Terrace extraction based on remote sensing images and digital elevation model in the loess plateau, China

Abstract

Terraces are important objects of soil and water conservation and land use surveys in the Loess Plateau and considerably impact soil erosion and agricultural production. Obtaining the spatial distribution of terraces has great significance for soil and water conservation research and government policy-making. A new method of terrace extraction was proposed in this study based on remote sensing images and digital elevation model (DEM) used as experimental data. The extraction method combines Fourier transform and digital terrain analysis. First, by analyzing the spatial and frequency domain characteristics of the terraced image, we determined the analysis window to be 50 × 50, whereas the area having a direction maximum energy value greater than 0.25 and image standard deviation greater than 2000 was considered as the terraced candidate. Then, the positive terrain was automatically extracted from the DEM based on the regional growth method, and the terraces on the positive terrain were obtained. After a series of morphological image processing methods were conducted, the final terraces were extracted. The results showed that the accuracy of drawing and user accuracy of terrace extraction were 79.0% and 73.5%, respectively. Compared with the object-oriented method, it was found that the proposed method was more reliable and accurate. This method developed in this study has the characteristics of simple operation, strong universality, and high precision and can be applied to soil and water conservation monitoring and land use surveys in the Loess Plateau.