Research And Application Of Spectrum Energy Characteristics Of Typical Landforms

Earth's surface is an important place for human production and living.Quantitative analysis and scientific expression of geomorphology has important theoretical and practical value.Traditional terrain research method is mainly filed survey combined with indoor manual drawing,which is often affected by the natural environment,technology level,human,etc.The emergence of DEM provides a new way of thinking for geoscientific research.Current digital terrain analysis based on DEM is being concentrated on terrain characteristic elements extraction,geographical model analysis,etc.Terrain parameters and their statistical characteristics(such as the mean and variance)are often used to quantitatively analyze terrain characteristics.From the perspective of frequency domain,this paper quantitatively studies the spectrum energy characteristics of the typical landforms using Fourier transform spectral analysis method,based on DEM data of Fujian Province.And then,the spectrum power characteristics of the typical landforms are used in the process of Fujian province landform classification.The main concents and achievements are summarized as follows:(1)Typical landforms Fourier power spectrum characteristicsAfter Fourier transform,the power spectrums of DEM data represent the terrain energy at each frequency position.Two-dimensional power spectrum indicates that the energy of terrain are mainly concentrated in the low frequency region,it is the skeleton information of terrain.The relationship between power spectrum of of typical landforms and radial frequency power is linear in double logarithmic coordinate,which means spectrum power decreases with the increase of radial frequency.(2)Energy variation characteristics of typical landformsFor further quantify the energy variation characteristics of typical landforms,this paper puts forward the terrain energy change rate index.The results showed that:Overall,the energy change rate of plains,hills,mountainous in turn increases,and the differences of energy change rate among three landforms are quite clear.Different landforms have obvious characteristics in the region of the medium frequency.High frequency part is mainly affected by terrain discontinuity and data error,which makes the spectrum energy violently swing.(3)Differences of typical landforms energy distributionBy comparing the spectra with the statistical properties of spectra derived from randomly generated landform,this paper find that terrain is not randomly surface,but has characteristic spatial scales.Then the characteristics of terrain are extracted by using filter technology and Fourier inverse transformation.(4)Landform classification of Fujian ProvinceSpectrum energy features of typical landforms are refined into frequency domain parameters.The landform classification of Fujian Province is done using remote sensing image supervised classification method based on frequency domain parameters and spatial parameters.Experiments show that the overall accuracy of classification can achieve 78.64%,Kappa coefficient is 0.72,prove the terrain spectrum energy characteristics can be used to guide landform classification.