基于BLOB区域和边缘特征分析的准直图像双光学目标识别方法  

Dual Optical Target Recognition Method for Collimated Images Based on BLOB Region and Edge Feature Analysis

作　　者：何文轩 王拯洲[1] 魏际同 王力 弋东驰 HE Wenxuan;WANG Zhengzhou;WEI Jitong;WANG Li;YI Dongchi(Xi′an Institute of Optics and Precision Mechanics,Chinese Academy of Sciences,Xi′an 710119,China;University of Chinese Academy of Sciences,Beijing 100039,China)

机构地区：[1]中国科学院西安光学精密机械研究所,西安710119 [2]中国科学院大学,北京100039

出　　处：《光子学报》2024年第2期244-266,共23页Acta Photonica Sinica

基　　金：国家自然科学基金(No.61705254);陕西省重点研发计划(No.2020GY-005)。

摘　　要：针对光路对接准直目标识别算法对双目标粘连状态无法判别的问题,提出了基于二进制大对象(Binary Large Object,BLOB)区域和边缘特征分析的准直图像双光学目标识别方法。首先,对二值化图像进行数字形态学处理,计算全图各BLOB区域的面积、中心、轴长、区域、有效BLOB区域个数等信息。其次,对有效BLOB区域个数大于1的完全分离双目标准直图像,统计各BLOB区域中心分别为位于两个面积最大的BLOB区域内的BLOB数量,数量小的候选BLOB区域为主激光目标,数量大的候选BLOB区域为模拟光目标。然后,对于有效BLOB区域个数等于1的待识别图像,从左、右、上、下4个方向分别提取模板边缘图像的有效坐标序列和待识别边缘图像坐标序列,搜索有效坐标序列和待识别边缘图像坐标序列的最大相关系数对应的有效坐标序列。当4个方向的相关系数全部大于0.95时,待识别图像为模拟光目标;当4个方向的相关系数都小于0.95时,待识别图像为主激光目标;否则待识别图像为粘连图像。实验结果表明:提出的双光学目标识别算法,不仅能够识别完全分离的模拟光目标和主激光目标,误差小于3个像素,处理时间小于1 s,而且能够判别处于粘连状态的光学目标和单个独立的光学目标,满足光路对接准直图像识别算法对于自适应性、精度和效率的要求。In order to solve the problem that the collimated target recognition algorithm of optical path docking cannot distinguish the adhesive state of double targets,a new method of collimated image dual optical target recognition based on Binary Large Object(BLOB)region feature analysis is proposed.There are two optical targets in the optical alignment image,that is,the simulated optical target and the main laser target.In the initial beam control stage,the positions of the two optical targets are random and uncertain,and there is a possibility of the two optical targets sticking together,which causes great difficulties in beam control.Therefore,optical path alignment needs to solve the image recognition problem in two cases:1)In the initial beam control stage,when the main laser beam and the analog beam are just introduced,the adhesion recognition algorithm needs to be used to identify the adhesion state of the two optical targets.If the two optical targets are in the adhesion state,the two targets need to be completely separated by adjusting the 2D frame BM6XY motor;2)In the case of two optical targets completely separated,it is necessary to distinguish between the analog light target and main laser target in the two optical targets.Firstly,the binary image is processed by digital morphology to calculate the area,center Cxy,axis length lenxy and region Reginxy,of each BLOB region in the whole image.Secondly,the number of valid BLOB regions vblobcount is counted,and the distance between the two maximum connected domains dir is calculated.When vblobcount>1 and dir>100,the collimation image is the completely separated double target image,otherwise it is the adhered image.Then,for the completely separated dualtarget image,the number of BLOBs located in the two largest BLOB regions with the center of each BLOB region is counted.The small number of candidate BLOB regions is the main laser target,and the large number of candidate BLOB regions is the analog light target.Finally,for the adhered image,when dir<100,the two opti

关 键 词：光路对接准直 BLOB区域 边缘特征分析 双光学目标识别 粘连图像识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]