Polished stone inspection
Despite improvements in stone processing technology, the control of surface quality, arguably the most significant consideration in the polishing of stone for aesthetic purposes, has, perhaps somewhat surprisingly, remained a manual operation.
This absence of automated surface inspection may partly be attributed to specific limitations in currently available inspection technology, particularly in relation to the relatively unstructured nature of both the operating environment and the product itself. The latter being apparent in the form of the complex stochastic patterns and features often present in natural stone. Limitations in quality control performance represent a crucial obstacle in the future evolution of the stone processing industry. As such, the introduction of automated surface analysis is of pivotal importance in realising improved product quality, greater operational efficiency, and a reduction in the environmental impact of stone processing.
This project aims to develop the technological know-how needed to effect the automated inspection of polished stone slabs and tiles. In addition, the automatic and objective inspection of the condition of the stone surface will lend itself to an analysis of the manufacturing process by such techniques as SPC (statistical process control), resulting in the opportunity for the development of procedures for the optimisation of processing parameters (including the design and materials of the processing tools). This will ultimately lead to improved productivity in stone processing. The introduction of standards will offer opportunity for an enhanced distribution network, greater acceptability of the product by end-users, and higher productivity at the stone processing stage.
Different types of surface defects require different detection algorithms. A principal aim of our work is the distinction between natural defects, such as fissures in marble veins, and process-induced defects caused by worn or damaged tools. Process-induced defects are not only less acceptable to prospective customers, they also indicate a process problem that needs to be addressed promptly. The most obvious distinction, from the point of view of a software algorithm, is that process-induced defects are geometrically definable whereas natural defects are not. To analyse the stone surfaces a bright filed illumination technique was used. The following descriptions are based on the processing of bright-field images, although they could just as well be applied to dark-field acquisitions. The raw image can be converted into a useful format with the application of standard machine vision routines.
The process involves:
- Reducing the amount of data in the image
- Removing noise
- Increasing the prominence of scratches
- Further data reduction
- Identifying process-induced defects
- and a novel 'hypothesise and test' circle detection algorithm used