Research On DEM Resolution Determination And Sacling Conversion Methods

As the basic national geographic information data, Digital Elevation Model(DEM) has been the fundemental data of National Spatial Data Infrastructure(NSDI), with its important applications in national economy and national defense construction. Under the situation that DEM of different scales, different resolutions and different precision coexist at the same time, the issue about DEM scale become an urgent problem to solve, which not only provide the basic guarantee for its integration with geoscience model but also is the key of its promotion and application. Therefore, the calculation and conversion of DEM resolution have become important research propositions for DEM producers and departments in use.This paper starts from the raw scatter data to explore the calculating method of DEM resolution in order to meet the requirements of terrain expressing. When the calculated resolution does not match with practical applications, it’s necessary to convert the DEM scale. To solve this problem, this paper is based on the mathematical theory of fractal and wavelet, taking the terrain expression of DEM with terrain analysis into consideration, then conducts a systematic research on DEM resolution calculation and scale conversion. The main findings are as follows:(1) The calculation model of DEM resolution based on fractal theory is proposed. DEM contains the expression of terrain, which should reveal maximum amount of information about terrain. This requires that research should start from raw data to explore the caculating method of DEM resolution. This paper establishes relationships between DEM resolution and fractal dimension through using fractal theory to find quantitative expressioin of terrain’s self-similarity and characteristics of terrain’s complexity. It is able to determine the horizontal resolution by seeking the maxmium inflection point which describles terrain information, according to the difference of line’s slope.(2) The scaling-up algorithm for DEM based on multi-band wavelet decomposition is studied. Combined the actual demand of the DEM in practice, basic principles of scaling-up alogrithm for DEM have been established. Inhence a scaling-up algorithm for DEM based on random number have been proposed considering the precision factor of scaling-up. In addition, another alogirthm to scale-up DEM based on multi-band wavelet have been provided by using the multiresolution of multi-band wavelet decomposition, namely use the low frequency part of DEM as the low resolution of DEM through decomposition. This paper also makes a contrast between the proposed method and the common resampling methods.(3) An algorithm for scaling down to DEM based on multi-band wavelet and interpolation is proposed. Get the high-frequency part through wavelet decomposition and bilinear interpolation with original data, regard these as the low-frequency part, and bulid the scaling-down DEM through inverse wavelet transform.This paper also make an objective and subjective evaluation of the experimental results.(4) A sacling-down algorithm for DEM based on multi-band wavelet and filtering is proposed. Taking the directions of multi-band wavelet into account, put the filtering effect of direction on DEM data to build the high-frequency part of DEM, combine this with DEM data obtained by multi-band wavelet decomposition to reconstruct the scaling-down DEM data. The comparison of this algorithm with previous one is conducted and analyzed.(5) On the basis of the analysis of learning-based ultrahigh resolution image reconstruction algorithm, this paper proposes a DEM scaling-down algorithm concerning neighborhood reconstruction, which is based on the nonlocal similarity of DEM constraints. Based on similarity and compatitbility of sub-region, the DEM data of high-resolution section has been gained. In DEM data of low resolution section, find the area of non-local similar sub-region being used to reconstruct the ultrahigh resolution data and map the data of high resolution section into the corresponding area according to the level of similarity, thus accomplish the purpose of scaling down to DEM. Also the comparison with previous algorithms include the one based on interpolation and the one based on multi-band wavelet reconstruction is discussed and analyzed.