Container number detection method based on deep neural network

Abstract

The invention discloses a container number detection method based on a deep neural network, and the method comprises the following steps: firstly obtaining an RGB image at the rear side of a container, inputting a container image containing a container number into a trained character segmentation neural network model, and obtaining an image set of container number character segmentation; performing perspective transformation and binarization processing on the obtained picture set, and inputting the picture set into a trained character recognition deep neural network to obtain character information in each character box; and finally, screening and combining the obtained text information to obtain an accurate container number. According to the method, the container number can be quickly andaccurately obtained, and higher recognition accuracy and recognition rate are realized.

Description

technical field [0001] The invention designs an image processing method, belongs to the technical field of image processing, and in particular refers to a container number detection method based on a deep neural network. Background technique [0002] With the advancement of science and technology and the increase of the country's import and export trade, the logistics and transportation industry is in an environment of rapid development year by year and a large increase in demand, and the transportation volume of containers is also increasing year by year. For containers, it plays an important role in maritime transportation. With the vigorous development of maritime transportation, a more intelligent and rapid method is needed to obtain the container number information, so as to realize the informatization, intelligence and modernization of containers for large-scale containers. . [0003] The container number is the unique identification code of each container, which is r...

AI Technical Summary

Problems solved by technology

In this way, the container number recognition model based on the deep neural network can identify the container number very well, but the box number positioning still has the defects of traditional algorithms

[0006] The effect of precise positioning of text by deep neural network is far superior to traditional methods, but based on deep neural network end-to-end direct input of container pictures and output container numbers, although the container number can be output quickly, but because end-to-end model training requires a lot of Container data and labeling are not easy to train, and other data sets are quite different from container characters

At the same time, there are still two important processes that are not easy to directly obtain the container number model end-to-end: container number positioning and container number identification division adjustment

While the end-to-end model wants to adjust one of the processes, it may affect the other process.

Images