3.4 - Local Analysis of Honed Surface in Microscopic Images
- Event
- SENSOR+TEST Conferences 2011
2011-06-07 - 2011-06-09
Nürnberg - Band
- Proceedings OPTO 2011
- Chapter
- O3 - Measuring technologies
- Author(s)
- L. Wang, F. Puente León - Universität Karlsruhe (Germany)
- Pages
- 79 - 84
- DOI
- 10.5162/opto11/o3.4
- ISBN
- 978-3-9810993-9-3
- Price
- free
Abstract
The honed surface of cylinder liners plays an essential role for oil consumption and longevity of combustion engines. The presence of two bands of honing grooves is intended and desired in the manufacturing processes, as they improve the lubrication between the cylinder wall and the piston. Due to the imperfectness of the metal working, honing grooves are smeared and interrupted by the folded metal. These undesired material defects increase the frictional losses and accelerate the wear of the piston.
This paper focuses on the image analysis based optical inspection of the folded metal. Image data were acquired by a scanning electron microscope (SEM) as the defective structures influence the surface function in a roughness scale. The local analysis of the honing texture was performed for the segmentation of such defective edges. Two approaches were presented. The first one distinguished the folded metal from other surface components with respect to the structural features. The statistical analysis of the gradient distribution was exploited to describe straight edges, rough edges and smooth surfaces. This operation generated some feature images, in which rough edges were detected as the folded metal. The second one considered the inspection task as a problem of texture suppression. Obviously, groove textures are irrelevant for the detection of the folded metal. A filter based technology was used for the identification of groove regions. After the elimination of groove textures surface defects were localized by edge detection.
To evaluate the performance of these methods, the inspection results were compared with the accurate manual drawings of the folded metal. Experiments manifested that both approaches feature low error rates. Furthermore, a new measure derived from the segmented image data was utilized to evaluate the surface quality in terms of the defect severity. The correctness of this measure was evidenced by comparison with the visual sorting of surface samples (according to the order from “seriously defected” to “slightly defected”). The test demonstrated that objective and reliable quality assessments of cylinder liners were achieved by the automated inspection.