| 摘要: |
| 为了有效分类空间目标，提出了一种航天器图像分类模型快速学习方法。分类模型的学习过程利用了分治思想，首先利用无监督聚类方法将图像数据集散列为类别桶，然后利用每个类别桶中的图像样本训练支持向量机完成学习过程。分类时利用支持向量机对待分类图像样本进行分类。实验结果表明，所提方法具有较好的实时性和准确率，能够为空间应用打下良好基础。 |
| 关键词: 图像分类 航天器识别 图像聚类 支持向量机 |
| DOI： |
|
| 基金项目: |
|
| A Fast Learning Method for Spacecraft Image Classification |
| YE Zhipeng,JIA Rui,YANG Yong,QI Huan,LIANG Hao |
| (Beijing Institute of Astronautical Systems Engineering) |
| Abstract: |
| A fast learning method is proposed for spacecraft image classification. Dividing and conquering strategy is introduced at the learning process. Images are firstly hashed into different bins by the unsupervised clustering method. Then support vector machine(SVM) is trained by the bins to finish the learning progress. SVM is used to directly classify the images. Experimental results indicate that the proposed method features real time classification and high accuracy, which provides a good fundamental for future spacial application. |
| Key words: Image classification Spacecraft recognition Image clustering Support vector machine(SVM) |