Research On Fast Work-Piece Location Method Based On Template Supervision

With the continuous enhancement of China’s manufacturing capacity,high-end intelligent manufacturing has gradually become the key to accelerate industrial development.As one of the non-contact measurement technologies,work-piece positioning system can play a key role in many automated production lines.In the industrial manufacturing field,the flexible system needs to adapt to the change of work-piece types and the requirements of real-time positioning.The fast work-piece positioning algorithm based on template supervision proposed in this paper can solve the complex positioning task of work-piece shape and texture change,and meet the flexible requirements of the system.The work-piece location algorithm based on template supervision consists of two parts:feature extraction and work-piece matching.In the feature extraction stage,the purpose is to extract template features that can resist shape and texture changes and improve the accuracy of work-piece localization.The purpose of the work-piece matching stage is to quickly and accurately locate the position and rotation Angle of the user input template image in the image to be tested by using the template features,mainly to solve the problem of search efficiency.In the feature extraction stage,this paper first proposes a template image enhancement method to enhance the user input template image to obtain the fused template image.The fused template image has the common characteristics of the current batch of work-pieces,and weakens the interference and noise caused by surface roughness,defects and scratches in the user input template image.Secondly,a feature point selection method based on boundary detection and gradient direction is designed.According to the gradient and distribution position of the edge points,the edge points in the rotation change prominent area and the template contour area are selected from the edge detection results to construct the feature point set,which improves the positioning accuracy of the algorithm in the work-piece with different surface characteristics.The problem of positioning error due to the change of shape and texture of work-piece in flexible machining is solved.In the work-piece matching stage,we propose a location search strategy based on pyramid down-sampling,which divides the positioning task into two stages:low-resolution image positioning and original resolution positioning,thus reducing a large number of unnecessary calculations.Secondly,the rotation feature of gradient distribution statistics is proposed,which can quickly determine the most similar Angle between the window region and the template image by calculating the similarity scores of only a few key angles during the Angle matching process,shortening the Angle matching time and significantly improving the search efficiency.Finally,experiments and analysis are carried out in the work-piece image data set composed of the work-piece images provided by the enterprise and the images collected by the laboratory hardware equipment,and the effectiveness of the proposed method is verified.