Classifying Land Cover Using Hyperspectral Image and LiDAR Data

This paper proposes an advanced approach to land cover classification by integrating hyperspectral and LiDAR data to leverage their complementary strengths. Hyperspectral sensors capture detailed reflectance across numerous spectral bands, providing unique spectral signatures that differentiate various land cover types. On the other hand, LiDAR technology offers high-resolution elevation data, enabling accurate characterization of surface structures. This paper details the process of converting unorganized LiDAR point clouds into organized formats for effective analysis, followed by a robust data fusion technique that combines spectral and spatial information. Utilizing a land cover classification methodology on the MUUFL Gulfport dataset, collected in November 2010, we evaluate the effectiveness of our approach by comparing classification accuracies of multiple machine learning algorithms, including kernel classification model, support vector machines, and others. Our results demonstrate significant improvements in classification accuracy when hyperspectral and LiDAR data are fused, highlighting the efficacy of this integrated methodology for complex land cover analysis.