Explore the vast range of titles available.

In order to solve the difficult problem of feeding process in the process of disorderly logistics package stacking entering automatic sorting equipment in logistics package sorting industry, an algorithm based on 3D vision to determine the normal vector of three points-projection to determine the position of the central point is proposed, which is used to obtain the position information of the physical and geometric central point on the projection plane of spatial logistics package. The mathematical model of the algorithm is established. The method of calculating the normal vector and the attitude of the center point of the algorithm and the workflow of the algorithm are expounded. In order to verify the accuracy of the proposed algorithm, an experimental method of determining the center point with dual lasers and the spatial pose with dual probes is proposed, and an experimental scene based on 3D vision, industrial robot, dual lasers and PC (Personal Computer) is constructed. The experimental results show that the algorithm can solve the position and attitude information of the physical and geometric center of the projection plane of the target logistics package in the disorderly logistics package stacking, and the maximum positioning error is , and the average error is . The average positioning time is 17.55 ms. The algorithm of determining the normal vector of three points and determining the position of the center point by projection can solve the position information of the physical geometric center of the projection plane of spatial logistics packages, which provides reference for the research of disorderly logistics package sorting technology.

Accurate parcel identification in disordered logistics environments poses significant challenges due to varying package sizes, materials, and orientations. This study presents an improved YOLOv8-Efficiency algorithm tailored for such complex scenarios. The proposed algorithm introduces the C2f-OR module to reduce parameters and computation, the Conv-Ghost module for efficient feature extraction, and the HIoU loss function to enhance identification accuracy. By constructing a dataset of 4689 photos, experiments demonstrate the algorithm's effectiveness, achieving a 93.2% mAP, a 1.6% recall rate improvement, and a significant reduction in computational complexity (9.9% decrease in FLOPs). This work provides a robust solution for real-time parcel identification in disordered logistics, facilitating automation and efficiency in logistics operations.