Research And Application Of Visual Positioning And Posture Recognition Methods For Small Circuit Breaker Components

Miniature circuit breakers(MCBs)are a kind of switchgear performing overload and short circuit protection in electrical circuits.Because of the small size,easy operation and low price,they are widely used in various places such as industrial,commercial,and residential buildings.With the continuous development of people’s life in recent years,the demand for MCBs grows explosively.Meanwhile,higher requirements have been placed on the efficiency and accuracy of the production process.Traditional manual assembly is inefficient and the quality is uneven,while conventional automatic assembling techniques with vibrating plate loading limit the flexibility in manufacturing.To solve the problems above and to meet future market demand,a visual positioning and attitude recognition method for miniature circuit breaker components is proposed.Based on this method a flexible system using machine vision for MCB assembly is built.The system builds a visual recognition module that identifies the category,position and posture of an MCB component by the VGG-16 deep learning classifier and feature-template matching,and sends the recognition result to the industrial robot.which is guided to flexibly switch the corresponding robot jaw and perform different assembly motions.The main research contents of this thesis are:1)According to the needs of MCB automatic assembly,the hardware platform of the visual recognition system is built.The hardware of the visual recognition system includes basler industrial cameras,Fuji lenses,camera brackets,light sources and light source controllers,and bases for parts placement areas.2)Based on the hardware platform of the visual recognition system,the source image of the part to be identified is obtained.According to the shape characteristics of the internal parts of the small circuit breaker,the corresponding image preprocessing algorithm is studied to edge the source image of the part to be identified.Detection,filtering binarization,corrosion expansion and other operations will cut out all parts included in the picture,and then identify the type and placement status of each part through a cascade classifier based on deep learning,and then use feature-template matching to obtain the position and posture.3)The visual recognition system converts the pixel coordinates of the gripping point to the coordinates of the robot coordinate system through camera calibration and coordinate conversion,and sends the processed parts type and placement status,gripping point position,and rotation angle information to the industrial robot controller through socket communication.The experimental results prove the feasibility of the method proposed in this paper,meets the precision requirement of flexible MCB assembly.