Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

YOLO and SGBM Integration for Autonomous Tree Branch Detection and Depth Estimation in Radiata Pine Pruning Applications

Yida Lin, Bing Xue, Mengjie Zhang, Sam Schofield, Richard Green

TL;DR

This work addresses safe, cost-effective autonomous pruning of radiata pine by integrating YOLO-based branch detection with SGBM stereo depth estimation, eliminating the need for LiDAR. It demonstrates superior branch segmentation performance over Mask R-CNN, achieving high mAP scores, and validates sub-second processing with reliable 3D localization within a 2 m range. The study presents a complete pipeline from data collection to 3D localization, including data augmentation, depth refinement via WLS, and robust spatial registration. The results show the approach is feasible for real-time autonomous pruning on consumer-grade hardware, offering safety enhancements and economic viability for forestry operations. Overall, the framework advances autonomous forestry by delivering accurate, low-cost perception and depth estimation suitable for branch-aware navigation and manipulation.

Abstract

Manual pruning of radiata pine trees poses significant safety risks due to extreme working heights and challenging terrain. This paper presents a computer vision framework that integrates YOLO object detection with Semi-Global Block Matching (SGBM) stereo vision for autonomous drone-based pruning operations. Our system achieves precise branch detection and depth estimation using only stereo camera input, eliminating the need for expensive LiDAR sensors. Experimental evaluation demonstrates YOLO's superior performance over Mask R-CNN, achieving 82.0% mAPmask50-95 for branch segmentation. The integrated system accurately localizes branches within a 2 m operational range, with processing times under one second per frame. These results establish the feasibility of cost-effective autonomous pruning systems that enhance worker safety and operational efficiency in commercial forestry.

YOLO and SGBM Integration for Autonomous Tree Branch Detection and Depth Estimation in Radiata Pine Pruning Applications

TL;DR

This work addresses safe, cost-effective autonomous pruning of radiata pine by integrating YOLO-based branch detection with SGBM stereo depth estimation, eliminating the need for LiDAR. It demonstrates superior branch segmentation performance over Mask R-CNN, achieving high mAP scores, and validates sub-second processing with reliable 3D localization within a 2 m range. The study presents a complete pipeline from data collection to 3D localization, including data augmentation, depth refinement via WLS, and robust spatial registration. The results show the approach is feasible for real-time autonomous pruning on consumer-grade hardware, offering safety enhancements and economic viability for forestry operations. Overall, the framework advances autonomous forestry by delivering accurate, low-cost perception and depth estimation suitable for branch-aware navigation and manipulation.

Abstract

Manual pruning of radiata pine trees poses significant safety risks due to extreme working heights and challenging terrain. This paper presents a computer vision framework that integrates YOLO object detection with Semi-Global Block Matching (SGBM) stereo vision for autonomous drone-based pruning operations. Our system achieves precise branch detection and depth estimation using only stereo camera input, eliminating the need for expensive LiDAR sensors. Experimental evaluation demonstrates YOLO's superior performance over Mask R-CNN, achieving 82.0% mAPmask50-95 for branch segmentation. The integrated system accurately localizes branches within a 2 m operational range, with processing times under one second per frame. These results establish the feasibility of cost-effective autonomous pruning systems that enhance worker safety and operational efficiency in commercial forestry.

Paper Structure

This paper contains 25 sections, 3 equations, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Object Detection and Segmentation
  4. Stereo Vision and Depth Estimation
  5. Computer Vision in Forestry
  6. Methodology
  7. Data Acquisition and Preprocessing
  8. Branch Detection and Segmentation
  9. Stereo Depth Estimation with SGBM
  10. Spatial Integration and 3D Localization
  11. Performance Evaluation
  12. Experimental Results and Analysis
  13. Branch Detection Performance
  14. Stereo Depth Estimation Results
  15. Integrated System Performance
  16. ...and 10 more sections

Figures (4)

  • Figure 1: System architecture integrating YOLO-based branch segmentation with SGBM stereo matching for accurate 3D branch localization.
  • Figure 2: Spatial integration process: (a) Branch regions detected by YOLO segmentation, (b) Corresponding SGBM depth map enabling precise 3D branch localization.
  • Figure 3: Comprehensive depth estimation pipeline demonstrating systematic improvement through processing stages: (a,b) Original stereo pair with natural lighting variations, (c,d) Preprocessed images with enhanced contrast and noise reduction, (e) Raw SGBM disparity output showing initial depth estimates, (f) WLS-filtered disparity map with preserved edge structure and reduced noise artifacts.
  • Figure 4: Integrated system performance analysis across operational ranges: (a--c) Depth maps at 1 m, 1.5 m, and 2 m distances showing consistent branch detection and depth estimation, (d--f) Corresponding depth measurement distributions demonstrating precision characteristics and statistical reliability for autonomous pruning applications.