| Due to the El Nio phenomenon and the changes of various climatic factors,the flood disasters in the south of Jiangxi Province have been frequently occurring,which has brought great economic losses to the south of Jiangxi Province.In order to reduce losses,more accurate water body monitoring is necessary.With the wide application and research of remote sensing technology in water body monitoring,two methods for improving the accuracy of water body extraction have emerged: one is based on image quality,and the other is based on the efficiency and precision of the extraction method.From the aspect of image quality,water monitoring by a single remote sensing image sometimes cannot meet people’s needs.For example,multi-spectral image data has multi-spectral information,but it is easy to receive interference from weather factors such as cloud,while radar images can make up for it,the speckle noise and shadow in the radar image reduce the accuracy of water extraction.From the aspect of extraction method,the threshold method is fast,but the threshold selection is difficult.Some scholars propose super pixel segmentation before threshold extraction to improve the threshold selection accuracy,but the results are not satisfactory.In this context,this paper conducts research on the above two methods,combine the high score one image with the high score third image to improve image quality,combining the improved SLIC algorithm with the threshold method improves the accuracy of threshold extraction,and achieves more accurate water information extraction.The main research work and results of the thesis as follows.(1)The high-scoring 1 and high-score 3 images of the study area are combined with common fusion methods for image fusion,including Brovey transform,Gram-Schmidt transform,HSV transform,PC Spetral sharpening and wavelet transform.The fusion results were qualitatively and quantitatively evaluated,and the fusion images most suitable for this paper were selected.From the fusion results analysis,it can be concluded that the HSV transformation method combines the high score one and the high score three images,and the fusion image phase Better spectral information and details than other fused images.(2)The fusion image is compared with the water extraction effect of the original image.Compared with the original image and the fusion image water body extraction results,the accuracy of the original image water extracted by NDWI method is 78.23%,the water body extraction error rate is 12.13%,and the FDWI method extracts the fusion image water body accuracy to 85.94%,the error rate is 9.41%.The results show that the fusion image water extraction effect is better than the original image.(3)An improved SLIC algorithm for introducing texture features is proposed.The algorithm is combined with the classical OTSU threshold method to study water body extraction and compared with the classical SLIC algorithm.From the perspective of water extraction area,the accuracy of water extraction by classical method is only 85.10%,and the accuracy of water extraction by improved method is 90.50%.It is concluded that the improved SLIC algorithm is better than the classical algorithm. |
