Automatic Spatial Clustering in Hyperspectral Imagery Using Qutrit-Based Grey Wolf Optimizer Algorithm

Hyperspectral imaging offers a powerful means of ground cover analysis. Absence of ground truth data makes correct number of clusters detection a challenging task in real-time environment. The rich spectral content also makes processing difficult and autonomous cluster detection methods an essential research paradigm. In this work, a qutrit based Grey Wolf Optimizer (GWO) is introduced to handle highdimensional hyperspectral scenes. Classical GWO is a robust swarm intelligence tool, yet it often struggles with stagnant diversity and premature convergence. To overcome these limitations, qutrit Hadamard gate is used to enhance population diversity. The premature convergence problem is addressed using qutrit NOT gates. Rigorous comparative analysis is done with classical GWO, its modified variant, and qubit variant of GWO. F-score is used for cluster quality evaluation and the Adjusted Rand Index is used as the objective function to determine optimal number of clusters. Results demonstrate that the qutrit integrated framework consistently outperforms existing methods across various benchmarks.