Research On Extraction Of Typical Land Cover Types With PolSAR

In recent years, Polsar system has been rapidly developed. It is meaningful to research theextraction technology of typical land cover types with Polsar data for the resource investigation,environment protection and disaster prevention in some places with the bad weather conditions.The thesis focuses on the extraction technology of water body with fully polarimetric sar data.The research contents mainly includes four parts：Polsar data preprocessing, the analysis forPolsar data feature of typical land cover types, the extraction of water body based on channelrelationships and the extraction of water body based on dynamic threshold SAR image segment.The main works are summarized as follows:(1) With the PALSAR data, the thesis elaborated Polsar data grades and productionprocesses; the decompress method of SIR-C MLC data was adopted and the Polsar images wereachieved; the causes of SAR image speckle noises were analyzed and several denoises methodswere introduced and realized.(2) By analyzing of the intensity features of full Polsar data, the thesis found that theintensity channels of T11, T22, T33are more available than that of HH, HV, VV ondiscriminating small area water bodies from bare fields, and the Cloude decomposition weretested and found that the water bodies with Bragg scattering types own lower value of scatteringentropy,whereas, the water bodies with mirror reflection have high scattering entropy, so Cloudeentropy is an important feature of discriminating different water types.(3) According to scattering types and scattering entropy, the thesis divided the water bodiesinto small area water body and large area water body. Two kinds of water bodies’s imageidentification features in T11, T22, T33channels are analyzed. For the small area water bodies,the thesis designed an intensification method of water information. By the method we caneffectively extract small water bodies and disable the influence of some land cover types.Experiment proves that the reinforce factorA1is more robust than others, and improvement ofraw SAR images’ light are more beneficial for discriminating small water bodies from bare fields.An interacted extraction method was designed based on balance light difference in water bodywith scattering entropy. Relative test proves the effectiveness of method.(4)For the large area water bodies with dark stripes, the thesis designed an auto-extractionmethod based on dynamic threshold image segment method. The extraction method includedynamic threshold segment and deleting regular land cover types with Hough transform theory.Relative test proves the effectiveness of the extraction method.