摘要
指出了传统边缘检测算子算法的不足,提出了一种利用基于BP神经网络的数字图像边缘检测算法,即利用传统边缘检测算子检测出来的图像中像素的灰度的不同比例作为学习训练图像,进行神经网络的学习训练,改变神经网络的结构参数得到神经网络的模型参数,最后给出了BP神经网络实现图像边缘检测的实验研究结果。从实现中可发现,将人们关于边缘特征的先验知识包含在内进行数字图像的边缘检测,能够取得比较好的效果。
The paper points out the limitation of the traditional algorithm and introduces a new algorithm of edge detection algorithm based on BP neural networks in digital image, i.e. we train the neural network with the proportion of the pixel in the edge image by the traditional algorithm. In addition, it gives the results of the experiments with the new algorithm; what's more, it gives the neural network's parameter through experiments. From the experiments, we know if making use of the prior knowledge about edge character, we can process edge detection well.
出处
《西安科技大学学报》
CAS
北大核心
2005年第3期372-375,382,共5页
Journal of Xi’an University of Science and Technology
关键词
神经网络
边缘检测
BP
训练
neural network
image edge detection
BP
train