狠狠撸

Master Thesis
Star Tracker for Satellite – CubeSat platform
Implemented by: Nguyen, Ngoc An
Academic Supervisor: Prof. Dr-Ing. Frieder Keller
Work place Supervisor: Prof. Dr. Hien Vo
Work place: Universidad del Turabo – Puerto Rico
Puerto Rico, May, 2016

Contents
? Introduction
? Existing Designs
? Hardware Design
? Algorithm Design
? Implementing Algorithm on FPGA
? Debug, Testing and Result

Introduction
Cube Satellite (CubeSat)
Dimension: 10 x 10 x 10 cm (1 litter)
Mass: < 1.33kg
Commercial off-the-shelf (COTS)
components Electronics Devices
CubeSat QB50P1
[http://www.amsat.org/?page_id=2944]

Introduction
Star Tracker
Image
Sensor
Star Detect &
Centroid
Calculation
Star Identification
Algorithm
Attitude
Determination
Star
Catalog

Existing Design
Star Tracker Accuracy Update
Rate
Size Power
Consumption
ST-200 30” 1 Hz 50g
30x30x38 mm
220 mW
NST 7” 5 Hz 500g
50x50x100 mm
500 mW
STELLA ~ 2” 4 Hz 120g
60x46x58 mm
200 mW
ST-16 7” 2 Hz 90g
59x56x32.5 mm
500 mW

Existing Design
Star Tracker STELLA
[University of Wuerzburg, Germany]

Existing Designs
Star Tracker CubeStar
[Stellenbosch University, South Africa]

System-level design
Real time
Star Detect
Module
FPGA
Image
Sensor
External
memory (Star
Catalog)
Integrated
Processor
core
External
Interfacing
(e.g. SPI, I2C)
Star Detection step is speed up by
hardware module.
?Processor core is free from
handling with large amount of image
data.
? Do not need to buffer image data
before process.
Image
Sensor
Star Detect &
Centroid
Calculation
Star Identification
Algorithm
Attitude
Determination
Star
Catalog
Thesis works:
?Hardware Design
? Design Real time Star Detect Module

Hardware Design
Brightness of Stellar Object
m1: apparent magnitude of star 1
m2: apparent magnitude of star 2
I1: illuminance of star 1 [W/cm2/s]
I2: illuminance of star 2 [W/cm2/s]
Star Vega is used as brightness reference
? Magnitude of star Vega is 0

Hardware Design
Effect of Magnitude cut-off
Two Schemes to maintain Sky coverage:
_ Small FOV, high magnitude cut-off ? more accurate.
_ Large FOV, low magnitude cut-off ? smaller star catalogs / smaller
lens size

Hardware Design
Image Sensor
MT9P031
(ST-16 star
tracker)
MT9M001 AR0130 VITA 1300 Python 0.5
Optical
format
1/2.5" 1/2" 1/3" 1/2" 1/3.6"
Resolution
2592 x 1944
= 5 MP
1280x1024
= 1.3 MP
1280x960
= 1.2 MP
1280 x 1024
= 1.3 MP
800 x 600
=0.5 MP
Pixel size 2.2 um 5.2 um 3.75 um 4.8 um 4.8 um
Sensitivity 1.4 V/lux.sec 4.8 V/lux.sec 6.5 V/lux.sec 4.6 V/lux.sec 7.7 V/lux.sec
ADC
Resolution
12-bit 10-bit 12-bit 10-bit 10-bit
Dynamic
Range
70.1 dB 68.2dB 82dB 60dB >60dB
SNR 38.1 dB 45 dB 44 dB 41dB 40dB
QE (@
550nm)
55% 55% 77% 55%
Power
Consumption
381 mW 363 mW 270 mW 290mW 375 mW
Operating
temperature
-30 to 70 0 to 70 -30 to 70 -40 to 85 -40 to 85
Candidate Image Sensors

Hardware Design
Lens choosing
f-number 2.0 1.4 2.8
Focal length 12mm 16mm 8mm
Field-of-view
Sensor AR0130
18.2 x 12.8 degree 12.78 x 9.0 degree 27 x 19.16 degree
Magnitude cut-off
(expected)
5.16 5.9 4.4
Number of Star in
Catalog
1977 4531 821
Lens available on
market
Edmund Optics
#66-893
Lensagon
BHR16012S12
Edmund Optics
#66-892

Hardware Design
_ Processor: FPGA Cyclone IV
EP4CE6
_ Image Sensor: AR0130
_ Lens: Edmund Optics #66-892
_ Resolution: 1024 x 720
_ Field-of-view: 27 x 19.16 degree
Star Tracker Prototype (without baffle)

Hardware Design
Baffle Design
Star Tracker <0.5U ?
Vane 0 Vane 1 Vane 2
Height [mm] 4.73 8.03 11.84
Distance from
outermost vane [mm]
0 11.24 24.24
?Baffle radius r = 20 mm
?Baffle length s = 35 mm

Star Detect and Centroid
Calculation
Star can be considered as a point-source light
?if the lens is right focus, star intensity occupy 1 pixel
on image sensor.
? Centroid Calculation cannot sub-pixel accuracy
? Cannot separate star image and hot pixel.
Solution:
Slightly defocus the image ? Star occupy 3x3 to 5x5 group of pixels.
Star Centroid is calculated by intensity weighting
1 1
1 1
( , )ROIend ROIend
center
x ROIstart y ROIstart
x I x y
x
DN
? ?
? ? ? ?
?
? ? ?
1 1
1 1
( , )ROIend ROIend
center
y I x y
y
DN
? ?
? ? ? ?
?
? ? ?
1 1
1 1
( , )
ROIend ROIend
DN I x y
? ?
? ? ? ?
? ? ?
ROI (Region of Interest)

Star Detect and Centroid
Calculation
Star valid conditions (method 2):
_ Median value of 3x3 window > threshold value
Star valid conditions (method 1):
_ Center pixel > threshold value 1
_ Sum of 3x3 window > threshold value 2
Additional condition: Center pixel = max of the window
? Avoid re-detect the same star spot

Noise in AR0130 sensor
( , ) ( , ) ( )read signal background colI x y I x y I x?? ? ?
Vertical strip noise is dominant

Noise Correction
720
1024
8
I(1,1) I(1,2) I(1,3) … I(1,1024
)
I(2,1) I(2,2) I(2,3) … I(2,1024
)
… … … … …
I(8,1) I(8,2) I(8,3) … I(8,1024
)
Mean
Column
1
Mean
Column
2
Mean
Column
3
… Mean
Column
1024
Every pixel will be subtracted by the mean value of its column
Noise-correction

Star Detect & Centroid
Calculation Block

Row Buffer (Internal RAM
FIFO)
FIFO)
FIFO)
FIFO)
Data
Pix_clk
FV
LV
Pixel
Counter
x_pixel
y_pixel
Centroiding
Block
: D flip-flop
Note:
Condition
Checking
Block
Window
data
en_tick
x_centroid y_centroid

Testing
Accuracy of centroiding is evaluated by inter-star angular distance
?With distortion correction the averaged error of inter-star distance is
0.0536 degrees (~ 3’12.96”)
? Dimmest star can be detected is σ Orionius (Magnitude 4.0)
Testing scene

Conclusion
Current achievement:
? Build Real time Star Detect Module.
? Build de-noise Pre-processing Module
? Can detect star with magnitude up to 4.0
Future Works:
? Integrated hardware modules with a processor
core into one FPGA chip.
? Implement Star Identification algorithm and
Attitude Calculation on processor core.
? Build Lab Test bench.
? Testing whole system.

狠狠撸

AnNguyen_MSThesis

Recommended

More Related Content

What's hot (20)

Viewers also liked (15)

Similar to AnNguyen_MSThesis (20)

AnNguyen_MSThesis