Research On Technologies Of Multi-view Online Detection Of Surface Defect In 3D Printing Based On Machine Vision

Fused Deposition Modeling 3D printing technology has developed very rapidly in recent years,however,there are many defects that may affect the surface quality,usability in the printing process,which causes the waste of time and materials and seriously affects the further development of FDM process.On the basis of investigating the development status and latest research results of 3D printing defect detection technology at home and abroad,this paper proposes a new multi-view online vision inspection scheme for the outer surface defects of 3D printing.Aiming at the problem of misidentification between model's own features and real defects,a virtual-real matching defect detection algorithm oriented to model's self-features is proposed to reduce the false detection rate and improve the detection accuracy.The main research contents and innovations of this paper are as follows:Firstly,the hardware structure and control system of multi-view vision inspection system for robot 3D printing are designed.Based on the principle of slice and path planning in 3D printing,a multi-view visual defect detection method is proposed in this paper.The camera perspective is adjusted to the vertical direction of the printing path to ensure that the camera is always perpendicular to the current printing surface during the printing process,and provides a good perspective for image acquisition.Aiming at the 3D printing scene and the surface characteristics of parts,an image detection system is developed with VC++ and OpenCV.On the basis of image preprocessing algorithms such as histogram equalization,LBP and median filtering,a mask filtering method is proposed according to the overlapping texture characteristics of 3D printing parts to enhance the detectability of features and increase the contrast between foreground and background.Then,a defect recognition and analysis algorithm is designed according to the geometric shape of the defect in 3D printing.Based on the cascade invariance of 3D printing and the mapping relationship between the world coordinate system and the image pixel coordinate system,a binuclear defect recognition technology based on image morphological segmentation is proposed.Aiming at the problem of discreteness and disorder of defects,a defect region expansion method based on position relationship is proposed.The defect mathematical expression is formed by the central coordinates,aspect ratio,area and distribution of the fused defect contour,which is convenient for defect analysis and quality evaluation of parts.Due to the fact that the existing visual inspection algorithms can only detect parts with gentle surface and they often misidentify their own features as defects for the surface with large gradient change,this paper proposes a virtual-real matching defect detection algorithm based on the model's own features.The defect and self-feature of the model can be recognized by comparing the detection results of the experimental image with the theoretical results of self-feature projection of the model under the same mapping relationship.Based on the position relationship and continuity between points,the visible points are selected from the discrete points cloud,and the feature points are extracted according to the visible continuity and the angle change between the visible points.These feature points are used to form theoretical projection images by mapping relations and distorted models from monocular calibration.Finally,the weighting function of the similarity between the contour parameters in the theoretical and experimental pictures is used to distinguish their own features and defects,so as to reduce the false detection rate and improve the detection accuracy.