276
11: Teerasit Kasetkasem, Manoj K. Arora, Pramod K. Varshney
Table 11.4. Classification accuracy of sub-pixel maps
Initial SPM
Kappa
coefficient
0.4126
11.5
Summary
95% Confidence
Interval
(0.37280.4524)
Resulting SPM
Kappa
coefficient
0.5511
95% Confidence
Interval
(0.48940.6128)
In this chapter, a new sub-pixel mapping algorithm based on the MRF model
is proposed. It is assumed that a sub-pixel map has MRF properties, i. e., two
adjacent pixels are more likely to belong to the same class than different classes.
By employing this property of the model, the proposed MRF based algorithm
is able to correctly classify a large number of misclassified pixels, which often
appear as isolated pixels. In our experimental investigations, the efficacy of the
proposed algorithm has been tested on both multi and hyperspectral datasets
obtained from IKONOS and HyMap sensors respectively. The results show that
for both the datasets, a significant improvement in the accuracy of the subpixel map over that produced from the conventional and most widely used
maximum likelihood estimation algorithm can be achieved. The approach has
been able to successfully reduce a significant number of isolated pixels thereby
producing a more connected and smooth land cover map.
References
Aplin P, Atkinson PM (2001) Sub-pixel land cover mapping for per-field classification.
International Journal of Remote Sensing 22(14): 2853-2858
Atkinson PM (1997) Mapping sub-pixel boundaries from remotely sensed images. Innovation in GIS 4: 166-180
Bezdek JC, Ehrlich R, Full W (1984) FCM: the fuzzy c-means clustering algorithm. Computers and Geosciences 10: 191-203
Binaghi E, Brivio PA, Ghezzi P, Rampini A (1999) A fuzzy set-based accuracy assessment of
soft classification. Pattern Recognition Letters 20: 935-948
Bremaud P (1999) Markov chains Gibbs field, Monte Carlo simulation and queues. Springer
Verlag, New York
Brown M, Lewis HG, Gunn SR (2000) Linear spectral mixture models and support vector
machines remote sensing. IEEE Transactions on Geosciences and Remote Sensing 38:
2346-2360
Foody GM (1996) Approaches for the production and evaluation of fuzzy land cover classifications from remotely sensed data. International Journal of Remote Sensing 17:
1317-1340
Foody GM (1998) Sharpening fuzzy classification output to refine the representation of subpixelland cover distribution. International Journal of Remote Sensing 19(13): 2593-2599
Foody GM (2000a) Mapping land cover form remotely sensed data with a softened feed forward neural network. Journal ofIntelligent and Robotic System 29: 433-449
11: Teerasit Kasetkasem, Manoj K. Arora, Pramod K. Varshney
Table 11.4. Classification accuracy of sub-pixel maps
Initial SPM
Kappa
coefficient
0.4126
11.5
Summary
95% Confidence
Interval
(0.37280.4524)
Resulting SPM
Kappa
coefficient
0.5511
95% Confidence
Interval
(0.48940.6128)
In this chapter, a new sub-pixel mapping algorithm based on the MRF model
is proposed. It is assumed that a sub-pixel map has MRF properties, i. e., two
adjacent pixels are more likely to belong to the same class than different classes.
By employing this property of the model, the proposed MRF based algorithm
is able to correctly classify a large number of misclassified pixels, which often
appear as isolated pixels. In our experimental investigations, the efficacy of the
proposed algorithm has been tested on both multi and hyperspectral datasets
obtained from IKONOS and HyMap sensors respectively. The results show that
for both the datasets, a significant improvement in the accuracy of the subpixel map over that produced from the conventional and most widely used
maximum likelihood estimation algorithm can be achieved. The approach has
been able to successfully reduce a significant number of isolated pixels thereby
producing a more connected and smooth land cover map.
References
Aplin P, Atkinson PM (2001) Sub-pixel land cover mapping for per-field classification.
International Journal of Remote Sensing 22(14): 2853-2858
Atkinson PM (1997) Mapping sub-pixel boundaries from remotely sensed images. Innovation in GIS 4: 166-180
Bezdek JC, Ehrlich R, Full W (1984) FCM: the fuzzy c-means clustering algorithm. Computers and Geosciences 10: 191-203
Binaghi E, Brivio PA, Ghezzi P, Rampini A (1999) A fuzzy set-based accuracy assessment of
soft classification. Pattern Recognition Letters 20: 935-948
Bremaud P (1999) Markov chains Gibbs field, Monte Carlo simulation and queues. Springer
Verlag, New York
Brown M, Lewis HG, Gunn SR (2000) Linear spectral mixture models and support vector
machines remote sensing. IEEE Transactions on Geosciences and Remote Sensing 38:
2346-2360
Foody GM (1996) Approaches for the production and evaluation of fuzzy land cover classifications from remotely sensed data. International Journal of Remote Sensing 17:
1317-1340
Foody GM (1998) Sharpening fuzzy classification output to refine the representation of subpixelland cover distribution. International Journal of Remote Sensing 19(13): 2593-2599
Foody GM (2000a) Mapping land cover form remotely sensed data with a softened feed forward neural network. Journal ofIntelligent and Robotic System 29: 433-449