Cross-regional sample generation based on Cropland Data Layer for large-scale winter wheat mapping: A case study of Huang-Huai-Hai Plain, China
The limited availability of training samples presents a major challenge in using remote sensing images for large-scale winter wheat mapping. Currently, several countries have published long-term land-cover products that offer dependable accuracy and comprehensive spatial distribution information of...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-08-01
|
| Series: | International Journal of Applied Earth Observations and Geoinformation |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S156984322500411X |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The limited availability of training samples presents a major challenge in using remote sensing images for large-scale winter wheat mapping. Currently, several countries have published long-term land-cover products that offer dependable accuracy and comprehensive spatial distribution information of crops. Given that crops within similar latitudinal ranges exhibit analogous temporal growth patterns, we propose a Cross-Regional Sample Generation (CrossRS) method utilizing the Cropland Data Layer and time-series images. A three-step strategy is used to generate pure winter wheat samples. The first step involves applying the Random Forest algorithm to transfer the mapping rules from the source area to the target area while determining the probability similarity of each pixel in the target area to winter wheat in the source area. During this process, a winter wheat pixel pool is initially created by defining a probability interval. The second step involves extracting phenological information of crops during critical growth periods based on field samples gathered from the target area. This information is then used to further exclude other winter crop pixels from the winter wheat pixel pool. The third step involves applying a spatial filter to the winter wheat pixels to eliminate easily confused pixels near boundaries. In this study, we choose the Huang-Huai-Hai Plain in China as the target area and Kansas in the United States as the source area. In comparison, the winter wheat map obtained using the CrossRS method significantly outperforms existing products in terms of accuracy and spatial detail. |
|---|---|
| ISSN: | 1569-8432 |