Research On FPGA-based Bispectral Image Fusion Technology

For the weapon system based on the laser guidance technology,the accuracy of laser pointer is an important factor that affects the performance of weapon system.Therefore,it is very important to obtain the laser spot position accurately for the performance evaluation of the irradiator itself and the comprehensive performance improvement of the weapon system.First of all,this topic compared and analyzed several hardware platforms for image processing,selecting the ZYNQ-XC7Z035 as image data processor of bispectral spectrum image fusion system,and writing driver to realize the acquisition of laser spot and the image of target board by the bispectral spectrum camera.This research realized the fusion and display of infrared image and visible light image on the architecture of ZYNQ system "ARM+FPGA".Secondly,the non-coaxial telephoto optical system was simply elaborated,which has the problem of non-coaxiality of the dual-spectral camera and various errors of in the processing and assembly of the measurement system.To solve the above problems,the SIFT algorithm was used to complete the registration of bispectral images,which provides the prerequisites for the output of high-quality fusion images.The next,the image fusion algorithm has been deeply researched.A variety of image fusion methods that can run on FPGA platforms and their performance are compared,which was simulated by MATLAB software.Combined with the characteristics of the FPGA platform and the needs of the subject,the subject decided to use the lifting wavelet transform algorithm to achieve image fusion.The subject improved the hardware implementation structure of lifting wavelet transform image fusion algorithm,using the advantages of Vivado HLS software that short development cycle simple operation,strong code portability and low algorithm development difficulty,using the C/C++ language to complete the writing of the core algorithm,to realize hardware platform fusion image output based on ZYNQ.Finally,based on the bispectral image fusion system built in this paper,an image fusion experiment with a distance of nearly one kilometer was carried out.The obtained fusion image was compared with the fusion image obtained by using the PC-side MALAB software.The comparison results show that the image detail information in the subjective evaluation is basically similar,and the objective evaluation data are close to each other,which fully proves that the quality of the fusion image output by the system meets the needs of the subject.