<![CDATA[This study proposes a Color Threshold-based object detection and calibration system using an Orbbec Astra Pro Plus Stereo Vision camera on a wheeled soccer robot (KRSBI-B) to improve the accuracy and efficiency of real-time ball detection. The camera calibration process is performed using the built-in automatic feature of the Orbbec Astra OpenNI SDK, without manual calibration. Validation results show that the camera depth system has a high level of accuracy with a maximum error of 2% at a distance of 50 cm and an average error of 1.73%, making it suitable for use in object distance estimation. The HSV (Hue, Saturation, Value)-based Color Threshold method was calibrated interactively using color sliders with an optimal range of Hue (5–20°), Saturation (120–255), and Value (100–255). Testing was conducted under synthetic lighting conditions of 500 lux with a distance variation of 50–300 cm. The best results were obtained at a distance of 250 cm, with 0% measurement error and full detection on binary images. The system also showed high stability at distances of 100–300 cm with an average error of only 0.63%, while objects were not detected at a distance of 50 cm due to the camera's field of view limit. This approach provides high processing speed, good detection stability, and low computational load compared to deep learning-based methods. The integration of stereo calibration and the HSV method results in an efficient, accurate, and adaptive vision system, making it highly suitable for real-time applications on wheeled soccer robots.]]>
Copyrights © 2025
