Research On Precise Location Method For Composite Reference Hole Guided By Texture Boundary

During robot drilling holes on aircraft structural parts,visual sensors are usually used to detect and locate pre-drilled reference holes to correct the origin of workpiece coordinate system online,which can correct the deviations of hole position caused by offline programming.With the increasing use of composite materials in aircraft manufacturing,how to achieve rapid detection of circular holes and precise positioning of reference holes under the composite texture background has become an urgent problem in robotic automatic hole-making technology.Aiming at this problem,after studying the techniques of texture description,texture segmentation,and target detection,this paper proposes a fast detection method for composite circular holes guided by texture boundary and a precise location method for composite reference holes.The main research contents and innovations of this article are as follows:Aiming at the problem that the composite texture has a strong interference on the detection task,texture segmentation is used to obtain the texture boundary to eliminate the texture interference.First,a texture segmentation method based on traditional texture features is proposed.This method extracts the texture contrast feature which is formed by Local Ternary Pattern(LTP)and the contrast of Gray Level Co-occurrence Matrix(GLCM)to perform the fast segmentation of textures inside and outside hole.Also,a texture segmentation method based on deep learning semantic segmentation model is proposed.This method uses Unet++ model to achieve texture segmentation inside and outside the circular hole,and uses attention enhancement on key position,model pruning and other techniques to improve the cost performance of the model.The texture contrast method is cost-effective and suitable for general application scenarios and CPU computing environment.The Unet++ method has good performance and large improvement space,and is suitable for demanding application scenarios and GPU computing environment.Aiming at the two major difficulties in detecting composite circular holes: the texture noise and the blur and deformation of hole boundary,a detection method for composite circular hole guided by texture boundary is proposed.This method uses the position information of texture boundaries firstly to remove most of the image edge points that belong to non-hole boundaries,which eliminate the interference of texture noise on circle detection.Then,it uses the connectivity information of texture boundaries to group the remaining image edge points,and finally detects an optimal circle from each group of edge points separately to enhance the ability to detect holes that have blurry deformed boundaries.This method has a high detection rate and a relatively fast speed,and shows high engineering application value.Aiming at the problem that the composite circular hole detection method has low recognition accuracy and cannot complete the precise positioning of composite reference holes,a multi-stage location method reference hole is proposed.This method extends the detection process of hole detection method guided by texture boundary into multiple stages,in which each stage performs a new round of hole detection under the guidance of the detection results generated in the previous stage,improving the hole location accuracy stage by stage and achieve a high location accuracy in the last stage.This method has good positioning performance of composite reference holes.