Our Technology

The Brown Platform for Autonomous Remote Sensing (B-PAR) is a suite of sensors carried by a helicopter drone.

Aircraft

AeroscoutThe aircraft is an Aeroscout B-330 helicopter. It is gasoline powered and can carry a 50 kg payload for 3 hours at sea level. We operate at flight altitudes from 100 – 500 m above ground at ranges up to 5 km beyond the ground-based pilot and ground-control station (outside the United States). This helicopter conducts autonomous flight under the control of a computer-based flight-control system that tightly maintains the motion and vibration environment required to operate remote sensing instruments.

Imaging spectroscopy and LiDAR

LaSelvaRGBThe B-PAR suite of sensors includes three lightweight imaging spectrometers designed by Headwall Photonics that jointly provide coverage of the 400 – 2500 nm region of the solar spectrum at varying spectral resolution. The visible and near-infrared (VNIR) sensor measures reflected sunlight in the 400 – 1000 nm region of the electromagnetic spectrum at a spectral sampling interval of 1.6 nm. The short-wave infrared (SWIR) instrument uses a Stirling-cooled HgCdTe detector that provides coverage of the 900 – 2500 nm region at a spectral sampling interval of 10.8 nm. The VNIR and SWIR instruments are aligned so that their full fields of view are closely matched. Spatial resolution of image data depends on flight altitude, but can be < 5 cm for the VNIR instrument and < 15 cm for the SWIR instrument at typical flight altitudes. These sensors provide information about species identity, leaf and flower pigments, water, and the carbon compounds cellulose and lignin.

The Chlorophyll fluorescence spectrometer provides the high-fidelity and very narrow spectral resolution required to observe solar-induced chlorophyll fluorescence. This instrument uses a cooled Si detector that is sensitive to the 670 – 780 nm region at a spectral sampling interval of 0.05 nm. First flights with the Chlorophyll fluorescence imaging spectrometer will occur in summer, 2017.

Measurements of canopy height are made using a Riegl VUX-1 LiDAR sensor. The VUX-1 is lightweight (3.5 kg) and can produce point densities exceeding 1,000 measurements per m2.

We also operate with a 24.3 megapixel digital camera that collects images with very high along-track and across-track overlap within and between flight lines. We process these images using algorithms from computer vision to produce geometrically correct image mosaics and high-density point clouds like the one above.

GPS and inertial navigation system

Accurate placement of sensor data in the real world depends on accurate and frequent measurements of the position and attitude of the aircraft during sensor operation. We use an Oxford Survey+2 coupled with a dual-antenna and Novatel GPS base station. Post-processed GPS accuracy is < 0.02 m.

Computing

We maintain a dedicated file server and high-performance computer to securely store and process large volumes of geospatial data. The file server is a 36-bay ZFS server supported by two Intel quad-core 2.4 GHz processors and 128 GB of DDR3 RAM. The system is configured into two partitions, one of which includes 24 × 4 Tb disks in a RAIDZ3 configuration that we use for data storage. The remaining 1 × 4 Tb disk is a hot spare. The second partition is a scratch space in a RAIDZ2 configuration and consists of 8 × 2 Tb disks. The RAIDZ3 and RAIDZ2 partitions have 3 and 2 disks dedicated to parity, such that multiple simultaneous disk failures will not result in data loss. Each unique file system on the server is snapshotted daily, and we retain daily snapshots for 30 days, which allows us to recover a data set that has been corrupted, accidentally deleted, or accidentally overwritten within 30 days. The entire digital file system is backed up monthly using a robotic tape backup system running the Bacula software package on an Ubuntu 14.04 server. Tapes are stored in a separate, climate-controlled building on the campus of Brown University.

Top