Grant: Discovery Grants Program - Individual

The proposed research introduces a computational framework for measuring resemblance between visual information granules. The utilization of associated sets and nearness distance functionals of feature vectors make it possible to design perception-based vision systems that support approximate reasoning (feature vectors provide descriptions of set members). This approach leads to information granule visualisation because such sets facilitate description- as well as location-based event detection in the design of artificial vision systems as well as broad-based technology transfer. The scientific approach of the proposed research is to develop the foundations and applications of associated sets in discerning affinities between descriptions of members of disjoint sets of visual objects. The proposed nearness quantification approach is accomplished via an associated set framework in discerning non-empty sets that are descriptively similar (or dissimilar) relative to the distances between the sets using various distance functions. In other words, we are seeking answers to the question: Is a pair of sets sufficiently near (sufficiently far apart) to be considered similar (or dissimilar)? A direct result of this research is its utility in a number of application areas such as visual surveillance, image authentication, trustworthiness of multimedia websites, product association displays in retailing, in education, and change detection and tile clustering in remotely sensed images from satellite and airborne camera systems. The novelty of the proposed research is in (i) finding patterns based on nearness (or apartness) of members of associated sets that are compared descriptively, (ii) using various metrics that measure the distance between sets to determine the degree of nearness or apartness of visual sets with high acuity and (iii) introducing feature vector-based distance functions to discern fine-grained nearness to (apartness from) members of associated sets.