Object-oriented multi-scale segmentation and multi-feature fusion-based method for identifying typical fruit trees in arid regions using Sentinel-1/2 satellite images DOI Creative Commons

Jiaxi Liang,

Mamat Sawut, Jintao Cui

и другие.

Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Авг. 6, 2024

Fruit tree identification that is quick and precise lays the groundwork for scientifically evaluating orchard yields dynamically monitoring planting areas. This study aims to evaluate applicability of time series Sentinel-1/2 satellite data fruit classification provide a new method accurately extracting species. Therefore, area selected Tarim Basin, most important fruit-growing region in northwest China. The main focus on identifying several major species this region. Time images acquired from Google Earth Engine (GEE) platform are used study. A multi-scale segmentation approach applied, six categories features including spectral, phenological, texture, polarization, vegetation index, red edge index constructed. total forth-four extracted optimized using Vi feature importance determine best phase. Based this, an object-oriented (OO) combined with Random Forest (RF) identify To find identification, results compared three other widely traditional machine learning algorithms: Support Vector Machine (SVM), Gradient Boosting Decision Tree (GBDT), Classification Regression (CART). show that: (1) helps improve accuracy features, September window texture contributing identification. (2) RF model has higher than models, overall (OA) kappa coefficient (KC) 94.60% 93.74% respectively, indicating combination algorithm great value potential classification. can be applied large-scale remote sensing provides effective technical means areas medium-to-high-resolution images.

Язык: Английский

Object-oriented multi-scale segmentation and multi-feature fusion-based method for identifying typical fruit trees in arid regions using Sentinel-1/2 satellite images DOI Creative Commons

Jiaxi Liang,

Mamat Sawut, Jintao Cui

и другие.

Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Авг. 6, 2024

Fruit tree identification that is quick and precise lays the groundwork for scientifically evaluating orchard yields dynamically monitoring planting areas. This study aims to evaluate applicability of time series Sentinel-1/2 satellite data fruit classification provide a new method accurately extracting species. Therefore, area selected Tarim Basin, most important fruit-growing region in northwest China. The main focus on identifying several major species this region. Time images acquired from Google Earth Engine (GEE) platform are used study. A multi-scale segmentation approach applied, six categories features including spectral, phenological, texture, polarization, vegetation index, red edge index constructed. total forth-four extracted optimized using Vi feature importance determine best phase. Based this, an object-oriented (OO) combined with Random Forest (RF) identify To find identification, results compared three other widely traditional machine learning algorithms: Support Vector Machine (SVM), Gradient Boosting Decision Tree (GBDT), Classification Regression (CART). show that: (1) helps improve accuracy features, September window texture contributing identification. (2) RF model has higher than models, overall (OA) kappa coefficient (KC) 94.60% 93.74% respectively, indicating combination algorithm great value potential classification. can be applied large-scale remote sensing provides effective technical means areas medium-to-high-resolution images.

Язык: Английский

Процитировано

4