Norsk Elektro Optikk AS (NEO) HySpex Airborne Sensors System Overview

Transcription

1 Norsk Elektro Optikk AS (NEO) HySpex Airborne Sensors System Overview Trond Løke Research Scientist EUFAR meeting

2 Outline Norsk Elektro Optikk AS (NEO) NEO company profile HySpex Optical Design Calibration and Characterization Airborne Systems System architecture Timing and generating events Software Conclusions

3 Norsk Elektro Optikk AS (NEO) General overview: 1985: Established as a privately-owned Electro Optics R&D company by researchers from the Norwegian Defence Research Establishment. Owned by a Foundation supproting arts and science, profit goes back to research including high-risk research projects. Present number of employees: ~45 of which 20+ in R&D Hyperspectral imaging at NEO brief overview: : First hyperspectral project: HISS (Hyperspectral Imager for Small Satellites) definition study performed by NEO for European Space Agency : Various R&D projects 2006: HySpex introduced to the market : HYPER-I-NET: EU project for interdisciplinary research focusing on hyperspectral imaging activities (funding ~10-15 Ph.D s, one at NEO) : SYSIPHE: Development of a high end airborne hyperspectral imaging system for VNIR+SWIR (NEO), MWIR+LWIR (ONERA), including real time processing system

4 Incoming light Entrance aperture HySpex Optical Design Focusing mirror Spectral 1000nm 400nm Spatial Slit spatial Collimating mirror Grating Optical layout of the HySpex sensor. Lens optics -Custom optics -Aspheric mirrors -Spheric and aspheric lenses -High optical quality, low smile and keystone Detector Illustration of spatial and spectral dimension on 2d sensor.

5 HySpex spectrometer calibration and characterization Instrument calibration: Spectral: -Wavelength as a function of sensor row number (band number) Radiometric/sensor: -Dark signal (automatic shutter) -Pixel responsivity, nonuniformity -DN to radiance -Bad pixels Geometric: -FOV pr pixel (sensor model for -georeferencing) and total FOV Instrument characterization: Radiometric/sensor: -Linearity -Noise, SNR, NER -Dynamic range -Stray light Spectral: -Spectral resolution, spectral misregistration Geometric: -Spatial resolution, -Spatial misregistration (keystone) Calibration data for image calibration. Essential! Documenting system performance. Nice to have! All hyperspectral instruments are delivered with: -Calibration data -Detailed test/calibration report

6 Spectral calibration Equipment for spectral calibration: Integrating sphere (hollow sphere with highly reflective and diffuse surface) Sufficient number of narrow band sources with precisely determined wavelength (gas excitation lamps, laser sources) Measurement method: Narrow band sources illuminate the interior of integrating sphere Hyperspectral camera points into the integrating sphere Result: Whole FOV of camera is filled with light from the narrow band source(s) Easy to calculate center of gravity of the sources for the whole FOV Smile plots and spectral calibration Spatial

7 Radiometric Calibration Measurement setup: Calibrated integrating sphere with known spectral radiance output L(i,j) in SI units W/(m^2 nm sr). The output of the exit port is a spatially uniform scene of known broadband radiance During the measurement, the camera is pointed into the integrating sphere. The result is that all spatial pixels are exposed to the same radiance and it is thus possible to associate the DN levels with radiance. Exit port

8 Radiometric Calibration From the expressions on the previous page, it can be shown that the radiance for each spatial pixel and spectral band can be expressed as: The ultimate goal of the radiometric calibration procedure is thus to generate the RE(i,j) matrix and the QE(i) vector for the particular sensor. The BG(i,j) matrix is dependent on the integration time and FPA temperature, so this should be acquired for each image. In the HySpex system, this is done automatically before and after every image acquisition by means of an automatic electromagnetic shutter. In principle, the RE matrix and QE vector could be incorporated into the same term, but in order to perform real time image correction, it is useful to separate the spatial and spectral calibration terms due to the high dynamic range of the QE vector.

9 Geometric Calibration Goal: To find the angular pixel size (pixel FOV) for each pixel across the whole sensor FOV and thus the viewing angle for each pixel with respect to the optical axis as well as the total across track FOV. This is called a sensor model. Motivation: The sensor model is an essential input to the georeferencing of the image, as you need to know the viewing angle of each pixel with subpixel accuracy in order to georeference the images with sub pixel accuracy. Note: The viewing angle should ultimately be defined with respect to the IMU frame of reference by boresight calibration on real airborne images with ground control points (not covered here)

10 Geometric Characterization Measurement: The sensor is mounted on a high precision rotation stage with high angular resolution and with the scan direction along the slit direction (opposite to the normal scan direction) The sensor should be viewing an object with good contrast (a sharp black/white transition). By scanning the rotation stage while acquiring image data, the black white transition will move from one edge of the FOV to the other (see figure). The rotation movement is paused and scanned at high resolution at exactly every 0.5º Calculation: From this image we can calculate the middle point in the black white transition for each angular position with sub pixel accuracy and by relating this to the precise angular movement, one can calculate the pixel FOV for each pixel with sub pixel accuracy.

11 Geometric Characterization

12 Test report All the calibration and characterization measurements performed on the camera are described and documented in a test report that is delivered with every camera: Calibration: Spectral Radiometric Bad pixels Characterization: Geometric /Sensormodel Spectral misregistration Spatial misregistration Spatial and Spectral resolution (for different parts of the FOV and for different wavelengths) SNR, dynamic range and linearity Stray light

13 HySpex Applications: Airborne, lab, field and industrial The same instrument can be used for airborne/outdoor scanning, industrial applications as well as short range lab or field measurements, making it a very versatile instrument. Field: In this forum, we focus on AIRBORNE. Airborne: Laboratory: Industrial:

14

15 HySpex Airborne Systems Data acquisition units( DAU) Rack mount DAU (SSDs and/or normal HDs). Compact DAU for smaller aircraft (SSD) Mini DAU for Ultra-Light aircraft (SSD) Ruggedized «lunchbox» computer for field work (SSD) All with Hot-Swap HDD bays for easy data transfer to processing computer. Standard system: Other optional components: Passive vibration damped mount Actively stabilized mount Various IMU/GPS solutions Touch screen interface Field of view expander Filters, etc Ultralight system:

16 HySpex Airborne: Electrical interfaces

17 Generating and timing Events Frequency divider: To divide down the freuency for synchronization between multiple HySpex cameras To divide down the frequency for event input on navigation system(s) The frequency divider is located inside the data acquisition unit Event input(gps/imu system): Diffrent navigation systems have different requiremnets of the input event. Timing accuracy is normally 50ns ACCURATE and RELIABLE TIMING NO LOST FRAMES EASY POST PROCESSING

18 HySpex products; IMU/GPS HySpex cameras can be integrated with a range of IMU/GPS unit from high end manufacturers such as Novatel, imar, IGI and Applanix (and many more): Manufacturer Novatel imar IGI Applanix Model SPAN-CPT itracert-f200 AEROcontrol POS AV 510 Conditions DGPS DGPS Post Processed DGPS Post Processed Horizontal position accuracy (RMS) 0.45 m ± 0.02 m 0.05 m m m Velocity accuracy (RMS) 0.02 m/s 0.01 m/s m/s 0.05 m/s m/s Attitude accuracy (RMS) Roll Pitch Azimuth Data processing rate 100 Hz 200 Hz Max 256 Hz 250 Hz Temperature (operational) -40 to +65 C -30 to +63 C -20 to +40 C -20 to +55 C Mass 2.36 kg 2.42 kg 3.2 kg 2.6 kg Choice of system depends on budget and accuracy requirements.

19 HySpex Airborne softwares HySpex AIR: Airborne data acquisition (integration time, preview, sessions, etc) HySpex RAD: Radiometric calibration of acquired data HySpex NAV: Timing and resampling of IMU/GPS data to each scan line

20 In development: Real time processing system NEO is working closely with the Norwegian Defense Research Establishment (FFI) on hyperspectral imaging for target detection FFI has been in the forefront on exploiting GPUs (Graphics Processing Units, e.g. nvidia) instead of CPU for real time hyperspectral data processing and georeferncing over several years Due to their parallel nature, hyperspectral image data are well suited for GPU processing Typically times faster processing on a GPU than one a standard Intel Quad-Core processor NEO and FFI are currently testing a general purpose real time processing system based on GPUs, for airborne applications. This will enable real time classification and georeferencing. Estimated to be ready for end users at the end of For airborne applications this software will have real time coverage plots, flightplanning and pilot guidance.

21 In development: Real time processing prospects from NEO Preliminary screen shot of real time software:

22 In development: HySpex ODIN-1024 Part of SYSIPHE project (cooperation with ONERA) 1024 spatial pixels for both VNIR and SWIR Common fore optics for VNIR and SWIR ->Excellent co-registration State of the art sensors Superb optical quality Spectral range Spectral resolution Pixel FOV nm 6 nm* 0.25mrad Total across track FOV 15 Spatial 2000m F-number 0.5 m F/1.5 VNIR F/2 SWIR

23 SNR for Odin

24 CONCLUSIONS HySpex Instrument: High spectral and spatial resolution, high speed, low stray light and very good optical quality in a rugged and compact package Instruments have been successfully applied and the performance demonstrated in a broad range of applications Open policy towards our customers in terms of transparency with regards to test and calibration procedures/data HySpex airborne systems have been selected by several prestigious remote sensing institutions Hyperspectral activities at NEO: Produce airborne hyperspectral camera systems (including IMU/GPS and real time processing) Work closely with research institutions, system integrators or end users to deliver customized hyperspectral imaging systems/solutions (military, remote sensing, industrial, medical, scientific) Participate in international R&D projects to stay in the forefront technologically and apply our technology in existing and emerging high end application areas.

25 Thank you for your attention!