COLOR SELECTION METHODS IN THE HSV COLOR SPACE USING VISIBLE LIGHT CAMERA ON UNMANNED AERIAL VEHICLE

• Main Author: Kusnandar, Toni
• Format: Dissertations
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/83378
• Institution: Institut Teknologi Bandung

Research on color selection procedures for various purposes has been extensively conducted. Detecting defects in fabrics, calculating the color index of microbial communities, analyzing digital colors to facilitate fashion design processes, applying colors in artworks, calculating canopy coverage, and achieving other objectives have been the subjects of research. The results of these studies have produced methods that involve complex steps, calculations, and lengthy computational times. This dissertation aims to optimize the color selection method in the HSV color space based on a visible light camera mounted on an Unmanned Aerial Vehicle (UAV). In this dissertation research, a new strategy is presented to optimize the color selection process using the Heaviside Step Function and Hadamard Product techniques. Images obtained from the UAV camera using the RGB color space are converted to the HSV color space to separate the color channel from the saturation and intensity channels. The images in the HSV color space are optimized by selectively choosing the desired colors and setting threshold values for each color component, saturation, and intensity obtained from the test image segment. Threshold calculations for each channel are performed using the standard deviation value approach for constant lighting conditions and the minimum and maximum value approach for varying lighting conditions. The color selection results are stored in a zero-one matrix and then multiplied element-wise with the matrix obtained from the conversion to HSV without selection using the Hadamard Product technique. The optimization results show that the selected colors can be perfectly distinguished from other colors. Additionally, the proposed method uses a simple procedure and does not require feature extraction. Compared to previous research, an extraordinary 1,078.82 times faster computational speed has been achieved. The computational speed of color selection in this research has the potential to be embedded in UAVs for real-time color selection. The research findings are not only applicable to plant images but also to all cases that require color selection in the visible light spectrum.