This article presents a novel method for superpoint segmentation in multispectral point clouds, called GSI-SS (spatial-spectral joint metric superpoint segmentation). The method aims to improve the efficiency and accuracy of processing multispectral point clouds by integrating both spatial geometry and spectral information. The key idea is to group similar points into superpoints, which can reduce computational demands and enhance processing efficiency. However, existing methods often focus only on spatial geometry, leading to inconsistent spectral features within superpoints, which degrades the performance of subsequent tasks. To address this, GSI-SS introduces a similarity metric that combines spatial geometry and spectral information to ensure consistency within superpoints. Additionally, an intersuperpoint point exchange mechanism is proposed to further refine the segmentation. The method was tested on two real multispectral point cloud datasets, achieving higher recall, precision, and F-score, as well as lower global consistency and feature classification errors. The results demonstrate that GSI-SS outperforms several state-of-the-art methods. The study highlights the importance of integrating spatial and spectral information in superpoint segmentation for multispectral point clouds. The main contributions include the use of a K-D tree to model multispectral point clouds, the development of a new similarity metric combining spatial and spectral information, and the introduction of an inter-superpoint point exchange mechanism to refine segmentation. The proposed method improves the accuracy and efficiency of multispectral point cloud processing, making it more suitable for applications such as 3D land cover classification, vegetation identification, building extraction, ground modeling, and surface reconstruction.This article presents a novel method for superpoint segmentation in multispectral point clouds, called GSI-SS (spatial-spectral joint metric superpoint segmentation). The method aims to improve the efficiency and accuracy of processing multispectral point clouds by integrating both spatial geometry and spectral information. The key idea is to group similar points into superpoints, which can reduce computational demands and enhance processing efficiency. However, existing methods often focus only on spatial geometry, leading to inconsistent spectral features within superpoints, which degrades the performance of subsequent tasks. To address this, GSI-SS introduces a similarity metric that combines spatial geometry and spectral information to ensure consistency within superpoints. Additionally, an intersuperpoint point exchange mechanism is proposed to further refine the segmentation. The method was tested on two real multispectral point cloud datasets, achieving higher recall, precision, and F-score, as well as lower global consistency and feature classification errors. The results demonstrate that GSI-SS outperforms several state-of-the-art methods. The study highlights the importance of integrating spatial and spectral information in superpoint segmentation for multispectral point clouds. The main contributions include the use of a K-D tree to model multispectral point clouds, the development of a new similarity metric combining spatial and spectral information, and the introduction of an inter-superpoint point exchange mechanism to refine segmentation. The proposed method improves the accuracy and efficiency of multispectral point cloud processing, making it more suitable for applications such as 3D land cover classification, vegetation identification, building extraction, ground modeling, and surface reconstruction.