A two-dimensional posture measurement system was developed to evaluate the risks of work-related musculoskeletal disorders(MSDs) easily on various conditions of work. The posture measurement system is an essential tool to analyze the workload for preventing work-related musculoskeletal disorders. Although several posture measurement systems have been developed for workload assessment, some restrictions in industry still exist because of its difficulty on measuring work postures. In this study, an image recognition algorithm was developed based on a neural network method to measure work posture. Each joint angle of human body was automatically measured from the recognized images through the algorithm, and the measurement system makes it possible to evaluate the risks of work-related musculoskeletal disorders easily on various working conditions. The validation test on upper body postures was carried out to examine the accuracy of the measured joint angle data from the system, and the results showed good measuring performance for each joint angle. The differences between the joint angles measured directly and the angles measured by posture measurement software were not statistically significant. It is expected that the result help to properly estimate physical workload and can be used as a postural analysis system to evaluate the risk of work-related musculoskeletal disorders in industry.