Knowledge-Based Occluded Object Recognition Based on New Interpretation of MFI and Floating Point Genetic Algorithm

Abstract

In this chapter we design a knowledge-based occluded object recognition scheme using Fuzzy Relational Calculus (FRC), which was proposed by Pedrycz (Pattern Recogn 23(1/2):121–146, 1990), and floating point Genetic Algorithm (GA) (Michalewicz in Genetic algorithm + data structures = evolution programs. Springer, 1994). In the course of doing this we consider the same framework of pattern classification as stated in the previous chapters. We introduce a new interpretation of Multidimensional Fuzzy Implication (MFI) to represent our knowledge about the model objects. The new interpretation is basically a set of one dimensional fuzzy implications. The consequence of all one dimensional fuzzy implications are finally collected through one intersection operator ‘∩’. We also consider the notion of fuzzy pattern vector, which is formed by the cylindrical extension of the antecedent clause of each one dimensional fuzzy implication. Thus, we get a set of fuzzy pattern vectors for the new interpretation of MFI and represent the population of training patterns, obtained from the knowledge about the model objects, in the pattern space. During the construction of the recognizer, based on FRC, we use fuzzy linguistic statements (for fuzzy membership function to represent the linguistic statement) to represent the values of features (e.g. featute F _i is small/medium/big etc. ∀i) for a population of patterns obtained from the model objects. Note that the construction of the recognizer essentially depends on the estimate of a fuzzy relation ℜ_i between the antecedent clause and the consequent clause of each one dimensional fuzzy implication. For the estimation of ℜ_i we use GA. Thus, a set of fuzzy relations is formed from the new interpretation of MFI. This set of fuzzy relations is termed as the core of the object recognizer. Once the object recognizer is constructed the non-fuzzy features if a scene, consist of the model objects, can be recognized. We use the concept of fuzzy masking to fuzzy the non-fuzzy feature values of the test patterns of the scenes. The performance of the proposed scheme is tested through recognition of scenes formed by the occluded objects.