摘要
为了能够快速而精确地提取图像中扩展目标的直线边缘特征,提出了一种基于块遍历的直线提取方法。该算法先通过块遍历整个图像,然后按照直方图把块分类并进行二值化,再通过链码得到相同单像素直线的多个块表示,然后通过这些直线的斜率和截距进行分类,最后对分类结果求均值。仿真结果表明,该方法不仅能正确地提取直线特征且提取的直线可以达到亚像素精度,其角度定位精度超过0.01 rad。从运算量和复杂性来看,其综合性能要优于经典霍夫变换方法,可以得到很好的直线提取结果,并为硬件实现打好基础。
In order to extract the line-features of extended object more quickly and more accurately, a line edge extraction method is proposed based ergodic block. With proposed algorithm,a whole image is traversed by blocks, then the blocks with double peaks is obtained according to the property of grayhistogram. After processing those blocks by gray threshold and chain-edge extraction, the blocks without line characters are eliminated, the others with line characters are sorted by slope and intercept. Finally,the lines are got by computing the mean of similar slopes and intercepts. Experiment resuits show that proposed algorithm can extract line in subpixel precision and the precision of angle is higher than 0.01 rad. The complexity and computation of proposed algorithm is more better than that of Hough transform, which provides a way for hardware-implementation.
出处
《光学精密工程》
EI
CAS
CSCD
北大核心
2009年第3期662-668,共7页
Optics and Precision Engineering
基金
国家863高技术研究发展计划资助项目(No.2007AA802401)
