In this study,the impact of the training sample selection method on the performance of fuzzy-based Possibilistic c-means(PCM)and Noise Clustering(NC)classifiers were examined and mapped the cumin and fennel rabi crop....In this study,the impact of the training sample selection method on the performance of fuzzy-based Possibilistic c-means(PCM)and Noise Clustering(NC)classifiers were examined and mapped the cumin and fennel rabi crop.Two training sample selection approaches that have been investigated in this study are“mean”and“individual sample as mean”.Both training sample techniques were applied to the PCM and NC classifiers to classify the two indices approach.Both approaches have been studied to decrease spectral information in temporal data processing.The Modified Soil Adjusted Vegetation Index 2(MSAVI-2)and Class-Based Sensor Independent Modified Soil Adjusted Vegetation Index-2(CBSI-MSAVI-2)have been considered to minimize soil background effects,enhancing vegetation detection accuracy,particularly in areas with sparse vegetation cover.The MMD(MeanMembership Difference)and RMSE(RootMean Square Error)approaches were used to measure the study’s accuracy.To illustrate that the classifier successfully describes classes,cluster validity(SSE)was also performed,and the variance parameter was computed to handle heterogeneity within cumin and fennel crop fields.For the calculation of RMSE,Sentinel-2 data was used as classified,whereas PlanetScope satellite data was utilized as the reference data set.The best result was obtained using the NC classifier with“individual sample as mean”using CBSI-MSAVI-2 temporal indices.For Fuzziness Factor(m)=1.1,the RMSE,MMD,Variance,and SSE values for the NC classifier using“individual sample as mean”on the CBSI-MSAVI-2 temporal indices for cumin were 0.00098,0.00162,0.02857,and 0.97143,respectively and for fennel were 0.00025,0.00248,0.10420,and 3.54286,respectively.展开更多
文摘In this study,the impact of the training sample selection method on the performance of fuzzy-based Possibilistic c-means(PCM)and Noise Clustering(NC)classifiers were examined and mapped the cumin and fennel rabi crop.Two training sample selection approaches that have been investigated in this study are“mean”and“individual sample as mean”.Both training sample techniques were applied to the PCM and NC classifiers to classify the two indices approach.Both approaches have been studied to decrease spectral information in temporal data processing.The Modified Soil Adjusted Vegetation Index 2(MSAVI-2)and Class-Based Sensor Independent Modified Soil Adjusted Vegetation Index-2(CBSI-MSAVI-2)have been considered to minimize soil background effects,enhancing vegetation detection accuracy,particularly in areas with sparse vegetation cover.The MMD(MeanMembership Difference)and RMSE(RootMean Square Error)approaches were used to measure the study’s accuracy.To illustrate that the classifier successfully describes classes,cluster validity(SSE)was also performed,and the variance parameter was computed to handle heterogeneity within cumin and fennel crop fields.For the calculation of RMSE,Sentinel-2 data was used as classified,whereas PlanetScope satellite data was utilized as the reference data set.The best result was obtained using the NC classifier with“individual sample as mean”using CBSI-MSAVI-2 temporal indices.For Fuzziness Factor(m)=1.1,the RMSE,MMD,Variance,and SSE values for the NC classifier using“individual sample as mean”on the CBSI-MSAVI-2 temporal indices for cumin were 0.00098,0.00162,0.02857,and 0.97143,respectively and for fennel were 0.00025,0.00248,0.10420,and 3.54286,respectively.
