基于泊松分布的日盲紫外电晕检测  

Corona Detection of Solar-Blind Ultraviolet via Poisson Distribution

作　　者：闵超波 顾燕[2] 杨锋[2] MIN Chaobo;GU Yan;YANG Feng(College of Internet of Things Engineering,Hohai University,Changzhou 213000,China;North Night Vision Technology Corp.,Ltd.,Nanjing 211102,China)

机构地区：[1]河海大学常州校区物联网工程学院,江苏常州213000 [2]北方夜视技术股份有限公司,江苏南京211102

出　　处：《红外技术》2020年第8期715-721,共7页Infrared Technology

基　　金：国家自然科学基金(61901157)。

摘　　要：针对日盲紫外电晕探测噪声大、信号微弱等特点,本文提出了一种利用泊松分布的日盲紫外电晕检测方法。为了表征紫外电晕信号的时空域特性,根据光电探测原理,构建了基于泊松分布的紫外电晕目标检测模型。针对紫外图像序列,首先根据暗噪声统计模型进行有无信号的判断,然后利用所建立的紫外电晕目标检测模型得到紫外图像的泊松概率映射图,最后基于改进的最大类间方差法(Otsu)实现紫外电晕目标检测与提取。本文方法在多组紫外图像序列中进行测试与对比,实验证明该方法可以准确地检测出紫外图像中电晕目标,方法结构简单,检测精度高,且性能鲁棒。A novel model is proposed for the object detection of solar-blind ultraviolet corona using the Poisson distribution and based on the photoelectric detection theory.This model can be used to represent the spatiotemporal characterization of solar-blind ultraviolet corona.In ultraviolet image sequences,initially,the existence of ultraviolet corona is determined by the statistical model of dark noise.Subsequently,the Poisson distribution map of the ultraviolet image is obtained using the proposed model.Finally,an improved Otsu’s method is applied to detect and extract the object of ultraviolet corona.The proposed method is successfully tested over many ultraviolet image sequences and compared with the other two methods.The experiment results demonstrate that the proposed method performs better in object detection and has higher accuracy in ultraviolet image sequences.

关 键 词：日盲紫外 电晕探测 目标检测 泊松分布 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]