In order to improve the sorting accuracy of packaging production line, a visual tracking system was designed with Delta robot as the research object. The structure of sorting system of packaging production line was introduced in detail, including robot module, control module, vision module and conveyor belt module. A translation calibration algorithm was proposed, which could simplify the whole calibration process, improve the accuracy and reduce the calibration time. The algorithm could better reduce the influences of space restriction on the tracking system. An image processing method was described, which could realize material identification, positioning and capture through the parameters such as the area, roundness and rectangle of the connected area. The feasibility and effectiveness of the system were verified by specific experiments. Delta robot visual tracking system has a missing capture rate of less than 0.1% and a false capture rate close to 0, which significantly improves the accuracy. In the whole process, the robot runs stably without any impact or vibration, and the system has high reliability, which can fully meet the requirements of packaging and other related industries.