高光谱图像HSI（Hyperspectral Image）可以捕获地物成分的光谱特征，但缺失三维信息，而激光雷达LiDAR（Light Detection And Ranging）可以捕获地物的距离高度信息，两类数据相互补充可以有效提升地物识别分类的精度。Mamba模型具有远程特征学习和高效运算的优势，但目前在多模态遥感数据融合分类中的研究较少，存在多源特征信息缺失，融合不足等问题。基于此，本文提出了一种基于Mamba结构的高光谱和LiDAR数据自适应融合协同分类网络。该网络包括一个可堆叠的基于Mamba结构的双通道协同注意力模块，利用参数共享促进了多源特征之间的相互学习，从而在分类任务中实现更高的分类精度和更好的泛化能力。实验结果表明，所提算法在Trento，Houston2013和MUUFL数据集上的总体精度分别达到了99.33%，91.74%和94.94%，能够更为高效地提取、融合多源特征。

Hyperspectral images (HSI) provide spectral features beyond human visual recognition, but they lack three-dimensional information. Light detection and ranging (LiDAR), on the other hand, offers detailed 3D data, and the combination of the two can significantly improve the accuracy of land-cover classification. In deep learning, Transformer models have become the mainstream classification framework due to their ability to capture long-range dependencies and support parallel processing. However, the quadratic computational complexity of the self-attention mechanism makes Transformer less efficient in processing hyperspectral data. To address this issue, this paper proposes an adaptive fusion collaborative classification network based on the Mamba architecture, designed to integrate hyperspectral and LiDAR data. The network incorporates a stackable dual-channel collaborative attention module, built upon the Mamba structure, as the core of the model. By introducing a parallel training and inference architecture, the computational bottleneck is effectively alleviated, while parameter sharing fosters mutual learning between multi-source features. This leads to higher classification accuracy and better generalization in classification tasks. Experimental results demonstrate that the proposed algorithm achieves overall accuracies of 99.33%, 91.74%, and 94.94% on the Trento, Houston 2013, and MUUFL datasets, respectively, efficiently extracting and fusing multi-source features while significantly improving computational efficiency and remote sensing image classification accuracy.