抄録
Multi-temporal analysis for texture features for the Thematic Mapper (TM) of Landsat-5 corresponding to ground cover classes has been conducted. The textural information content of TM data has been compared to that of MSS data. Furthermore, this information content has been enhanced by using seasonal change of the textural features for TM classification. Consequently, the effects of improvements of the spatial resolution and the quantization level on the textural features have been clarified. An example using multi-temporal TM data of Ottawa shows that a 32.9% improvement of the weighted mean percentage classification accuracy can be achieved by using multi-temporal TM spectral data compared to the uni-temporal case, and by adding the multi-tempomal texture features to the original spectral feather 1.6 to 4.7% improvements can be achieved.
