Currently

the anti-collision safety modules of industrial robots are all contact modes

which highlight the shortage of personnel protection. In order to further prevent incidents of personnel injury during industrial robot operations

a monocular vision safety assistant system is designed

which can provide non-contact safety warning for industrial robots. The system uses the monocular vision ranging to measure the distance between persons and device

and identifies persons by using Face Recognition library

and then completes the security determination by the obtained valid information. Considering that the working environment of industrial robots is often accompanied by interference brought in by high-power devices

the system adopts median filtering and Wiener filtering to deal with the pepper and Gaussian noise caused by interference for ensuring the accuracy of distance measurement and person identification

a set of safety strategy is proposed to fit multiple applications at the same time

which is combined with the running states of industrial robot. The test case shows that the system can make security predictions based on the set of security range

and the selection of safety strategy level and processed image information can provide effective safety warning for industrial robots control systems.