Design And Implementation Of Aerial Image Recovery Algorithm

In recent years,UAV(Unmanned Aerial Vehicle)has been widely used in resource exploration,battlefield detection,biochemical detection and other fields because of its ability to fly in complex environment and obtain ground image at low altitude.However,in the process of image acquisition,it will be affected by motion blur and noise,so that the captured image is blurred.In addition,when the UAV flight to a certain height,it will lead to the acquisition of aerial images in the pixel point containing a large scene size,low image resolution.Aiming at these two problems,this thesis puts forward an aerial image recovery algorithm.The main work content is as follows:Aiming at the problem that UAV aerial image is disturbed by motion blur and noise in the imaging process,an aerial image debluring algorithm based on low rank matrix recovery and normalized sparse priori is proposed.Considering the influence of noise on aerial image,the algorithm decomposes the aerial image into R,G,B three channels and carries on the low rank matrix recovery processing to the three channels respectively to suppress the influence of noise on the aerial image.Normalized sparse prior is introduced as a regular term to deal with aerial image debluring.Experimental results show that the image quality recovered by this method is better.Aiming at the problem of low resolution of UAV aerial shooting image,an improved superresolution reconstruction algorithm of aerial image based on neighborhood embedding is proposed.Considering that aerial images have abundant diagonal features,this method uses wavelet transform combined with first-order two-step feature to extract the feature of lowresolution image patches.It uses the method of low rank matrix recovery to process the matrix composed of eigenvectors in order to remove the irrelevant noise.High resolution images are obtained by neighborhood embedding.Experimental results show that this method is obviously superior to the traditional method in objective index and has a good effect on the recovery of diagonal details.An aerial image deblurring algorithm is implemented on TMS320C6748 platform.Firstly,the aerial image deblurring algorithm is transplanted to the DSP platform based on the development process.Secondly,the standard C code is optimized in the light of the characteristics of the DSP.Finally,the experiments are carried out in the integrated development environment of CCS with hardware simulation tools.The experimental results show that the effect of aerial image processed by DSP is basically the same as that simulated by Matlab.