A highly efficient temporal-spatial probability synthesized model from multi-temporal remote sensing for paddy rice identification

This article develops a temporal-spatial probability synthesized model (TSPSM), in which a metric describing the characteristic of temporal and spatial information is defined to map paddy rice distribution. The purpose is to reduce the effect of cloud contamination on classification. The error matrix and Kappa were used as accuracy measurement. Results showed that TSPSM obtained higher accuracy with significant difference from error matrices of the other two conventional methods, post comparison classification with post-classification comparison and majority voting. Moreover, smaller window was suitable for the area with higher fragmentation, while the larger was suitable for the area with lower fragmentation. It was concluded that TSPSM could help to improve the potentials of temporal optical image to map crops.