A New CBAM-P-Net Model for Few-Shot Forest Species Classification Using Airborne Hyperspectral Images

该论文以高峰林场为研究区，通过在原型网络的卷积块之间插入卷积块注意力模块，构建CBAM-P-Net模型。该模型能够提取有效特征、过滤冗余特征或者降低其权重，从而提高原型网络的分类精度。卷积块注意力模块可以基本取代降维方法，该论文使用原始高光谱影像作为模型输入，实现高精度森林树种分类制图。

This paper takes Gaofeng Forest Farm as the study area. By inserting the Convolutional Block Attention Module (CBAM) between the convolutional blocks of the prototypical network, a CBAM-P-Net model is constructed. This model can extract effective features, filter redundant features or reduce their weights, thereby improving the classification accuracy of the prototypical network. The Convolutional Block Attention Module can basically replace dimensionality reduction methods. This paper uses raw hyperspectral images as the model input to achieve high-precision forest tree species classification mapping.

搜集汇总

背景与挑战

背景概述

该数据集聚焦于利用机载高光谱图像进行少样本森林物种分类，提出了CBAM-P-Net模型以提高分类精度。数据集属于人工林资源监测研究项目，包含4.5 MB的文档格式数据。