Frontiers in Ecology and Evolution (Mar 2023)

A new rice identification algorithm under complex terrain combining multi-characteristic parameters and homogeneous objects based on time series dual-polarization synthetic aperture radar

  • Hao Ma,
  • Lihua Wang,
  • Lihua Wang,
  • Lihua Wang,
  • Weiwei Sun,
  • Songling Yang,
  • Yanghua Gao,
  • Li Fan,
  • Gang Yang,
  • Yumiao Wang

DOI
https://doi.org/10.3389/fevo.2023.1093454
Journal volume & issue
Vol. 11

Abstract

Read online

Accurate mapping of rice-growing areas is essential to ascertain the spatial distribution of rice fields, and ensure food security. It is a challenging task to timely and accurate identify rice under the complex terrain due to its diversified land cover, small- or middle-sized rice fields with fragmented distribution. In this paper, the time series VV and VH backscatter coefficient datasets were first constructed based on 411 sentinel-1 synthetic aperture radar (SAR) images in Chongqing city with complex terrain. Then, the rice multi-characteristic parameters, including SAR backscatter features, composite features, rice phenological parameters, texture features and topographic features, were generated. On this basis, the homogeneous image objects were produced. Furthermore, a rice identification algorithm combining multi-characteristic parameters and homogeneous objects based on time series dual-polarization SAR (MPHO-DPSAR) was established. The research demonstrated that the MPHO-DPSAR algorithm can achieve accurate mapping of small and medium-sized and fragmented rice fields in regions under complex terrain according to the accuracy evaluation at three levels and the comparison with other three classical rice identification methods. The suitability and limitations of proposed MPHO-DPSAR algorithm were also discussed from the aspects of SAR data temporal and spatial resolution, rice phenology, and surface landscape complexity.

Keywords