基于矢量加权线检测算子的虹膜环状线条检测

发布时间：2019-03-28 08:33

【摘要】：由于虹膜环状线条与背景相比信号强度比较弱,且背景图像纹理异常丰富,环状线条灰度不完全连续,在灰度图像上直接用边缘检测算子进行虹膜环状线条检测时,会丢失很多有效信息,误检率及漏检率较高,为了解决这个问题,在对虹膜环状线条特征进行分析的基础上,设计了一种适用于虹膜环状线条检测的矢量加权线检测算子。首先由于环状线条在虹膜上的位置比较固定,选取虹膜环状区域左右两部分区域作为感兴趣区域;其次,设计矢量加权线检测算子,依据优势信号出现在不同通道的随机性自适应的进行加权,由此将矢量图像转变为边缘信息最为突出的单通道图像,再根据环状线条的线特征分布特点设计线检测矩阵,对预处理后的感兴趣(ROI)区域进行检测;最后在二值化图像上根据环状线条的形状设计区域形状因子排除非环状线条,实现环状线条的检测。该方法在图库中人工标定的1 921条环状线条的检测正确检出率达到91.78%。实验结果表明,本方法与经典的边缘检测算子相比,更适用于虹膜环状线条的检测。
[Abstract]:Because the iris ring line is weak compared with the background signal intensity, and the background image texture is abnormally rich, and the ring line gray scale is not completely continuous, the edge detection operator is used to detect the iris ring line directly on the gray scale image. In order to solve this problem, a vector weighted line detection operator suitable for iris ring line detection is designed on the basis of analyzing the features of iris ring lines. Firstly, because of the fixed position of the ring line on the iris, the left and right parts of the iris ring region are selected as the region of interest. Secondly, the vector weighted line detection operator is designed and weighted adaptively according to the randomness of the dominant signals in different channels, thus the vector image is transformed into the single channel image with the most prominent edge information. Then the line detection matrix is designed according to the characteristic distribution of the line features of the ring line to detect the (ROI) region of interest after pretreatment. Finally, according to the shape of circular lines, the shape factor of non-ring lines is excluded from the binary image, and the detection of ring lines is realized. The correct detection rate of this method is 91.78% for the 1 921 circular lines manually calibrated in the image library. The experimental results show that this method is more suitable for iris ring line detection than the classical edge detection operator.
【作者单位】：沈阳工业大学视觉检测技术研究所;沈阳化工大学计算机科学与技术学院;
【基金】：国家自然科学基金(61271365)项目资助
【分类号】：TP391.41

【参考文献】