Switch machine automatic on-off device moving contact driving depth identification method

Abstract

This invention provides a method for recognizing the driving depth of the moving contact of an automatic switch machine. The method involves acquiring the original color image of the moving contact from the side using an image sensor, then sequentially performing grayscale conversion, Gaussian filtering for noise reduction, and gradient calculation to obtain gradient magnitude and direction data. Subsequently, non-maximum suppression, dual-threshold classification, and hysteresis boundary tracking are applied to extract the real-time edge position of the moving contact. By matching and aligning with preset calibration edge data, the included angle is calculated to obtain the initial driving depth value and dynamic sway data set. The edge coordinates, depth, and sway data are then fused to construct a feature tensor, which is mapped to a Riemannian manifold space through Wasserstein distance calculation and multi-dimensional scaling transformation. Finally, the standard sample with the highest similarity is selected, and correction parameters are obtained through comparison. The initial data is then compensated and corrected to obtain accurate driving depth and dynamic sway data, improving the accuracy and stability of the recognition results.

Images