Research On Motion Blur Removal Algorithm For Aerial Image

Aerial image refers to the aerial target information obtained by aerial vehicle which photographs the ground objects in the air by carrying photographic equipment.It is widely used in many important fields such as landform mapping,military reconnaissance and so on.Therefore,the research on the clearness of aerial images has very important practical significance.This paper studies the motion blur problem in aerial images from two aspects:the traditional algorithm based on image degradation model and the depth algorithm based on neural network model.The research process of traditional algorithm is as follows: Firstly,the image motion length and the image motion angle of motion blur kernel are estimated based on the theoretical analysis of aerial image imaging characteristics;Secondly,combined with the visual and spectral characteristics of aerial image and the theory of guided filtering,a non blind deblurring process suitable for aerial images is designed,and a motion deblurring algorithm suitable for embedded devices is proposed;Then,on the basis of the above algorithm,this paper uses DSP development board to build and implement an aerial image deblurring system,which is simple in structure,convenient in operation and suitable for the embedded DSP platform with limited resources;Finally,from the subjective and objective evaluation,the effectiveness of the proposed algorithm and the integrity of the designed system function are verified by algorithm verification and system verification.The research process of depth algorithm is as follows: According to the feature of abundant edge information in aerial images,convolution block attention mechanism is introduced into the conditional countermeasure generation network,and the network structure of overlapping residual block and attention mechanism is adopted to enhance the learning ability of the network for edge information.On this basis,this paper collates and makes five sets of aerial image datasets,which are used to train the proposed algorithm.Then,the results of each aerial datasets in the proposed algorithm are compared and analyzed,and the results of the proposed algorithm and other algorithms in aerial datasets and ordinary datasets are compared respectively.Through subjective and objective evaluation,it is proved that the proposed algorithm has a very significant deblurring effect for aerial images.Finally,an real motion blurred dataset of aerial images is captured by UAV,and the subjective evaluation verifies the effectiveness of the proposed algorithm for real motion blurred images.