The 6-DOF Pose Estimation Algorithm Based on YOLO11-BiFormer-Robust ICP Model

In the fields of robotic manipulation and automated manufacturing, accurately estimating the pose of target workpieces in complex stacked environments is a challenging and critical task. This paper presents a 6-DOF pose estimation method that utilizes an RGB-D camera to capture color and depth images of the workpiece. By introducing the BiFormer module into YOLO11, the model’s ability to detect and segment occluded targets in complex stacked environments is enhanced, without significantly affecting the model’s operational efficiency. Specifically, the BiFormer module improves feature extraction capabilities through a dual-layer routing attention mechanism, significantly increasing detection and segmentation accuracy in complex environments. The mask obtained from the RGB image is projected onto the depth image to extract the corresponding segmented point cloud data. Subsequently, the Robust Iterative Closest Point (Robust ICP) algorithm is employed for 6-DOF pose estimation, which performs exceptionally well in noisy environments. Experimental results demonstrate that, compared to the original YOLO11 model, the model with the BiFormer module shows improvements in both segmentation accuracy and recognition precision. This method exhibits high accuracy and robustness in workpiece recognition and pose estimation, making it suitable for real-time industrial applications.