shendb2022 / DSC

A Dual Sparsity Constrained Approach for Hyperspectral target detection, IGARSS, 2022
Apache License 2.0
7 stars 0 forks source link

How do you select the target prior? #1

Open medoyounis opened 1 year ago

medoyounis commented 1 year ago

If I wanted to try the algorithm on a different his images that contain a predefined target, how do I initialize the value for the target_prior_coordinates in the code? for example, in the sandiego image, the target_prior_coordinates is (22,69), so how were those numbers chosen? is 22 the y axis coordinates and 69 is the x axis coordinates?

how do I set kt(number of atoms) for other HSI images?

shendb2022 commented 1 year ago

First, selecting the target spectrum is an important issue in hyperspectral target detection. The target spectrum should be representative and reflect the spectral curve of targets. Generally speaking, the average spectrum of target pixels, the pixel most close to the other target pixels calculated by L2-norm or SAM, and the prior-known spectral signature from the library can be used as the target spectrum.

Second, in this work, we just randomly select a target pixel as the target spectrum. It can be replaced by the average spectrum or the pixel most close to the other target pixels. 22 and 69 represent the coordinates on the x-axis and y-axis, respectively.

Third, for other HSIs, you just need to do a loop and select the best Kt. You can constrain the range in [10,150] with stride 10.