Research On Express Package Separation Technology Based On Deep Learnin

As a labor-intensive industry,the express delivery industry has an urgent demand for intelligent transformation in industrial upgrading and express speed increase.With the development of express sorting technology,the dynamic sorting system of express parcels has become increasingly mature,but at present,the express parcel sorting system still relies on manual supply,which hinders the improvement of the operation efficiency of the sorting system,while the express parcel separation technology as an alternative measure to manual supply needs improvement.At present,the separation technology of express packages is still in its infancy and mainly focuses on the regular package.The separation method has a significant error and the accuracy of separation cannot be guaranteed.In this thesis,the following work has been done to solve the problem of separation of the express package:1.A deep learning-based courier parcel detection method is proposed.First,a data acquisition device is set up in the post office to collect images,and image synthesis technology is used to simulate complex actual scenes to enrich data sets.Then,network training is carried out based on the constructed express parcel data set so that it can automatically identify express parcels.An adaptive sample mining method is proposed to solve the problem of the low recognition rate of a few irregular packages.Aiming at the problem of inaccurate positioning of the algorithm package,a boundary frame repositioning method is proposed to solve the problem of boundary overlap and poor accuracy.The biggest advantage of this algorithm is accurate recognition,high efficiency,and strong robustness,do not need a high-precision depth camera and auxiliary pull packet section.When using this algorithm to identify and express parcels,the accuracy remains above 98.6%.2.A phased edge extraction method is proposed.Given the complex field conditions of the postal center bureau,the belt texture,uneven reflection of the belt,and other factors have a great impact on the traditional edge detection method,resulting in a poor edge detection effect.Based on the express parcel detection method,a phased parcel edge extraction method is proposed.The advantages of this method are to solve the problems of low detection accuracy,poor robustness,and excessive consumption of computing resources of traditional edge detection methods,and can extract the parcel edge more accurately.The accuracy of the algorithm is 92%,and the recall rate can be maintained at about 93%.The maximum F-measure is 0.926,and the average absolute error is only 0.038.In summary,this thesis focuses on the research of express parcel separation technology based on deep learning,and the proposed algorithm has been tested in multiple rounds.The experimental results show that the accuracy of the above algorithm in identifying express parcels is over 98%,and the accuracy of the parcel edge is 92%.It can accurately and quickly solve the problem of single piece separation of various specifications of dense parcels,especially special-shaped parts.It has good separation accuracy and robustness,and can well meet the requirements of separate express parcels.