High Efficiency Hyperspectral Imager

1
High Efficiency Hyperspectral Imager

• Sponsored by Florida Space Institute
• Group 4

Andrew Mayer David Desrochers Chae Ku O David
Johnson
2
Project Overview

• Acquiring images of earth and storing them in
  volatile memory
• Researching and purchasing all necessary
  components (camera, main board, etc.)
• Building software for controlling acquisition
• Automating the acquisition with one or two
  switches
• Power distribution to the system, including the
  on/off switch for the astronauts
• Build the software, which is capable of analyzing
  the data returned by HEHSI using the fast Fourier
  transform

3
Project Description

• Fly in space using NASAs small payloads program
• Prototype utilizing a new type of interferometer
• Hyperspectral imaging taking images in numerous
  spectra at once
• Includes electrical, mechanical, and optical
  engineering

4
Project Motivation

• Innovative method of hyperspectral imaging
• Efficiency of using progressive scan technology
  over line scan technology
• Low cost solution to current methods using COTS
  components
• Potential project uses
• Vegetation fuel identification
• Target detection
• Soldier identification system

5
Project Objective

• Design hardware and software for system to
  automate image acquisition with minimal human
  intervention
• System must survive conditions of space and
  return with unharmed data
• Design electrical components necessary for DC
  operation of system
• Design image processing software
• Achieve this with minimal budget

6
Project Specifications

• Space environment extreme conditions
• Stress vibration 3.5 gs lateral and axial
• Temperature 100C, depending on shuttle
  orientation and orbit altitude
• Image capture speed 120 frames per second
• Hard drive space as large as possible as budget
  permits, as many scenes as possible
• RAM space enough to hold one scene of data,
  256 MB

7
Block Diagram of System
8
Computer System Hardware
9
Computer Hardware

• Design tradeoffs
• Size, cost, performance, physical limitations
• Two available industrial interfaces
• CompactPCI
• PC/104
• Two fast digital image transfer interfaces
• RS-422
• RS-644 (LVDS)

• CCD vs. CMOS sensors
• Hardware decisions
• Mobile Pentium III
• Disk on chip
• DMA
• Solid-state hard drive
• Storage space
• Function of velocity and altitude of orbiter
• Worst case (i.e. max. altitude, avg. velocity)

10
Storage Space
11
Hardware Overview
12
Hardware Overview

• Framegrabber
• NI PXI-1422
• 16-bit digital image capture _at_ 120 fps
• RS-644 (LVDS)
• 80-100MB/s transfer speed
• Camera
• Pulnix TM-6710
• 648 x 484 resolution
• 120 fps operation
• 8-bit greyscale CCD
• RS-644 (LVDS)
• Backplane
• Kaparel PS1361, 2 slot passive
• 64-bit CompactPCI

• CPU Board
• PEP CP302, Mobile PIII 400MHz, 384MB RAM
• 16MB Disk on Chip
• Integrated video
• USB, keyboard/mouse
• UDMA/66 IDE controller
• -40C - 85C operating temperature
• Hard Drive
• UDMA/33 or UDMA/66
• Minimum of 300MB, but will acquire largest that
  budget permits
• Solid-state (non-moving)

13
Hardware Diagram
Backplane
Framegrabber
Camera
System dimensions 100mm x 160mm x 30mm
CPU board
Camera dimensions 46mm x 140mm x 30mm
Solid State Hard Drive
Camera picture taken from Pulnix TM-6710
document, others from PICMG.org
14
Electrical System
15
Power System

• 48 D Cell Batteries as power source
• Batteries have 1.5V and 2 Amp per hour power
  rating
• DC/AC power supplies is used for calibration
  source
• DC/DC power supplies is used for rest of the
  system
• Use Photocell as relay

16
Power Source
17

• Batteries will be divided into two groups.
• These two groups will be divided into three
  parallel strings.
• The power source and power supplies will be
  connected to P-400 JT06RE-16-6-S power
  connector.
• Connector points A and B will be connected to
  power source.
• Connector points C and D will be connected to
  power supplies.
• Connector points E and F will be used as ground.

18
DC Power Supply

• One input and four outputs
• Input 12 V
• Output 12V, -12V, 5V, and 5V
• Flyback topology
• Benefits of using Flyback
• Cost and Space
• Can be used for DC input voltage low as 5V
• Good for multi-output

19
DC Power Supply Schematic
20
Flyback Power Supply

• Good for multi-output supply.
• Transistor On, energy is supplied from output cap
  to load, transformer stores energy.
• Transistor Off, energy is supplied from
  transformer, and cap stores energy.
• Transistor is controlled by error signal.
• Error signal is obtained by comparing fraction of
  output voltage to reference voltage.

21
AC Power Supply

• Used for Calibration lamp
• Converts 12VDC to 210VAC
• When lamp is connected to ground, out2 is 10mA
• When lamp is connected to ground, out1 is 5mA

• Schematic is courtesy of UVP

22
Relay System

• NASA provides only three relays for GAS can.
• One of the relays is reserved for removing all
  power from the payload in case of power failure.
• Astronaut turns on the system while GAS can is
  closed.
• Calibration lamp turns on and computer performs
  calibration.
• Relay is used to shut off lamp when there is
  light, and turn on the camera and computer.
• After one scene, system is reset.

23
FlightSoftware
24
Dataflow Diagram
Use Take images and store on disk
25
Requirements

• Must be developed to have minimal human interface
• Must control power relays
• Software size cannot exceed 16 MB
• Complete software package will store numerous
  scenes on non-volatile memory
• Scenes for calibration
• Scenes for analyzing

26
Contents of Software

• Embedded Windows NT OS
• Batch files to automate system
• Compression executables
• Camera drivers
• NI-IMAQ API for camera control and image
  acquisition
• All necessary dll files for custom software
• dlls for Borland C, API, drivers

27
Cycle
Cycle is controlled by a batch file
28
Image Acquisition

• NI-IMAQ API
• Functions used for acquiring the images
• Developed through Borland C Builder
• Will take a total of 988 frames at a time
• 988 frames / 120 Hz 8.23 sec
• Complete scenes will be dumped into volatile
  memory as bitmaps.
• Ensuring no frame loss

29
Compression

• JPEG Possible lossy choice
• Lossless compression choices
• Huffman O(n)
• Almost as good as LZW, and fairly fast
• Hard to implement
• Lempel-Ziv-Welch (LZW) algorithm O(n)
• Best compression
• Either slow or have very high memory requirements
  depending on implementation
• Run-Length O(n)
• Very fast
• Easy to implement
• Worst compression

30
Compression of Choice

• Compression to be used is Huffman
• Assigns the most frequently used character the
  least amount of bits, second most second least,
  etc.
• Larger the file, better the compression ratio
• To better ratio, data from one byte to the next
  will be subtracted
• Ex 32 2F ED B3 B3 B3 12 12 12 12
• 32 3 -BE 3A 0 0 A1 0 0 0
• Hence, higher frequency of a character
• This gives approximately 25 better compression
  ratio
• Compression ratio is approximately ¼ original
  size for bitmaps

31
Storage

• Compressed in volatile memory and slowly stored
  on hard drive
• Assuming 75 compression ratio, with the fixed
  size of 295 MB/scene and fixed throughput of 1024
  KB/s for HD,
• (295MB/scene)(1-.75) /(1024KB/s)73.75 sec
• Should give 2 or 3 minutes for compression
  total of 3-4 minutes to store
• Still better than (295MB/scene)/(601024KB/s)4.92
  minutes
• Use multi-threaded program
• While images are being acquired, the storage
  would free up volatile memory
• However, might cause frame loss

32
Signal to Relay and Shutdown

• After storage is complete, signal must be sent to
  relay circuit.
• Signal must be sent through USB
• Signal will cut power for the camera and
  calibration lamp (if needed)
• Software shutdown takes place
• If any exceptions occur, watchdog timer on
  mainboard will reset computer

33
Ground ProcessingSoftware
34
Information

• Purpose process raw data captured by the system
  from space into useful information for scientific
  research

35
Design Overview

• Software utilized
• Microsoft Visual C 6.0
• Image Processing Library 98
• MATLAB 6 Image Processing Toolkit
• Functionality
• Uncompress Raw Data Set
• Correlate Images to Calculate Error
• Correct Images in Data Set
• Perform Fast Fourier Transform (FFT)
• Recompress Processed Data Set

36
Software Comparison

• MS Visual C
• Complete Control of Over User Interface
• Application-specific Algorithms
• Object-oriented design
• Fast Processing Speed

• MATLAB 6
• Easy Code Development (high-level script files)
• Use of Existing Algorithms
• Java GUI Developer (GUIDE)
• Slow Processing Speed

37
Design Details

• Decompressing Raw Data
• Utilize Algorithms Written to Compress Data Set
  in Flight
• Convert Single Compressed File To Individual
  Bitmap Images
• Individual Bitmaps Stored in an Array ? Three
  Dimensional Image Set
• Correlating Images (Calculating Error)
• User Selects One Pixel From Each of Two Images in
  Data Set
• Software Calculates Difference in Points Compared
  to Index of Image in Set

38
Index 1
Index 200
39
Design Details (cont.)

• Correcting Error
• Images Resized (10X) Before Correction for 1/10th
  Pixel Accuracy
• Resized Images Cropped in Vertical Direction Due
  to Data Loss
• Horizontal Error Corrected By Scaling Images to
  Achieve Spatially Square Pixels

40
Example of error in the vertical direction.
Example of error in the horizontal direction.
41
Design Details (cont.)

• Fast Fourier Transform (FFT)
• Applied To Each Vector in the ? Direction of the
  Data Cube
• FFT Translates Path Difference Information Into
  Different Wavelengths (?)

42
Challenges

• Developing Windows NT Embedded Operating System
• Including NI-IMAQ drivers and manual registry
  inputs
• Donation of Windows NT Embedded
• No USB programming experience
• Implementing relay controller

43
Milestone Chart
44
Budget
45
Project Progress