- 5
- 0
- 约1.89万字
- 约 3页
- 2017-12-21 发布于天津
- 举报
神经网络方法应用于轮廓定位-计算机工程与应用.PDF
Computer Engineering and Applications 计算机工程与应用 2011 ,47 (19) 179
神经网络方法应用于轮廓定位
肖 平,周智恒
XIAO Ping ,ZHOU Zhiheng
华南理工大学 电子与信息学院,广州 510641
College of Electronics Information Engineering ,South China University of Technology ,Guangzhou 510641 ,China
XIAO Ping ,ZHOU Zhiheng.Neural network method applied to contour location.Computer Engineering and Applica-
tions ,2011 ,47 (19):179-181.
Abstract :For the object contour location in the images ,traditional active contour models and self-organized mapping neural
network based algorithms are prone to trap into local minimum in their optimizations and depend on the initial contour se-
lection.A contours location algorithm based on RBF neural network is proposed in this paper.Firstly ,image feature points are
obtained by adaptive gradient threshold method.After the feature points are clustered ,a group of radial basis functions are
constructed.Using the pixels ’intensities and gradients as the input vector ,the final object contour can be obtained by the pre-
dicting ability of the neural network.Compared with traditional algorithm ,simulation results show that the proposed contours
location algorithm can realize the object contour detection in high efficiency.
Key words :contour location ;neural network ;adaptive threshold
摘 要:关于从图像中定位物体轮廓的问题,目前所采用的活动轮廓模型和基于自组织神经网络的算法,存在能量泛函优化容易
陷入局部极值和演化过程依赖于初始轮廓的选取等问题。提出了一种基于RBF 神经网络的轮廓定位算法。首先,通过自适应梯
度阈值方法来获取图像特征点。然后,通过特征点的聚类建立一组基函数,把图像像素点的像素值和梯度构造输入向量空间,在
网络权值训练完成后,利用网络的预测功能来准确判断物体轮廓。与传统算法相比,仿真结果表明提出的轮廓定位算法可以高
效地实现目标轮廓定位。
关键词:轮廓定位;神经网络;自适应阈值
DOI :10.3778/j.issn. 1002-8331.2011.19.050 文章编号:1002-8331(2011 )19-0179-03 文献标识码:A 中图分类号:TP391
1 引言 廓的依赖性很强,假如物体的内部还存在轮廓而且初始轮廓
轮廓提取是计算机视觉中非常重要的课题。活动轮廓模 在物体的外部,算法是失效的。
型首先是由Kass[1]在1988年提出。这个模型从一个初始轮廓 针对活动轮廓模型和基于自组织映射的算法存在的问
开始演化,直到能尽量近似实际目标轮廓为止。初始轮廓是 题,
原创力文档

文档评论(0)