Traditional deep learning target detection algorithms face significant accuracy limitations in automated warehousing systems, especially in pallet stacking and occlusion scenarios. Based on this, we propose an innovat...
详细信息
Traditional deep learning target detection algorithms face significant accuracy limitations in automated warehousing systems, especially in pallet stacking and occlusion scenarios. Based on this, we propose an innovative transfer learning approach that achieves efficient multi-scale feature extraction and fusion through the improved YOLOv8s-pose architecture and BDEM (boundary detail extraction module)-enhanced FFDPN (feature focus diffusion pyramid). Combined BA Wing loss (boundary-aware Wing loss) and APT-TAL (adaptive power transformation task-aligned labelling) strategy improves detection accuracy at 12 keypoints of pallet E-section. The system achieved 94.4% detection accuracy and 93.2% keypoint positioning accuracy at a processing speed of 104.2 FPS (frames per second). Finally, the lmeds (least median of squares)-based position optimisation solution further improves reliability by controlling the pallet inclination and distance measurement errors to within 4.0 degrees and 29 mm, providing a practical solution for automated logistics systems.
The computer vision approach involves a lot of modeling problems in preventing noise caused by sensing units such as cameras and projectors. In order to improve computer vision modeling performance, a robust modeling ...
详细信息
ISBN:
(纸本)9781509026784
The computer vision approach involves a lot of modeling problems in preventing noise caused by sensing units such as cameras and projectors. In order to improve computer vision modeling performance, a robust modeling technique must be developed for essential models in the system. The RANSAC and least median of squares (lmeds) algorithms have been widely applied in such issues. However, the performance deteriorates as the noise ratio increases and the modeling time for algorithms tends to increase in actual applications. In this study, we propose a new lmeds method based on fuzzy reinforcement learning concept for modeling of computer vision applications. The performance of the algorithm is evaluated by modeling synthetic data and camera homography experiments. Their results found the method to be effective in improving calculation time, model optimality, and robustness in modeling performance.
暂无评论