Research on the Method of Universal Forklift Pallet Detection and Pose Estimation Based on Visual and 3D Point Cloud Fusion

As logistics demands surge and labor costs rise, the industry turns to intelligent automation. Autonomous forklifts, key in this shift, autonomously transport goods to cut costs and boost efficiency. Pallet detection is undoubtedly a key step in achieving automatic forklifts. Traditional pallet detection methods, relying on single data modalities like images or 3D point clouds, face challenges with accuracy due to lighting, occlusion, and real-time processing. This paper introduces a novel approach for pallet detection and pose estimation on forklifts, merging visual data with 3D point clouds. Using a depth camera, the method refines pallet location via improved YOLOv8 detection and fuses color and depth data for precise 3D localization. The results show high performance and reliability, promising for real-world industrial applications.