Design And Implementation Of Military Night Vision System Based On DSP

Video image shot in nightfall or night is not easy to observe and distinguish because of details covered by a large number of connected dark areas. The paper researches image real-time enhancement techniques and designs a low illumination video image processing system which meet the requirement of individual soldier night vision equipment based on DSP.The article analyzes the degradation reason of night video image and deeply researches basic principle of gray stretching algorithm, histogram equalization algorithm, using dark channel prior enhancement algorithm and single scale Retinex algorithm. Codes of full algorithms which is mentioned in preceding part are given out on personal computer with Open CV library and the advantages and problems in each of algorithms are analysed through processing effects of comparing single image.The paper adopts ADSP-BF609 as the hardware core of the battlefield night vision system, and the function modules of video capture, preprocessing and output are refined according to the idea of modular design. Space conversion and template convolution are completed by a pipe connected PIXC and PVP after configuration parameters of OV3640, EPPI, CH7024, PIXC and PVP. The driver of OV3640, PIXC, EPPI and other modules is programmed with CCES development environment provided by ADI. The four algorithms discussed in this paper are parallelization and implemented on DSP system, in view of the difference between DSP and PC.In order to improve the system performance, hardware resource allocation and software structure are optimized, L2 cache resource and calculated amount of master/slave cores are allocated reasonably, the double buffer and structure words between two cores communication are set up, which makes the gray stretch, histogram equalization, and single scale Retinex three algorithms achieve the requirement of real-time. Images enhanced through the designed system are improved significantly in the resolution of distant view, near view and person comparing with images obtained from ordinary equipment on the same conditions, and has great improvement in four objective evaluation indexes that are average luminance, contrast ratio, informationentropy and definition comparing with a battlefield TV equipped for combat division unit in army troops.