Left-handed66 @ 01.22.18, 20:24
Someone can translate this page into Russian?
http://joe-c.de/pages/posts/mapcal_198.phpGoogle translation
the introduction
For the most accurate capture of thermal images, I use point-to-point calibration (description in English) for each pixel, this process is called Mapcal, because the result is a map with calibration values.
You need 2 evenly heated surfaces, one of which is as cool as possible, and one as warm as possible.
For quick tests, a wall is usually sufficient as a cold link, and a palm is like a hot link.
It is not very important what surface temperatures have exactly, but it is very important that they are uniformly warm, or at least a part in front of the camera lens radiates the sensor evenly.
calibration
To capture raw camera data are accepted. First, the “cold link” is recorded. That is, the camera should be aimed at the object of the cooler.
It then produces a camera image that ideally displays the measured value for the link characteristic of each pixel.
If this still image of the camera is saved, a “hot link” is executed again.
Then 2 camera images are available. The next step will be the calculation, the success is usually the same automatically.
DeathPixelMap:
DeathPixelMap is first created (it will be a “Cold Reference” “taken for this”, it’s basically just a binary array ... either the pixel is good or bad. All pixels whose reading is in a certain range of numbers are considered good first, all the rest (shorts or open spots on the Sensor, for example) are marked as bad
immediately GainMap
Now the change is detected for each pixel (which was not marked as bad), as a result of which the “hot” values ​​are divided into “cold”, and the result is the so-called gain value. If it is too large, the pixel is wrong, if it is too small, if it is is defective or “blind pixel”, both marked as bad in
Deathpixelmap
Now all (good) pixels of the “cold” frame are calculated together, forming the average. Then each pixel is subtracted from this average and an offset map is created.
As soon as 3 maps are created, the function will be ready and Mapcal can be used. How Mapcal works:
The original image is captured by the camera.
each pixel is multiplied by the corresponding pixel value of the map with the extension
For each pixel, the corresponding value of the offset map pixel is calculated.
then all strongly marked pixels are replaced by the average of all nice neighboring pixels.
The result is a camera image that only needs to be colored for display and with which you can measure temperature.
if in your own words but uses
search for dead pixels (snapped into maximum brightness) on a uniform dark background (cold wall),
search for dead pixels (snapped into minimum brightness) on a uniform light background (warm hand),
and 2 point calibration
two-point-calibrationConsiders dead pixels on average on live around,
and for the living, it does two-point-calibration,
and eventually builds an offset map for all pixels.
on this amendment map, then the desired image is corrected.