Research And Implementation Of Image Defog Based On Fpga

Severe weather such as rain,snow,and haze can cause severe atmospheric scattering,which can cause color distortion and lower contrast of video images acquired by shooting equipment,resulting in road monitoring systems such as the Skynet system based on the extraction of internal features of the image and the more widely used visual recognition systems cannot work properly.Therefore,in order to make these video acquisition systems work normally and stably,and restore the minutia information in the foggy image,it is indispensable to efficaciously remove the interference caused by the weather effect in the image,reproduce the key information of the acquired foggy video image,Such preprocessing is of great significance to improving the reliability and robustness of the vision system.With the advent of the era of intelligence and big data,people have put forward more miniaturization and real-time requirements for video processing equipment.Firstly,this thesis probes the considerable merit and meaning of defogging research,and illustrates the research dedication that have been made in this domain domestic and overseas particularly,along with the encountered conundrum and solving scheme in the research process.Based on the classic algorithm,this thesis has made a few ameliorations which are appropriate for hardware graft,on the basis of the modified algorithm,the overall framework and hardware graft blue print of the defogging system are given,and the working tenet and implementation procedure of the sub-modules are illustrated at length.Secondly,in order to better transplant the algorithm to the hardware,this thesis has made the following improvements to the original algorithm: 1.Use the method of thresholding the dark channel gray distribution histogram to calculate the atmospheric light value,avoiding the complicated sorting problem;2.Reduce the calculation complexity of transmittance refinement by changing the size of the filter window,and trade a certain degree of fineness for faster disposing speed;3.Use the histogram equalization algorithm to ameliorate the effect of the regained image.Due to the continuity and comparability of video images,therefore,it is necessary to further improve the algorithm that can only process a single image to make it meet the requirements of this thesis for processing video.The optimization ideas are mainly proposed in two aspects: 1.Optimize the calculation of atmospheric light value A,solve the flicker problem,and enhance the instantaneity of the calculation;2.Optimize the calculation of transmittance and boost the calculation ratio of workpiece.Finally,experiments on the Matlab platform test and verify the feasibility of the meliorated algorithm.Finally,the thesis uses Xilinx’s Zynq-7020 development kit to transplant the image defogging system,on the PL(Progarmmable Logic)side,the IP core(Intellectual Property Core)design and packaging of the key steps of the defogging algorithm and the construction of each module are implemented,while the PS(Processing System)side mainly completes the control work,including module initialization and some auxiliary program design.Observing the final experimental results and evaluating the defogging effect,resource occupancy,and time-consuming aspects,it can be seen that the defogging processing system in this thesis has achieved the expected results.Comparing the defogging system with the software implementation method,it is not only faster,but also smaller in size and power consumption,and has higher practicability.