Estimation of Class Mixture Proportion in Mixel Considering Component Class Number by Using Fuzzy Reasoning

Abstract

This paper presents a new fuzzy estimation method for the class mixture proportion considering the number of component class in the mixed pixel (mixel) on a remote sensing image. The method consists of two steps. First step is estimation of the number of the landcover classes in the mixel. Second step is estimation of the class mixture proportion based on the above class number. The distribution of the training data was defined as a fuzzy set on a spectral space. It was assumed that the spectral characteristics of the mixel was regarded as a linear function of the reflection level of the pure pixels corresponding to the component class. Both the number of component class and the class mixture proportion were estimated by the fuzzy simplification reasoning method in the proposed method.
The simulation results indicated the reasonable distinction as the number of component class. That is, the threshold value for two, three and four classes were obtained as 0.90, 0.97 and 0.97, respectively. It was observed that the proposed method gave low TRMS (the Total Root Mean Square error) to the simulation data that was produced by the random sampling data of each training class. Therefore, it was confirmed that the proposed method was an effective and practical technique to estimate the class mixture proportion in the mixel.