基于最大类间方差算法的雨滴特征提取研究  

Feature Extraction of Raindrops Based on Ostu Algorithm

作　　者：行鸿彦[1] 黄春虹[2] 

机构地区：[1]南京信息工程大学,气象灾害预报预警与评估协同创新中心,南京210044 [2]南京信息工程大学,江苏省气象探测与信息处理重点实验室,南京210044

出　　处：《气象科技》2017年第3期409-414,共6页Meteorological Science and Technology

基　　金：国家自然科学基金(批准号:61671248);江苏省产学研联合创新资金计划(批准号:BY2013007-02);江苏省高校自然科学研究重大项目(批准号:15KJA460008);江苏省“六大人才高峰”计划和江苏省“信息与通信工程”优势学科资助

摘　　要：为了获得雨滴粒子的特征值,采用阈值处理方法对雨滴粒子及其背景图像进行分割,将复杂的雨滴图转化为只有雨滴粒子真实边界的二值图像,通过均值算法标记雨滴粒子质心位置,利用雨滴长轴信息建立雨滴粒子形状模型,结合全局阈值处理中最大类间方差法和局部阈值法,提出一种基于最大类间方差的雨滴特征值检测方法。试验结果表明该方法能够有效地从图像数据中提取雨滴粒子特征值,粒子数目和雨滴形状是测量降雨强度信息不可或缺的参数。In order to obtain the eigenvalues of rain particles. In this paper, the threshold processing method is adopted to segment raindrops and the background image. The complex raindrop figure is converted to a binary image of the real boundary. The raindrop＇s centroid is marked by using the mean algorithm, and the raindrop shape model is established by using the long axis information of raindrops. Combined with the global thresholds of Otsu and the local threshold method, a method for detecting the characteristic values of raindrops by Ostu is proposed. Through the experiment, it is shown that the method can effectively extract the characteristic values of raindrops from the image data. The number of particles and the shape of raindrops are indispensable parameters to measure the intensity of rainfall.

关 键 词：阈值分割 最大类间方差法 雨滴形状 

分 类 号：P426[天文地球—大气科学及气象学]