Spatial correlation guided prototype distillation small sample classification method

• Main Authors: ZHANG ZHICHENG, CHEN LIANG, SHOTANE, NI RUNFENG, ZHOU DYLAN, LIU YUQING, ZHANG TONG, CHEN HE
• Format: Patent
• Language: Chinese; English
• Published: 01.12.2023
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; PHYSICS

The invention discloses a prototype distillation small sample classification method guided by spatial correlation, and the method comprises the steps: carrying out the self-supervision pre-training of a CNN-ViT double-branch mixed heterogeneous network model through an unlabeled optical remote sensing image, so as to obtain a CNN-ViT double-branch mixed heterogeneous network model after the self-supervision pre-training; performing meta-training on the network model by using mode features, context features and spatial features of the labeled optical remote sensing image to obtain a CNN-ViT double-branch hybrid heterogeneous network model for small sample classification; and inputting the to-be-classified optical remote sensing image into the CNN-ViT double-branch hybrid heterogeneous network model for small sample classification to obtain a classification result of the to-be-classified optical remote sensing data. According to the method, potential spatial discriminant semantic information can be effectively