零件多源图像特征提取和识别的研究

提出了基于小波变换的零件多源图像融合和提取零件图像特征的方法。首先,应用小波变换对多源图像进行多尺度分解,利用小波分解系数融合零件多源图像。然后,对融合图像进行多尺度边缘检测,被检测的图像分成若干个子区域并分别统计其中的边缘像素量,各区域中的相对边缘像素系数作为零件图像特征。最后,应用神经网络和网络技术,进行远程零件多源图像识别。实验结果表明,文中提出的方法是有效的。

Research on features extraction and recognition of parts multi-source image

A method to fuse part multi-source image and to extract part image feature based on wavelet transform is presented. Firstly, the part multi-source image is analyzed using wavelet multi-scale transform to obtain the coefficients of wavelet transform, which fuses a part multi-scale image. Then, the edges from fused part multi-source image is detected using wavelet multi-scale edge detection, edge image divides into several areas and counts edge pixels in these areas, the ratio of edge pixels in an area to total pixels in the area is part image feature. Finally, the part multi-source image is realized pattern recognition using neural networks and network technology. Experiment results that the proposed method can efficiently recognize multi-source parts.