Real-Time Forest Inventory from a $2,200 Sensor Payload
Replacing $100,000 LiDAR rigs with an automated mobile perception cart that maps 100% of trees under dense canopy with zero GPS.
Traditional forest inventory relies on slow manual calipers or expensive laser scanners. Funded by the USDA-NIFA STTR Program, CUTMAP pairs a solid-state LiDAR and stereo cameras with edge AI on an NVIDIA Jetson—measuring tree diameters with sub-centimeter accuracy in seconds. Developed by Spectral Sciences, Inc. in partnership with Dr. Bogdan Strimbu at the Oregon State University College of Forestry.
Interactive 3D Understory Scans
Drag to spin the 3D trees. Use the top buttons to switch between stands or toggle the elevation height colormap.
The Numbers at a Glance
How CUTMAP Works in the Field
Traverse Without GPS
A forester, UTV, or robotic rover rolls the low-cost sensor cart through the stand. Real-time LiDAR-inertial SLAM maps the terrain with zero satellite dependency.
Automatic Breast-Height Slicing
Onboard algorithms calculate ground elevation under the cart, isolating a thin LiDAR slice at exact 1.37 m breast height while ignoring low brush and ground clutter.
Instant Digital Inventory
In under 5 seconds per stand, cylinders lock onto individual trees, outputting exact diameters, stem coordinates, and GIS-ready shapefiles for immediate analysis.
From the Bench to Harvard Forest
A 56-second overview of the hardware development, laboratory calibration, and zero-GPS understory mapping in Harvard Forest.
Real-Time Edge Perception on Jetson AGX Orin
Direct screen capture from the sensor cart showing 17 tree cylinders fitting in 3.9 seconds as the cart traverses a native pine stand.
System Architecture & Processing Pipeline
From multi-sensor odometry to sub-pixel SAM bark boundary segmentation and geometry-constrained 3D RANSAC cylinder fitting.
Algorithmic Pipeline & System Architecture
A 3-stage computational pipeline converting raw spatial photon returns into operational silvicultural harvest prescriptions.
Additive Point Cloud Creation
Fuses prior aerial laser scanning (ALS) with continuous terrestrial mobile LiDAR-inertial odometry and stereo depth.
Input Data Stream
Computational Engine
FAST-LIO2 SLAM
Iterated error-state Kalman filter tight-coupled to IMU, mapping dense understory geometry without GPS ephemeris.
Multi-View Disparity Engine
Hardware-synchronized disparity matching isolating foreground vegetation and sub-pixel trunk boundaries.
Aerial-to-Terrestrial Coregistration
Progressive multi-resolution alignment registering terrestrial stem manifolds to top-of-canopy crowns.
Empirical Deliverables
Four Ways to Team With Us
Spectral Sciences, Inc. is actively structuring Phase II teaming agreements for joint field pilots and commercial licensing.
Timberland Managers & TIMOs
Replace slow, expensive 5% manual cruise samples with continuous 100% stem census data (<$5/acre) ready for ArcGIS, Remsoft, and carbon registries.
Machinery & Autonomous OEMs
Integrate a compact, low-power (<40W) perception payload into harvester cabs, forwarders, tracked rovers, or quadruped scouts.
Federal Co-Funding (Phase II STTR)
Partner on upcoming USDA-NIFA Phase II STTR grants to co-develop customized sensor packaging and regional timberland validation at low financial risk.
Commercial Licensing & Exclusivity
Secure preferential commercial licensing terms and territory rights for Spectral Sciences' proprietary multi-sensor calibration and forest AI stack.
Discuss a Phase II Field Pilot
Connect directly with Principal Investigator Dr. Cory Glenn Garms.
Peer-Reviewed Publications
Calibrating a Low-SWaP-C Lidar–Stereo Platform: Staged Procedures, Failure Coupling, and Operational Validation
Demonstrates how multi-sensor calibration integrity achieves sub-millimeter precision from low-cost solid-state perception.
CUTMAP: A Low-SWaP-C Mobile Mapping Platform for Forest Mensuration: Calibration Integrity, Sensor Fusion, and Ground-Truth Prototyping
Field validation against 65 logger's-tape stems across native forest stands in Burlington, MA.
Impact of stem lean on estimation of Douglas-fir diameter and volume using mobile lidar scans
Formulates non-destructive 3D cylinder fitting to eliminate systemic bias in merchantable volume calculations.
Read paper →