Energy storage power station safety monitoring method and device based on three-dimensional point cloud and voiceprint, and medium

The invention discloses an energy storage power station safety monitoring method and device based on three-dimensional point cloud and voiceprint, and a medium, and the method comprises the steps: carrying out the noise reduction of input image data through employing an improved K-SVD algorithm, transmitting a video image to a YOLOv8 model improved by an EMA attention mechanism, and carrying out the image processing, the infrared three-dimensional point cloud is transmitted to an LSKA large separation convolution improved YOLOv8 model for training, and final image information and infrared three-dimensional point cloud imaging information are obtained; a voiceprint sensor collects sound data, voiceprint features are extracted by using an artificial neural network algorithm optimized by an EVO algorithm, the processed data are input into an improved Transform model for voiceprint processing, and final voiceprint data are obtained; and fusing the Transform model and the YOLOv8 model by adopting a low-rank multi-