Mathematical morphology and bottom-hat filtering approach for crack detection on relay surfaces

In every industry, the detection of defects in the surface texture is an important part of many quality control applications. The detection of features and classification based on it in a digital image is the key requirement in quality control systems in production. Conventionally, humans were engaged to detect defects in the surfaces and they used to present report sheets based on their assessment. But, this process is a time consuming and expensive one. An inspection system to replace human inspectors should be capable of detecting flaws such as scratches, stains, (textural defects) and dents, cracks, blow holes (structural defects) occurring in various shapes and sizes. The problems may arise by these types of defects and reduces the production chain. The products which are defective is not suitable in the industries and they must be rejected during the production process. This gave way to numerous automated techniques for the detection of defects. So, researchers were trying for some alternate method that would detect defects, hence minimizing the human involvement and at the same time detecting the defects precisely. This paper deals with the structural defects (i.e. crack) on metal disc surfaces in relays. The acquired metal disc images were enhanced by the pre-processing techniques. To extract the cracks in metal disc surfaces two methods were proposed. The first approach used mathematical morphology and watershed segmentation. The second method makes use of mathematical morphology and bottom-hat filtering. The result shows that the later approach is superior for crack detection.