基于二维小波分析的景观特征尺度识别

采用3种常用小波基（Haar、Daubechies和Symlet），探讨了二维小波分析在沈阳地区城市、城乡交错区和农村3种景观类型特征尺度识别中的有效性.由于二维小波分析的变换尺度必须为2的整数次幂，这会导致景观中一些特征尺度无法精确识别，故本文将各景观类型影像像元大小分别重采样为3、3.5、4和4.5 m，以增加分析时的尺度密度.结果表明：应用二维小波分析可清晰地识别景观特征尺度；Haar、Daubechies、Symle可分别作为城乡交错区景观、城市景观和农村景观特征尺度识别的最优小波基；Haar和Symlet均可应用于农村景观精细特征尺度的识别和城乡交错区景观边界的检测；Daubechies和Symlet可分别用于城市景观和农村景观的边界检测.

Recognition of landscape characteristic scale based on two-dimension wavelet analysis

Three wavelet bases, i.e., Haar, Daubechies, and Symlet, were chosen to analyze the validity of two-dimension wavelet analysis in recognizing the characteristic scales of the urban, peri-urban, and rural landscapes of Shenyang. Owing to the transform scale of two-dimension wavelet must be the integer powerof 2, some characteristic scales cannot be accurately recognized. Therefore, the pixel resolution of images was resampled to 3, 3.5, 4, and 4.5 m to densify the scale in analysis. It was shown that two-dimension wavelet analysis worked effectively in checking characteristic scale. Haar, Daubechies, and Symle were the optimal wavelet bases to the peri-urban landscape, urban landscape, and rurallandscape, respectively. Both Haar basis and Symlet basis played good roles in recognizing the fine characteristic scale of rural landscape and in detecting the boundary of peri-urban landscape. Daubechies basis and Symlet basis could be also used to detect the boundary of urban landscape and rural landscape, respectively.