基于散射光谱的材质分类识别研究

有效的甄别、掌握空间碎片的状态,是合理利用太空资源和在轨航天器规避风险的前提,为了安全、持续地开发和利用空间资源,基于散射光谱,使用夹角余弦和欧式距离嵌入K临近法(KNN-AC-ED),并与经典的朴素贝叶斯分类器作为对比,对实验室测得的空间碎片中常用的四种材质进行分类识别。分类识别的结果显示,KNN-AC-ED法比经典的朴素贝叶斯分类器总体分类精度高4%。研究表明:朴素贝叶斯分类器需要提取每个光谱曲线的三个特征,而特征提取费时费力;KNN-AC-ED法利用光谱线型和强度两种信息,不仅全面的反应了光谱的信息,且计算相对更快捷。通过对空间碎片常用材质的分类,为进一步研究空间碎片的尺寸、材质等信息提供一定的借鉴意义。

Study on Classification and Recognition of Materials Based on Scattered Spectrum

Effective screening, to grasp the status of space debris, is a reasonable use of space resources and on orbit spacecraft to avoid the risk of the premise. In order to develop and utilize space resources safely and continuously,four kinds of materials commonly used for the test of space debris in the lab are classified and recognized based on the scat-tered spectrum,with the combination of the AC-ED and the utilization of KNN,Naive Bayes classifier. The result of classification and recognition shows that the classification accuracy with KNN-AC-ED is 4% higher than that with the classical Naive Bayes classifier. The study shows, compared with the high complexity and manual intervention of the Naive Bayes classifier extracting features, the physical model is clear and the calculation is quick through KNN-AC-ED with the utilization of spectral linearity and strength. Through the classification of common materials of space debris, the size, materials and the other information of space debris can be further studies and be a significant reference.