Extracting area information of paddy rice based on stratified multiple endmember spectral mixture analysis

To resolve the serious pixel un-mixing problem produced by coarse spatial resolutions sensors, and improve the extraction accuracy of plant area for paddy rice, the stratified multiple endmember spectral mixture analysis (SMESMA) method was proposed in this paper. The complexity of landscape will be mitigated using stratified classification method, and the number and types of endmembers are allowed to vary in a per-pixel basis by multiple endmember spectral mixture method, which can overcome the spectral variations within classes. The accuracy of classification was improved significantly by combining these two methods. In this study, the HJ-1B CCD image was stratified into three stratifications. A landscape will be removed from the image after extracted, and the next classification will run based on the new stratified image. Multiple endmember spectral mixture analysis was applied to map the stratification images, and the optimized endmembers was determined by EAR、MASA and CoB methods. The results showed that that SMESMA had better classification accuracy of 85.78% and kappa coefficient of 0.85 than that of 79.1% and 0.78 by per-pixel based maximum likelihood classifier (MLC), which indicated that SMESMA was a useful classifier and method for paddy cultivation area extracting with coarse spatial resolution image.