一种检测金属表面划痕的多尺度多方向形态学方法

A Multi-Scale and Multi-Directional Morphological Method for Metal Surface Scratch Detection

在精密仪器制造中,金属表面的细浅划痕是最难检测的,却又是不可忽视的,采用机器视觉技术对其进行自动检测对于提高产品质量和生产效率有重要意义。在分析细浅划痕形状特征的基础上,设计一组多尺度多方向的形态学结构元素,并通过多尺度多方向的Top-hat变化构建形态学划痕指数,以突显划痕,且抑制背景,进而实现细浅划痕的检测。实验结果表明,这里方法能够有效地检测出各种形状的划痕,能够很好地克服光照条件和噪声的影响,具有较强的鲁棒性。

Inthe manufacture ofprecision instruments, the fine and shallow scratch on metal surface ismost difficult to be detected and cannot be ignored. Usingmachine vision technology, automatic detection of metal surface scratch is important for improve product quality and production efficiency. Based on the analysisof shape characteristic of the fine and shallow scratch, it designs a set of multi-scale and multi-directional morphological structure elements, and buildsthe morphological scratch index （MSI） by top-hat transform. Using the MSI, the scratch can be enhanced and the background can be suppressed, which is useful for scratch detection.As experimental results shown, this method is effective for scratches with various shapes, and robust for lighting conditions and noise.