Agent Based Sketch Recognition Platform

There are always many recognizers in a sketch recognition system, and they concurrently output their own recognition results. These recognition results are not consistent. It is a difficult problem to model and solve the consistency problem. Agent based sketch recognition method models the problem as a multi-agent coordination problem, and then solves it using a certain coordination method.This thesis will implement an agent based sketch recognition platform. The platform can provide a coordination framework for users which can be reused to make an agent based sketch recognition system. The framework includes two parts: the coordination method and the agent architecture. This thesis will design and implement the two parts.The mountain climbing based method and the graph based method are two multi-agent coordination methods for sketch recognition. This thesis analyzes the two methods, and then designs a non-serial dynamic programming based multi-agent coordination method for sketch recognition.This thesis analyzes how to create a reactive and a proactive agent in the ABLE environment, and finally implements the platform.