基于变分水平集方法的浒苔绿潮面积信息提取  

Information extraction of enteromorpha green tide area based on variational level set method

作　　者：张永梅[1,2] 潘振宽[3] 曹丛华[1,2] 端金鸣[3] 逯京格 

机构地区：[1]国家海洋局北海预报中心,山东青岛266061 [2]山东省海洋生态环境与防灾减灾重点实验室,山东青岛266061 [3]青岛大学计算机科学技术学院,山东青岛266071

出　　处：《海洋学报》2017年第9期121-132,共12页

基　　金：国家自然科学基金项目(41306028)

摘　　要：绿潮面积信息提取是绿潮遥感监测中极其重要的环节,其结果将直接影响后续的统计分析和预测预警工作。目前科研人员一般基于传统阈值方法对绿潮面积信息进行提取,其提取结果具有不稳定、效率低、人为因素影响大等缺点。针对上述问题,本文基于变分水平集的对偶方法和分裂Bregman投影方法对绿潮面积信息进行了提取,并提出一种对上述两种方法提取出的绿潮面积信息结果进行量化的新方法。分别基于传统阈值方法、变分水平集的对偶方法和分裂Bregman投影方法进行了3幅影像的绿潮信息提取实验,对提取结果进行了比对分析。实验结果表明,对分辨率较高的卫星遥感数据,无论从运行效率还是从绿潮面积信息提取结果的精确性及稳定性上,基于变分水平集的对偶方法和分裂Bregman投影方法均优于基于传统阈值方法。Green tide area information extraction is an important link of the remote sensing monitoring, its resultwill directly affect the subsequent statistical analysis and the early warning prediction. Now researchers generally extract the green tide area information based on the traditional threshold method, and this approach has many dis- advantages such as low efficiency, unstable result and human factors. Against to above-mentioned problem, the dual method and split Bregman projection method based on variational level set method was studied used to the green tide area information extraction in this paper. A new quantization method was proposed, which was used to deal with the green tide information result extracted by the two mentioned methods. Based on the traditional threshold method and the dual method and split Bregman projection method, the experiments of three images were respectively carried out, and extraction results were compared and analyzed. To higher resolution satellite remote sensing data, the experiment results show that not only the extraction efficiency but the accuracy and stability based on variational level set method are all superior to the traditional threshold method.

关 键 词：绿潮 信息提取 图像分割 变分水平集方法 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]