A novel waveform-based pixel decomposition method for farmland shelterbelts width extraction using Sentinel−2 multi-spectral and UAV RGB images

As vital components of the Three-North Shelter Forest Program, farmland shelterbelts play a critical role in enhancing agricultural resilience. Among their structural attributes, shelterbelt width stands out as a key parameter that directly influences their ecological effectiveness. This study proposed a novel waveform-based pixel decomposition method (WBPD) for extracting the shelterbelt width. By integrating Sentinel−2 images with UAV-derived orthophoto maps, we extracted fractional coverage (FC) waveforms and enhanced width estimation through refined pixel decomposition. The main findings are as follows: (1) The dimidiate pixel model accurately estimated FC (R² = 0.764, RMSE = 0.15), with the highest accuracy observed during the protection maturity stage (R² = 0.767) and robust performance maintained in the pre-maturity stage (R² = 0.675). (2) The proposed WBPD method demonstrated strong agreement with ground-based width measurement (R² = 0.758, RMSE = 2.12 m, MAE = 1.78 m), and exhibited stable accuracy across varying shelterbelt orientations (MAE = 1.58 m). (3) Spatial analysis revealed that the average shelterbelt width in the study area was 28.68 m, with 91.76% of widths falling within the 20–40 m range. These results provide an effective solution for monitoring shelterbelt width and offer essential baseline data for assessing the protective functions of shelterbelts in agricultural landscapes.