Color-based Object Categorization Model Using Fuzzy HSV Inference System

Color is one of the most identifiable properties of objects and wide range of colors need to be detected. Images captured in RGB color space are hard to extract more than three colors from them due to their cubic shape of merging luminance, saturation, and color spectrum. This paper presents a color-based Object Categorization model that utilizes the HSV color space. The process starts by converting the RGB space into three main components, Hue (wavelength of colors), Saturation (purity of colors), and Value (grayness level or lightness), Unfortunately, Hue causes some sort of fogginess when color classification is applied, due to the gradient between each color. For that, the hue fuzzy sets are used which categorize colors in images. They are used to identify nine deferent object's color and categorize them into the nine common pure colors. The categorization process is then carried out by a PUMA 560 robot in nine different trajectories based on the colors extracted by the fuzzy sets. In order to illuminate bright and dark objects and putting them into the tenth trajectory of the unwanted objects, we modified the saturation and value as fuzzy sets. The system is simulated, tested and proved its effectiveness in detecting colored objects due to their gradient and variations for environments with normal or highly structured light settings, using medium or high-resolution camera.