Reliable perception of orchard scenes is needed for agricultural automation. Fruit detection, counting, and localization support yield prediction, harvest planning, and precision fertilization, while 3D point clouds provide geometric information with less dependence on ambient illumination than conventional RGB images. Orchard analysis also requires predictions at more than one scale: individual fruits and trunks must be separated, but the fruits must still be associated with their parent tree. Most existing 3D panoptic segmentation methods operate at a single instance granularity and do not represent this hierarchy. We propose HAE-Net, a hierarchy-aware segmentation method built on a shared encoder and multiple decoders. At the encoder output, the Hierarchical Feature Enhancement (HFE) module supplies global aggregation for tree-level grouping, multi-scale context for semantic classification, and local boundary information for standard-instance separation. The Cross-Decoder Attention Gate (CDAG) regulates the encoder features transferred at each skip connection. Its spatial, channel, multi-scale, and pixel-level components allow the three decoders to use different mixtures of the shared representation. A Hierarchical Consistency Loss (HCL) links the two instance granularities by constraining the predicted centers of component instances and their parent tree instance. We evaluate HAE-Net on HOPS (Hierarchical Orchard Panoptic Segmentation), which includes point clouds acquired with terrestrial laser scanners, uncrewed aerial vehicles, uncrewed ground vehicles, and SfM reconstruction. Quantitative comparisons and ablations use the labeled TLS validation split. HAE-Net achieves an mIoU of 71.63% and an mPQ of 53.96%, exceeding HAPT3D by 9.81 and 5.43 percentage points, respectively. Because ground-truth labels for the four test subsets are not publicly accessible, their predictions are used only for qualitative inspection.
The automation and standardization of hyperspectral imaging, particularly at high spatial resolution, are essential for advancing plant phenomics and supporting diverse plant science research. We developed HyperBird, a hyperspectral microscopic imaging robot, to automate the acquisition of up to 351 leaf-disc samples in a single tray within approximately 2.4 h and thereby support scalable plant experiments. System calibration established a spatial resolution of 24.3 μm (full width at half maximum; FWHM), a spectral resolution of 1.78 nm (FWHM), and a depth of field of 4.53 mm, producing hyperspectral data cubes of 2195 × 2000 × 950 pixels across the 400–1000 nm spectral range. System-level and biological-sample consistency assessments confirmed stable spectral measurements during extended scanning sessions and across independently prepared trays. Thermal evaluation confirmed that the 150 W illumination design introduced minimal sample heating during scanning. HyperBird was first evaluated using grape downy and powdery mildews, where no consistent measurable effect of repeated imaging on pathosystem development was detected under the tested experimental conditions. Subsequently, HyperBird was used to characterize spatiotemporal spectral progression in grapevine leaves inoculated with Plasmopara viticola. An automated regional spectral tracing pipeline was developed to isolate spectra from retrospectively traced regions and compare them with whole-leaf averaged spectra across 0–9 days post-inoculation (DPI). Categorical mixed-effects modeling showed that these spatially resolved regional spectra exhibited significant disease-associated spectral change by 5 DPI, with a sharp transition at 6 DPI, whereas whole-leaf spectra showed delayed and weaker disease-aligned changes beginning at 7 DPI. These results demonstrate that HyperBird enables high-throughput, spatially resolved hyperspectral imaging for quantifying localized plant disease progression and provides a scalable platform for studying spectral biology across plant phenotyping applications.
In precision agriculture, assessing missing seedlings is essential for variable-rate reseeding and yield estimation. However, missing plants represent a semantic absence rather than explicit visual targets, making them difficult to define using conventional bounding-box annotations. To address this challenge, we propose a video-based, detection-driven gap inference framework for UAV wheat seedling phenotyping. Specifically, a semi-supervised Dual-Stream Anchor-Refinement Tracking (DART) mechanism is introduced to generate reliable pseudo-labels for visible seedlings from sparse temporal anchors, reducing dependence on dense manual annotations. To address visual fragmentation and weak target responses, a Masked Trajectory Inpainting Network (MTI-Net) combines object-aware masking with history-guided feature reconstruction to recover incomplete seedling representations using temporal context. The refined visible-seedling detections are then converted into agronomic gap indicators through a Dual-Scale Inertial Metric (DS-IM), which incorporates mechanical-sowing kinematic regularities and local growth-space constraints. Evaluations on the UAV-WheatSeedling dataset demonstrate positive mean gains when the proposed modules are embedded into RT-DETR and multiple YOLO-family backbones. When integrated with YOLO12-S, the method achieves 92.92±0.17% mAP@.50, 69.33±0.22% mAP@[.5:.95], and 86.59±0.41% Recall. Under 50% sparse-anchor supervision, the proposed framework approaches the 100% sparse-anchor YOLO12 baseline in mAP@[.5:.95], while achieving higher mean Recall. For downstream gap inference, DS-IM achieves an F1-score of 84.53% and a frame-level count MAE of 0.229 gaps/frame. This study provides a data-efficient and interpretable framework for low-altitude UAV-based early seedling assessment.
Plant counting functions as a central role in the quantification system of plant phenotyping. Over the years, the field has evolved through successive technological paradigms, and its modern form has been strongly influenced by advances in visual counting from computer vision. In this review, we synthesize the historical origins of plant counting, assess its current fragmented landscape, and outline a roadmap toward standardized, universal plant counting systems. We propose a coherent conceptual framework— the four-level hierarchy of plant counting, which characterizes the environment, platform, sensor, and counting entity in biological organization —grounded in the need of high-throughput plant phenotyping. This framework aims to guide the development of plant counting systems that are not only accurate on individual dataset, but also reusable, comparable, and trustworthy across modern plant phenotyping scenarios. We argue that existing plant counting approaches are constrained by species-specific designs, sensor-dependent assumptions, and local, region-bound assessments. We hope this review can serve as a reference for building cross-species, cross-modal, and cross-scale visual plant counting systems.
