Research On Real-time Panoramic Remote Sensing Imaging And Image Processing Technology For UAV

Because UAV remote imaging systems are light weight,small,inexpensive,reusable,fast highly mobile,low risk have low loss and high resolution,they have been successfully applied by the military,such as in reconnaissance missions and in early warning systems,and for resource exploration,emergencies and meteorological oceanography.Measurements in other non-military fields have broad application prospects.Due to the influence of flight altitude and camera focal length,single UAV remote sensing images cannot cover the required large field of view.Currently,UAV panoramic remote sensing imaging mainly adopts two technical systems: image mosaics and fisheye lenses.The major advantage of panoramic imaging based on image mosaicking is its high imaging resolution.However,due to the registration process before mosaicking,the imaging speed is slow,the real-time performance is poor,the amount of data is large,the mosaic algorithm is cumbersome,and the cost is high.Fisheye lenses can obtain images with large fields of view around the hemisphere,but they also produce serious and difficult to correct barrel distortion and have low imaging resolutions,complex system structures and are expensive.This paper focuses on real-time panoramic remote sensing imaging and image processing technology based on a micro-UAV platform.Based on optical imaging,UAV kinematics,remote sensing,image processing and other theories,the research is structured as follows:(1)A catadioptric real-time panoramic remote sensing imaging system for a UAV is established.Based on the principle of catadioptric real-time panoramic imaging and related concepts,a catadioptric real-time panoramic optical imaging system mounted on a micro-UAV platform is designed by using a parabolic reflector.The spherical panoramic image is projected into a cylindrical image.The adaptive wireless transmission system and the transmission method of UAV-borne aerial image based on the data buffer occupancy ratio are studied.(2)The image denoising algorithm is studied.Because the image will inevitably be affected by noise in the process of acquisition and generation,it is necessary to denoise the image to improve the image quality.A series of new image denoising algorithms based on singular value decomposition(SVD)and wavelet transforms(WT)are proposed based on the in-depth study of the multi-resolution characteristics of a wavelet transform and the row-column characteristics of singular values.Specifically,the following algorithms are utilized: 1)an image denoising algorithm using wavelet transform and singular value decomposition to enhance directional features,2)an image denoising algorithm based on the singular value difference in the wavelet domain,and 3)an image denoising algorithm based on block-rotation singular value decomposition in the wavelet domain.(3)The restoration algorithm of the blurred image is studied.To solve the problem of blurred image distortion caused by platform motion,the point spread function of the blurred image is estimated by extracting the edge function or the singular value function of the blurred image,and the blurred image is restored by a Wiener filter with an optimal window.The final high-definition and high-resolution restored denoised image is obtained.(4)The abovementioned processing methods are applied to land/water segmentation,and the water surface and land are recognized by gray-scale features.Based on a UAV real-time remote sensing imaging device and image processing technology,an adaptive atmospheric horizontal visibility detection system is constructed.A theoretical model of the atmospheric visibility calculation is given,and the hardware and software mentioned above are verified.The results show that the real-time panoramic remote sensing detection imaging equipment and the subsequent image processing technology can obtain real-time,omni-directional and high-quality remote sensing images.The UAV real-time panoramic remote sensing imaging system and its image processing method are the basis of meteorological and marine environment monitoring applications.They can be used to monitor the meteorological and hydrological support information related to weather forecasts and military operations,such as all-sky clouds,fog,waves and tides.It can also be applied to manned aerial vehicles,micro air vehicles,robots and other platforms,as well as security monitoring sites,such as customs departments and banks,to form a real-time omni-directional monitoring system.The research results can lay a technical foundation for acquiring real-time,high-quality UAV images with meteorological and hydrological information imaging equipment and provide a technical approach for the realization of real-time panoramic remote sensing imaging systems of micro UAVs.