Design of inertial measurement unit aided infrared imaging system

Infrared (IR) imaging systems are frequently used in civil and military imaging applications. Infrared image sensors are used in the focal plane array to detect electromagnetic emissions in the wavelengths of the infrared region of the electromagnetic spectrum. Infrared radiation, which is focused through various optical elements and fallen onto the focal plane array, is absorbed by the infrared sensor during a certain exposure time, and is then read as an analog electrical signal by using the readout circuits. One of the conditions required for the IR image to not be blurred is that the relative motion between the imaged object and IR imaging system must be zero or remain within an acceptable limit, during the integration time. The motion of the imaged object or the IR imaging system during the integration time will cause blurring of the IR image. Within the scope of this thesis, the blur caused by the IR imaging system movement has been examined and modeled. In order to quantify the blur, an inertial measurement unit is used to measure the movement of the IR imaging system. Thus the software for calculating PSF for each pixel was developed using the IMU data of the movement of IR imaging system during the integration time. After calculating the blur due to the IR imaging system movement, the resulting blur is tried to be eliminated by using image processing techniques. In order to eliminate the blur, image processing techniques used in the literature were examined and simulations were made about their performance. In order to get the applicability of the obtained deblurring algorithms on FPGA in real time, the necessary hardware and software resource were tried to be determined. Deblurring algorithm was implemented in the Xilinx System Generator environment using FPGA IP cores for real-time working on FPGA. In case of use of SDRAM as a memory in real-time image deblurring, the problems that will arise due to matrix transposition to be used in 2D-FFT application are examined. Using this memory, it was found that image transpose would be long enough to slow down the entire real-time image processing flow. The Corner Turn Matrix method is designed in the System Generator environment for faster transpose of the image and more efficient use of SDRAM memory. Deblurring algorithm designed in System Generator environment is simulated and the results are compared with MATLAB results. Some critical sub-IP cores of the deblurring algorithm have been tested in real time using the development board that carries the Xilinx Zynq SoC (Sytem-on-chip). MATLAB simulations were performed for other deblurring algorithms which were not applied in real time. As a result of the thesis study, the blur caused by the movement of the IR imaging system in a scenario during the integration time can be calculated quantitatively. Then this blur can be eliminated by using the deblurring algorithm designed in the System Generator environment if it meets certain requirements. For some types of blur, the performance of the System Generator application may remain low. For these types of blur, images can be clarified by using the algorithm based on the iterative least squares method designed in MATLAB.