I think it can be very helpful for various applications to add specific time series summary features to our set of built-in reducer functions. There are various of such features. For example, you have ones that tell you something about the complexity/entropy, skewness, randomness, symmetry, seasonality or spectral density of the time series, or ones that e.g. count the number of peaks above a certain value.
Although I am not a remote sensing expert I can imagine such measures can be very helpful to define if a given time series belongs to a certain land use type (e.g. forest having high complexity, agricultural land strong seasonality, etc).
A lot of such features are implemented in a Python library that I just found: https://tsfresh.readthedocs.io/en/latest/text/list_of_features.html The library seems to be compatible with numpy and pandas, so it should be pretty straightforward to implement from a technical perspective.
@whisperingpixel @dtiede @augustinh22 You probably have better knowledge about which measures are useful/common for RS purposes
Description
I think it can be very helpful for various applications to add specific time series summary features to our set of built-in reducer functions. There are various of such features. For example, you have ones that tell you something about the complexity/entropy, skewness, randomness, symmetry, seasonality or spectral density of the time series, or ones that e.g. count the number of peaks above a certain value.
Although I am not a remote sensing expert I can imagine such measures can be very helpful to define if a given time series belongs to a certain land use type (e.g. forest having high complexity, agricultural land strong seasonality, etc).
A lot of such features are implemented in a Python library that I just found: https://tsfresh.readthedocs.io/en/latest/text/list_of_features.html The library seems to be compatible with numpy and pandas, so it should be pretty straightforward to implement from a technical perspective.
@whisperingpixel @dtiede @augustinh22 You probably have better knowledge about which measures are useful/common for RS purposes