Pose Detection Of Hole Via Visual System Based On Image Super-Resolution

In the field of mechanical automation assembly and docking,there are many scenes of shaft hole assembly.By detecting the circular hole on the part,some parameter such as the center coordinate and the circular hole normal vector can be obtained,that is,the spatial posture of the circular hole is obtained,which providing accurate navigation for the automated assembly process.In recent years,visual inspection technology has developed rapidly.Due to its high efficiency,non-contact,high precision,it has been widely used in the detection of circular hole pose during shaft hole assembly.In the visual inspection process of the circular hole position,there are some non?ideal detection scenes such as unstable illumination and small round hole in the camera field,which makes it impossible to detect effective circular hole information,so it is necessary to improve the image quality of the round hole detection process.The traditional method is to improve the resolution of the image by improving the hardware performance.Considering that some visual inspection devices need to be deployed on a large scale in a mechanical automation assembly scenario,the economic cost of upgrading the hardware is high,and the improvement of image quality is limited.In this paper,the image super-resolution reconstruction technique is applied to the circular hole pose visual inspection task in the industrial shaft hole assembly process to solve the problem of insufficient detection image resolution and improve the accuracy of the round hole pose detection.The main research contents of this paper are as follows:(1)An image super-resolution reconstruction model for edge enhancement of industrial parts is constructed.Firstly,the industrial parts sample library is established and preprocessed.Then the edge distribution enhancement feature distribution loss function and the residual block simplification method are used to improve the deep recursive residual super resolution reconstruction model,so that the model training results are more suitable for super resolution reconstruction of the industrial round hole parts image.(2)A method for visual inspection of the position and orientation of the hole based on super-resolution reconstructed images is proposed.Firstly,based on the super-resolution reconstructed image of the circular hole,the phase consistency edge extraction is performed,which reduces the influence of disturbance factors such as illumination instability and noise.Then the feature parameters,parallel chord theorem and least squares median method are used to fit the hole parameters,which reduces the influence of outliers and improves the accuracy of parameter fitting.Finally,the camera observation model based on super-resolution reconstructed image is adapted to the image scale transformation,and the pose estimation of the circular hole is realized based on stereo matching.(3)A method for estimating the pose of parts based on porous combined visual inspection is proposed.For the parts with multiple circular hole features,the part pose estimation method based on the porous combined visual inspection is used to select the effective target circular hole and the relative position determination according to the part's own attributes.The pose estimation of the part is realized based on the porous combination relationship..(4)A circular hole position and vision detection system based on image super-resolution reconstruction is built.According to the method proposed in this paper,the circular hole pose visual detection system based on image super-resolution reconstruction is developed,the system hardware environment is built,and the system software is developed,which realizes the integration of various functions of the system and is applied to the AC charging adapter of electric vehicle.The pose estimation results verify the effectiveness of the proposed method.