Complex Scene Random Bin Picking Algorithm Based on YOLOv8-TripletAttention

In the fields of robotic operations and automated manufacturing, random bin picking is a challenging and crucial task. This paper proposes a method for workpiece pose estimation based on the combination of YOLOv8 and the Triplet Attention module, designed for random bin picking in complex scenarios. The method utilizes an RGB-D camera to capture color and depth images of the workpiece. By introducing the efficient and lightweight Triplet Attention module, the model achieves a significant improvement in detection and segmentation accuracy without compromising the running speed. Specifically, the Triplet Attention module enhances the spatial attention mechanism by employing three-way rotational convolutions, leading to notable improvements in detection and segmentation precision. The segmented mask is applied to the depth image to extract the corresponding point cloud data, and a six-degree-of-freedom pose estimation is performed using the Normal Distributions Transform (NDT) algorithm, which excels in the presence of noise and partial occlusions. Experimental results show that, compared to the original YOLOv8 model, the model with the Triplet Attention module demonstrates improvements in segmentation precision and recognition accuracy. This method exhibits high accuracy and robustness in workpiece recognition and pose estimation, making it suitable for real-time industrial applications.