Point Set Alignment-Based Image Stitching In Unmanned Aerial Vehicle Transmission Line Inspection

With the intelligence of electric power inspection,UAV inspection has the advantages of high efficiency,flexibility and adaptability to changing environment,and is increasingly used in electric power inspection.In order to improve the efficiency and accuracy of UAV inspection,it is an important research direction to obtain accurate panoramic images to analyze the target state.In this paper,the point set alignment algorithm research of image stitching is carried out for the power inspection images taken by UAV aerial photography,and the main research contents are as follows:(1)Improved point set alignment algorithm research based on motion consistency clustering.(1)Improved point-set alignment algorithm based on motion-consistent clustering.Aiming at the problem of losing local feature information in non-rigid alignment based on the traditional hybrid Gaussian model,a clustering algorithm based on spatial motion consistency is introduced to improve the problem of losing local structure feature information when the motion vectors in multiple local spaces are not in the same direction.An improved point set matching via Motion Consistency Clustering(P-MCC)algorithm is proposed.Through theoretical analysis and simulation experiments,it is verified that the improved P-MCC algorithm has good performance and robustness in the case of data degradation in the form of deformation,outlier points and occlusion.However,in the case of data degradation in the form of rotation and noise,some feature points in the local area still have the problem of misalignment.(2)Improved point set alignment algorithm based on spatial distance information and local feature information.To address the problem that the improved P-MCC algorithm does not fully exploit the local structure feature information,and therefore is more influenced by noise and rotation in the alignment process,the weighted spatial distance is introduced to obtain the neighbourhood of each motion vector,and the weighted spatial distance is combined with shape context features,Hungarian algorithm and metric to define the scale factor of the hybrid model with global features and local features,and the improved point set alignment algorithm based on spatial distance information and local feature information is proposed.The Robust point set registration algorithm based on neighborhood and local structural features(PR-NLS)is proposed.Through theoretical analysis and simulation experiments,it is verified that the improved PR-NLS algorithm fully guarantees the integrity of local structural information,and has good accuracy and robustness in non-rigid alignment problems with data degradation in the form of deformation,outliers,occlusion,noise,rotation,etc.(3)Application of image stitching method based on PR-NLS algorithm in UAV inspection.The traditional image stitching algorithm has a large number of mis-matching problems that lead to low accuracy of subsequent image stitching.The image stitching method based on the improved PR-NLS algorithm is applied to the image stitching of transmission line towers and corridors measured by UAVs.Through local image alignment and local image stitching,a panoramic image with good performance of image overlay and stitching is obtained,which better preserves the information integrity of transmission line images in power inspection.In summary,this paper proposes an improved P-MCC algorithm and an improved PR-NLS algorithm,respectively,which describe the global and local features of the images using spatial distance clustering,weighted spatial distance,shape contextual features,Hungarian algorithm and metrics to obtain a point set alignment algorithm with good performance,while the image stitching method designed using the improved algorithm is validated for the measured transmission line pole and The effectiveness and accuracy of the proposed method is verified for the measured transmission line towers and corridor images.