Research On Multi-exposure Fusion For Missile-borne Image Recorder

In recent years,with the development of China’s aerospace industry,the demand for image acquisition in the development of missiles and rockets is increasingly extensive,which plays a significant role in monitoring the flight status of missiles and the external environment.Under complex lighting conditions,it is difficult for the image sensor to satisfy the wide dynamic range of the test scene because overexposure or underexposure will appear by a single exposure setting.Based on the existing experimental conditions,in order to adapt to the high-speed motion scene of the missile-borne image recorder,this paper considers the requirements and problems comprehensively to propose a combination of stereo matching parallax fusion multi-sensors and multi-exposure fusion algorithm.The fusion result is a high dynamic range(HDR)image.The multi-exposure fusion algorithm needs to input the multi-exposure image sequence.When shooting a moving object,the faster the object moves due to the exposure gap,the more serious the ghosting phenomenon of the fused image will be.HDR image acquisition of moving objects is a natural defect of multi-exposure fusion algorithm.In order to solve this problem,this paper firstly starts with image acquisition,abandonsthe previous mode of time-sharing multiple exposures with a single sensor,and instead adopts a single exposure with multiple sensors at the same time.Three OV4689 phase modules with the same parameters are arranged in parallel to form a new matrix phase module,which ensures the consistency of the world coordinates of the measured scene in time dimension.In order to correct the trinocular camera module and eliminate the installation error to ensure the camera coplanar and alignment,this paper proposes to obtain the internal and external parameters of each camera first by using Zhang’s correction method,and then to complete the scene image correction by using Boguet stereoscopic calibration algorithm.In view of the inconsistency of coordinates between corresponding pixels caused by different camera coordinate systems,this paper proposes to integrate the parallax information of multi-eye vision into the multi-exposure fusion algorithm.The improved HAD-Census stereo matching algorithm is used to accurately obtain the parallax of the measured scene under different exposure conditions.In this paper,the weight map is constructed by quality factor evaluation and the idea of image pyramid decomposition is adopted to select the well-exposed pixels from the multi-exposure image sequence in the form of multi-resolution reconstruction.In this method,the multi-exposure fusion algorithm and stereo matching algorithm are integrated,and the corresponding parallax is eliminated when the pixels are fused to obtain the final HDR image.In fact,the relationship between the distance of the measured object and the image Angle should also be studied to analyze the influence of multiple perspectives on the multi-exposure fusion results.In this paper,static scenes and dynamic scenes are tested respectively,and multi-exposure image sequences and HDR images obtained by fusion are compared and analyzed according to gray histogram.The results show that the dynamic range of the fused image is wider,each pixel is well exposed and there is no obvious ghost phenomenon.Under the experimental conditions,the moving pedestrian in the dynamic scene is compared with the flight process of the projectile image,which proves the scientificity and feasibility of this paper.