Positioning radiata pine branches requiring pruning by drone stereo vision

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

cs.CVApr 12, 2026

Positioning radiata pine branches requiring pruning by drone stereo vision

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

Organizations: Centre of Data Science and Artificial Intelligence & School of Engineering and Computer Science, Victoria University of Wellington, PO Box 600, Wellington, New Zealand · Computer Vision Research Lab & Department of Computer Science and Software Engineering, Canterbury University, PO Box 8041, Christchurch, New Zealand

Abstract

This paper presents a stereo-vision-based system mounted on a drone for detecting and localising radiata pine branches to support autonomous pruning. The proposed pipeline comprises two stages: branch segmentation and depth estimation. For segmentation, YOLOv8, YOLOv9, and Mask R-CNN variants are compared on a custom dataset of 71 stereo image pairs captured with a ZED Mini camera. For depth estimation, both a traditional method (SGBM with WLS filtering) and deep-learning-based methods (PSMNet, ACVNet, GWCNet, MobileStereoNet, RAFT-Stereo, and NeRF-Supervised Deep Stereo) are evaluated. A centroid-based triangulation algorithm with MAD outlier rejection is proposed to compute branch distance from the segmentation mask and disparity map. Qualitative evaluation at distances of 1-2 m indicates that the deep learning-based disparity maps produce more coherent depth estimates than SGBM, demonstrating the feasibility of low-cost stereo vision for automated branch positioning in forestry.

Figures & tables

Figure 1: The drone, equipped with a ZED Mini camera for stereo vision and a robotic arm, autonomously detects and prunes branches of radiata pine. The ZED Mini camera enables the drone to accurately identify the branches, while the robotic arm precisely prunes them.

Figure 2: Triangulation using Two Cameras to Obtain the Depth Map. The point $(u_{l},v_{l})$ represents the projection of point $p(x,y,z)$ in three-dimensional space onto the image plane of the left camera, whereas point $(u_{r},v_{r})$ corresponds to the projection of the same point onto the right camera’s image plane. The variable $b$ denotes the baseline distance separating the left and right cameras.

Figure 3: Triangulation using Two Cameras to Obtain the Depth Map. $f$ represents the camera’s focal length. With the values of $f$ , $b$ (the baseline distance between the cameras), and the disparity of point $p(x,y,z)$ between the left and right camera images, we can calculate the distance $z$ from the pixel representing point $p$ in space to the camera.

model	COCO datasets		Branches
	mAP box50–95	mAP mask50–95	mAP box50–95	mAP mask50–95
Mask R-CNN R50-C4	39.8	34.4	76.86	0.06
Mask R-CNN R50-DC5	40.0	35.9	77.54	9.16
Mask R-CNN R50-FPN	41.0	37.2	79.19	6.75
Mask R-CNN R101-C4	42.6	36.7	88.05	0.05
Mask R-CNN R101-DC5	41.9	37.3	79.12	9.94

Table 1: Performance Comparison of Mask R-CNN with Different Backbones and Various YOLO Segmentation Models. For Mask R-CNN, the FPN backbone, using ResNet with conv and FC heads, offers the best speed-accuracy tradeoff. The C4 backbone, based on ResNet conv4 with a conv5 head, follows the original Faster R-CNN design, while the DC5 variant, with dilations in ResNet conv5, is inspired by Deformable ConvNets. The mAPbox and mAPmask metrics represent the model’s average accuracy in detection and segmentation tasks across IoU thresholds from 50% to 95%.

Figure 4: Block Matching illustration. $L$ denotes the template window and $T$ the search scan line. The left camera image $E_{l}$ serves as the reference, and the corresponding pixel is located in $E_{r}$ .

Panel of Figure 5 — Figure 5: SGBM disparity map generation pipeline: (a, b) original left and right images, (c, d) pre-processed images, (e) SGBM-generated disparity map, and (f) WLS-refined disparity map.

Panel of Figure 6 — Figure 6: Comparison of depth maps generated by MiDaS and Depth Anything models at branch distances of 1m, 1.5m, and 2m from the camera

Panel of Figure 7 — Figure 7: Comparison of PSMNet Fine-tuning Results Across Different Pretrained Models and Training Epochs

Panel of Figure 8 — Figure 8: Comparison of Stereo Matching Results Across Different Neural Network Models

Panel of Figure 9 — Figure 9: Illustration of the centroid-based branch localisation process: (a) predicted points surrounding the branch, (b) grouping the closest points into triangles, and (c) computing their centroids.

Panel of Figure 10 — Figure 10: Branch depth detection results: (a) detected points overlaid on the depth map using the centroid method, and (b) the corresponding points displayed on the RGB image.

Panel of Figure 11 — Figure 11: Comparison of Depth Estimation Using SGBM and NeRF at Varying Distances

Panel of Figure 12 — Figure 12: Comparison of Depth Estimation Using SGBM and NeRF at Varying Distances

Figure 13: Training process of the depth map model

Stereo Vision

Stereo vision uses images from two or more viewpoints to estimate depth, recover 3D structure, and support spatial perception.

See all 43 papers

Low-Cost Stereo Vision for Robust 3D Positioning of Thin Radiata Pine Branches in Autonomous Drone Pruning
May 6, 2026Yida Lin, Bing Xue, Mengjie Zhang +2Aerial Robotics Stereo Matching
Centre of Data Science and Artificial Intelligence & School of Engineering and Computer Science, Victoria University of Wellington, PO Box 600, Wellington, New Zealand · Computer Vision Research Lab & Department of Computer Science and Software Engineering, Canterbury University, PO Box 8041, Christchurch, New Zealand
Training Deep Stereo Matching Networks on Tree Branch Imagery: A Benchmark Study for Real-Time UAV Forestry Applications
Feb 23, 2026Yida Lin, Bing Xue, Mengjie Zhang +2Stereo Matching Stereo Depth Estimation
Centre for Data Science and Artificial Intelligence, Victoria University of Wellington, Wellington, New Zealand · Department of Computer Science and Software Engineering, University of Canterbury, Canterbury, New Zealand
Progressive Per-Branch Depth Optimization for DEFOM-Stereo and SAM3 Joint Analysis in UAV Forestry Applications
Feb 24, 2026Yida Lin, Bing Xue, Mengjie Zhang +23D Reconstruction Depth Estimation
Centre for Data Science and Artificial Intelligence Victoria University of Wellington, Wellington, New Zealand · Department of Computer Science and Software Engineering University of Canterbury, Canterbury, New Zealand

May 6, 2026J/K move · Enter open · S save

Low-Cost Stereo Vision for Robust 3D Positioning of Thin Radiata Pine Branches in Autonomous Drone Pruning

Yida Lin, Bing Xue, Mengjie Zhang, Sam Schofield+1

Centre of Data Science and Artificial Intelligence & School of Engineering and Computer Science, Victoria University of Wellington, PO Box 600, Wellington, New Zealand · Computer Vision Research Lab & Department of Computer Science and Software Engineering, Canterbury University, PO Box 8041, Christchurch, New Zealand

Aerial Robotics Stereo Matching Robotics Semantic Segmentation

Open paper AskPDF arXiv

Manual pruning of radiata pine, a species of major economic importance to New Zealand forestry, is hazardous, labour-intensive, and increasingly constrained by workforce shortages. Existing autonomous pruning platforms typically rely on expensive sensors such as LiDAR and are limited to thick branches, which restricts their wider adoption. This paper investigates whether a single low-cost stereo camera mounted on a drone can provide sufficiently accurate branch detection and three-dimensional positioning to support autonomous pruning of branches as thin as 10 mm, thereby removing the need for auxiliary depth sensors. The proposed pipeline comprises two stages: branch segmentation and depth estimation. For segmentation, Mask R-CNN variants and the YOLOv8 and YOLOv9 families are compared on a custom dataset of 71 stereo image pairs captured with a ZED Mini camera; YOLOv8 and YOLOv9 are selected as representative state-of-the-art real-time segmentors at the time of data collection, and the framework is designed to remain compatible with newer YOLO releases. For depth estimation, a traditional method (SGBM with WLS filtering) and deep-learning-based methods (PSMNet, ACVNet, GWCNet, MobileStereoNet, RAFT-Stereo, and NeRF-Supervised Deep Stereo) are evaluated, including cross-dataset fine-tuning experiments that expose the domain gap between urban driving benchmarks and natural forestry scenes. The main novelty of this work lies in coupling stereo segmentation with a centroid-based triangulation algorithm and Median-Absolute-Deviation outlier rejection that converts a segmentation mask and disparity map into a single robust branch-to-camera distance, addressing the challenges of sparse texture, thin structures, and noisy disparity values typical of forest scenes. Qualitative evaluations at distances of 1-2 m show that the learning-based stereo methods produce more coherent depth es...

Abstract

Stereo Vision

Explore similar work

Low-Cost Stereo Vision for Robust 3D Positioning of Thin Radiata Pine Branches in Autonomous Drone Pruning

Training Deep Stereo Matching Networks on Tree Branch Imagery: A Benchmark Study for Real-Time UAV Forestry Applications

Progressive Per-Branch Depth Optimization for DEFOM-Stereo and SAM3 Joint Analysis in UAV Forestry Applications