Research On Motion Blur Restoration Technology Of UAV Visual Inspection Image

In recent years,the UAV visual inspection have been increasingly utilized in various applications.With the developments of applications,the requirements for the quality of visual inspection images are getting higher and higher.However,when there is relative motion between the UAV platform and the detected target during the exposure time,the quality of the captured image is degraded due to motion blur which affects target recognition and measurement accuracy.The use of high-precision image stabilization equipment can suppress the influence of high-frequency vibration,but it is difficult to solve the problem of image blur caused by airflow disturbance.In view of this,this thesis is devoted to researching motion blur restoration methods for UAV visual images.The main contents herein of this thesis is summarized as follows:1)Establishment of motion blur model.In this thesis,the causes of motion blur from the physical mechanism are firstly analyzed,and the mathematical model is established to describe the process of image degradation with a point spread function.Secondly,the various methods of image restoration model based on image degradation model are described,and the advantages and disadvantages of each method are compared.2)Research on image restoration method based on the spectrum analysis.Aimed at the small scale motion blur in the conventional UAV road inspection,the feasibility of simplifying the three-dimensional camera motion into two-dimensional motion is firstly analyzed.Secondly,two point-spread function models are established for linear motion blur and rotational motion blur respectively.In addition,the information of point-spread function is estimated using spectrum analysis method.Finally,the traditional method is exploited to restore the blurred image.The experimental results show that this method has a good effect on small-scale motion blur restoration,but the performance in terms of large-scale motion blur and composite motion blur restoration is not satisfactory.3)Research on convolutional neural network image restoration method based on motion information.Aimed at the problem of large-scale motion and composite motion blur,the mapping relationship between three-dimensional camera motion and two-dimensional pixel motion is firstly derived.This mapping relationship is used to calculate the motion of the pixels in the image.Furthermore,the convolutional neural network based on motion information is designed.This network provides the motion constraint to each pixel,which can effectively solve the problem of different pixel motion scales in composite motion.Secondly,in order to overcome the problem that the existing training set cannot provide motion information,the method for generating blurred images based on camera motion is proposed.Finally,experiments are performed using simulation images and real-world images,and the results show that the method can effectively solve the blurring problem caused by composite motion and improve the accuracy of UAV visual inspection.