Analysis of compatibility between lighting devices and descriptive features using Parzen's Kernel: application to flaw inspection by artifical vision

We present a supervised method, developed for industrial inspections by artificial vision, to obtain an adapted combination of descriptive features and a lighting device. This method must be implemented under real-time constraints and therefore a minimal number of features must be selected. The method is based on the assessment of the discrimination power of many descriptive features. The objective is to select the combination of descriptive features and lighting system best able to discriminate flawed classes from defect-free classes. In the first step, probability densities are computed for flawed and defect-free classes and for each tested combination. The discrimination power of the features can be measured using the computed probability of error. In the second step, we obtain a combination that gives a low probability of errors. This leads us to choose each feature individually and then build a multidimensional decision space. A concrete application of this method is presented on an industrial problem of flaw detection by artificial vision. The results are compared with those given by classical multiple discriminant analysis (MDA) to justify the use of our method.

[1]  L. Lebart,et al.  Statistique exploratoire multidimensionnelle , 1995 .

[2]  Fabrice Meriaudeau,et al.  Machine vision for the control of reflecting nonplane surfaces , 1997, Other Conferences.

[3]  B. Silverman,et al.  Choosing the window width when estimating a density , 1978 .

[4]  P. Bolon,et al.  Analyse d'images: filtrage et segmentation , 1995 .

[5]  M. Unser.  Local linear transforms for texture measurements , 1986 .

[6]  P. Gorria,et al.  Real time defect detection using image segmentation , 1997, ISIE '97 Proceeding of the IEEE International Symposium on Industrial Electronics.

[7]  F. Truchetet,et al.  A lighting characterisation by a reliable method. Application to defect detection by artificial vision in industrial field , 1998, IECON '98. Proceedings of the 24th Annual Conference of the IEEE Industrial Electronics Society (Cat. No.98CH36200).

[8]  H. Ventsel,et al.  Théorie des probabilités , 1973 .

[9]  Christophe Dumont,et al.  Lighting study for an optimal defects detection by artificial vision , 1997, Electronic Imaging.

[10]  Y. Lechevallier,et al.  Règles de décision de Bayes et méthodes statistiques de discrimination , 1996 .

[11]  Vijay V. Raghavan,et al.  Feature Selection and Effective Classifiers , 1998, J. Am. Soc. Inf. Sci..

[12]  B. Lamalle,et al.  Simulations of lighting for an optimal defect detection by artificial vision , 1997, Proceedings of IEEE/ASME International Conference on Advanced Intelligent Mechatronics.

[13]  Mark Michael,et al.  Experimental Study of Information Measure and Inter-Intra Class Distance Ratios on Feature Selection and Orderings , 1973, IEEE Trans. Syst. Man Cybern..

[14]  Yves Lechevallier,et al.  Méthodes de segmentation non paramétriques , 1982 .

[15]  Richard O. Duda,et al.  Pattern classification and scene analysis , 1974, A Wiley-Interscience publication.

[16]  Luc Van Gool,et al.  Automated Visual Inspection , 1990 .