Research On The Method Of Exposure Fusion For Aerial Visible Light Camera Based On Estimating And Adjusting Exposure

As an important aerial photoelectronic imaging device,aerial visible-light cameras can acquire a large amount of real and exhaustive information on ground features,thus playing an important role in aerial reconnaissance.Due to the influence of weather,solar altitude and ground scenery change,the complexity of scene brightness and exposure inaccuracy occurs frequently when an aerial visible-light camera takes photographs for ground features.The existing automatic exposure method based on average gray value for engineering use can't meet the demand of aerial reconnaissance in a complex scene.Automatic exposure is the first step of image acquisition and is vital to later image processing.It is of great research value and application significance to study the method of automatic exposure in a complex brightness scene.When the dynamic range of a scene is much larger than that of a camera,automatic exposure can't help the photographed images fully reflect the scene brightness levels.In this case,it is necessary to fill the gap in automatic exposure by means of image fusion.This paper explores the principle of camera imaging in depth,analyzes the exposure function of CCD imaging system,discusses the exposure fusion process and its key points,and focuses research on the automatic exposure method in a complex brightness scene and the exposure fusion method in a high dynamic range scene.The main research efforts of this paper are as follows:1.The scene is constantly changing when the aerial camera taking photographs.To solve the problem that the method of average gray value can't obtain an image with accurate exposure in a complex brightness scene,this paper proposes an automatic exposure method based on histogram analysis.First,conduct initial exposure according to the information near the gray levels 0 and 255 in the gray histogram of the current frame,in order to avoid histogram overflow and keep the exposure basically normal.Then analyze the gray histogram information,and calculate the proportion of non-ROI region in the image(including the regions with little gray level and less information).Assign different weights to the non-ROI region by using the weight distribution curve according to the non-ROI proportion in the whole image,so that the photometry can focus on the region of interest(ROI).Finally,calculate the weighted average gray value,and use it as a feedback variable to control the exposure of the next frame.2.To meet the volumetric and real-time requirements of an aerial visible-light camera,FPGA is chosen as the hardware platform of automatic exposure algorithm owing to its merits such as repeatable programming,strong real-time performance and small volume.The automatic exposure algorithm proposed by this paper is moved,module by module,to FPGA hardware platform,and then various modules are invoked through the top module for joint simulation.The simulation result shows that,in both dark and bright backgrounds,the proposed method can improve the information entropy of ROI effectively compared with traditional methods.Moreover,it can meet the real-time requirement by processing a 1024 ×768 image in 15.76 ms at the pixel clock of 50 MHz.3.To solve the problem that an image with a variety of brightness levels can't be obtained through automatic exposure in a scene with a high dynamic range,this paper proposes the use of exposure fusion to fill the gap in automatic exposure.As the vibration of an aerial camera shooting multi-exposure images is unavoidable,the camera must be registered before fusion.This paper analyzes the features and registration requirements of aerial multi-exposure images,adopt SURF for extracting and descripting the feature points.Euclidian distance is used as similarity measure and K-D tree is used as searching strategy.RANSAC is used for elimination of mismatches.Select affine transformation as the spatial transformation model.4.In order to make up the shortcoming of automatic exposure in high dynamic range scene,this paper proposes an exposure fusion method combined with brightness analysis in complex wavelet domain.First,analyze the relationship between image brightness and exposure,and set the function of 3 assigned weights.Then use the Dual-Tree Complex Wavelet Transform(DT-CWT)to decompose multi-exposure source images into low-frequency and high-frequency components.For low-frequency components,assign their weights according to the brightness analysis result,and then average the weights to obtain the fused low-frequency components,thus brightness of result is generally accurate and moderate.Use both the absolute value of high-frequency coefficient and the weight of the corresponding low-frequency components to fuse high-frequency components so as to effectively keep the detail information of source images and reduce the impact of noise.Finally,the fused image is obtained by subjecting the low and high frequency coefficients to inverse DT-CWT.In conclusion,this paper analyzes the theories for capturing an image with accurate brightness through an aerial visible-light camera,and studies key issues about automatic exposure method and exposure fusion in complex brightness scenes.It has made some achievements that can be used for reference as a theoretical basis for capturing an image with various brightness levels through an aerial visible-light camera.