Point Cloud Segmentation And Dimensional Detection Of Oil Distribution Pan Surface Based On Fringe Projection

Hydraulic system plays an important role in industrial production,aerospace and other fields.As one of the key parts commonly used in the hydraulic system,the overall performance of the hydraulic system is determined by the oil distribution pan’s processing accuracy.The oil distribution pan working face is the spherical area in the surface of it and it’s the core area of the size measurement of it.The machining accuracy of the parts of the oil distribution pan is directly affected by the measurement of it’s size parameters.Due to it’s fast detection speed and high efficiency,the non-contact measurement method can meet the requirements of high-precision and large-batch detection of oil distribution pan production line.However,due to the low accuracy of existing methods for three-dimensional point cloud segmentation in the measurement process,the measurement accuracy is low.Therefore,this paper proposes a method to obtain three-dimensional point cloud of the surface by grating projection based on the segmtioned surface image of the oil distribution pan,and then simulates the three-dimensional point cloud of the surface into a sphere to calculate the sphere diameter.The main work content of this paper is as follows:Firstly,since the segmentation accuracy of three-dimensional point cloud depends on the edge segmentation accuracy of the spherical part of the oil distribution pan image,this paper improves the edge segmentation accuracy of the oil distribution pan image by means of noise reduction,surface highlight preprocessing and edge enhancement.Firstly,bilateral filtering method is selected to denoise the oil distribution pan image.After processing,the image surface noise is less,and the surface details are well preserved.Then,several low dynamic oil distribution pan images at different exposure times were fused to remove the highlight points,and the logarithmic transformation method was used to process the image contrast,which significantly improved the gray value of the dark area and retained more fringe information.Finally,Sobel operator is used to enhance the image edge of the oil distribution pan.The sharpening effect is good,the edge is clear,and the edge details are well retained,which improves the contour extraction accuracy of the spherical part of the image of the oil distribution pan.Secondly,the calculation principle and process of 3D point cloud of oil distribution pan under digital grating projection are studied.The four-step grating phase-shift fringe method is used to extract the phase principal value of the oil distribution pan image.Since the phase principal value is wrapped in one cycle,the two-frequency heterodyne method is used to calculate the wrapped phase principal value.Then,considering the constraints such as the position relationship of each device in the measurement system,a flexible measurement model without constraints is proposed in this paper.The relationship between the phase principal value of each point in the image coordinate system and the coordinate of the point in the corresponding camera coordinate system is determined,and the relevant system calibration parameters are solved by a series of operations.Finally,a high precision 3D surface batch detection automatic platform for hydraulic components is built.The maximum inter-class variance method was used to binarize the collected image of the oil distribution pan,extract the contour of the spherical part of the oil distribution pan and make a mask.Through the mask,the four-step phase shift diagram of the spherical part was divided.The four-step phase shift method was used to solve the phase principal value,and the dual-frequency heterodyne method was used to expand the phase,and the three-dimensional point cloud data of the spherical part of the oil distribution pan was calculated by combining the system calibration parameters.The least square method was used to synthesize the point cloud data to measure the sphere diameter with a precision of 99.238%.The results were compared with the sphere diameter fitting results based on Z-axis height segmentation,regional growth segmentation,RANSAC segmentation and artificial frame selection of point cloud data.The measuring accuracy of the above four methods are 97.996%,93.679%,99.130% and 79.535%,respectively.It is analyzed and verified that the proposed method can realize high-precision batch detection of oil distribution pan,realize point cloud segmentation and size measurement of oil distribution disc under grating projection,and improve the detection level of three-dimensional surface.It can provide reference value for accurate measurement of hydraulic components.Figure [41] Table [6] Reference [64]...