Hyperspectral Images Change Detection Based on Dense Multi-scale Attention for Land Resource Auditing

Land resource audit is an important aspect of natural resource conservation, but achieving full coverage is challenging due to the vast geographical area. The development of hyperspectral imaging technology in recent years has effectively addressed the difficulties in land resource audit. However, further research is needed on how to accurately extract change information from hyperspectral images. Over the past period, Deep learning methods have been widely applied in tasks pertaining to hyperspectral images change detection, yielding commendable outcomes. This article introduces a dense multi-scale attention-based approach for detecting multiple classes of changes in hyperspectral images. This method makes the most of spectral details regarding hyperspectral images and introduces multi-scale structures and dense links in spatial modules, effectively improving change detection results. Accurate. The outcomes from experiments conducted across various hyperspectral datasets certify that approach surpasses most present methodologies in multi-category change detection.