This research offers an innovative strategy for monitoring the status of panel indicator lights in intelligent power plants. We utilize the YOLOv5face model to locate panel indicator lights, overcoming the issue of positioning. Faced with the challenge of sparse abnormal statuses of panel indicator lights in power plants, we propose an original image synthesis solution to address the problem of data imbalance. Considering the potential irregularities of images during the implementation of security monitoring, we conduct necessary image corrections. Following the four-corner positioning of the YOLOv5face model, we trim the corrected images and then segment map them into 0-1 matrices.