| Summary: | 碩士 === 元智大學 === 光電工程研究所 === 94 === Color is one of the most important information for vision. Human vision can be described in various color model forms, e.g. RGB color model, YIQ color model and HSV color model. The HSV color model is closer to the description of human sight sensation. Therefore, we propose a two-channel joint transform correlator based on HSV color model to achieve pattern recognition.
In the thesis we consider different intensities on target pattern for recognition in various situations. The target pattern is transformed from RGB to HSV color space. We choose hue and saturation components to be the target pattern for recognition in the input plane. Afterward, we recognize the target pattern, which may be rotated, added noise and in a realistic background. The numerical results indicate that the device have nice recognition capability. On the other hand, the proposed architecture has an important result -- the intensity variation on target pattern will have no effect on recognition ability.
|
|---|
