Incorporating Shadows In Impervious Surface Mapping And Analysis With High Spatial Resolution Remote Sensing Data

Impervious surface is one of the key indicator for urban ecology and development, and its landscape pattern is an effective measurement for urbanization. Therefore, impervious surface mapping and its landscape pattern evaluation show great importance in real applications. Remote sensing image plays an important role in impervious surface mapping, since it could provide observations fast and repeatedly over large areas. Medium and low spatial resolution remote sensing images are commonly used in impervious surface mapping at global and regional scales. Comparatively, high spatial resolution remote sensing data is an effective option for impervious surface mapping at local scale. Nevertheless, there are many shadows in high spatial resolution remote sensing images, which bring the challenge for impervious surface mapping and its landscape pattern analysis. In order to solve the problems above, we aimed at developing an object-based shadow detection method, and also extracting impervious surface in consideration of shadows problems with high spatial resolution remote sensing data. Furthermore, we discuss and evaluate the impact of shadows on landscape pattern for impervious surface. In detail, our main work and conclusions are as follows:1) Development of an object-based shadow detection method with high spatial resolution remote sensing image, and evaluation and optimization for segmentation parameters within object-based shadow detection method.Shadows commonly exist in high spatial resolution images, which bring challenges for image interpretation. Object-based approach is an effective method for high spatial resolution image interpretation, whereas its segmentation parameters directly affect the final detection accuracy. Therefore, we developed an object-based shadow detection method in fusion of multi-features. And then, segmentation parameters'impact on final shadow detection results was studied, followed by segmentation parameters optimization. Specifically, multi-resolution segmentation was conducted to acquire the initial objects. Then, shadows were detected based on the unit of object according to the criteria in fusion of multi-features with Dempster-Shafer theory. On the other hand, one-at-a-time screening method and orthogonal experimental design were both used to evaluate the impact of segmentation parameters on the final shadow results, which provide a key step in the parameter combination optimization. Experiments show that our object-based shadow detection method is effective, and interaction effect between segmentation parameters on the final shadow results cannot be overlooked in the optimization.2) Impervious surface mapping based on multi-temporal high spatial resolution remote sensing image and LiDAR dataCombining high spatial resolution image and LiDAR data is an effective way to solve the shadow problems in impervious surface mapping. However, it is hard to acquire the high spatial resolution image and LiDAR data produced at the same time in the real applications, since the LiDAR data is expensive and complex. Thus, we need to solve the observation errors between multi-temporal high spatial resolution remote sensing image and LiDAR data when they are fused for impervious surface mapping, including real landscape change, mis-registration caused by high buildings, impervious surface mappping covered by shadows, and missing information in LiDAR data. In summary, we proposed a strategy to map impervious surface combined with high spatial resolution remote sensing image and LiDAR data. Specifically, multivariate alteration detection was utilized to acquire the observation errors between the multi-temporal and multi-source data, followed by multi-source land-cover classification fusion in consideration of object information from independent segmentation. Finally, impervious surface was mapped with a high accuracy. Experiments prove that our strategy could conquer the observation error resulted from multi-temporal and multi-source data to map the impervious surface accurately. 3) Evaluation of shadows'effect on landscape pattern analysis for impervious surface. Landscape pattern for impervious surface could reflect urbanization process, and landscape metrics are the main method to analyze the landscape pattern. Impervious surface at local scale can be mapped accurately through high spatial resolution remote sensing image. However, the impact of shadow on landscape pattern for impervious surface is rarely involved in the previous research. Thus, we proposed an efficient method to evaluate the impact of shadow on landscape pattern for impervious surface in this thesis. Specifically, we began with the calculation of the difference of landscape metrics before and after impervious surface restoration under shadows. Then, the correlation between the shadow areas and the difference of landscape metrics was studied. Experiments proved that shadows have significant impact on landscape pattern analysis.