Postal parcel text detection method and device based on deep learning

Abstract

The invention relates to a postal parcel text detection method and device based on deep learning, and the method comprises the following steps: an image adjustment step of obtaining a to-be-detected postal parcel image, detecting the position of a bar code in the image through a pre-trained target detection model, and performing rotation correction on the postal parcel image based on the inclination angle of the bar code to obtain a corrected image; a receiving address position detection step of taking the corrected image as the input of the target detection model again, and obtaining a receiving address key information position box; and a text detection step of carrying out text detection on the addressee address key information position box by utilizing a trained text detection model. Compared with the prior art, a deep learning method is used, the problem that the receiving address is difficult to detect when the postal parcels are automatically sorted is solved, then the method is in butt joint with a postal parcel receiving address text recognition task, and finally automatic sorting of the postal parcels is achieved.

Description

technical field [0001] The invention relates to the field of computer vision, in particular to a deep learning-based postal package text detection method and equipment. Background technique [0002] Nowadays, with the standardization of the courier form, many courier companies have completed the transformation from handwritten courier forms to printed electronic forms. At present, most express companies mainly sort express mail packages by identifying form barcodes, that is, using networked automatic sorting machines or networked mobile devices held by couriers to scan the barcodes on the express parcels and identify the corresponding barcodes. The tracking number, and then search and match all the information corresponding to the identifying tracking number in the database, and obtain the recipient's address information from it. However, there are certain problems in the sorting method by locating barcodes. First, this method can only be carried out online. Once there is a...

AI Technical Summary

Problems solved by technology

However, there are certain problems in the sorting method by locating barcodes. First, this method can only be carried out online. Once there is a problem with the network or the database fails, the sorting will not be possible. Second, when the barcode is blocked, polluted or missing At this time, the sorting system will not be able to work, and it can only be completed manually by the courier, which is time-consuming and laborious; in addition, there are many text information on the courier package waybill, and a large number of irrelevant forms, patterns and text areas make it difficult to detect the receiving address. The location of the receiving address of different types of waybills is different on the waybill, which increases the difficulty of detection

Images