Feature Selection for Cross-Scene Hyperspectral Image Classification via Improved Ant Colony Optimization Algorithm

Hyperspectral images (HSIs) generally contain a large amount of spectral bands (features), and the redundant information in them will cause Hughes phenomenon classification process. And feature extraction selection are two main existing methods to effectively reduce redundancy field HSIs classification. Compared with methods, can preserve most features original data without losing their valuable details. However, based on single scene (domain) perform poorly some scenes (domains) insufficient labeled samples. Therefore, how adopt an efficient method select optimal subsets source target use sample assist so as improve accuracy much possible is still very challenging. In order solve above problem, this paper proposes new cross-scene algorithm: Improved Ant Colony Optimization Algorithm-Based Cross-Scene Feature Selection Algorithm (IMACO-CSFS). obtain more accurate scenes, IMACO-CSFS priority sorting-based ant colony strategy make subsequent search process focus global solution (optimal subset) found previous iteration. addition, further accelerate convergence speed solution, elite ants proposed efficiently for training classifier. Furthermore, simultaneously considers overall accuracies both dynamically adjusts scale ensure consistency selected between attenuating effect shift achieving higher image scene. Experimental results three HSI pairs demonstrate that superior