Defect inspection for curved surface with highly specular reflection

Abstract

Highly specular reflection (HSR) curved surfaces and their inspection in most manufacturing processes mainly depends on human inspectors whose performance is generally subjective, variable, and therefore inadequate. An automatic vision inspection system offers objectivity, better reliability, and repeatability and is able to carry out defect measurement to evaluate the industrial part's quality. Thus, it is vital to develop an automatic vision system to monitor surface quality online. The main purpose of this chapter is to introduce a new defect inspection method capable of detecting defects on HSR curved surfaces, in particular, to create a complete vision inspection system for HSR curved surfaces (e.g., chrome-plated surfaces). In the first part of this chapter, reflection analysis of HSR curved surface is performed. And a new method is introduced to measure reflection properties of our inspection object. Then, a method is introduced to avoid the loss of defects and solve these challenges which result from various defects and complex surface topography on HSR curved surface. A set of images are captured under different illumination directions. A synthetic image is reconstructed from the set of images. The synthetic image appears more uniform intensity compared with the original image because those specular areas have been completely removed. Furthermore, all defects are integrated in the synthetic image. In particular, for more complicate curved surface, an improvement method is proposed and experiments also validate the method. Finally, a complete vision defect inspection system has been created. The lighting system with side and diffuse illumination is selected for our inspection system and it succeeds in reducing the specular reflection from a curved surface, although some brightness appears at the edge. System parameters and object pose are determined by comparing defect expressivity and specular ratio in the image. Moreover, all defects can be quickly extracted by combining template matching and morphology techniques. The presented automatic vision defect inspection system has been implemented and tested on a number of simulation images and actual parts consist of HSR curved surfaces.