欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
针对在目标识别过程中切距离的线性逼近、易陷入局部最优的局限性，提出了一种新的仿射变换不变距离度量，即多分辨率迭代切距离（MITD），并将其用于仿射变形下的目标识别。MITD将迭代切距离嵌入多分辨率框架，计算模板图像和实时图像之间的变换不变距离，以扩大算法的收敛域，提高目标识别算法的识别概率。实验结果表明，所提出的基于MITD的目标识别方法较基于欧氏距离（ED）和切距离（TD）的算法具有明显优势，优于基于迭代切距离（ITD）的算法，具有较高的鲁棒性。
英文摘要：
A new transformation invariant metric proposed based on analyzing the limitation of that was multi-resolution iterative tangent distance was tangent distance, which approximate the real distance between object and template linearly and prone to fall into local optimum limitations in the process of target recognition. The iterative tangent distance was embedded into a multi-resolution framework and the real transform invariant distance was obtained between template and real image through iterations in order to expand the domain of convergence of the algorithm, and improve the probability of identification of the target recognition algorithm. Experiments of object recognition show that the proposed algorithm outperforms existing algorithms based on Euclidean distance, tangent distance and iterative tangent distance, and it has strong robustness.
同期刊论文项目
同项目期刊论文
期刊信息
位置：