仪器仪表学报
儀器儀錶學報
의기의표학보
CHINESE JOURNAL OF SCIENTIFIC INSTRUMENT
2014年
8期
1890-1899
,共10页
陈健%郑绍华%潘林%余轮
陳健%鄭紹華%潘林%餘輪
진건%정소화%반림%여륜
瞳孔检测%边缘检测%椭圆拟合%参数传递
瞳孔檢測%邊緣檢測%橢圓擬閤%參數傳遞
동공검측%변연검측%타원의합%삼수전체
pupil detection%edge detection%ellipse fitting%parameter passing
瞳孔检测是提高眼底照相机自动化程度的关键技术。针对大遮挡情况下瞳孔检测准确度降低的问题,提出结合椭圆拟合及参数传递的大遮挡瞳孔精确检测方法。该方法首先引入下采样、二值化、最大连通域及扫描线法实现瞳孔区域的提取与修复。为提高椭圆拟合的准确性,提出闭合边缘检测基础上的伪边缘点去除算法,并使用结合最小二乘法的椭圆拟合实现瞳孔参数的获取。最后,提出结合帧间参数传递的修正算法,实现对瞳孔参数的修正。实验结果表明,与现有方法相比,单帧图像瞳孔检测方法可快速地完成瞳孔参数获取,而修正算法则提高了大遮挡情况下瞳孔参数的准确性。
瞳孔檢測是提高眼底照相機自動化程度的關鍵技術。針對大遮擋情況下瞳孔檢測準確度降低的問題,提齣結閤橢圓擬閤及參數傳遞的大遮擋瞳孔精確檢測方法。該方法首先引入下採樣、二值化、最大連通域及掃描線法實現瞳孔區域的提取與脩複。為提高橢圓擬閤的準確性,提齣閉閤邊緣檢測基礎上的偽邊緣點去除算法,併使用結閤最小二乘法的橢圓擬閤實現瞳孔參數的穫取。最後,提齣結閤幀間參數傳遞的脩正算法,實現對瞳孔參數的脩正。實驗結果錶明,與現有方法相比,單幀圖像瞳孔檢測方法可快速地完成瞳孔參數穫取,而脩正算法則提高瞭大遮擋情況下瞳孔參數的準確性。
동공검측시제고안저조상궤자동화정도적관건기술。침대대차당정황하동공검측준학도강저적문제,제출결합타원의합급삼수전체적대차당동공정학검측방법。해방법수선인입하채양、이치화、최대련통역급소묘선법실현동공구역적제취여수복。위제고타원의합적준학성,제출폐합변연검측기출상적위변연점거제산법,병사용결합최소이승법적타원의합실현동공삼수적획취。최후,제출결합정간삼수전체적수정산법,실현대동공삼수적수정。실험결과표명,여현유방법상비,단정도상동공검측방법가쾌속지완성동공삼수획취,이수정산법칙제고료대차당정황하동공삼수적준학성。
Pupil detection is the key technology for improving the automation of fundus camera.Aiming at the problem of low pupil de-tection accuracy under large occlusion condition,an accurate pupil detection method based on ellipse fitting and parameter passing is presented.Firstly,down-sampling,binarization,maximum connected region and sweeping line methods are introduced to realize the ex-traction and patching of the pupil area.The spurious edge point eliminating algorithm based on closed edge detection is presented to im-prove the ellipse fitting accuracy,and the pupil parameters are obtained using ellipse fitting algorithm combining with least square meth-od.Finally,the pupil parameter correction algorithm based on inter-frame parameter passing is presented to correct the pupil parame-ters.Experiment results show that compared with existing methods,the proposed single frame image pupil detection method could obtain the pupil parameters quickly,and the pupil parameter correction algorithm could improve the pupil parameter accuracy under large occlu-sion condition.