A kinematic calibration approach for improving the absolute positioning accuracy of industrial robots using a binocular vision system is proposed. Reflective marker balls attached to the robot’s end-effector enable precise positional data acquisition through a binocular stereo vision measurement system. A comprehensive error model based on the Modified Denavit-Hartenberg (MD-H) parameters is developed, and the Levenberg-Marquardt optimization algorithm is applied to identify and compensate the kinematic parameters. Experimental validation conducted with an ELITE EC66 robot demonstrates enhancements in positioning performance. The validation set data shows that the maximum absolute positioning error, average error, and standard deviation were substantially reduced from 3.56 mm, 0.73 mm, and 0.69 mm to 0.50 mm, 0.15 mm, and 0.12 mm, respectively. These results validate the efficacy of the proposed calibration method, highlighting its potential for practical applications due to its simplicity, efficiency, and cost-effectiveness.