search

jkeifer / pyHytemporal

An in-progress python library and command line tools for hypertemporal image classification
MIT License
0 stars 0 forks source link

Major issues finding fits when multidate image contains negative DOYs #10

Open jkeifer opened 10 years ago

jkeifer commented 10 years ago

The algorithm no longer seems able to deal with pixel signatures with negative DOY values. That is, it has extreme difficulty finding the...I don't know, it just doesn't work. Trying to find the fits of an image starting DOY -13 did not work. Telling the process that the image started DOY 0 instead resulted in an acceptable output. I have no idea what is wrong or where.

jkeifer commented 10 years ago

I believe this is related to the fact that in the minimized function, the tshift parameter is ADDED to the DOY from the pixel. That is, if the pixel's signature is supposed to be earlier, a seemingly negative tshift, the max DOY could be something like DOY 33 where in the ref sig the max DOY is DOY 121. Thus, the tshift is: i + tshift = 33 + tshift = 121 -> tshift = +88.

This is not intuitive. I am wondering if this should be changed to subtraction, so the user would be shifting the ref sig to the pixel, not the other way around.

jkeifer commented 10 years ago

I think this should be changed to subtraction before 1.0.