Research On Low-light Digital Camera Imaging Metho

Low-light-level night vision technology is a technology that uses photoelectric imaging to help overcome the limitations of the human eye in low light conditions,and can enhance the recognition ability of the human eye at night.This technology has played a huge role in military,industry,scientific research and other fields,especially in the military,the demand for this technology is a huge driving force for the development of low-light night vision.In this paper,two sets of digital camera imaging systems are developed based on CMOS image sensors,and on this basis,improvements and implementation schemes of low-light enhancement algorithms are proposed to meet the requirements of low-light imaging.The thesis first investigated the development status of low-light-level night vision technology and digital industrial cameras.On this basis,the two imaging system schemes of this article were determined based on the principle of the yellow duck imaging system.The ISP chip and the FPGA chip are used as the core image processing devices.Select the image sensor,image buffer chip and equalization chip model on the basis of the system scheme,and complete the drawing of the schematic diagram according to the chip manual.The program design of high-definition video signal transmission is completed by analyzing the interface standard,transmission format,encoding format of the SDI video signal,the principle of the GTP transceiver and the reference clock.On this basis,the basic function of the camera module is debugged,including the FPGA drive sensor,ISP drive sensor and interface board output debugging.So far,the hardware circuit design is completed.In order to meet the requirements of low-light-level imaging,the low-light-level enhancement algorithm is transplanted to the system with FPGA as the core image processing device,and the algorithm principle is improved according to the parallelism and limitations of FPGA.Algorithms include Bayer interpolation,automatic white balance,color space conversion,histogram equalization and 3D noise reduction algorithms,and provide RTL views and resource analysis.Finally,the two low-light level digital cameras developed were analyzed for imaging results and lateral comparison experiments.The results showed that both digital cameras can perform real-time imaging under 10-2lx illuminance,which meets the requirements of low-light level night vision imaging.The horizontal comparison shows that the root mean square error and structural similarity of the image after ISP processing are larger than those after FPGA processing,and the peak signal-to-noise ratio is smaller than that after FPGA processing.It shows that FPGA is better than ISP in denoising and brightness,and ISP processing is better in color saturation and contrast.