par Cao, Zhengjun
;Markowitch, Olivier 
Référence IEEE transactions on geoscience and remote sensing, 64, 5514917
Publication Publié, 2026-04-13
;Markowitch, Olivier 
Référence IEEE transactions on geoscience and remote sensing, 64, 5514917
Publication Publié, 2026-04-13
Article révisé par les pairs
| Résumé : | The signature-constrained two-stage framework for hyperspectral target detection (HTD) (IEEE Transactions on Geoscience and Remote Sensing (IEEE TGRS), 5514917, 2025) needs to predict whether the label of each encoded feature vector belongs to the target category. The process computes the inner product between the feature vector and each signature feature vector. It claims that the product reflects the similarity between the two vectors, with higher values indicating greater similarity. We find this claim is generally not valid without further assumptions. We stress that the general inner product cannot be used to indicate the Euclidean distance between two vectors. We show that the distance-compatible similarity score can reflect the true similarity between any two vectors with the same dimension. |



