An Intelligent Construction Method for Vibration Signal Dataset Based on Optimized YOLOv5 Algorithm

The use of neural network algorithms for recognizing vibration signal recognition requires a dataset for training. However, traditional manual dataset creation is time-consuming and labor-intensive. This article proposes a computer-based intelligent method for constructing vibration signal datasets using the YOLOv5 target detection algorithm. The surveillance videos captured by the camera are processed using frame difference calculation, Gaussian filter, binarization, inflation, and contour recognition. The YOLOv5 algorithm is then used to detect targets and map the corresponding vibration event categories to the vibration signals captured by the geophone, resulting in the dataset. The experimental results indicate that this method produces a vibration signal dataset with distinctive features and enhances efficiency.