Vehicle front windshield vin code recognition method and system based on deep learning

Abstract

The invention provides a vehicle front windshield vin code recognition method and system based on deep learning, and the method comprises the steps: a region detection module in a target detection model introduces a text detection algorithm DB based on micro-binarization into a prediction network CPM which is sensitive to context; a vin code text recognition module adds an adversarial generation network GAN on the basis of a text recognition algorithm CRNN algorithm, so that the vehicle front windshield vin code recognition method and system based on deep learning are especially suitable for effective vin code detection and recognition of images with small vin areas and low vin code text resolution.

Description

technical field [0001] The invention relates to the field of vehicle detection, in particular to a method and system for recognizing a vin code on a front windshield of a vehicle based on deep learning. Background technique [0002] The vin code, the full name of the vehicle identification number (Vehicle Identification Number), is a set of seventeen letters or numbers, and is a unique set of numbers used to identify the vehicle. The vin code plate will be fixed on the left side of the dashboard and at the lower left of the front windshield. In vehicle identification, it is often necessary to identify the vin code to obtain detailed vehicle information. [0003] Taking vehicle recognition in the field of unmanned car wash as an example, the current common method is to set up a high-resolution camera at the set position of the unmanned car wash equipment. The image of the corresponding vin code on the glass, and upload the acquired image to the cloud for image processing, in...

AI Technical Summary

Problems solved by technology

However, since the camera on the unmanned car washing equipment is far away from the vin code area of ​​the vehicle, and the size of the vin code area is small, the vin code area on the image captured by the camera on the unmanned car washing equipment is smaller than the entire image. The words are very small, which makes it more difficult for the cloud to detect the vin code area; even if the cloud segment obtains the vin code area, the resolution of the vin code area will be very low, which also increases the difficulty of identifying vin by the current deep learning model

[0004] Of course, there are still some technical solutions that manually collect images of clear vin code areas and use deep learning models to identify them. This technical solution requires manual participation and does not meet the application requirements of unmanned car wash equipment.

Images