摘要
介绍了一种基于目标灰度门限和目标之间灰度距离门限的区域自动阈值检测法,用该方法检测提取出灰度图像中包含目标的小区域。然后利用传统的门限自动选择方法找到合适的门限,利用该门限值对所获目标区域进行二值化以得到目标,然后采用改进不变矩方法提取目标特征并采用优化BP神经网络进行识别。该方法经实验验证,效果较好。
This paper presents a new method based on threshold of gray levels and distance of gray levels of different target. The small targets in the gray-level images can be extracted by applying this new method. Then the right gates are obtained by using traditional method of adaptive threshold. The target can be obtained by binary transformation based on these gates. Then the method of features extraction based on invariant Moment theory is used to extract the features and the targets are recognized based on improved BP neural network. This approach has been proved feasible experimentally.
出处
《装甲兵工程学院学报》
2006年第2期38-41,共4页
Journal of Academy of Armored Force Engineering
关键词
图象分割
不变矩
BP神经网络
image segmentation
invariant moment
BP neural network