CAJ | 학술논문

针对视频测量建筑物健康监测获取海量影像序列数据快速、准确识别和跟踪目标点的需求，提出基于影像块技术的椭圆形目标点识别和跟踪完整算法.该算法采用影像分块技术降低数据处理量，实现椭圆形目标点跟踪，集成数学形态学和椭圆几何属性特征，消除图像块边缘检测的非椭圆边缘信息，实现椭圆轮廓的提取，并采用最小二乘法拟合椭圆中心实现亚像素定位，快速、准确地实现视频测量建筑物健康监测椭圆形目标点的识别与跟踪.试验结果表明该方法获取的椭圆中心点像素坐标的RMS 残差优于0．025个像素，且相对于随机 Hough 变换和模板识别算法，跟踪效率提高5倍以上.
침대시빈측량건축물건강감측획취해량영상서렬수거쾌속、준학식별화근종목표점적수구，제출기우영상괴기술적타원형목표점식별화근종완정산법.해산법채용영상분괴기술강저수거처리량，실현타원형목표점근종，집성수학형태학화타원궤하속성특정，소제도상괴변연검측적비타원변연신식，실현타원륜곽적제취，병채용최소이승법의합타원중심실현아상소정위，쾌속、준학지실현시빈측량건축물건강감측타원형목표점적식별여근종.시험결과표명해방법획취적타원중심점상소좌표적RMS 잔차우우0．025개상소，차상대우수궤 Hough 변환화모판식별산법，근종효솔제고5배이상.
In order to satisfy the requi rement of identification and tracking the el l iptical artificial targets fast and accurately for the image sequences from videogrammetric measurement for structural health monitoring,this paper proposes a systemic algorithm to identify and track the el l iptical targets using the image block technique.The proposed method extracts the image block from original images to reduce the amount of data processing for the oval targets tracking.The mathematical morphology and el l iptical geometric characteristics are integrated to el iminate the non-el l iptical edge information to extract the el l iptical contour in the range of image block.At last,the sub-pixel center location for el l iptical artificial targets is acqui red by the least square algorithm.The experimental results show that RMS error of 0.025 pixel can be achieved by the proposed method,furthermore,compared with the random Hough transform and template recognition algorithm,the tracking efficiency is improved over 5 times.