Abstract: The relationship between the staff and the perimeter of the substation is changeable, and the workload of monitoring and maintenance is heavy. Aiming at the problems of strong system interference, high false alarm rate and missing alarm rate, and low image clarity of violations, a three-dimensional virtual perimeter monitoring method for digital violations of the substation based on machine vision is studied. The laser and camera are installed in the operation area of the substation to build a closed loop area. The laser alerts operators' violations and captures the images of operators' violations. After image contrast enhancement, segmentation and matching, the static violations are transformed into three-dimensional dynamic images. After convolution processing of the convolution layer, the characteristics of three-dimensional dynamic violations are strengthened. Through sub sampling layer, pooling layer after dimension reduction, compression and integration of the characteristics of the full connection layer, the Softmax layer is used to identify violations of operators and monitor the perimeter of violations of operators. The experiment proves that the perimeter detection device has the advantages of high sensitivity, long detection distance, strong anti-interference, low false alarm rate and missing alarm rate, and digitally presents the monitoring results to protect the personal safety of operators.