| Along with the development of remote sensing technology and the further research of forestry production,the use of remote sensing in the forest has been widely applied.It covers the forest resources dynamic monitoring,the forest biological physical parameter inversion,the forest ecosystem carbon cycle simulation and so on.High spatial resolution image contains abundant spatial information,which reflects the difference of surface features.It also has a clear geometric structure,which can be distinguished by the feature of shape and size,and the texture is clear.It can reflect the complex type of vegetation and forest conditions in the forest resource management.Object-based image analysis is widely used in the classification of high resolution remote sensing image.But this method is constrained only to qualitative analysis and rarely used in quantitative research.We take Anji County as the study area.The object-based multi-scale structure is used to classify land-use types of Anji County,using SPOT6 imagery,and the distribution information of Moso bamboo were extracted;A multi-scale carbon storage estimation model is constructed by combining the data of sample and the features of multi-scale objects.Finally,carbon storage of bamboo forest in the county were estimated..It is expected that the method we proposed in the study can break through the single scale(pixel scale)carbon reserves estimation method.It mainly contains the following aspects:1.Object-based image segmentation of SPOT6 imagery in Anji County by using polygon boundary and a multi-hierarchy structure system is constructed.According to the characteristics of features,the classification is carried out at different scales and combine the different classification results of different scales by using the context relationship of objects.The classification map of land-use types of Anji County using object-based method is compared with the maximum likelihood method based on pixel.2.An advantage of object-based approach is the hierarchical structure between objects.Through the scale size of image segmentation,we can get not only the information of small scale objects but also the information of large scale objects.Combined with GIS spatial analysis tool,quantitative information transferred from large scale to small scale can be realized.In this paper,the hierarchical structure of the object method is used to transfer the large scale remote sensing information to the small scale objects.Combining the sample information,a regression model of multi-scale carbon storage estimation based on object is constructed.Based on the classification results of Moso bamboo,carbon storage of Moso bamboo forests in Anji County was estimated.Through this present study,conclusions can be drawn as follows:1.Object-based image analysis approach can be used to extract object information at different scales,with the total accuracy of 83.87%,and the producer’s accuracy of Moso bamboo forest of 88.89%,which is higher than the maximum likelihood classification.2.In the process of mapping of bamboo forest,Features including the average of the NIRed,the contrast texture of the blue band,the standard deviation texture of the blue band and the mean value of the NDVI vegetation index played the most important part.3.Based on the multi-scale remote sensing characteristics of the irregular sample,the object-based carbon storage model of the bamboo forest is constructed,which break through the traditional single scale(based on pixel scale)method to estimate the carbon storage and resulted the better estimation results.4.Combined with the sample information,a regression model of multi-scale carbon storage estimation based on object is constructed.The prediction accuracy is higher than the regression model constructed by using the single scale features and get the carbon storage with better accuracy in bamboo forest based on object metrics. |
PDF Full Text Request