基于改进LSD的立体组合靶标自动编码算法研究

发布时间：2018-04-08 08:42

本文选题：机器视觉　切入点：立体组合靶　出处：《科学技术与工程》2016年32期

【摘要】：传统的基于机器视觉空间目标姿态测量技术,由于多采用平面靶,故存在受安装空间限制等问题,提出了立体组合靶的设计;并在靶标中设计不同数目的直线,从而实现靶标的快速自动编码。由于传统的LSD直线检测算法存在断线、派生直线过检测等问题,增加了候选区域包含边缘像素点数筛选,剔除部分伪直线边缘,提高了算法的检测效率;并通过设置候选区域质心间距阈值和方向阈值,合并隶属于同一直线的候选区域,避免了同一直线过检测的情况;增加了相似断线间的合并操作,解决了该算法对相交直线检测的缺陷,实现了立体组合靶标的快速自动编码。实验结果表明:改进后的LSD算法在运行速度、检测效果方面均优于传统的LSD算法,具有较好的工程应用价值。
[Abstract]:Machine vision pose measurement technology based on the traditional, because of using the planar target, so there are limited installation space restrictions, put forward the three-dimensional design of composite target; linear and design different number in target, fast automatic encoding so as to achieve the target. By LSD algorithm to the existence of the traditional offline derivative straight line detection problem, increase the candidate regions including edge pixels screening, eliminate the pseudo linear edge, improves the detection efficiency of the algorithm; and by setting the threshold of spacing and direction of the candidate region centroid threshold, the candidate region with belonging to the same line, to avoid the same line detection is increased; the merge operation is similar to break between the solution to the defects of the intersection line detection algorithm, to achieve a rapid automatic stereo encoding combined target. The experimental results show that the improved LSD The algorithm is superior to the traditional LSD algorithm in the running speed and the detection effect. It has good engineering application value.

【作者单位】：军械工程学院火炮工程系;
【基金】：国家自然科学基金(51307183) 军内科研基金资助
【分类号】：TP391.41

【相似文献】