Research On Multi-source Image Perspective Restoration Technology In Haze Environment

In recent years,the application of outdoor vision systems such as road video surveillance and defense military surveillance has become more and more widespread with the rapid development of image processing technology and computer vision.However,the outdoor vision system is very sensitive to environment,especially in haze environment with low visibility,as a result,the system cannot obtain complete,accurate and clear video images.In order to solve this situation,multi-source perspective restoration technology was used,this technology processes multi-sensor acquired multi-source images to obtain a more complete,clear,and accurate new image by using a series of image processing techniques.This article uses visible light images and infrared images as multi-source images.The visible image can express the details very well,such as shape,color and texture of the target,but the clarity is low,and some target bodies may be lost due to the influence of haze.The infrared image is the expression of heat radiation so it has a strong ability to penetrate the haze,also can show the target body clearly.However,it will lose the details in image.Therefore,this paper pretreats the visible image by removing haze firstly.During this process,the method based on the improved dark channel prior algorithm is used,then we can obtain the visible haze-free image.The second step is the registration of the visible haze-free image and the infrared image through an improved SURF algorithm.Finally,the registered images will be fused,by the method based on the improved weighted average fusion algorithm.From the simulation results of the algorithm,the new image obtained by fusion can show the advantages of the both two images and achieve the optimal state of the image,which proves the feasibility and effectiveness of the proposed algorithm.Based on the research of multi-source image perspective recovery technology used in this paper,this paper builds a hardware development platform based-on FPGA——multi-source image perspective recovery system based-on FPGA.This paper adopts a modular design method and pipeline technology,and completes the FPGA design of the system through algorithm analysis,algorithm optimization,and algorithm verification and other processes.From the result of hardware simulation and system debugging,the whole system can achieve the purpose of this design.It is also stable and has high real-time performance.