Current Issue Cover
CACIS2019基于多标签的高光谱地物分类

张晶,王亦斌,方帅(合肥工业大学计算机与信息学院)

摘 要
在高光谱地物分类中,混合像元在两个方面给单标签分类带来了负面影响:单类地物在混入异类地物后,其光谱特征会发生改变,失去独特性,使类内差异变大;多类地物在混合比例加深的情况下,光谱曲线会互相趋近,使类间差异变小。为了解决这一问题,本文将多标签的技术运用在高光谱分类中。方法 基于高光谱特性,本文将欧氏距离与光谱角有机结合运用到了LIFT算法的类属属性构建中,形成了适合高光谱多标签的方法。基于标签地位的不相等,本文为多标签数据标注丰度最大标签,并在kNN算法中为丰度最大的标签设置比其余标签更大的权重,完成本文对最大丰度标签的分类。结果 在多标签分类与单标签分类的比较中,多标签表现更优,且多标签在precision指标上表现良好,优于单标签0.5%~1.5%。在本文多标签方法与其余4种多标签方法的比较中,本文方法在2个数据集上表现最优,剩余1个数据集上表现第二。在最大丰度标签的分类上,本文方法表现优于单标签分类,本文方法在数据集Jasper Ridge上在总体分类精度上提高0.2%,在混合像元分类精度上提高0.5%。结论 多标签分类的技术在高光谱地物分类上的应用是可行的,可以提升分类的效果。本文方法根据高光谱数据的特性对LIFT方法进行了改造,在高光谱多标签分类上表现优异。高光谱地物的多标签分类中,每个像元的多个标签的地位是不同的,在分类中可以通过设置不同的权重的方式体现该性质,提升分类的精度。
关键词
Multi-label based hyperspectral feature classification

Zhang Jing,Wang Yibin,Fang Shuai(School of Computer Science and Information Engineering,Hefei University of Technology)

Abstract
Objective There are a number of ground features in the range represented by a mixed pixel.The spectral curve of the mixed pixel is affected by the combination of these features, and the nature corresponds to multiple categories of labels. In the single label classification, the label information of a mixed pixel is represented by a single label, which causes loss of information.In the classification of hyperspectral features, mixed pixels have negative effects on single-label classification in two aspects: When a single type of ground feature is mixed with a different type of ground feature, its spectral characteristics will change, losing its uniqueness and making the difference within the class become larger.When the mixing ratio of multiple ground features increases, the spectral curves tend to be closer to each other, so that the difference between class become smaller.In order to resolve the conflict between single label frame and mixed pixel, it is proposed to apply the technique of multi-label classification to the classification of hyperspectral features. Applying the technique of multi-label classification to hyperspectral feature classification, a pure pixel is assigned a corresponding label, and a mixed pixel is assigned a corresponding set of labels. The application of multi-label technology distinguishes between pure pixels and mixed pixels, so that the spectral curves of pure pixels of different features do not have large intra-class differences and can maintain their uniqueness. And in a multi-label framework, pixels with a deeper mixing ratio can have multiple labels, without being forced to retain only one label.Under the labeling of multiple labels, mixed pixels can retain most of the label information, and can also avoid the precision degradation caused by single label classification. Method In the multi-label classification of hyperspectral features, it is necessary to consider the characteristics of hyperspectral data and the unequal status of mixed pixel labels. There are label-specific features in hyperspectral multi-label classification, that is, there is a band feature combination with strong discrimination ability for a certain label. And the similarity between hyperspectral curves can be measured by Euclidean distance and spectral Angle. Based on hyperspectral characteristics, this paper organically combines Euclidean distance and spectral Angle to build LIFT algorithm"s label-specific features.Thus, a new LIFT method suitable for hyperspectral multi-labels was born.Based on the inequality of label status, this paper marks the label with the maximum abundance for the multi-label model, and sets a greater weight for the label with the maximum abundance than other labels in kNN method, so as to form the prediction of the maximum abundance label. Result The following results are obtained by comparing multi-label classification with single-label classification.On the dataset Samson and dataset Jasper Ridge, multi-label classification performs better in all indicators than single-label classification. On the dataset Urban, multi-label classification performs better than single-label classification in indicator precision and indicator ,and is not as good as single-label classification in indicator accuracy and indicator recall. Multi-label classification has achieved good classification results on hyperspectral images, and it performs well on precision indicators. In the comparison between the multi-label algorithm in this paper and the other four multi-label algorithms, the algorithm in this paper has the best performance in two data sets and the second performance in the remaining one.. And on the three data sets, the overall performance of the ML-kNN algorithm and the BR algorithm is poor. In the prediction of the maximum abundance label, the method in this paper is superior to the single label prediction on the three data sets. Conclusion The application of multi-label classification techniques in the classification of hyperspectral features is feasible and can improve the classification effect. Different from the traditional multi-label classification, the characteristics of hyperspectral images are curves formed by reflectivity of hundreds of continuous narrow bands. And the multiple labels of hyperspectral mixed pixels represent the ground feature information belonging to the pixel. Due to the different abundance of ground features, these labels of ground features have different status. Therefore, in the multi-label classification of hyperspectral images, it is necessary to consider the sample characteristics and the inequality on the labels. The multi-label method in this paper considers the label-specific features of hyperspectral multi-labels, and applies the similarity measurement of hyperspectral curves to the construction of label-specific features, successfully improves the classification effect on the basis of the characteristics of hyperspectral data. In this paper, the method of maximum label abundance prediction is based on the consideration of label inequality, and the prediction result is better than that of single-label classification. The proposed methods based on the two special properties of hyperspectral multi-labels improve the classification effect respectively. However, the improvement based on the two properties is split, and the future can combine the two aspects. In the future, when studying the multi-label classification of hyperspectral, it is necessary to analyze the spectral curves of pixels in different aspects, deeply study the relationship between labels and the information contained in them to design an algorithm suitable for the attributes of hyperspectral multi-label classification.
Keywords
QQ在线


订阅号|日报