荧光磁粉探伤中裂纹自动检测方法研究

发布时间：2019-02-15 07:55

【摘要】：工件裂纹检测方法多种多样,相比价格比较昂贵的超声波探伤仪,荧光磁粉裂纹检测因成本低、灵敏度高以及检测速度快等优点而被广泛使用。由于传统的荧光磁粉表面裂纹检测主要依靠人工判别来实现的,这种方式检测的效率较低且判断精确度不高或者是因视觉疲劳而出现误检,同时人体长期在紫外光环境下工作易危害身体健康。因此,研究基于数字图像处理和模式识别技术的荧光磁粉裂纹自动化检测方法具有重要的理论意义和实用价值。本文重点研究了荧光磁粉图像裂纹自动检测的图像预处理、图像分割、特征提取以及裂纹识别等4个关键步骤。首先,采用加权有向平滑滤波去除噪声的方法对图像进行预处理,旨在对图像进行去噪的同时能够尽可能的保留边缘信息,为后期的灰度-梯度共生矩阵最大熵图像分割提供有利条件。实验结果表明,与传统的均值滤波和中值滤波等去噪方法相比,此方法去噪效果更佳,更好地保留边缘信息,避免边缘模糊化;其次,给出基于遗传算法的灰度-梯度共生矩阵最大熵分割改进算法,有效解决了传统灰度-梯度共生矩阵最大熵分割产生的裂纹断裂现象,提高了分割运算速度。为了消除基于遗传算法的灰度-梯度共生矩阵最大熵改进算法对荧光磁粉图像分割引入的噪声,提出了点邻域搜索方法,完善图像的分割质量;再次,根据裂纹与非裂纹图像在形状、灰度以及梯度信息的差异性,本文采用Hu矩不变量和梯度方向直方图(HOG)描绘子作为分类器的输入样本,并针对HOG描绘子给出基于霍夫变换的改进方法;最后,针对支持向量机算法中人为选取惩罚因子和核参数的缺点,给出基于粒子群寻优(PSO)的SVM荧光磁粉裂纹检测改进算法,进一步提高裂纹的识别率。
[Abstract]:Compared with the expensive ultrasonic flaw detector, the fluorescence magnetic particle crack detection is widely used because of its advantages of low cost, high sensitivity and high detection speed. Because the traditional detection of surface cracks of fluorescent magnetic powder mainly depends on manual discrimination, the efficiency of this method is low and the accuracy of judgment is not high, or because of visual fatigue, there is false detection. At the same time, the human body long-term work in ultraviolet light environment is likely to endanger the health of the body. Therefore, it is of great theoretical significance and practical value to study the automatic detection method of fluorescent magnetic particle crack based on digital image processing and pattern recognition technology. This paper focuses on four key steps of image preprocessing, image segmentation, feature extraction and crack identification for automatic detection of cracks in fluorescent magnetic particle images. Firstly, the weighted directed smoothing filter is used to pre-process the image, which can keep the edge information as much as possible while de-noising the image. It provides favorable conditions for gray-gradient co-occurrence matrix maximum entropy image segmentation. The experimental results show that compared with the traditional methods such as mean filter and median filter, this method has better denoising effect, better preserving edge information and avoiding edge blurring. Secondly, an improved gray-gradient co-occurrence matrix maximum entropy segmentation algorithm based on genetic algorithm is presented, which effectively solves the crack fracture phenomenon caused by the traditional gray-gradient co-occurrence matrix maximum entropy segmentation, and improves the segmentation operation speed. In order to eliminate the noise caused by the gray gray-gradient co-occurrence matrix maximum entropy improved algorithm based on genetic algorithm, a point neighborhood search method is proposed to improve the image segmentation quality. Thirdly, according to the difference of shape, grayscale and gradient information between crack and non-crack image, we use Hu moment invariant and gradient direction histogram (HOG) descriptor as the input sample of classifier. An improved method based on Hough transform is given for HOG descriptors. Finally, aiming at the disadvantage of artificially selecting penalty factor and kernel parameter in support vector machine algorithm, an improved algorithm of SVM fluorescence magnetic particle crack detection based on particle swarm optimization (PSO) is proposed to further improve the crack identification rate.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TB302.5;TP391.41

【相似文献】