An improved calibration method for robotic visual flexible measurement systems was proposed and a mathematical model for hand-eye errors and kinematic parameter errors of a robot was established. The robotic flexible visual measurement system was set up with a line-structured laser sensor mounted on the robot end and a standard sphere was fixed in the robot workspace as the calibration target.In the calibration, the robot was controlled to measure the coordinates of the standard sphere center at different poses. The initial hand-eye transformation was firstly identified by theoretical kinematic parameters of the robot, then, based on constraint of sphere center,the exact hand-eye transformation and real kinematic parameters were identified with an iterative algorithm at the same time. A calibration experiment for the robotic flexible visual measurement system was performed based on an ABB IRB2400 robot and the accuracy verification experiment was conducted by a laser tracker. The results show that the distance standard deviations of the robotic visual flexible measurement system before and after calibrations have been reduced from 0.566 mm to 0.173 mm, which verifies the validity and practicability of the method proposed. The method improves the accuracy of the hand-eye relationship and does not need external measuring devices, so is suitable for industrial applications.