A Contour-Directional Detection for Deriving Terrace Ridge From Open Source Images and Digital Elevation Models

Wen Dai,Guanghui Hu,Peng Zhang,Xin Yang,Guoan Tang,Nan Huang

doi:10.1109/access.2019.2940437

Wen Dai, Guanghui Hu + Show 4 more

Open Access

https://doi.org/10.1109/access.2019.2940437

Copy DOI

Abstract

Agricultural terraces, as typically artificial landforms, have maintained an important role throughout human history. Terrace ridges, which form the main constructive part of terraces, face the greatest risk from gully or gravity erosion. Mapping terrace ridges can benefit agricultural production management and soil-and-water conservation. This study presents a contour-directional detection method that automatically maps terrace ridges by considering the image features and topographic features of terraces. First, the terrace areas were extracted via object-based image analysis. Second, the Canny edge detector was employed to extract all edges from the image, and a digital elevation model (DEM) was used to generate the contour direction. The contour direction and edge image were then overlaid to implement the contour-directional detection method, which detects ridges along the contour direction. Finally, those candidate terrace ridges which lengths exceed an appropriate threshold were identified as terrace ridges. Five study areas located in the Loess Plateau of China were used to validate the proposed method. The results showed that the correctness, completeness, and quality ranged from 82.57% to 94.92%, from 82.51% to 88.39%, and from 70.28% to 84.40%, respectively. However, the proposed method is widely available and can be easily applied by using open global images and DEMs, such as Google Earth images and ASTER GDEM.

Highlights

Agricultural terraces, as typically artificial landforms, have maintained an important role throughout human history
Those edges that are distributed along the contour direction can be regarded as terrace ridges
A length threshold was used to filter out the false positives

Summary

Introduction

Agricultural terraces, as typically artificial landforms, have maintained an important role throughout human history. Given that transforming sloping fields into terraced fields improves agricultural production and facilitates soil-andwater conservation [1]–[5], terraces are being widely built all over the world. Field investigation [12], [13] and visual interpretation [14], [15] are traditional methods for terrace mapping given their simplicity, directness, and accuracy. These methods are time consuming, costly, and unsuitable for the long-term monitoring of large areas, thereby motivating researchers to pursue automatic methods for terrace mapping

Objectives

Methods

Results

Conclusion