摘要
提出了基于小波变换提取零件图像特征和用自组织特征映射神经网络实现识别的方法,首先,对零件图像进行小波多尺度边缘检测,提取零件图像的边缘轮廓;然后将被检测的边缘轮廓图像分成若干个子区域并分别统计各子区域的边缘像素量,各子区域中的相对边缘像素系数作为零件的特征,将这些特征作为神经网络的输入样本,由自组织特征映射神经网络实现识别。实验结果表明该方法是有效的。
Based on wavelet - neural networks, it presents a method to extract part image feature and recognize parts. This method detects the edges from part image using wavelet multi - scale edge detection, divides edge image into several sub- areas which possesses edge pixels respectively. It defines the relative edge pixel coefficient in each sub- area as its feature and uses the features as the input of neural network to realize pattern recognition. Experiment results that the method can efficiently distinguishes the parts.
关键词
小波变换
自组织特征映射
神经网络
模式识别
Wavelet Transform
Self - organizing Feature Map
Neural Networks
Pattern Recognition