Research Of 3D Arc Reconstruction Based On Multi-vision

Multi-vision-based 3D reconstruction is regarded as an important research problem in computer vision related fields due to its wide application prospects.The use of 3D reconstruction algorithms to achieve small sample 3D reconstruction of arcs is the focus of this paper.Because the shape of the arc has an important reference value for analyzing the life and capacity of switching appliances,and whether it will cause power failures,etc.,in order to ensure the reliability and safety of the product,the shape of the arc generated between the switch contacts of low-voltage appliances is comprehensively analyzed.Analysis has extremely important research significance.First,a high-speed camera is used to collect the two-dimensional image of the arc.Because the arc generation process is relatively fast,it is necessary to filter the images with unclear boundaries or incomplete arc generation during the acquisition process.Among them,the two-dimensional images of different angles that meet the conditions of the three-dimensional reconstruction.There are only 2 pictures,so this article proposes a corresponding solution for 3D reconstruction based on small samples.Secondly,this article analyzes the existing 3D reconstruction technology,and realizes the reconstruction of the arc through feature extraction and matching,global optimization,and dense point cloud generation.For the arc camera position calibration,edge detection and feature matching,it has found practical and efficient.The method uses the optimization and improvement of the motion recovery structure algorithm,and uses the beam adjustment method(Bundle adjustment)to adjust the parameters of multiple cameras in the world coordinate system and the three-dimensional coordinate parameters,thereby reducing the error of the positioning attitude of the camera mark and the point cloud.And proposed a block threshold processing module to determine the edge of the arc,so as to reduce the interference of noise points on the entire three-dimensional model.Thirdly,this paper proposes a new algorithm based on the original Sift registration method,that is,the random consensus algorithm is used to improve the relative matching accuracy of the matching points,and the matching results are filtered for mismatches,thereby improving the arc image matching It also proposes a similar feature point matching filter module to improve the matching efficiency.Finally,the visualization of the three-dimensional arc reconstruction results is realized,which is convenient for observation and in-depth study of the arc shape,which can greatly improve the design level of electrical appliances and ensure the safety of electrical appliances.Since the arc from igniting to arc extinguishing is very short,it cannot be directly observed and judged,and its physical length cannot be measured,and the two-dimensional arc will face some problems such as missing information.Therefore,it is necessary to carry out the moment when the arc is generated.Three-dimensional reconstruction and visualization is the focus of this article.Through in-depth research on the arc shape and mechanism,the electrical circuit design,material selection and safety of electrical appliances can be improved.Therefore,the in-depth study of this subject has a comprehensive theoretical basis and objective practical value.