Aplikasi rangkaian neural dalam pengesanan simpang bagi penterjemah lakaran pintar

Abstract

A corner detector is one of the components for feature extraction in a sketch interpreter engine. Many conventional corner detectors nowadays are based on mathematical models and equations. This research developed a corner detector without using complicated mathematical models and equation. A chain code was used as image or data representation for corner detector data source. Two chain codes applied in this research included Freeman chain code (FCC) and Vertex chain code (VCC). The performance and suitability of these chain codes usage were compared This research focused on the production of an intelligent engine for a sketch interpreter, hence a neural network was chosen to be applied in this corner detector. The neural network package in Matlab software was used by this artificial intelligent method to develop a neural network classifier. Back propagation neural network algorithm was used to develop and produce the classifier for corner detector algorithm. Two dimensional sketch line drawing was involved as an important input in developing and producing the classifier. This research produced the framework for developing chain code corner detector classifier, the neural network classifier, and the development process of VCC from rectangular cell. The algorithm of neural network classifier corner detector for FCC and VCC were also produced. Based on the analysis, the algorithm of neural network classifier corner detector for FCC was found to have more potentials in terms of accuracy of corner detection and suitability of the chain code. This research involved three main components - corner detector, chain code, and neural network. The integration of these components produced a corner detector algorithm and this algorithm can be used in the engine of an intelligent sketch interpreter