The Study of Nickel Anode Plate Copper Ear Wire Overlapping Scene Segmentation Based on Mask R-CNN

A segmentation method combining Mask R-CNN (Mask Region-Based Convolutional Neural Network) and monocular depth estimation is proposed to identify and segment the copper ear wire closest to the robotic arm on nickel anode plates, enabling automatic threading of copper ear wires into conductive rods. The method utilizes depth maps to provide spatial distance information for each pixel, enhancing the Region Proposal Network's (RPN) ability to generate accurate candidate regions in complex scenes, thereby improving feature alignment. Experimental results demonstrate the method achieves an average precision of 88.82% across various overlapping scenarios, with an online detection correct segmentation rate of 89.40% and a processing time of 0.41 s per image. Compared to the traditional Mask R-CNN, the method improves precision by 5.51% and increases the recall rate by 2.39%, offering efficient technical support for the automatic threading system of nickel anode plates.