摘要
相关识别是图像识别的重要方法和手段,广泛应用于各个领域。通过设定阈值来寻找相关峰的方法有其本身固有的缺陷,它只利用了相关峰的峰值大小信息,因而一般用来处理比较理想的情况。对于那些有噪声或背景的目标,很难做到正确识别。在相关峰的识别过程中,如何利用相关峰的形状信息,将起到决策作用。提出在阈值判断的基础上,结合BP神经网络对相关峰形状进行判断,可提高复杂情况下的目标识别率。
Correlation recognition is an important method in image recognition, which has been widely used in many respects. The usual object recognition is to find the correlation peaks, but this method has the flaw in itself, which only takes the advantage of the size of the correlation peaks. For those objects with noise or background, it is hard to do in this way. During the process of recognition of correlation peaks, it is very important to use the shape of the correlation peaks. On the basis of threshold judgment, this article judges the shape of the correlation combined with BP neural network, which can increase the recognition ratio under the more complex condition.
出处
《实验室研究与探索》
CAS
北大核心
2009年第4期56-58,74,共4页
Research and Exploration In Laboratory
关键词
相关识别
阈值
相关峰
MATLAB
BP神经网络
识别率
correlation recognition
threshold
correlation peak
MATLAB
BP neural network
recognition ratio
