Design of a Full-Space Tracking and Measurement System Based on the Light Pen with Deep Learning Methods

Abstract

Visual measurement, a critical component in industrial production, often faces accuracy challenges due to image distortion at the edges of the field of view, particularly in large spaces. This paper capitalizes on the robust feature-learning capabilities of deep-learning algorithms tailored for industrial robots in factory settings, and introduces a comprehensive tracking and measurement system, grounded in deep learning, specifically designed for light pen industrial robots. The approach begins with an analysis of the industrial robot's workspace, leading to the installation of a binocular vision sensor on a rotating platform. This setup forms the basis for a novel full-space tracking and measurement device for industrial robots, utilizing a light pen. The system integrates a target detection algorithm with a rotation platform control algorithm, ensuring the light pen remains centered in the camera's field of view in real-time. This innovation addresses the limitations of traditional light pens, particularly their restricted measurement range and reduced accuracy. Furthermore, the transformation matrix between the pixel and world coordinate systems is established by analyzing the relationship between the binocular vision sensor and the precision rotating platform. In the process of pinpointing the sphere center coordinates of the light pen probe, the system establishes a mapping relationship between the marker point and the probe's sphere center. It then employs the cross-ratio invariance principle and the least squares method for precise calculations. Finally, the system's accuracy is validated using the actual dimensions of standard gauge blocks and balls as references. Experimental results demonstrate that the geometric measurement error in the proposed tracking and measurement system on these standard objects is less than 0.1mm. These results offer an alternative for full-space tracking and measurement in complex industrial environments.