4. Conclusion
We have presented an oyster shape analysis and categorization method. We include detail image processing steps for contour extraction, shape representation, and shape matching. A fast and accurate shape matching method called turn angle cross-correlation similarity measure is introduced. It uses the turn angle cross-correlation value to align two contours represented by their turn angle functions to achieve rotation, translation, and scaling invariance for shape matching. It also uses the turn angle crosscorrelation value as the similarity measure between the test contour
and the shape models generated by a simple training procedure for each shape category. Experimental results have shown very high shape grading accuracy compared to human grading results. The design work presented in this paper shows that it is a simple, accurate, and low-cost solution for commercial use.