License plate recognition model establishing method and device
A license plate recognition and model technology, which is applied in the field of image recognition, can solve the problems that the license plate recognition method cannot adapt to the complex and changeable environment, and the recognition process is time-consuming.
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0040] Embodiments of the present invention provide a method for establishing a license plate recognition model, such as figure 1 shown, including:
[0041] S11. Acquire license plate images of multiple vehicles. The multiple license plate image samples include positive samples and negative samples. The positive samples are real license plate images, and the negative samples are non-real license plate images. That is, the negative samples can be images of other parts of the vehicle or incomplete license plate images, and the massive samples are marked. The label content includes the license plate number and / or the character length of the license plate, and all characters of the fake plate are marked as 0. Finally, all the samples are scaled to the size of the input layer of the neural network model, and the size of the input layer of the neural network model in this embodiment is 128x128. The acquisition of the license plate image uses the ACF detection algorithm to locate t...
Embodiment 2
[0056] An embodiment of the present invention provides a device for establishing a license plate recognition model, such as figure 2 shown, including:
[0057] An acquisition unit 21, configured to acquire license plate images of a plurality of vehicles;
[0058] The license plate image sample acquisition unit 22 is used to expand the license plate image to a preset size to obtain a license plate image sample;
[0059] The training unit 23 is used to use a plurality of license plate image samples and the license plate information in the license plate image samples as training data to train the neural network model until the recognition rate of the license plate information in the license plate image samples by the neural network model is greater than the first preset The threshold or the loss value of the loss function of the neural network model converges to a preset value.
[0060] Preferably, the neural network model is a convolutional neural network model.
[0061] Pre...
Embodiment 3
[0072] An embodiment of the present invention provides a method for identifying license plate information, such as image 3 shown, including:
[0073] S31, acquiring the license plate image to be recognized;
[0074] S32, input the license plate image to be recognized into the model established by the method described in Embodiment 1, and determine the license plate information of the license plate image to be recognized. For example, a license plate whose information is 京A·AB123 and the sixth character "1" is trained and recognized by the machine, polling and recognizing digits 0-9, the maximum posterior probabilities obtained are 0.001, 0.001, 0.99, 0.001, 0.001, 0.001, 0.001, 0.001, 0.001, 0.001, 0.001, 0.001, when the number 1 corresponding to 0.99 is recognized, it is the sixth character of the license plate, and the information of other characters of the license plate is obtained by the method of maximum posterior probability. , the recognition result of each digit of ...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More