A REBPNet-based IR radiometric calibration method considering ambient temperature change

Abstract

Considering the calibration problems of the gray-level drift, complicated calibration procedures, numerous calibration points and high time cost in the infrared radiation measurement system when the ambient temperature changes, a REBPNet-based IR radiometric calibration method is proposed. Firstly, a unified database is generated by integrating ambient temperature compensation coefficient, different gray-level images, neutral density attenuators, integration times, and theoretical radiations under the deduced infrared physical model. Then, the calibration coefficients are automatically learned through the relative error backward propagation network. Furthermore, a multi-stage learning strategy is provided for the training of the complex calibration models, including the segmented training database and the learning rate setting rule. Finally, the radiometric calibration experiments are performed on an infrared radiation measurement system. Results show that the method is effective in improving the calibration accuracy when the ambient temperature changes. Additionally, the method is capable of measuring the internal stray radiation and its drift of the optical system and attenuators.