研究目的
To test three operationally available speckle filters at multiple window sizes on data from three regions with differing average field sizes, to determine if there was any improvement to overall classification accuracy with changing filter type and/or changing window size.
研究成果
The adaptive Touzi filter resulted in the highest overall classification accuracies for small and large field sizes, though differences between the highest accuracies obtained by all three filter types were not statistically significant. There is an indication that classification accuracies improve for small and medium average field sizes with changing window size. Continued testing and assessment of the algorithms' global availability and computational processing requirements will help determine the best practices for speckle filter selection in SAR data set development.
研究不足
The field training and validation data were limited to three general crop types (maize, soybean, and wheat). The study also noted that the overall crop classification accuracies drop significantly when multiple crop types are considered, and the numerical difference between data sets reduces.
