At present, most of forging automation production lines in China are restricted by the risk factors that sticking and ejector cannot be recovered during the forging process, and it is often necessary to manually judge whether the workpiece position satisfies the continuous processing conditions before machining. To solve this problem, a forging location detection system based on machine vision was proposed, and the industrial CCD camera was used to replace the human eye to determine whether the workpiece position met the requirements of plastic grasping for robot grabbing or pressing machine. Then, the hardware composition, detection principle and detection algorithm of the detection system were described, and a method for detecting whether the hydraulic ejector bar could be recovered without rigid connections designed. The experimental results show that the single detection time of the detection system is less than 500 ms which has no missing result, and the error rate is less than 2%. Furthermore, the causes of false results were analyzed. And the system is of high detection speed and high rate of accuracy. Finally, by applying the above system, the number of detection operators reduces 2-3.