Spot tracker: 'IVcurves.csv' is empty

[PhilippW1111]

I used the Spot tracker and exported the results. Unfortunately, the file 'IVcurves.csv' is empty every time.

Here are the data I used: https://drive.google.com/drive/folders/1g5VKPdfWo55Gpztxf6XqE8K0-w9ClsFU?usp=sharing

[schmid-iap]

Hi Phillipp,

there is one trivial case that can create an empty IVcurves.csv file: If there are no curves that could be successfully tracked. I'll see whether I can prevent saving in such a case.

There is also a possibility that there are only NaNs (not a number values) in a the ...IVcurves.csv: If I0 smoothing is set to a number of points that is larger than the number of data points (=number of images in the stack), the smoothed I0 values become NaN, and so do the intensities after division by I0. The ...rawInt.csv are not affected in that case. This bug will be fixed in the next release, when a new smoothing algorithm will be implemented.

Concerning NaN values, there is also a bug in several recent ImageJ versions; it should be fixed tomorrow, I hope (Help>Update ImageJ). The bug was present since version 1.53m. This bug can make the plots in the I(V) Curve Editor invisible, by setting the y axis range to NaN. In some recent versions it can also cause an exception.

There are a few problems with your data set:

• The stack on your gl drive folder contains only RGB color images; the Spot Tracker only works with grayscale.
• If your images were obtained as JPG, they (1) have a highly nonlinear response (pixel values are roughly the square root of the intensity, but not for low intensities) and (2) the images contain compression artifacts because JPG is a lossy compression. One can't use images taken with a consumer camera for LEED I(V), maybe with the exception of images stored in raw format (I have not analyzed whether these are linear; this may also depend on the camera type). [In any case, a black&white camera has better efficiency than a color camera: A color camera uses half of the area for red and blue pixels, which block the photons emitted from the LEED screen. We are almost finished with our data acquisition based on TheImagingSource GigE industrial cameras; the 33GX174e currently seems to be the best one and is reasonably priced].
• The energies must be ascending; the stack has descending energies (Image>Stack>Tools>Reverse reverses it).
• The images are unnecessarily large, which makes spot tracking a pain. I would strongly suggest reducing the size by about a factor of about 1/6 in x&y (First convert to 32-bit floating point, then Image>Adjust>Size; make sure you have "average when downsizing" checked).
• The dark frame does not look like one; it contains just a few blobs and is zero otherwise. Since the background of the data is about 1, there is no need for a dark frame, unless you also use a flat field (which would be advisable to reduce the impact of the grid non-uniformity and dust on the grid).
• Of course, for real data acquisition, you need a finer energy grid (1.0 or 0.5 eV).

By the way, what is the structure? It looks a bit like a (7x7), but a (7x7) grid does not really fit. Even if I define all the integer spots at the highest energy, I need to increase the search radius to get all spots. In my experience, this only happens if the pattern is incorrect, or if the grids of the LEED screen are badly bent. Is it incommensurate (some moiré)?

--Michael