In order to reduce the influence of optical data imaging quality in the process of crop extraction,Senti-nel-1 and Sentinel-2 data were used based on Google Earth Engine platform to identify wheat and rape respective-ly in four phenological periods:overwintering period,regreening period,booting period and maturity period.A feature set is constructed by using a random forest method to preferentially select the optimal features from a set of 34 features consisting of spectral features,vegetation index features,red edge index features,texture features and polarisation features.It also compared the recognition results of four classifiers,namely,minimum distance,decision tree,support vector machine and random forest,in the four phenological periods to determine the opti-mal classifier;and also verified the influence of polarization features on the recognition results in the four pheno-logical periods.The results showed that the optimal classifier was random forest in all four phenological periods;the phenological periods with the highest to lowest recognition accuracy were wheat booting,maturity,regreen-ing and overwintering,with OA of 92.91％,91.93％,90.24％and 87.69％,and Kappa coefficients of 91.00％,89.92％,87.61％and 84.53％;The inclusion of polarisation features in each of the four phenological periods im-proved the accuracy of identification,more so in the two phenological periods of wheat,greening and maturity.
维普
万方数据