Hyperspectral imaging based multidimensional Terracotta Warrior extraction algorithm

The Terracotta Warriors of the Qin Shi Huang Mausoleum hold immense historical significance. However, during the excavation process, accurately locating and extracting the Terracotta Warriors presents substantial challenges. We conducted research to develop an extraction algorithm. (1) A hyperspectral model was established to extract features. (2) A visual and distance-based feature extraction model was established to simulate visual and distance information, enabling the extraction of multidimensional features. (3) A parallel deep network was constructed, integrating spectral information, simulated visual data, and distance information to accurately extract the Terracotta Warriors. To verify the effectiveness of the algorithm, we collected 51 sets of large-size hyperspectral data on site. Experimental results demonstrate that the algorithm achieves a Combination Measure of 93%, outperforming the traditional Graph-FCN algorithm with a score of 12%. The algorithm innovatively incorporates distance information to classify the Terracotta Warriors, achieving good results.