Research on position and attitude localization of light pen based on sub-pixel edge detection

Addressing the viewing angle deviation problem in light pen-based coordinate measuring systems in industrial measurement, In this paper, a light pen position positioning method based on subpixel edge detection is proposed. Since the angle between the camera imaging plane and the light pen feature point plane causes measurement errors, high-precision pose positioning technology is urgently needed. This paper conducts research in three aspects: First, it uses a monocular vision system to capture images of light pen feature points and establishes optimal pose determination criteria based on geometric models. On this basis, the Roberts edge detection algorithm and RANSAC random sample consensus method are used for feature extraction, and the accurate calculation of feature center pixel coordinates is achieved through a least squares fitting algorithm based on subpixel edge detection. Furthermore, feature matching and Kalman filtering algorithms are combined to realize dynamic prediction and real-time positioning of pose states, completing pose positioning by driving the servo pan-tilt platform through pixel coordinate deviation. Experimental verification shows that the system demonstrates excellent performance within a 500 mm*500mm*200mm measurement range: pose positioning error within 0.06 mm, angle deviation not exceeding 0.1°, average positioning time less than 2 seconds, and real-time image processing frame rate reaching 30 FPS, providing a new technical solution for high-precision industrial measurement.