基于BP神经网络的铁谱磨粒图像识别方法研究

Research on Recognition of Wear Particles Based on BP Neural Network

在磨粒识别过程中,铁谱磨粒图像预处理和特征参数提取是关键。应用图像形态学的处理方法对磨粒图像进行预处理,结果表明,利用开运算、闭运算的图像形态学处理方法对铁谱磨粒图像进行预处理,可以消除图像二值化后留下的孤立小碎点、孔洞以及边界断点。通过磨粒图像的统计特征参数和傅里叶特征参数建立BP神经网络,并对磨粒进行识别,结果表明:采用该方法能正确识别磨粒图像,辨别磨损机制。

It is the key for ferrographic wear particle image preprocessing and characteristic parameter obtaining during the recognition of wear particles. The wear particles image was preprocessed using morphological operation, result shows that the morphological operation, such as opening and closing operation, can remove noise points, fill small holes and link boundary breakpoints from a binary image. The BP neural network was established by statistical and Fourier characteristic parameters of the wear particles image, and was used to recognize the wear particles. The result shows that ferrographic wear particle images and relevant wear mechanism can be recognized correctly through this method.