Real Time Surface Inspection: A Multivariate Approach

The increasing demand of quality in industrial manufacturing leads to automated solutions for quality assurance systems. Automated visual inspection is one important approach for reliable detection of faults that happen during the manufacturing process. Visual inspection in industry have hard requirements in terms of real time capability, reliability and costs. Surface inspection is one important discipline in visual inspection, where defects in the surface of products have to be detected - like scratches, colour mismatches or irregularities in patternsextures. Generally, surface inspection is a complex topic. Only a minor category of potential problems is satisfactory solved for industry. The major trouble here is that many strategies for realizing surface inspection work well (offline) in the lab but are not subject for real time implementation, which is mandatory for industrial acceptance. The project aims to investigate a new method for surface inspection which enables the implementation of efficient real time solutions. The method is based on Multivariate Calibration (MC). The most important advantage of this method is that only vector - matrix multiplication is performed during runtime which can be implemented very efficiently on DSP platforms. The project not only has to show the suitability of the selected method but also to find important extensions (high discriminative power, reliability, position invariance, non-linearity) to it for surface inspection purposes. After a theoretical investigation, a real time implementation has to verify the usability of the selected method in practice.