Related Experiment Video
Updated: Jun 17, 2025
08:47
Author Spotlight: UAV Remote Sensing for Efficient Invasive Plant Biomass Estimation
Published on: February 9, 2024
1.4K
Object-oriented multi-scale segmentation and multi-feature fusion-based method for identifying typical fruit trees in
Jiaxi Liang1,2,3, Mamat Sawut4,5,6, Jintao Cui1,2,3
1College of Geography and Remote Sensing Sciences, Xinjiang University, Urumqi, 830046, Xinjiang, China.
Scientific Reports
|August 6, 2024
Summary
Accurate fruit tree identification using Sentinel satellite data and Random Forest machine learning is now possible. This method enhances orchard yield evaluation and planting area monitoring in regions like the Tarim Basin.
Area of Science:
- Agricultural Remote Sensing
- Machine Learning Applications in Ecology
- Geospatial Data Analysis
Background:
- Precise fruit tree identification is crucial for scientific yield evaluation and dynamic monitoring of orchard areas.
- Existing methods may lack the accuracy and scalability required for large-scale agricultural management.
- The Tarim Basin, a key fruit-growing region, presents a suitable study area for developing advanced identification techniques.
Purpose of the Study:
- To evaluate the applicability of time series Sentinel-1/2 satellite data for fruit tree classification.
- To develop and present a novel method for accurate extraction of fruit tree species.
- To compare the performance of the Random Forest (RF) model with other machine learning algorithms for this task.
Main Methods:
- Utilized time series Sentinel-1/2 satellite images from the Google Earth Engine (GEE) platform.
- Applied multi-scale object-oriented (OO) segmentation combined with Random Forest (RF) classification.
- Extracted and optimized 44 features, including spectral, phenological, texture, polarization, vegetation index, and red edge index, using the Vi feature importance index.
Main Results:
- Object-oriented segmentation improved the accuracy of fruit tree identification features.
- September satellite images provided the optimal time window for identification, with spectral, phenological, and texture features being most influential.
- The RF model achieved superior accuracy (Overall Accuracy: 94.60%, Kappa Coefficient: 93.74%) compared to SVM, GBDT, and CART algorithms.
Conclusions:
- The combination of object-oriented segmentation and the RF algorithm offers significant potential for accurate fruit tree identification and classification.
- This methodology is suitable for large-scale remote sensing classification of fruit trees.
- The study provides an effective technical approach for monitoring fruit tree planting areas using medium-to-high-resolution satellite imagery.
