Extraction method of dispensing track for components based on transfer learning and Mask-RCNN

In the process of dispensing, the traditional dispensing robot generally obtains the component pad profile according to the Mark point assisted positioning, and directly uses the profile as the dispensing profile. However, due to the influence of welding and other factors, the posture of the components often changes after welding, which easily causes the actual dispensing contour to be difficult to completely match the pad, so there is a certain deviation in the dispensing. Moreover, component recognition based on convolutional neural network requires a large number of samples for training, which is not conducive to the expansion of dispensing components. This paper focuses on the high-precision dispensing task. Based on the indirect positioning components, this paper uses Mask RCNN to extract complex component dispensing track in different environments. Compared with traditional methods, this method has higher robustness and dispensing accuracy. At the same time, the transfer learning method is used to train the neural network, so that the algorithm has better scalability and flexibility when facing the detection and segmentation tasks of new components. The experimental results show that the component dispensing track extraction method proposed in this paper has higher precision and flexibility than the traditional method.