摘要
针对当前激光遥感超像素区域目标信息提取方法中存在的目标检测能力较差、信息提取精度较低,信息提取过程耗时较长等问题,提出基于显著图与相似性度量的激光遥感超像素区域目标信息提取方法。获取图像超像素区域目标的静态灰度特征与动态特征,分别构建静态与动态显著图,将两个显著图融合得到目标区域时空显著图;对该显著图进行修正,获得更准确的目标区域,将其作为信息提取的训练样本,构造对应的特征向量子集,运用层次化处理方法进行特征层次划分,结合图像二值化算法计算各层次信息隶属度,据此进行各层次信息相似度检测,得到激光遥感超像素区域目标信息提取结果。实验结果表明,所提方法的目标信息提取精度与运行效率明显高于实验对比方法,具有较好的鲁棒性。
Aiming at the problems of poor target detection ability,low information extraction precision and long time of information extraction in the current laser remote sensing super-pixel region target information extraction method,a target information extraction method in laser remote sensing super-pixel region based on saliency map and similarity measure is proposed.Acquiring the static gray feature and dynamic feature of the target in image super-pixel region,then constructing static and dynamic saliency maps respectively,and fusing the two saliency maps to obtain the spatio-temporal saliency map of the target region.Correcting the saliency map to obtain more accurate target region and set as the training sample for information extraction.Constructing the corresponding feature vector subset,using the hierarchical processing method to classify the feature hierarchy,combining the image binarization algorithm to calculate the membership degree of each level of information,and the information similarity detection of each level are performed.Finally,the target information extraction result in laser remote sensing super pixel region are obtained.The experimental results show that the extraction accuracy and operational efficiency of the proposed method are significantly higher than the experimental method,and it has better robustness.
作者
李燕
赵志云
LI Yan;ZHAO Zhiyun(Department of Mining Engineering,Lvliang University,Lvliang Shanxi 033000,China)
出处
《激光杂志》
北大核心
2019年第9期117-120,共4页
Laser Journal
基金
吕梁学院校内基金项目(No.ZRQN201603)
关键词
激光遥感技术
超像素区域
目标检测
信息提取
laser remote sensing technology
super-pixel region
target detection
information extraction
作者简介
李燕(1988-),女,山西离石人,硕士在读,助教,研究方向:遥感数据应用;赵志云(1988-),男,山西方山县人,硕士研究生,讲师,研究方向:3S技术应用。
