Abstract
Visual sorting of express parcels in complex scenes has always been a key issue in intelligent logistics sorting systems. With existing methods, it is still difficult to achieve fast and accurate sorting of disorderly stacked parcels. In order to achieve accurate detection and efficient sorting of disorderly stacked express parcels, we propose a robot sorting method based on multi-task deep learning. Firstly, a lightweight object detection network model is proposed to improve the real-time performance of the system. A scale variable and the joint weights of the network are used to sparsify the model and automatically identify unimportant channels. Pruning strategies are used to reduce the model size and increase the speed of detection without losing accuracy. Then, an optimal sorting position and pose estimation network model based on multi-task deep learning is proposed. Using an end-to-end network structure, the optimal sorting positions and poses of express parcels are estimated in real time by combining pose and position information for joint training. It is proved that this model can further improve the sorting accuracy. Finally, the accuracy and real-time performance of this method are verified by robotic sorting experiments.
Highlights
With the prosperity of the e-commerce industry, the rapid development of the logistics industry has given rise to massive orders for logistics companies, which has brought a heavy load for the sorting work of logistics systems
Two deep learning methods are used for object detection and semantic segmentation, and an ICP-based [5] object model matching algorithm is used to estimate the pose of the object
To overcome the aforementioned challenges, in this paper, we propose a rapid sorting method based on multi-task deep learning
Summary
With the prosperity of the e-commerce industry, the rapid development of the logistics industry has given rise to massive orders for logistics companies, which has brought a heavy load for the sorting work of logistics systems. Two deep learning methods are used for object detection and semantic segmentation, and an ICP-based (iterative closest point) [5] object model matching algorithm is used to estimate the pose of the object. Shao et al [18] combined ResNet with the U-net [19] structure, a special framework of convolution neural networks (CNNs), to predict the picking region without recognition and pose estimation, which makes the robotic picking system learn picking skills from scratch This type of system more directly determines the grasping points or candidate grasping areas. The multi-layer shallow feature map and the final feature map are merged to extract more detailed candidate grasping areas, and a cascaded convolutional optimal grasping position detection network based on key points is used to achieve real-time estimation for target objects.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
