A Multi-Feature Integrated Type Classification And Posture Recognition Method For Miniature Circuit Breaker Components

In existing assembly schemes of the automatic manufacturing for miniature circuit breakers(MCBs),components of the same type are arranged to a predefined posture by a dedicated mechanism and fed to the corresponding mechanism that accomplishes assembly.Since each mechanism set are only designed for one type of components,its expansibility,universality and flexibility are limited.When the product is updated,redesign of the entire assembly equipment is required.Therefore,existing schemes cannot accommodate the fast development of modern manufacturing.A device applying industrial robotic and machine vision technologies is proposed to achieve mixed loading and flexible assembling with multi-type random-postured components.The device can significantly improve the assembly flexibility and meet the requirement of rapid product iteration.However,in the process of visual-guided loading,both shape and type complexities of components cause remarkable difficulties in type classification and posture recognition.To resolve the problems,a new method for type classification,posture recognition and grasping point locating is proposed by comprehensively analysis of color,area,contour and imaging features.The proposed method firstly acquires images of the components by a camera,and then obtains the region of each component by image segmentation.Secondly,the method performs a two-stage classification by the comparisons of characteristics and imaging differences of the components.Type classification is achieved via the color and area information in the first stage,and posture classification is achieved in the accordance of the area,color and structure information in the second stage.Thirdly,the component rotation is computed by key-point matching which is reached by analysis on significant structure features.Thus,the posture estimation of each type of components is accomplished.Finally,the coordinates of components are calculated by applying the coordinate and affine transform matrices.The experimental results show that the classification accuracy of the proposed method reaches higher than 99%,and the errors of rotation angles and grasping coordinates are within the ranges of ±0.8° and±0.3mm respectively.Therefore,it is proven that the proposed method meets the requirements of flexible MCB assembly.