Abstract: With the rapid development of information technology and earth observation technology, remote sensing technology is playing an increasingly important role in our life and economic construction. The accuracy of class-ification model and its anti-jamming capability are crucial for the accurately mapping of complex land cover and land-use categories. This paper deals with the limitations of insufficient labeled samples of large-scale high-resolution remote sensing images. Based on the low-rank representation and semi-supervised learning methods, a large-scale remote sensing image classification method is proposed. In order to reduce the exponentially growing computational complexity of low-rank method, the image is divided into blocks and the low-rank representation is implemented for each block respectively. The results implemented on WorldView-2 remote sensing data show that the proposed method achieves relatively high classification accuracy with nearest neighbor classifier under few labeled samples. Therefore, the proposed remote sensing image classification method can effectively improve the accuracy of land cover classification, and has high efficiency in remote sensing image classification.

Key words: remote sensing image, low-rank representation, classification

摘要： 随着信息技术和对地观测技术的飞速发展，遥感技术在社会生活和经济建设中发挥着越来越重要的作用。分类模型的准确性和抗干扰能力对精确绘制复杂的土地覆盖和土地利用分类至关重要。针对大规模遥感数据难以获取大量标记数据的问题，基于低秩表示模型和图的半监督学习方法，提出了基于分块低秩图的大规模遥感图像半监督分类应用。为了解决低秩表示计算复杂度高的问题，将预处理后的图像按像素进行分块处理，并在每个块上实现低秩表示。在WorldView-2影像上的分类结果表明，在少量标记样本下，该方法利用简单的最近邻分类器即可实现对城市地物的精确分类。因此，该方法有效地提高了土地覆盖的分类精度，在遥感图像分类中具有较高的效率。

关键词: 遥感图像, 低秩表示, 分类