摘要
城市的快速发展产生大量建筑垃圾,引发了城市污染,产生“垃圾围城”现象。本文以北京市某区域为研究区域,研究多特征组合方法对高光谱影像的建筑垃圾分类精度的作用,并利用ASD QualitySpecTrek手持式光谱仪实地测量的光谱数据和珠海一号高光谱影像进行建筑垃圾分类试验;提取了光谱特征、植被指数、水体指数和纹理特征等90种特征变量,利用平均不纯度减少的方法对其进行特征重要性排序,选取其中重要性综合结果相对较高的特征变量构成分类多特征组合向量,利用随机森林算法对多特征组合向量进行建筑垃圾分类。试验结果表明,采用多特征组合的随机森林分类方法比传统的随机森林方法效果更好,总体分类精度达85.86%,Kappa系数为0.80,而原始的随机森林方法总体分类精度仅有83.48%,Kappa系数为0.75,这表明了多特征组合随机森林算法的有效性。
The rapid growth of urban areas has led to a significant rise in construction waste,causing urban pollution and waste accumulation issues.This study aims to investigate the impact of multi-feature combination methods on the accuracy of construction waste classification in hyperspectral images,focusing on a specific region in Beijing.Construction waste classification experiments are conducted using on-site spectral data collected with the ASD QualitySpec Trek handheld spectrometer and Zhuhai-1 hyperspectral images.A total of 90 feature variables,including spectral features,vegetation index,water index,and texture features,are extracted and ranked based on their importance using the mean decrease impurity.Feature variables with relatively high importance scores are selected to create a multi-feature combination vector for classification.The random forest algorithm is then employed to perform construction waste classification experiments on this vector.The experimental results reveal that the random forest classification method using multi-feature combination outperforms the traditional random forest method,achieving an overall classification accuracy of 85.86%and a Kappa coefficient of 0.80.In comparison,the original random forest method achieves an overall classification accuracy of 83.48%and a Kappa coefficient of 0.75,indicating the effectiveness of the random forest algorithm using multiple feature combinations.
作者
张戴新月
刘扬
高思岩
ZHANG Daixinyue;LIU Yang;GAO Siyan(School of Geomatics and Urban Spatial Informatics,Beijing University of Civil Engineering and Architecture,Beijing 102616,China;Beijing Future Urban Design Innovation Center,Beijing University of Civil Engineering and Architecture,Beijing 100044,China;Zhengyuan Geographic Information Group Co.,Ltd.,Beijing 101300,China;Beijing Engineering Research Center of Intelligent Pipe Network Assessment and Operating Regulation,Beijing 101300,China)
出处
《测绘通报》
CSCD
北大核心
2024年第6期59-64,共6页
Bulletin of Surveying and Mapping
基金
国家自然科学基金(42271478)
国家重点研发计划(2018YFC0706003)。
关键词
建筑垃圾
多特征组合
随机森林
高光谱影像
construction waste
multiple feature combination
random forest
hyperspectral images
