hello
i am working with hyperspectral pixel mixtures and i want to create a netowkr that maps the abundance values of the mixtures to the spectral signatures.
Basically my input vector is 1x11 and output vector is 1x91
1x11 represents the mixture values for example [0.25,0,0,0.5,0,0,0,0.25,0,0,0] and the output label is the 1x91 spectral values or a 1x91 dimensional vector of normalized values
As you can see the input values contain many zeros and i want to create a mapping between these input mixture values to the output reflectance values.
Should i use dense layers in this case or approach it in another way?
Hi Muneeb,
As very often, several architectures are possible to address your problem.
I suggest you the following publications for a quick but detailed overview:
- Dowler et al., " Abundance Guided Endmember Selection: an Algorithm for Unmixing Hyperspectral Data ", 2010 - https://www.researchgate.net/profile/Mark_Andrews6/publication/224200905_Abundance_guided_endmember_selection_An_algorithm_for_unmixing_hyperspectral_data/links/54d6c6dd0cf246475811a636/Abundance-guided-endmember-selection-An-algorithm-for-unmixing-hyperspectral-data.pdf
- Keshava et al., " Spectral Unmixing ", 2002 - http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.653.8400&rep=rep1&type=pdf
- Licciardi et al., " Pixel Unmixing in Hyperspectral Data by Means of Neural Networks ", 2011 - http://www.gipsa-lab.fr/~giorgio-antonino.licciardi/transaction_licciardi_delfrate.pdf
- Su et al., " Nonnegative Sparse Autoencoder for Robust Endmember Extraction from Remotely Sensed Hyperspectral Images ", 2017 - https://www.umbc.edu/rssipl/people/aplaza/Papers/Conferences/2017.IGARSS.Autoencoder.pdf
- Zhou et al., " A Gaussian Mixture Model Representation of Endmember Variability in Hyperspectral Unmixing ", 2018 - https://arxiv.org/pdf/1710.00075.pdf
- Ozkan et al., " Improved Deep Spectral Convolution Network for Hyperspectral Unmixing with Multinomial Mixture Kernel and Endmember Uncertainty ", 2019 - https://arxiv.org/pdf/1808.01104.pdf
Regards
- Badges
- Science topic
- Similar topics
- Vectorization