The invention discloses an
infrared thermal
imaging array real-time optimization method based on an AI model, relates to the technical field of
infrared imaging optimization, and is used for solving the problem that a real-time image optimization path is not clear. According to the method, a frame-level
state vector and a region-level
feature set are used as input, a region-level module activation
mask is dynamically generated,
exposure time, a
dynamic range mapping curve and a confidence coefficient identifier are output through an AI model, parameter gating and backspacing are carried out under the constraints of a
frame rate,
delay and
power consumption, and a region-level module activation
mask is obtained. Non-uniformity correction, motion compensation space-time
noise reduction and detail enhancement are selectively executed according to regions, a
processing path and an
image quality index are recorded, then non-interrupted self-supervised
fine tuning is triggered under an approximate field-averaging condition, a
fixed pattern noise template and related parameters are updated, and the
image quality is improved. Therefore, key area priority and low-value area early exit are realized, redundancy calculation is reduced,
infrared imaging contrast and
dynamic range utilization rate are improved, and
temperature measurement consistency and long-term stability in multiple scenes are enhanced.