Motion Blur Image Of Inspection Robot Restoration Research

Regular inspections of substations are an important means to ensure the safe and effective operation of the power system.The investment of inspection robots greatly reduces the difficulty of manual inspections,eliminates safety hazards in manual inspections,and improves work efficiency.Obtaining high-quality video images by the inspection robot vision system is the prerequisite and key to complete the inspection task of the inspection robot.However,complex working roads and collision obstacles will jitter,which makes the inspection robot’s vision system produce inevitable motion blur images,which seriously affects the subsequent detection work.Therefore,it is very important to remove motion blur and restore the sharpness and details of the captured image as much as possible.Filtering the motion blur image from the video image of the inspection robot is the prerequisite for subsequent deblurring.The artificial judgment screening is time-consuming and laborious and there is the possibility of misselection;it is estimated that the blur kernel of the motion blur image is the core and foundation of deblurring.Its accuracy will directly affect the quality of image restoration;the ultimate goal of deconvolution restoration to obtain high-quality images,inefficient convolution algorithms often lose the edge details of the image,and even produce ringing effects.In view of the above problems,the restoration methods of motion blurred images are studied:First,the reason for the motion blur image is analyzed,the degradation model of the motion blur image is established and the restoration process is determined.The image quality evaluation is applied to the motion blur determination,and it is found that a perceptual,no-reference objective image clarity method can correctly evaluate images with different backgrounds and different degrees of blur,which is in line with the application background of inspection robots;this article combines experiments For verification,it is proposed to set the definition threshold as the criterion for the fuzzy image.On this basis,the motion blurred image is screened out by combining the characteristics of the motion blurred image spectrogram with parallel dark stripes.The determination of fuzzy screening and motion blur greatly reduces the possibility of false screening and misselection,and lays a solid foundation for subsequen restoration.Then,in order to improve the quality of the blur kernel estimation,this paper first uses an improved three-dimensional block matching algorithm to denoise the motion blur image,and uses guided filtering to enhance the edge details of the image;the image is reconstructed using the initially estimated blur kernel,based on the reconstructed image Continue to estimate the fuzzy kernel,modify the accurately estimated fuzzy kernel after repeated iterations,and use the super Laplace algorithm to restore it.Finally,according to the super Laplacian algorithm,the local statistical properties of the image are ignored,and the restored image often loses edge details and even produces ringing effects.The concept of dictionary learning is introduced,and based on the fuzzy kernel estimation,it is combined with the super Laplacian algorithm to restore the image together.After experimental verification,the restored image effectively reduces the ringing effect and improves the image quality.