Scene-aware adaptive labeling template generation method and system

Abstract

The present application relates to the technical field of data analysis, and is a scene-aware adaptive labeling template generation method and system, specifically comprising: constructing a dynamic grid visualization framework of a perception node to which a target to be labeled belongs in a high-temperature industrial scene; obtaining historical multi-modal perception information data of the target to be labeled in the high-temperature industrial scene, and real-time multi-modal perception information data of the target to be labeled; constructing a machine learning model, performing prediction calculation of the asymmetric confidence of the real-time multi-modal perception information data, and performing quality level judgment of the labeling template generation; generating a corresponding labeling template generation strategy identifier according to the labeling template quality level, and outputting an adaptive labeling template. The present application solves the problem of labeling template distortion caused by the formation of a non-stationary refractive index field due to a severe temperature gradient or pressure gradient in the prior art.

Images