A Vision-based inventory method for stacked goods in stereoscopic warehouse

Inventory of stacked goods in the stereoscopic warehouse is important for modern logistics. Currently, this inventory task is completed by counting manually. With the advance of industry 4.0 and deep learning technology, automatic inventory based on machine vision comes true, greatly saving labor and material costs. In this work, we firstly collected WSGID, an image dataset about wine boxes stacked in a stereoscopic winey warehouse. Moreover, we presented an automatic inventory method based on machine vision, consisting of a stacked goods surface detecting model and a prior-based quantity calculating algorithm. To get a better detecting performance, we introduced STCNet, an improved detection network based on Swin Transformer. The final results of 86.7 mAP, 82.8 mAP, and 85.9 mAP on three sub-datasets are achieved and are higher than the baselines. To count the quantity of goods after detection, we proposed an adaptive and robust calculating algorithm. Our method got an accuracy of 85.71 on the largest sub-dataset. Extensive experiments on the WSGID and COCO benchmark demonstrate the effectiveness of our approach. Our work indicates that the machine vision method successfully facilitates inventory for stacked goods in the stereoscopic warehouse.