铁谱磨粒多重分形特征研究

为有效描述铁谱磨粒特征,提出用多重分形谱参数表达磨粒形态特征的新方法。选择盒计数法计算磨粒图像的多重分形谱,研究磨粒多重分形谱的有效性,分析磨粒多重分形谱参数的不变性和鲁棒性;确定磨粒图像预处理方法,并对4类典型磨粒的多重分形谱参数进行统计分析。结果表明:将多重分形谱参数应用于磨粒识别,总识别率为82.5%。磨粒具有明显的多重分形特性,可用多重分形谱参数来描述磨粒的形态特征;多重分形谱参数具有平移不变性,但对灰度变化和噪声干扰的鲁棒性较差,在提取多重分形谱参数时,需要对磨粒图像做严格的预处理。

Study on Multi-fractal Features of Ferrographic Wear Particles

To effectively describe the features of ferrographic wear particles,a novel method that utilizes multi-fractal spectrums parameters to depict features of wear particle images was proposed.The multi-fractal spectrum of wear particle images was computed by the box counting method to study the validity of multi-fractal spectrums of wear particle images,and the invariance and robustness of multi-fractal spectrum parameters were analyzed for wear particle images.The pre-processing steps of wear particle images were designed and the statistical analysis was carried on the multi-fractal spectrum parameters of 4 types of wear particles.The results show that,when applying the multi-fractal spectrum parameters for wear particle recognition,the total recognition rate is 82.5%.Wear particles have obvious multi-fractal characteristics and multi-fractal spectrum parameters can be employed to describe the morphological characters of wear particles.Multi-fractal spectrum parameters have translational invariance and poor robustness for gray changes and noise pollution,so preprocessing of wear particle images is needed when extracting multi-fractal spectrum parameters.