX 
—*— Winter wheat X 
Figure 4. Accuracy assessment of the crop classification 
Dense vegetation cover in July was mainly used to classify 
winter wheat, spring barley and peas with high percentage of 
identification (from 0.699 for winter wheat to 0.857 for spring 
barley). Kappa coefficient varied for fields with peas from 0.703 
to 0.839. It was related to different nitrogen fertilizer rate applied 
under crops. 
The overall accuracy for cereals and peas was 0.688, for the 
other crops - 0.608. Low accuracies were obtained for sugar beet, 
and sunflower. Because of the broad row-spacing and the within- 
field heterogeneity of crop growth, there was reduced 
classification accuracy. 
4. CONCLUSION 
It can be assumed that IRS-ID images acquired in three periods 
were classified to determine seven different crops. The overall 
accuracy for cereals and peas was 0.688 and for the other crops 
was 0.608. 
In comparison to the studies (Turker et al., 2005, Conrad et al., 
2010) the overall accuracy of classification was lower because of 
heterogeneity of crops within some fields, differences in 
agricultural technologies applied to the same group of crops, 
spatial variability of soil nutrients and moisture. 
Additional image acquisition and database development of 
spectral signatures based on the results received within the 
experimental stations using sensors with medium resolution can 
provide images with lower spatial resolution to determine certain 
crops within the larger territories.