Abstract: As it is all known, it is difficult to measure high temperature directly in complex industrial environment. Thus, a new temperature soft-measuring method based on Support Vector Regression (SVR) was proposed. SVR model was used to fit the complex nonlinear mapping relationship between the feature values of color images of the high temperature object and its temperature. And then the trained model could predict the temperature by inputting the features of colorimages. Simulation results demonstrate that the improved algorithm has excellent generalization ability and predictive ability. What’s more, this model needs less support vectors and learns faster.

Key words: SVM regression, nonlinear relationship, modeling and simulation, temperature soft-measuring