GPU-based infrared sensor model capable of operating in real time

This work describes a Graphic Processing Unit (GPU) based Infrared (IR) Sensor Model Software capable of operating in real time which has been recently developed by us. This module, which is based on physical models, is capable of simulating the images registered by an infrared imager. The software is also capable of accounting for the image degradations (blur and noise) by applying these effects on synthetic images generated by an IR Scene Generator Software at high frame rates. In this module, in order to enable processing of large format images at high frame rates, the parallel computations are carried out on the GPU. In this work, we achieved a performance increase when compared to the results of an earlier and similar work presented by us. In order to decrease the execution time, some routines (especially those pertaining to between GPU with CPU data transfer) have been improved. In order to determine the runtime performance of the module, the module has been run in a closed loop where IR Scene Generator and Target Detection / Tracking Modules also exist and maximum achievable frame rates corresponding to different image sizes have been recorded. We believe this study will support the efforts for developing closedloop simulation systems employing the method of signal injection.