基于模板匹配和不变矩的浮游动物快速测量方法  

The method of rapid measurement of marine zooplankton based on template matching and invariant moment

作　　者：徐帅 杨俊毅[1,2] 郑旻辉[1,2] 谢尚微 

机构地区：[1]国家海洋局第二海洋研究所,浙江杭州310012 [2]国家海洋局海洋生态系统与生物地球化学重点实验室,浙江杭州310012

出　　处：《海洋学研究》2017年第4期69-75,共7页Journal of Marine Sciences

基　　金：国家自然科学基金项目资助(41376169);国家重点研发计划项目资助(2016YFC0302403);国家"863"计划项目资助(2012AA092102-1)

摘　　要：近年来基于图像处理方法的浮游动物分类技术逐渐用于海洋生态系统的研究中,相关检测仪器也从实验室处理向原位现场分析发展,这对检测算法的效率和处理速度提出了更高要求。本文根据海洋浮游动物显微图像的形状特点,提出将模板匹配方法与不变中心矩描述方法相结合,先利用模板匹配限定目标初步范围,再比较不变矩确定最终目标动物,并统计动物数量和尺寸。本文方法不受动物方位旋转和尺寸缩放的影响。实验验证结果表明,该方法与传统识别方法相比简单高效,处理速度快,误差范围小,适用于浮游动物的实时原位观测。Zooplankton is widely distributed in the marine ecosystem,which plays an important role in global climate change.Zooplankton analysis based on image processing needs rapid development.According to the shape of marine zooplankton microscopic image characteristics,a hybrid method based on template matching and invariant central moment description method was proposed in this study.Template matching was used to define the target initial range,and the final target organisms were determined by comparing the moment invariants.The amount and size of the target were calculated.This method was not affected by the rotation of biological azimuth and the scaling of size.The verified experiments show that the proposed method is fast and efficient,which is suitable for situ zooplankton monitoring in real time.

关 键 词：浮游动物 模板匹配 不变矩 

分 类 号：Q958.8[生物学—动物学] TP391.41[自动化与计算机技术—计算机应用技术]