Choquet integral-based morphological operators with applications to object detection and information fusion

In this research, we defined Choquet integral based-morphological operators. These operators generalize both gray-scale morphological operations and Choquet integrals. Based on these operators, we developed a novel methodology to select important information sources. We applied the methodology to information fusion and image feature extraction. These operators were also used to improve the speed of MSNN.; We showed how the Choquet integral can be interpreted in terms of mathematical morphology in the case that the fuzzy measure is defined on subsets of the two-dimensional plane. We showed that by interpreting the domain of the fuzzy measure as the shape of interest, the Choquet integral can be naturally viewed as computing the degree to which an input shape fits into the shape of interest, and therefore naturally defines fuzzy morphological operations.; We, then, formally defined a generalization of gray-scale morphological operations that are based on Choquet integral filters. This definition covers both binary and gray-scale morphology with flat and gray-scale structuring elements.; We developed a methodology to select (reject) important (unimportant) information sources. Some of the applications include feature selection and domain shape learning.; We generalized the MSNN using Choquet integral based morphological operators. We investigated the applicability of the CMSNN to a land mine detection problem. We compared the performance to the MSNN. The experiments showed that CMSNN has better performance than MSNN. We also showed that domain reduction (domain learning) may increase detection performance.