| Wetlands are zones of transition between terrestrial ecosystem and aquatic ecosystem,and they are also an ecosystem with many unique functions on the earth.The applications of advanced remote sensing technology not only can help human beings to perceive earth surface on a larger and wider scale,but also can help to provide accurate feedback on the wetland changes in a real-time manner.Such applications play a vital role during the process of ecological conservation and restoration.In this paper,the GF-1 remote sensing image is used as the data source.Based on the assessment of qualitative and quantitative indicators,the optimal segmentation scale is analyzed for the wetland information extraction in Changsha City.Three methods(including pixel-based maximum likelihood classification,support vector machine classification with object-based single scale and object-oriented hierarchical classification)are utilized to conduct the research of wetland area information extraction.By the combination of confusion matrix,the classification accuracy of the three methods is evaluated,and the spatial distribution of wetland information in the study area is analyzed,which provides a scientific basis for the management and protection of wetlands in Changsha City.The specific research results are as follows:(1)In terms of the optimal band combination,by the use of standard deviation,information entropy and optimal index of the image,the spectral feature information of different wetland types is analyzed based on GF-1 images.Results show that the image RGB channels of optimal band combination are near-infrared,red and green for wetland information extraction.(2)In terms of remote sensing image segmentation,based on the analysis of the texture,brightness,color and other features contained in the pixel,image segmentation was conducted by the combination of edge detection algorithm and Full Lambda-Schedule merging algorithm.Results show that the edge detection algorithm has a good recognition effect on the features at the junction of the waterway,and the segmentation effect can be further optimized by combining the merging algorithm.(3)In terms of the optimal segmentation scale,based on qualitative analysis and quantitative global scoring index analysis,the optimal segmentation scale suitable for various wetlands types are obtained.For river wetlands with strong integrity,the optimal segmentation scale is the largest with the scale of 70;for sporadic pits and ditches from artificial wetland,the optimal segmentation scale is the smallest with a scale of 30.(4)In terms of spectral feature analysis,by the analysis of spectral features(NDVI,CIWI,single band,etc.),texture features(variance,mean,etc.)and shape features(area,rectangular shape,etc.)for GF-1 images,the feature factors suitable for the extraction of various wetland types were selected.Results show that 11 quantitative indicators(such as NDVI,CIWI,bandl,band4,entropy,area,rectangular shape measurement,and extensibility)show stronger separation.(5)In terms of wetland information extraction,the object-oriented hierarchical classification wetland information extraction has the highest accuracy(overall accuracy and Kappa coefficient are 90.53%and 0.8754,respectively),followed by support vector machine classification with object-based single scale(overall accuracy 84.96%,Kappa coefficient 0.8115).The performance of pixel-based maximum likelihood classification is slightly worse.It indicates that for different extraction objects,setting a reasonable segmentation scale is conducive to the improvement of the overall accuracy.(6)In terms of spatial distribution statistics,according to the area statistics of various wetland types,the total wetland area of Changsha City is 24955.22 hm2(?0.3 hm2),which is in line with that of Changsha City in 2012(the statistical caliber is?8 hm2).In comparison,the area of wetlands above 8 hm2 decreased by 586.45 hm2,accounting for 3.62%of the total wetland area in 2012.If the reduction of wetlands below 8 hm2 is taken into account,the total reduction of wetland area will further increase,clearly showing that the wetland is constantly being occupied by the urbanization process.It shows that the urbanization process is the main factor affecting the wetland area of Changsha City. |
PDF Full Text Request