| UAVs carry photoelectric payloads for aerial photography on the ground,and have important applications in geographic mapping and intelligence reconnaissance.However,due to the adverse effects of weather conditions in the process of aerial photography by drones,especially the haze weather,the image quality will be greatly reduced.Therefore,the video taken under the haze weather conditions will be defogged and enhanced to improve the image quality.Improving the all-weather work of UAVs is of great significance.The field of view of a single image shot at the same time is limited.The field of view of the video or image is determined at the time of shooting,and a wider field of view cannot be presented at the same time.Therefore,the captured dual-light video images are integrated,and the redundant part of the image is spliced into The panoramic image and the fusion of the visible light image and the infrared image into one image are conducive to the analysis of information.However,due to the high-resolution imaging of aerial photography equipment,the storage capacity of a single image and the richness of information are increased,and real-time defogging enhancement and splicing and fusion processing are more challenging.Based on this,this article has done the following work:1.Aiming at the problem that the imaging quality of aerial images is affected by haze weather conditions,this thesis proposes a real-time defogging algorithm for aerial video.The algorithm analyzes and estimates the area where the atmospheric light value is located by combining the bright channel image with the mean filtered image of a specific window size.Secondly,it combines the dark channel image with the dark channel image after Gaussian convolution to refine the transmittance image to ensure that the Filter out the noise in the transmittance image while retaining edge information.Finally,the optical flow is used to track the A-value pixels to eliminate the flicker of the defogging video.2.Aiming at the problem of limited view of aerial images and redundancy of shooting information,this thesis proposes an image stitching and fusion algorithm based on airborne information.First,according to the onboard information,the coordinate transformation between the earth coordinate system and the camera coordinate system is used to calculate the latitude and longitude of the four corners of the image.Calculate the overlapping area of two adjacent images under the guidance of the latitude and longitude of the image corners.According to the overlapping area,the image area for feature registration can be greatly reduced and the stitching time can be shortened.At the same time,the perspective transformation matrix between infrared and visible light images can be calculated according to the latitude and longitude of the corner points,and the fusion of visible light images and infrared images can be further completed based on this.3.According to the proposed real-time defogging and image stitching fusion algorithm,supplemented by other image processing and recognition functions,a software system for UAV intelligence processing is designed.The software system includes basic image processing modules,such as image removal.Fog,image enhancement,brightness and contrast adjustment;stitching and fusion module,which performs panoramic stitching and fusion of input images;target recognition and tracking module,which uses deep learning algorithms to identify specific targets in aerial video images and track moving targets.The three processing modules can adapt to the three image formats of visible light image,infrared image and SAR image. |
PDF Full Text Request