1 of 93

Flash Detection Software

Workshop Section

Capture your own Impact Flash

�

Stefanos Achlatis, Georgia Christofidi

IAASARS, National Observatory of Athens, Greece

sachlatis@noa.gr

Project Duration: October 2021 – September 2022

�

ESA Contract No. 4000135574/21/NL/IB/gg

2 of 93

Installation

3 of 93

What Operating System do you have?

Our Software works for all 3 major OS

4 of 93

Online Detection Domain:

Visit FireCapture Site:

http://www.firecapture.de/

Scroll down and download FireCapture Version based on your OS
Then install FireCapture

5 of 93

Online Detection Domain:

Download our FDS Plugin

Download this jar
Name: FDS.jar

Open FireCapture Folder
Go to the “Plugins” folder (Fig. 1)
Open x64 or x86 folder depending on your installation (Fig. 2)
Create a folder (Fig. 3)

Name: FDS

Inside this folder place our plugin “FDS.jar” (Fig.4)

6 of 93

Simulations:

For this workshop we create some videos with impact flashes or other events.
Download these videos:

Here

Example of such video

7 of 93

Simulations:

Now we will place the simulations in Firecapture

Go to the FireCapture folder
Go to the “testImages” folder (Fig. 1)
Open the “moon” folder (Fig. 2)
Delete the existed “moon.avi” video
Take the first video

FDS_moon_1.avi

Place it in this folder
Rename the video (Fig. 3) :

Name: moon.avi

8 of 93

Let’s start

Open FireCapture and select Dummy Mode

If you had a camera, you would choose your real camera

9 of 93

Let’s start

Let’s set up our plugin
In the “Preprocessing” area press the “None” button
In the pop-up window select the FDS

10 of 93

Let’s start

Select a Threshold value
Then press the button “Press to start”

A proper threshold value is important. The program will capture the event only if at least one pixel of the frame has value greater than the threshold (in the difference image).

You can change this parameter, at any point during the observation.

Setting a big threshold value will lead to “miss” some events.

On the other hand, setting a small threshold value will cause false positives.

Set the proper value, based on the weather conditions and the system.

11 of 93

Logger

During the whole process make sure to look the “Logger”
Logger will print useful information about the process

Will inform you if you have captured any event
Will inform you about some stats of the event

12 of 93

Parameters

Select how many frames before and after the impact you would like to capture
In which format you would like to capture the frames: png, fits, or both

13 of 93

a-parameter

Weighted average
Same idea is used in NELIOTA software
Default value a=0.35

Frame 1

Frame 2

Average 1

(1-α) α

Average 1

Frame 3

Average 2

(1-α) α

Average N-1

Frame N+1

Average N

(1-α) α

1-st average

2-nd average

……

N-th average

α < 1.0

14 of 93

Location of output files

After your first detection you will see a new folder in the location C:\Users\stefo
The new folder is named “observations”
There sub-directories of the day will be created
In the directory of the day, you will find enumerated all the detected events

15 of 93

Output Files

Metadata
Sequence of frames:

5 (configurable parameter) before the event
The frames of the event
5 (configurable parameter) after the event

frame_005

16 of 93

Metadata

In the dummy mode, metadata are dummy too

Timestamp is dummy
FPS is dummy

Include:

Timestamp
Number of pixels
Location (of pixels in the frame)
Camera Information
Capturing Information
FPS

frame_005

17 of 93

Now we are ready to capture our first Lunar Impact Flash (provided by NELIOTA)

First easy example

18 of 93

First Video

Small duration video
The video will play on repeat, so you will see again and again the same impact flash
One frame impact flash (artificially set to be one frame)
Set the proper threshold

Around 100

frame_004

frame_005

19 of 93

Results

You will be notified by the logger that you have captured something
Go to the “writing path” and check what is written
Don’t let it run for much time, the video is small and will play on repeat thus you will capture many events

frame_005

20 of 93

Discussion

Event Info
Camera Info
Time Info

How many events did you captured?
Was this flash multi-frame?

frame_005

21 of 93

Second Event (provided by NELIOTA)

Multi-frame video

22 of 93

Set-up Second Simulation

Now we will place the second event in FireCapture

Go to the FireCapture folder
Go to the “testImages” folder (Fig. 1)
Open the “moon” folder (Fig. 2)
Hide the existed “moon.avi” video
Take the second video

FDS_moon_2.avi

Place it in this folder
Rename the video (Fig. 3) :

Name: moon.avi

23 of 93

Second Video

Multi-Frame Video
Now, experiment with different thresholds
How the threshold interact with multi-frame events?

frame_005

frame_006

24 of 93

Discussion

The duration of the event was N?
What threshold did you used?
Note that you capture: 5+N+5 frames

frame_005

25 of 93

Experiment:

Now set even smaller threshold.
What do we see?

frame_005

frame_006

26 of 93

Third Event (provided by NELIOTA)

Threading

27 of 93

Set-up Third Simulation

Now we will place the third event in FireCapture

Go to the FireCapture folder
Go to the “testImages” folder (Fig. 1)
Open the “moon” folder (Fig. 2)
Hide the existed “moon.avi” video
Take the third video

FDS_moon_3.avi

Place it in this folder
Rename the video (Fig. 3) :

Name: moon.avi

28 of 93

Third Video

This is a bigger video
We will use “threading”
Threading is enabled when a detection is occurred

Searching for flash

Writing flash

29 of 93

Experiment

In order to see the effect of threading set the parameters as shown in the figure
In this way we will make the writing process slightly slower, and you will see that the FireCapture will “lag” with the impact flash frame
Why?

Searching for flash

Writing flash

30 of 93

In which cases “threading” could be a problem?

Discussion

31 of 93

Discussion:

Threading could be a problem in the following cases:
When the moon is “shaking” due to bad weather or telescope movement if sunlit parts are included. This will lead to many fake detections.
What about satellites?

32 of 93

Question

If threading was enabled, will the movement of the lunar limb cause fake detection?

frame_005

33 of 93

Question

If threading was enabled, will a fast movement of the telescope cause a fake detection? Why?
If the threading was disabled, and no fast movement happened. Will we have a fake detection? Why?

frame_005

34 of 93

Why?

Frame 1

Frame 2

Frame 1

Frame 2

Before Processing

After Processing

35 of 93

Why?

Frame 1

Frame 2

Frame 1

Frame 2

36 of 93

Discussion

Let's see the results of the experiment. Open the frame_010.fits with a FITS viewer.
Point at the bright side, see that the value of the pixels in this region is 255.

frame_005

37 of 93

Fourth Event (provided by NELIOTA)

Impact Flash

38 of 93

Set-up Fourth Simulation

Now we will place the fourth event in FireCapture

Go to the FireCapture folder
Go to the “testImages” folder (Fig. 1)
Open the “moon” folder (Fig. 2)
Hide the existed “moon.avi” video
Take the fourth video

FDS_moon_4.avi

Place it in this folder
Rename the video (Fig. 3) :

Name: moon.avi

39 of 93

Fourth Video

This is a video with an impact flash with small duration
Set the proper threading and parameters

40 of 93

Discussion

What threading do you use?
Is a multi-frame impact or single-frame?

41 of 93

Offline Detection

42 of 93

Set-up

Firstly, unzip the folder DetectionStandalone.zip
Open the Detection Standalone app

43 of 93

Set-up

We will perform the task of offline detection in a satellite that was captured by NELIOTA
Name of the folder “FDS_offline_1”

frame_005

44 of 93

Set-up

We will perform the task of offline detection in an impact flash that was captured by NELIOTA
Name of the folder “FDS_offline_2”

frame_005

45 of 93

Set-up

We will perform the task of offline detection in an impact flash that was captured by NELIOTA
Name of the folder “FDS_offline_3”

frame_005

46 of 93

Set-up

Offline Detection could take the entire directory of the observations of the day, and not each event separately
Thus create a new folder and name it "FDS_Offline" and place all the above three events

frame_005

Each of these folders contains up to 50-200 events
Offline detection will inform us quickly which of them could be impact flashes, and which of them are satellite, cosmic rays

47 of 93

First Step

Offline Detection could take the entire directory of the observations of the day, and not each event separately
Thus create a new folder and name it "FDS_Offline" and place all the above three events
Select this directory and press "Start Detection"
Wait a bit...

frame_005

48 of 93

What will happen while waiting:

For each event, the program will read some essential information from the metadata file
Select a Region of Interest around the event
Perform Levenberg–Marquardt algorithm and fit a 2D Gaussian in the event
Depending on the characteristics of the Gaussian the program classify the event

49 of 93

Results

In the directory "FDS_Offline" you will find a .csv file with the results
Moreover, a folder named "Detection_Results" will be created
There you can find more information about each event detection results

50 of 93

Localization

51 of 93

First Event (provided by FDS team)

52 of 93

Set-up

We will perform the task of localization in an impact flash that we captured using this software on 03/06/22
Choose the folder “FDS_localization_1”

frame_005

53 of 93

Set-up

Click “Event Localization”
Select the Directory of the event
Input observatory information as shown in the figure
Press “Start Localization”

frame_005

54 of 93

Set-up

After pressing “Start Localization” you will wait a bit for the program to automatically find the lunar limb.
Accurately finding the lunar limb is important for the localization

frame_005

55 of 93

Results of automatically circle fitting

The results

Visual inspection of the results – change color if you want

These results are empty because no focal length was given

Location of the event

56 of 93

Are we satisfied with the suggested lunar limb?

View Circle

View Limb

57 of 93

If no, change parameters of the algorithm

58 of 93

If no, change parameters of the algorithm

59 of 93

Set sd=4 and press Retry

60 of 93

Set sd=6 and press Retry

61 of 93

Set Boost = 1%

62 of 93

Set Boost = 5% -> This is the best result

63 of 93

If you are still not satisfied with the result use Manual Fit

64 of 93

Manual Circle Fitting

65 of 93

Manual Circle Fitting

Select Point on the Image
Press "Add point"
Continue until you have selected enough points
Press "Manual Fit"

frame_005

66 of 93

The best result: Set Boost = 5%

See "Past attempts"
Choose the best fitting circle
Press continue
Wait a bit

67 of 93

Next Step

In this step you must answer two questions:

Is the non-sunlit lunar hemisphere the east or the west?
Is the image flipped?

Use the reference Image to answer these questions

frame_005

68 of 93

Next Step

Is the non-sunlit lunar hemisphere the east or the west?

West

Is the image flipped?

Then press "Continue without automatic rotation"
Wait... It will take a while

69 of 93

Next Step:�Automatic Correlation

frame_005

70 of 93

Automatic Correlation

The image must be rotated so that Mare Humorum is in the correct spot

frame_005

71 of 93

Suggested rotation

View rotation

View moon

72 of 93

Rotation: -31.53 deg

73 of 93

Mare Humorum is now correctly placed

If you are satisfied with the correlation press "Complete Localization"

74 of 93

Results on the event directory:

75 of 93

Results on the event directory:

This event was also captured by NELIOTA and we can compare our localization results

76 of 93

Second Event (provided by NELIOTA)

Event from Kryoneri telescope at 2020-06-25 18:28:18

77 of 93

Set-up

We will perform the task of localization in an impact flash from Kryoneri telescope at 2020-06-25 18:28:18
Choose the folder “FDS_localization_2”

frame_005

78 of 93

Set-up

Click “Event Localization”
Select the Directory of the event
Input observatory information as shown in the figure
This time, give the focal length of the telescope to calculate the pixel scale
Press “Start Localization”

frame_005

79 of 93

Fitting Circle

With the pixel scale, a suggested radius is calculated
Compare with radius found by circle fitting

frame_005

80 of 93

Next Step

Is the non-sunlit lunar hemisphere the east or the west?

West

Is the image flipped?

frame_005

Let's do automatic rotation

81 of 93

Automatic rotation

Automatic rotation found a rotation angle of 21.7 degrees

frame_005

82 of 93

Grimaldi crater is correctly matched

83 of 93

Comparing results with NELIOTA

84 of 93

Third Event (provided by NELIOTA)

Event from Kryoneri telescope at 2018-08-08 02:29:44

85 of 93

Set-up

We will perform the task of localization in an impact flash from Kryoneri telescope at 2018-08-08 02:29:44
Choose the folder “FDS_localization_3”

frame_005

86 of 93

Set-up

Click “Event Localization”
Select the Directory of the event
Input observatory information as shown in the figure
This time, give the focal length of the telescope to calculate the pixel scale
Press “Start Localization”

frame_005

87 of 93

Fitting Circle

With the pixel scale, a suggested radius is calculated
Compare with radius found by circle fitting
Clear image, better results

frame_005

88 of 93

Next Step

Is the non-sunlit lunar hemisphere the east or the west?

West

Is the image flipped?

frame_005

This time we will not do automatic rotation

89 of 93

Find the right rotation

This time try to find the right rotation
Use rotation of 1.2 degrees

frame_005

90 of 93

Shift Horizontal and Vertical

Horizontal shift by 6 pixels
Vertical shift by 3 pixels

frame_005

91 of 93

Mare Crisium is now correctly placed

If you are satisfied with the correlation press "Complete Localization"