Phenomics & high-throughput phenotyping

01Overview and value chain#

Markers EC: NACE 26.51 (Measuring, testing & navigating equipment) | OECD: Plant breeding and genetics | Regulator: USDA-APHIS (USA), EFSA (EU), MARA (CN)

High-throughput phenotyping uses advanced imaging (multispectral, hyperspectral, LiDAR, thermal) and machine learning to rapidly and non-destructively quantify plant traits such as architecture, biomass, and stress tolerance. These systems bridge the gap between genotype and phenotype, enabling faster breeding cycles and optimized crop management. Modern platforms process 1,000–10,000 plants per day against capital costs of $50K–$500K for greenhouse gantry systems, shifting the bottleneck from data acquisition to AI-driven analysis.

The key directions of high-throughput phenotyping are:

1. Multispectral & Hyperspectral Imaging: NIR, SWIR, and thermal bands capture biochemical properties such as chlorophyll content and water stress invisible to RGB cameras.
2. 3D Reconstruction: LiDAR and stereovision reconstruct plant architecture and biomass, yielding precise volumetric and structural traits.
3. High-Throughput Automation: Greenhouse gantries, conveyor belts, drones, and under-canopy rovers scale capture to thousands of plants per day.
4. AI Trait Extraction: Convolutional neural networks segment plants from soil and extract complex traits such as leaf area index, fruit count, and disease severity.

Sectoral value chain#

[Sensor Hardware] ──> [Automation Platforms] ──> [Image Processing] ──> [Data & AI]
                              │
                      (Trait Extraction)
                              │
                              ▼
[Crop Yields] <─── [Breeding Programs] <─────┘

Value chain levels#

Cross-cutting technologies of the sector:

• Machine Vision: Multispectral and hyperspectral cameras capture biochemical and structural plant properties.
• Deep Learning: CNNs segment plants from backgrounds and extract complex phenomic traits at scale.
• Field Robotics: Drones and under-canopy rovers extend phenotyping from greenhouses into open fields.

02US#

The United States focuses heavily on field phenomics and automation, combining under-canopy robotics with drone-based analytics.

Field robotics, drone analytics, automation#

• EarthSense: Deploys the TerraSentia under-canopy robot for scalable, in-field phenotyping of row crops.
• Sentera: Leads in drone-based analytics, delivering high-resolution crop trait maps to breeders and growers.
• Automation focus: US players push rugged field deployment to bring greenhouse-grade phenotyping outdoors.

03CN#

China rapidly adopts automated greenhouse phenotyping to support national food-security goals, with institutes partnering closely with local integrators.

Greenhouse platforms, trait extraction, food security#

• Shanghai AgriPheno: Provides automated greenhouse phenotyping platforms and trait-extraction services tied to national breeding programs.
• Institute partnerships: Research institutes collaborate with integrators to scale phenomic pipelines for staple crops.
• Food-security drive: State backing accelerates adoption to speed the breeding of resilient, high-yield varieties.

04EU#

Europe is a pioneer in advanced plant phenotyping infrastructure, supplying high-precision 3D and multispectral systems for research and breeding.

3D multispectral scanning, digital phenotyping, fluorometry#

• Phenospex: Delivers 3D multispectral field scanning via its PlantEye sensor for crop phenotyping.
• LemnaTec: Builds digital phenotyping systems (PhenoCenter, Scanalyzer) for greenhouse and field deployment.
• Photon Systems Instruments & Hiphen: Advance PlantScreen fluorometry platforms and drone-based image analysis respectively.

05Leading companies and research institutes#

06Tech stack and innovations#

Modern phenotyping fuses non-destructive sensor technologies with advanced computation; the critical bottleneck has shifted from data acquisition to data analysis.

1. Multispectral & Hyperspectral Imaging:
   • NIR, SWIR, and thermal bands capture biochemical properties such as chlorophyll content and water stress that RGB cameras miss.
   • Enables early, non-destructive detection of biotic and abiotic stress across thousands of plants per day.
2. 3D Reconstruction:
   • LiDAR and stereovision reconstruct plant architecture and biomass into precise volumetric models.
   • Captures structural traits (height, leaf angle, canopy volume) essential for architecture-driven breeding.
3. Deep Learning Trait Extraction:
   • Convolutional neural networks segment plants from soil backgrounds and quantify complex traits such as leaf area index, fruit count, and disease severity.
   • Turns terabyte-scale field scans into reproducible, breeder-ready trait data.

07Value chains and production pipelines#

Industrial pipeline of high-throughput plant phenotyping (ISO 9001 / breeding data standards)#

┌───────────────────────────┐      ┌───────────────────────────┐
│ 1. Sensor Imaging         │ ───> │ 2. Multispectral & 3D     │
└───────────────────────────┘      └───────────────────────────┘
                                                 │
                                                 ▼
┌───────────────────────────┐      ┌───────────────────────────┐
│ 4. AI Trait Extraction    │ <─── │ 3. High-Throughput Capture│
└───────────────────────────┘      └───────────────────────────┘
              │
              ▼
┌───────────────────────────┐      ┌───────────────────────────┐
│ 5. Predictive Analytics   │ ───> │ 6. Breeding Deployment    │
└───────────────────────────┘      └───────────────────────────┘

Stage 1: Sensor imaging

Standard RGB and thermal imaging capture simple morphological traits, establishing the baseline data layer for a phenomic pipeline.

Stage 2: Multispectral & 3D integration

NIR and SWIR bands plus LiDAR/stereo vision add biochemical indicators and precise volumetric architecture to the trait stack.

Stage 3: High-throughput capture

Greenhouse gantries, conveyor belts, drones, and under-canopy rovers scale acquisition to 1,000–10,000 plants per day, with CapEx ranging from $50K to $500K.

Stage 4: AI trait extraction

CNNs segment plants from soil backgrounds and extract complex traits (leaf area index, fruit count, disease severity) from gigabyte- to terabyte-scale scans.

Stage 5: Predictive analytics

Phenomic data is integrated with genomic and environmental models to predict performance, addressing the core bottleneck of data interoperability and reproducibility.

Stage 6: Breeding deployment

Predictive insights feed elite-seed development and field management, increasingly paired with digital twins that simulate plant performance under diverse climate scenarios to accelerate climate-resilient crops.