复杂背景下燃油喷雾图像中油束轮廓的提取  

Extraction of Oil Stream Contour in Fuel Spray Images Under Complex Background

作　　者：王超 王也 杨添淏 隆武强[1] Wang Chao;Wang Ye;Yang Tianhao;Long Wuqiang(Institute for Internal Combustion Engines,Dalian University of Technology,Dalian 116023,China)

机构地区：[1]大连理工大学内燃机研究所

出　　处：《内燃机学报》2019年第5期410-416,共7页Transactions of Csice

基　　金：国家自然科学基金资助项目(51779044)

摘　　要：提出了一种在复杂背景下燃油喷雾图像中前景油束轮廓提取的新算法,该算法包括油束前景区域预提取、油束轮廓细提取和油束轮廓修补3个步骤.在油束前景区域预提取阶段,利用具有自适应阈值的帧间差分法和形态学填充处理提取出完全覆盖真实油束的前景区域.在油束轮廓细提取阶段,结合像素点在时间维度和空间维度上的信息对预提取获得的油束前景区域内的每个像素点进行多重判断,得到油束轮廓.在油束轮廓修补阶段,结合前景点修补阈值和油束横向、纵向位置边界条件对在视窗污染位置处的油束轮廓进行预测,结果表明:该算法能够精确地提取出处于复杂动态背景下的前景油束轮廓.During the study of fuel spray characteristics by the schlieren-high speed photographic system,the irregular flow of gas at high temperature environment and the window pollution caused by multiple fuel injections could bring great difficulties to the identification and extraction of the foreground oil steam contour.Aimed to fix this problem,a new method for extracting oil stream contour in fuel spray images under complex background was presented.The method includes three steps:pre-extraction of oil stream foreground area,exact extraction of oil stream contour and oil stream contour complementation.In the process of pre-extraction of oil stream foreground area,a larger area,which completely covers the real spray area,was obtained by interframe difference method with self-adaptive threshold and morphological filling processing method.In the process of exact extraction of oil stream contour,each pixel in the foreground area was judged multiple times by using its information in time and spatial dimension,and then the accurate oil stream contour was obtained.In the process of oil stream contour complementation,the oil stream contour prediction was conducted at the window pollution area by combining the pre-attraction patch threshold and the oil stream lateral and longitudinal position boundary conditions.Experimental results show that the algorithm could extract the oil stream contour accurately in complex dynamic background.

关 键 词：油束轮廓提取 油束轮廓修补 算法 前景点多重判断 

分 类 号：TK402[动力工程及工程热物理—动力机械及工程]