Abstract
Vision-based perception is essential for robotic manipulator grasping because the accuracy of object localization directly affects grasp planning and motion execution. This paper presents a comparative analysis of computer vision algorithms for object detection and localization in a structured robotic pick-and-place scenario. Classical edge detection methods, including Canny, Sobel, Prewitt, and Roberts, were applied to the same RGB image of a manipulator workspace containing box-shaped objects. The algorithms were evaluated using detection accuracy, F1-score, centroid localization error, contour stability, and processing time. The results showed that Canny provided the most reliable performance, detecting all 12 target objects with the highest localization stability. The study confirms that classical edge detection can support lightweight vision-based robotic grasping.