Automatic infrared calibration based on physical guiding

Infrared thermal imaging detectors often suffer from temperature measurement errors due to various factors during operation, posing challenges for efficient and real-time temperature correction. Traditional calibration methods based on blackbody reference sources and linear models, while simple to operate, have significant limitations. On the one hand, they are constrained by the radiation range of blackbodies and the linear assumption, making it difficult to accurately calibrate the full response area of the detector. On the other hand, the calibration process relying on manual intervention is not only time-consuming and laborious but also susceptible to dynamic factors such as environmental temperature fluctuations and device heat accumulation, making it difficult to adapt to the needs of automated monitoring scenarios. To address this, we propose a multi-physical element coupling calibration algorithm based on deep learning. By constructing a high-dimensional joint feature space of integration time, blackbody temperature, and environmental parameters (such as ambient temperature, camera temperature, etc.), we utilize a multi-branch position encoding and conditional guidance mechanism to achieve dynamic modulation of physical parameters on network features. A hierarchical residual denoising module is introduced to suppress non-uniform noise, and deep separable convolution is used to achieve pixel-level response correction. A multi-task optimization strategy involving radiation field reconstruction, Planck's law constraints, and total variation loss is designed to force the reconstruction results to conform to infrared physical laws.