Research On Spatial Data Multi-Scale Modeling

After half a century of development, geographic information system(GIS) has been widely used in many fields, including: map navigation, urban planning, public health, defense, resource monitoring, logistics, et al. In all these GIS applications, spatial data is the basis of the service. And now spatial data has become massive data with the increasing of spatial information acquisition platforms. On the orther hand, mobile GIS has gradually developed into the main application. There is a huge contradiction between the massive data and the limited bandwidth or limited computing resources. Multiscale modeling technology is an effective way to resolve this contradiction. In all existing multiscale modeling methods, wavelet analysis has obvious advantages for its inherent multi-resolution features. But there are still many problems in the existing wavelet-based multi-scale modeling methods for spatial vector data, which impedes the construction of a unified multiscale modeling framework in practical application. Therefore, this paper focuses on the wavelet-based multi-scale modeling methods and related technologies for spatial data, to form a unified and practical wavelet-based multi-scale modeling framework.The main work of this paper is summarized as follows:To solve the problem that the local error can not be inconveniently controlled in existing wavelet analysis methods for line vector data, a line vector data wavelet analysis method with local error correction is presented. The high band coefficients are used to find these regions where local errors are too large. The local interpolations are performed on these regions by specific designed interpolation function to reduce the corresponding areas’ local error in low band representation. The original wavelet coefficients can be recovered from the correction coefficients by designed formula and the accurate reconstruction can be achieved. The resulting data can be easily organized and the amount of data is increased less in this algorithm. And the experimental results show that the large local errors in low band representation are effectively reduced.There are two main problems in existing Loop subdivision wavelet algorithms, which are existing large error in low band representation and having high complexity to get filter coefficients. To solve these problems, a smooth Loop subdivision wavelet algorithm is put forward. At first, a common subdivision wavelet lift structure is constructed. Then, the specific form of smooth Loop subdivision wavelet is determined by analyzing correction point. At last, the wavelet filter coefficients are solved with the criteria of reverse subdivision results having minimal change when disturbance is added. This algorithm can obtain the analytic filter coefficients which are related to vertex valence. And the experimental results show that the errors of low band representation are significantly reduced.To solve the low efficiency problem in existing Loop subdivision wavelet transform implementation methods, a fast Loop subdivision wavelet transform method based on neighboring position computation(FLSWTMNPC) is raised. The method consists four parts: the first part is a mesh segmentation method which adapts to subdivision and reverse subdivision; the second part is organizing all vertices on mesh patch by a unified ring form; the third part is the positions calculation the adjacent points; the fourth is a border stitching method to ensure the correctness of vertices on patch boundary. The FLSWTMNPC method can support subdivision and reverse subdivision calculation. The data organization is refined and the calculation efficiency is improved in this method.The GPU accelerated wavelet transform methods for spatial mesh data and raster data is studied. The existing GPU accelerated methods can only support subdivision computation and lots of work needs to proceed on CPU. Aimed at this problem, the FLSWTMNPC’s GPU accelerated algorithm is proposed. The subdivision wavelet transform can be totally implemented on GPU and the computation efficiency is improved. In the GPU accelerated wavelet transform methods for raster data, the block’s boundary wavelet coefficients is eliminated by boundary stitching and a multiphase form GPU accelerated wavelet decomposition method is proposed, which improve the efficiency of data organization and computation.