<p>Industrial robots with integrated vision sensors are extensively used in cutting, laser welding, painting, and gluing. This paper aims to explore an online tracking trajectory planning method for complex paths based on visual feedback to meet the increasing demands for flexibility and high efficiency in manufacturing systems. The experimental setup includes a 6-DOF industrial robot and a monocular industrial camera, which is installed in an eye-in-hand configuration. The trajectory planning method consists of two parts: the vision module sets the region of interest (ROI) to obtain the contour information of the target trajectory in real-time, and outputs the trajectory points after image processing and feature extraction; The real-time dynamic look-ahead module receives these points and uses non-uniform rational B-spline (NURBS) curves to dynamically construct and interpolate paths, ensuring a strong fit to unknown paths and generating smooth trajectories. In order to meet high-speed and high-precision manufacturing requirements, this study employs the IGH-EtherCat master platform for efficient multi-axis robot control and sets up data channels to synchronize robot motion states and visual feedback information. To verify the method’s effectiveness, three different paths were set up, and five experiments were conducted from different initial positions for each. Results showed consistent tracking performance, with high smoothness in robot operation trajectories and tracking errors below 0.4&#xa0;mm.</p>