摘要： 通过对经典边缘检测算法的研究，提出了基于前馈多层感知网络的边缘检测算法。在训练过程中结合边缘检测的特点对BP算法进行了改进。实验结果表明该算法具有很强的自适应性，能够在很大程度上避免背景及噪声对边缘检测结果的影响。

关键词: 边缘检测, 人工神经网络, 多层感知器, BP算法

Abstract: By analyzing the classical edge detection methods and the classification ability of nenral networks, an edge detector based on multi layer perceptron (MLP) is proposed. The network is training under supervision by BP method, the experiment result shows that the edge detector is adaptive and robust to the background noise.

中图分类号: 

• TP391.41