Algorithm of centerline extracted based on eigen decomposition of Hessian matrix
-
摘要: 在对动态目标闭环跟踪过程中,目标中心线的提取对目标的定位以及实时姿态判定有着重要影响,是实现高精度定位以及姿态判定的重要依据。根据Hessian矩阵在微分几何中的意义以及动态目标的成像特点,提出一种基于Hessian矩阵本征分解的目标中心线提取算法。算法在分割出目标的基础上,进行欧氏距离变换并建立关于目标的灰度Hessian矩阵,对Hessian矩阵进行本征分解,获取目标中心的初始点集,对这一点集采用随机采样一致算法进行野值剔除,最终拟合出目标的中心线。为了验证算法的有效性,设置了评价机制,并和前人提出的若干算法进行了对比,仿真实验表明该算法具有较高的稳定性和准确性。Abstract: In the premise of stabilized tracking for the dynamic target, the centerline extracted is very important to the high accuracy of target location and real-time pose decision, because it is the major judging for target location. According to the Hessian matrixs geometric sense in differential geometry and the 2D feature of target in image, the method of centerline extracted based on eigen decomposition of Hessian matrix was proposed. On the basis of targets segmentation, the Hessian matrix with gray of target in image can be constructed by Euclidean transformation. And then the initial gather of center points can be gained by the eigen decomposition of Hessian matrix. Finally the centerline can be extracted after the bad points were rejected with RANSAC algorithm. In order to prove the availability of this method, with the evaluation index being set the simulation experiment was developed. The results show this method has obvious stability and veracity.
-
Key words:
- centerline extraction /
- Hessian matrix /
- eigen decomposition /
- RANSAC
点击查看大图
计量
- 文章访问数: 1469
- HTML全文浏览量: 220
- PDF下载量: 237
- 被引次数: 0