Express waybill positioning correction method and device

Abstract

The invention provides an express waybill positioning correction method and device, and the method comprises the steps: carrying out the marking of an image containing an express waybill, enabling theexpress waybill in the image to be covered, marking the orientation of a text, and obtaining the marking data of the express waybill; taking the annotation data as a training set to train a convolutional neural network, so as to enable the convolutional neural network to predict the position and the text direction of the express waybill; and inputting a to-be-detected image containing the expresswaybill into the trained convolutional neural network to obtain a predicted position and a predicted text direction of the express waybill, and performing positioning correction according to the predicted position and the predicted character direction of the express waybill to obtain a positioned and corrected express waybill. The express waybill positioning and correcting method provided by theinvention can be used in a task of realizing express sorting automation, and makes up for the defect that no express waybill positioning and correcting technology exists in the prior art.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to a convolutional neural network-based method and device for correcting the location of express delivery orders. Background technique [0002] In recent years, the development of e-commerce has enabled the masses of people to buy products from all over the world, and these products will eventually be delivered to people by express delivery. Due to the huge number of couriers, the efficiency of manual sorting is low, and the automation of sorting work through computing-level vision and other methods can effectively improve the sorting efficiency. The positioning and correction of express bills is an indispensable part of the entire automation process. Only when the face bills are positioned correctly can the follow-up tasks be carried out accurately. [0003] At present, there are already many automatic identification technologies for express bills, such as: [00...

AI Technical Summary

Problems solved by technology

The problems with this method are as follows. First, the face sheet does not necessarily occupy the main part of the image, so there may be problems in the detection, making it impossible to obtain the required center point. Second, the corrected text is horizontal In this case, due to the upside-down problem of the image, it is possible to recognize 6 as 9 under the assumption that only numbers exist. If you consider the case of English letters, it is obvious that it will not work.

[0006] It can be seen that the above-mentioned CN108960148A ignores the importance of the positioning and correction of the courier surface, so that it cannot be effective when the surface is tilted and the surface occupies a small image area; CN110705486A also ignores the importance of the positioning and correction of the courier surface. When correcting, the image may be rotated 180 degrees (that is, the text should be facing up but the correction result is facing down), which makes this method ineffective. At the same time, when the face sheet occupies a small overall image area, this method also cannot detect the required textbox, blocking the overall flow

Images