Huanglong Scenic and Historic Interest Area

Hyperspectral identification of travertine state in Huanglong by the PSO-BPNN method

A novel hyperspectral identification method achieved high accuracy in distinguishing the state of travertine formations at Huanglong. The PSO-BPNN classifier, combining particle swarm optimization with backpropagation neural networks, demonstrated exceptional performance with an overall accuracy of 93%, F1-score of 92%, and Kappa coefficient of 97%. This method effectively categorized healthy travertine, blackened formations, those affected by algae erosion, and bare ground using hyperspectral data. The approach involved correlation analysis, sensitive band extraction, and real-world application on Micro-Hyperspectral imaging data, outperforming traditional BPNN classifiers.