An Improved Character Recognition Framework for Containers Based on DETR Algorithm.

Xiaofang Zhao,Liyun Xiao,Peng Zhou,Ke Xu

doi:10.3390/s21134612

Abstract

An improved DETR (detection with transformers) object detection framework is proposed to realize accurate detection and recognition of characters on shipping containers. ResneSt is used as a backbone network with split attention to extract features of different dimensions by multi-channel weight convolution operation, thus increasing the overall feature acquisition ability of the backbone. In addition, multi-scale location encoding is introduced on the basis of the original sinusoidal position encoding model, improving the sensitivity of input position information for the transformer structure. Compared with the original DETR framework, our model has higher confidence regarding accurate detection, with detection accuracy being improved by 2.6%. In a test of character detection and recognition with a self-built dataset, the overall accuracy can reach 98.6%, which meets the requirements of logistics information identification acquisition.

Highlights

Table 3and shows the comparison of the multi‐scale detection effect between the proposed model and and DETR
This paper proposes an end-to-end recognition framework for container characters based on the DETR-ResNeSt50-multi-scale location coding (MSLC) model
The test result in the self-built container character data set can reach 98.6%, which meets the requirements of the smart port for the detection accuracy of container numbers

Summary

Introduction

With the continuous development of international trade and the social economy, the port transportation industry has developed rapidly, and cargo throughput has increased dramatically. In order to ensure the real-time location of cargo and the safety of cargo during the logistics transportation process, containers are used as storage units to centrally detect, identify and track the information of the transported cargo. The accurate detection and identification of container characters has begun to become an important part of the intelligent information management system for port logistics, and it is an inevitable demand for the realization of comprehensive port automation. The traditional methods of port container character registration and identification are mainly based on manual identification, which is susceptible to large quantities of goods and bad weather and leads to low efficiency of number identification. With the object detection technology represented by convolutional neural networks constantly refreshing the highest detection index of each public image set, the application of machine vision in the industrial field has begun to expand [1,2]

Methods

Results

Discussion

Conclusion