The task of multispectral image segmentation (segmentation of images with
numerous channels/bands, each capturing a specific range of wavelengths of
electromagnetic radiation) has been previously explored in contexts with large
amounts of labeled data. However, these models tend not to
利用深度学习技术和强化学习框架，该研究探索了未标记数据识别和分类的方法，实现了在开放域中对新类别的发现，通过多模态信息提取和融合特征，利用自监督学习和聚类方法来增强模型训练，通过环境反馈的奖励调整网络参数，确保对未知数据类别的学习准确性。研究结果在 3D 和 2D 领域的实验数据集上表现出良好的性能。