CN117437626A

CN117437626A - License plate quality evaluation method and device

Info

Publication number: CN117437626A
Application number: CN202311443419.XA
Authority: CN
Inventors: 胡中华; 李冰茹; 陈炫憧
Original assignee: Beijing Signalway Technologies Co ltd
Current assignee: Beijing Signalway Technologies Co ltd
Priority date: 2023-11-01
Filing date: 2023-11-01
Publication date: 2024-01-23

Abstract

The application relates to a license plate quality evaluation method, device, equipment and storage medium. Belonging to the technical field of visual inspection, the method comprises the following steps: identifying license plate characters of the target license plate image to obtain character information of the target license plate image; according to character position information of each license plate character, capturing an effective character area image and a single character area image of each license plate character from a target license plate image; performing quality evaluation on the effective character area image, and determining a first quality evaluation result of the effective character area image; performing quality evaluation on each single character area image, and determining a second quality evaluation result of each single character area image; and determining license plate quality evaluation results of the target license plate image according to the first quality evaluation results and the second quality evaluation results. The method based on the application can reflect the quality condition of the license plate to be detected in whole and in part, and can evaluate the recognition effect of the license plate recognition device based on the license plate instruction evaluation result.

Description

License plate quality evaluation method and device

Technical Field

The application relates to the technical field of visual detection, in particular to a license plate quality evaluation method, a license plate quality evaluation device, computer equipment and a storage medium.

Background

With the development of license plate recognition technology, devices carrying the license plate recognition technology (namely license plate recognition devices) are spread at every corner of people's production and life, and great convenience is provided for people's working trips.

When the license plate recognition device recognizes the license plate, sometimes the license plate number cannot be accurately recognized due to various reasons, so that the recognition result is wrong. However, if the error of the identification result cannot be clearly identified, whether the error is related to the quality of the license plate is not clear, and a license plate quality evaluation method for evaluating the license plate identification effect of the license plate identification device is lacking at present.

Disclosure of Invention

In view of the foregoing, it is desirable to provide a license plate quality evaluation method, device, computer device, and storage medium that can accurately evaluate the quality of a license plate.

In a first aspect, the present application provides a license plate quality evaluation method. The method comprises the following steps:

identifying license plate characters of the target license plate image to obtain character information of the target license plate image; the character information comprises character position information of each license plate character in the target license plate image;

according to character position information of each license plate character, capturing an effective character area image and a single character area image of each license plate character from a target license plate image;

Performing quality evaluation on the effective character area image, and determining a first quality evaluation result of the effective character area image;

performing quality evaluation on each single character area image, and determining a second quality evaluation result of each single character area image;

and determining license plate quality evaluation results of the target license plate image according to the first quality evaluation results and the second quality evaluation results.

In one embodiment, performing quality evaluation on the effective character area image, determining a first quality evaluation result of the effective character area image includes:

performing gray level conversion processing on the effective character area image to obtain a first gray level image of the effective character area image;

performing HLS color conversion processing on the effective character area image to obtain a first HLS color image of the effective character area image;

determining a fuzzy evaluation result of the effective character area image according to the first gray level image and the first HLS color image;

and determining a first quality evaluation result of the effective character area image according to the fuzzy evaluation result of the effective character area image.

In one embodiment, determining a blur evaluation result of the effective character area image according to the first grayscale image and the first HLS color image includes:

Determining a first gray standard deviation corresponding to the first gray image and determining a brightness average value corresponding to the first HLS color image;

determining a ratio between the first brightness mean and the first gray standard deviation;

and determining a fuzzy evaluation result of the effective character area image according to the magnitude relation between the first gray standard deviation and the first threshold value and the magnitude relation between the ratio and the second threshold value.

In one embodiment, performing quality evaluation on each single character area image, determining a second quality evaluation result of each single character area image includes:

respectively carrying out gray level conversion processing and HLS color conversion processing on each single character area image to obtain a second gray level image and a second HLS color image corresponding to each single character area image;

determining a contrast evaluation result and a fuzzy evaluation result of each single character area image according to the second gray level image corresponding to each single character area image;

determining exposure evaluation results of the single character area images according to the second HLS color images corresponding to the single character area images;

and determining a second quality evaluation result of each single character area image according to the contrast evaluation result, the blurring evaluation result and the exposure evaluation result of each single character area image.

In one embodiment, determining the contrast evaluation result and the blur evaluation result of each single character area image according to the second gray level image corresponding to each single character area image includes:

determining a second gray standard deviation and a gray mean of a second gray image corresponding to each single character area image;

carrying out normalization processing on the gray average value to obtain a normalized value;

determining a contrast evaluation result of the single character area image according to the magnitude relation between the second gray standard deviation and the third threshold value;

and determining a fuzzy evaluation result of the single character area image according to the magnitude relation between the normalized numerical value and the fourth threshold value.

In one embodiment, determining the exposure evaluation result of each single character area image according to the second HLS color image corresponding to each single character area image includes:

determining the brightness average value of a second HLS color image corresponding to each single character area image;

and determining the exposure evaluation result of each single character area image according to the magnitude relation between each brightness average value and the fifth threshold value.

In one embodiment, the character information further includes character score information of each license plate character, and determining a second quality evaluation result of each single character area image according to the contrast evaluation result, the blur evaluation result, and the exposure evaluation result of each single character area image includes:

Extracting character feature vectors to be detected of each single character area image by adopting a feature extraction network;

determining cosine distances between the character feature vectors to be tested and the standard character feature vectors;

determining interference evaluation results of the images of the single character areas according to the magnitude relation between the cosine distances and the sixth threshold value and the magnitude relation between the score information of the characters and the seventh threshold value;

and determining a second quality evaluation result of each single character area image according to the contrast evaluation result, the blurring evaluation result, the exposure evaluation result and the interference evaluation result of each single character area image.

In one embodiment, the character information further includes character score information of each license plate character, and determining a license plate quality evaluation result of the target license plate image according to the first quality evaluation result and each second quality evaluation result includes:

determining a character score average value according to the character score information;

determining a third quality evaluation result of the target license plate image according to the size relation between the character score average value and the eighth threshold value;

and determining license plate quality evaluation results of the target license plate image according to the third quality evaluation result, the first quality evaluation result and each second quality evaluation result.

In one embodiment, the method further comprises:

correcting the original license plate image to obtain a target license plate image and a correction angle;

correspondingly, the character information also comprises character quantity information, and the license plate quality evaluation result of the target license plate image is determined according to the first quality evaluation result and each second quality evaluation result, and the method comprises the following steps:

determining a shielding evaluation result of the target license plate image according to the size relation between the character number information and the number threshold;

determining an inclination evaluation result of the target license plate image according to the shielding evaluation result of the target license plate image and the relation between the correction angle and the angle threshold;

and determining license plate quality evaluation results of the target license plate image according to the shielding evaluation results, the inclination evaluation results, the first quality evaluation results and the second quality evaluation results.

In a second aspect, the application further provides a license plate quality evaluation device. The device comprises:

the recognition module is used for recognizing license plate characters of the target license plate image to obtain character information of the target license plate image; the character information comprises character position information of each license plate character in the target license plate image;

The intercepting module is used for intercepting an effective character area image and a single character area image of each license plate character from the target license plate image according to character position information of each license plate character;

the first determining module is used for carrying out quality evaluation on the effective character area image and determining a first quality evaluation result of the effective character area image;

the second determining module is used for carrying out quality evaluation on each single character area image and determining a second quality evaluation result of each single character area image;

and the third determining module is used for determining license plate quality evaluation results of the target license plate image according to the first quality evaluation results and the second quality evaluation results.

In a third aspect, the present application also provides a computer device. The computer device comprises a memory storing a computer program and a processor which when executing the computer program performs the steps of:

In a fourth aspect, the present application also provides a computer-readable storage medium. A computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of:

In a fifth aspect, the present application also provides a computer program product. Computer program product comprising a computer program which, when executed by a processor, realizes the steps of:

The license plate quality evaluation method, the license plate quality evaluation device, the computer equipment and the storage medium are used for identifying license plate characters of the target license plate image to obtain character information of the target license plate image; the character information comprises character position information of each license plate character in the target license plate image; according to character position information of each license plate character, capturing an effective character area image and a single character area image of each license plate character from a target license plate image; performing quality evaluation on the effective character area image, and determining a first quality evaluation result of the effective character area image; performing quality evaluation on each single character area image, and determining a second quality evaluation result of each single character area image; and determining license plate quality evaluation results of the target license plate image according to the first quality evaluation results and the second quality evaluation results. The license plate quality evaluation result of the target license plate image is formed based on the first quality evaluation result of the effective character area image and the second quality evaluation result of each single character area image, the quality condition of the license plate to be tested can be reflected in whole and part, and the recognition effect of the license plate recognition device can be evaluated based on the license plate instruction evaluation result.

Drawings

Fig. 1 is an application environment diagram of a license plate quality evaluation method provided in this embodiment;

fig. 2 is a flow chart of a first license plate quality evaluation method provided in the present embodiment;

fig. 3 is a flowchart of a first quality evaluation result of determining an effective character area image according to the present embodiment;

fig. 4 is a schematic flow chart of a first determination of a second quality evaluation result according to the present embodiment;

fig. 5 is a schematic flow chart of a second determination of a second quality evaluation result according to the present embodiment;

fig. 6 is a flowchart of a first license plate quality evaluation result for determining a target license plate image according to the present embodiment;

fig. 7 is a flowchart of a license plate quality evaluation result of a second license plate image determination method according to the present embodiment;

fig. 8 is a flowchart of a second license plate quality evaluation method according to the present embodiment;

fig. 9 is a block diagram of a license plate quality evaluation device according to the present embodiment;

fig. 10 is an internal structural diagram of the computer device provided in the present embodiment.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application will be further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the present application.

The license plate quality evaluation method provided by the embodiment of the application can be applied to an application environment shown in fig. 1. Wherein the terminal 102 communicates with the server 104 via a network. The data storage system may store data that the server 104 needs to process. The data storage system may be integrated on the server 104 or may be located on a cloud or other network server. Specifically, the image acquisition device 102 sends the target license plate image to the server 104, and the server 104 identifies license plate characters of the target license plate image to obtain character information of the target license plate image. The character information comprises character position information of each license plate character in the target license plate image. The server 104 intercepts the effective character area image and the single character area image of each license plate character from the target license plate image according to the character position information of each license plate character. The server 104 performs quality evaluation on the effective character area image, and determines a first quality evaluation result of the effective character area image. The server 104 performs quality evaluation on each individual character area image, and determines a second quality evaluation result of each individual character area image. The server 104 determines a license plate quality evaluation result of the target license plate image according to the first quality evaluation result and each second quality evaluation result. Wherein the image capture device 102 may be, but is not limited to, a video camera, a still camera, or the like, that captures images. The server 104 may be implemented as a stand-alone server or as a server cluster of multiple servers.

In one embodiment, a license plate quality evaluation method is provided, and the method is applied to the server in fig. 1 for illustration, as shown in fig. 2, and includes the following steps:

s201, recognizing license plate characters of the target license plate image to obtain character information of the target license plate image. The character information comprises character position information of each license plate character in the target license plate image.

The target license plate image is a license plate image of a license plate to be detected. The license plate characters refer to characters on a license plate to be detected, and generally comprise Chinese characters, letters, numbers and the like. The character information refers to character information on a target license plate image obtained after the target license plate image is identified, and the character information comprises, but is not limited to, character position information of characters of each license plate. The character position information refers to the position information of each license plate character in the target license plate image.

An alternative implementation manner of this embodiment is as follows: inputting the target license plate image into an image recognition model, and recognizing license plate characters of the target license plate image by the image recognition model to obtain character information of the target license plate image. The image recognition module may be a neural network model configured inside the server.

Another alternative implementation of this embodiment is: the method comprises the steps of inputting a target license plate image into external image recognition equipment, recognizing license plate characters of the target license plate image by the external image recognition equipment, and feeding back character information of the target license plate image to a server.

S202, according to character position information of each license plate character, capturing an effective character area image and a single character area image of each license plate character from a target license plate image.

The effective character area image is an integral area image containing the identified effective license plate characters. The single character area image refers to a partial area image containing only a single character.

Optionally, in this embodiment, an image capturing tool may be used to capture, from the target license plate image, an effective character area image and a single character area image of each license plate character according to character position information of each license plate character. The image capture tool may be an existing software tool for capturing images.

S203, performing quality evaluation on the effective character area image, and determining a first quality evaluation result of the effective character area image.

An alternative implementation manner of this embodiment is as follows: and inputting the effective character area image into a neural network model, and outputting a first quality evaluation result of the effective character area image by the neural network model.

Another alternative implementation of this embodiment is: and acquiring index parameters of the effective character area image, performing quality evaluation on the effective character area image according to the index parameters, and determining a first quality evaluation result of the effective character area image. Among other index parameters, exposure, sharpness, contrast, and the like, are included but are not limited.

In this embodiment, according to each index parameter, the quality evaluation is performed on the effective character area image, and an alternative implementation manner of determining the first quality evaluation result of the effective character area image is as follows:

and comparing each index parameter with a corresponding threshold value, and determining a first quality evaluation result of the effective character area image according to the comparison result, for example, comparing the definition with a definition threshold value, and if the definition exceeds the threshold value, determining that the first quality evaluation result is clear.

S204, performing quality evaluation on each single character area image, and determining a second quality evaluation result of each single character area image.

An alternative implementation manner of this embodiment is as follows: and inputting each single character area image into the neural network model, and outputting a second quality evaluation result of each single character area image by the neural network model.

Another alternative implementation of this embodiment is: and acquiring index parameters of each single character area image, performing quality evaluation on each single character area image according to each index parameter, and determining a second quality evaluation result of each single character area image. Among other index parameters, exposure, sharpness, contrast, and the like, are included but are not limited.

In this embodiment, according to each index parameter, quality evaluation is performed on each single character area image, and an alternative implementation manner of determining the first quality evaluation result of each single character area image is as follows:

and comparing each index parameter with a corresponding threshold value, and determining a first quality evaluation result of each single character area image according to the comparison result, for example, comparing the definition with a definition threshold value, and if the definition exceeds the threshold value, determining that the second quality evaluation result is that the single character area image is clear.

S205, determining license plate quality evaluation results of the target license plate image according to the first quality evaluation results and the second quality evaluation results.

The license plate quality evaluation result refers to an evaluation result obtained by comprehensively determining according to the first quality evaluation and the second quality evaluation results.

An alternative implementation manner of this embodiment is as follows: inputting the first quality evaluation result and each second quality evaluation result into a trained neural network model, and determining the license plate quality evaluation result of the target license plate image by the neural network model.

Another alternative implementation of this embodiment is: and according to the evaluation indexes and the index threshold values in the second quality evaluation results, summarizing to obtain a comprehensive evaluation result, and determining index information of the license plate quality evaluation result of the target license plate image according to the comprehensive evaluation result and the first quality evaluation result. For example, the evaluation index is an interference index, the index content of the interference index in each single character area image is determined to be the number of images with interference by acquiring each second quality evaluation result, the number of images is compared with a number threshold, if the number threshold is exceeded, the index content of the interference index of the comprehensive evaluation result is license plate interference, if the first quality evaluation result is license plate blurring, the license plate quality evaluation result of the target license plate image is determined to be license plate interference according to the comprehensive evaluation result and the first quality evaluation result, and the license plate is blurring. If the number threshold is not exceeded, the index content of the interference index of the comprehensive evaluation result is that no interference exists on the license plate, and if the first quality evaluation result is that the license plate is fuzzy, the license plate quality evaluation result of the target license plate image is determined to be that no interference exists on the license plate according to the comprehensive evaluation result and the first quality evaluation result, and the license plate is fuzzy.

According to the license plate quality evaluation method, license plate characters of the target license plate image are recognized, and character information of the target license plate image is obtained; the character information comprises character position information of each license plate character in the target license plate image; according to character position information of each license plate character, capturing an effective character area image and a single character area image of each license plate character from a target license plate image; performing quality evaluation on the effective character area image, and determining a first quality evaluation result of the effective character area image; performing quality evaluation on each single character area image, and determining a second quality evaluation result of each single character area image; and determining license plate quality evaluation results of the target license plate image according to the first quality evaluation results and the second quality evaluation results. The license plate quality evaluation result of the target license plate image is formed based on the first quality evaluation result of the effective character area image and the second quality evaluation result of each single character area image, the quality condition of the license plate to be tested can be reflected in whole and part, and the recognition effect of the license plate recognition device can be evaluated based on the license plate instruction evaluation result.

In one embodiment, in order to accurately evaluate the quality of the effective character area image, as shown in fig. 3, an alternative implementation manner in S203 is as follows:

s301, gray level conversion processing is carried out on the effective character area image, and a first gray level image of the effective character area image is obtained.

The first gray level image is an image obtained by performing gray level conversion processing on the effective character area image.

Alternatively, in this embodiment, the image processing tool may perform gray-level conversion processing on the effective character area image, so as to obtain the first gray-level image of the effective character area image. The graphic processing tool can be a neural network model, image processing software and the like.

S302, HLS color conversion processing is conducted on the effective character area image, and a first HLS color image of the effective character area image is obtained.

The first HLS color image is an image obtained by performing HLS color conversion processing on the effective character area image, that is, converting the effective character area image into an HLS color image, where in the first HLS color image, an H component represents a hue, an L component represents a brightness, and an S component represents a saturation. The present application is heavily directed to utilizing the L component in HLS color images.

Optionally, in this embodiment, HLS color conversion processing may be performed on the effective character area image by using an image processing tool, so as to obtain a first HLS color image of the effective character area image. The graphic processing tool can be a neural network model, image processing software and the like.

S303, determining a fuzzy evaluation result of the effective character area image according to the first gray level image and the first HLS color image.

The fuzzy evaluation result refers to a result obtained by performing fuzzy evaluation on the effective character area.

Optionally, in this embodiment, a first gray standard deviation corresponding to the first gray image is determined, and a luminance average corresponding to the first HLS color image is determined. And determining a ratio between the first brightness mean and the first gray standard deviation, and determining a fuzzy evaluation result of the effective character area image according to a magnitude relation between the first gray standard deviation and the first threshold value and a magnitude relation between the ratio and the second threshold value.

Specifically, a first gray standard deviation corresponding to the first gray image may be determined according to the gray value of each pixel point in the first gray image, and a brightness average value of the HLS color image may be determined according to the L component of the first HLS color image. A ratio between the first luminance mean and the first gray scale standard deviation is determined. And determining a fuzzy evaluation result of the effective character area image according to the magnitude relation between the first gray standard deviation and the first threshold value and the magnitude relation between the ratio and the second threshold value. If the first gray standard deviation is smaller than the first threshold (for example, 10) and the ratio is larger than the second threshold (for example, 20), the blurring evaluation result of the effective character area image is determined to be image blurring. Otherwise, determining that the image is not blurred as a result of the blurring evaluation of the effective character area image.

S304, determining a first quality evaluation result of the effective character area image according to the fuzzy evaluation result of the effective character area image.

Optionally, in this embodiment, according to the fuzzy evaluation result of the effective character area image, the index content of the evaluation index of the first quality evaluation result is determined, so as to determine the first quality evaluation result of the effective character area image. The first quality evaluation result at least comprises an evaluation index corresponding to the fuzzy evaluation result of the effective character area image. For example, if the result of the fuzzy evaluation of the effective character area image is an image fuzzy, the index content of the evaluation index is an image fuzzy, and then the first quality evaluation result of the effective character area image is determined.

In this embodiment, gray-scale conversion processing is performed on the effective character area image, and a first gray-scale image of the effective character area image is obtained. Performing HLS color conversion processing on the effective character area image to obtain a first HLS color image of the effective character area image. According to the first gray level image and the first HLS color image, the fuzzy evaluation result of the effective character area image can be accurately determined. And determining a first quality evaluation result of the effective character area image according to the fuzzy evaluation result of the effective character area image. Based on the embodiment, the accuracy of the first quality evaluation result of the effective character area image can be effectively improved.

In one embodiment, in order to more accurately and comprehensively determine the second quality evaluation result of each single character area image, as shown in fig. 4, an alternative implementation manner in S204 includes:

s401, gray level conversion processing and HLS color conversion processing are respectively carried out on each single character area image, and a second gray level image and a second HLS color image corresponding to each single character area image are obtained.

The second gray level image is a gray level image obtained by performing gray level conversion on the single character area image. The second HLS color image refers to an image obtained after HLS color conversion processing is performed on the single character region image, that is, an image in which the single character region image is converted into HLS color.

Optionally, in this embodiment, gray conversion processing may be performed on each single character area image by using an image processing tool, so as to obtain a second gray image and a second HLS color image corresponding to each single character area image. The graphic processing tool can be a neural network model, image processing software and the like. It should be noted that, each single character region image needs to be subjected to a gray level conversion process and an HLS color conversion process, that is, after each single character region image is subjected to a conversion process, a second gray level image and a second HLS color image are obtained.

S402, determining a contrast evaluation result and a fuzzy evaluation result of each single character area image according to the second gray level image corresponding to each single character area image.

The contrast evaluation result is an evaluation result obtained according to the second gray level image corresponding to each single character area image. The fuzzy evaluation result is determined according to the second gray level image corresponding to each single character area image.

Optionally, in this embodiment, for each single character area image, a second gray standard deviation and a gray average value of a second gray image corresponding to the single character area image are determined. And carrying out normalization processing on the gray average value to obtain a normalized value, for example, normalizing the gray average value to be between 0 and 1 to obtain the normalized value. And determining a contrast evaluation result of the single character area image according to the magnitude relation between the second gray standard deviation and the third threshold value. And determining a fuzzy evaluation result of the single character area image according to the magnitude relation between the normalized numerical value and the fourth threshold value.

Optionally, in this embodiment, according to a magnitude relation between the second gray standard deviation and the third threshold, an alternative implementation manner of determining the contrast evaluation result of the single character area image is: if the second gray standard deviation is smaller than a third threshold (e.g., 10), the contrast evaluation result of the single character area image is determined to be low in contrast. Otherwise, if the second gray standard deviation is not smaller than the third threshold value, determining that the contrast evaluation result of the single character area image is that the contrast is normal.

Optionally, in this embodiment, according to a magnitude relation between the normalized numerical value and the fourth threshold, an alternative implementation manner of determining the fuzzy evaluation result of the single character area image is: if the normalized value is smaller than a fourth threshold (e.g., 0), determining that the blur evaluation result of the single character area image is image blur; otherwise, if the normalized value is not smaller than the fourth threshold value, determining that the fuzzy evaluation result of the single character area image is that the image is normal.

S403, determining the exposure evaluation result of each single character area image according to the second HLS color image corresponding to each single character area image.

The exposure evaluation result is determined to be the evaluation result related to the image exposure according to the second HLS color image corresponding to the single character area image.

Optionally, in this embodiment, a luminance average value of the second HLS color image corresponding to each single character area image is determined, and an exposure evaluation result of each single character area image is determined according to a magnitude relationship between each luminance average value and the fifth threshold value. Specifically, the brightness average value of each single character area image is determined according to the L component of the second HLS color image corresponding to each single character area image. If the brightness average value is greater than a fifth threshold value (for example, 200), determining that the exposure evaluation result of the single character area image is overexposure; otherwise, if the brightness average value is not greater than the fifth threshold value, determining that the exposure evaluation result of the single character area image is that the exposure is normal.

S404, determining a second quality evaluation result of each single character area image according to the contrast evaluation result, the fuzzy evaluation result and the exposure evaluation result of each single character area image.

Alternatively, in this embodiment, the index content of the evaluation index in the second quality evaluation result may be determined according to the contrast evaluation result, the blur evaluation result, and the exposure evaluation result of each single character area image, so as to determine the second quality evaluation result of each single character area image. The second quality evaluation result at least comprises a contrast index, a blurring index and an exposure index.

In this embodiment, gray conversion processing and HLS color conversion processing are performed on each single character region image, respectively, to obtain a second gray image and a second HLS color image corresponding to each single character region image. And determining a contrast evaluation result and a fuzzy evaluation result of each single character area image according to the second gray level image corresponding to each single character area image. And determining the exposure evaluation result of each single character area image according to the second HLS color image corresponding to each single character area image. According to the contrast evaluation result, the fuzzy evaluation result and the exposure evaluation result of each single character area image, the second quality evaluation result of each single character area image is determined, and the accuracy and the comprehensiveness of the second quality evaluation result are effectively improved.

In one embodiment, in order to further increase the comprehensiveness of the second instruction evaluation result, as shown in fig. 5, an alternative implementation manner in S404 includes:

s501, extracting character feature vectors to be detected of each single character area image by adopting a feature extraction network.

The feature extraction network is a neural network for extracting feature vectors of characters to be detected of a single character area image.

Optionally, in this embodiment, each single character area image is input to the feature extraction network, and the feature vector of the character to be detected of each single character area image is extracted.

Optionally, an optional implementation manner of the training obtaining feature extraction network in this embodiment is:

collecting a plurality of real license plate images, and intercepting a single character area sample image based on the real license plate images; wherein, the character types comprise Arabic numerals 0-9 for 10 kinds, capital English letters A-Z (without O and I) for 24 kinds, provincial (in the open city/autonomous region) first words for 31 kinds (without port Australian platform), and 9 kinds of learning, police, making, collar, hanging, port, australian, emergency and emergency for 74 kinds in total; the license plate types comprise 15 types of national standard cards such as yellow cards, blue cards, new energy cards and the like. And carrying out color reversal processing on the single character small images of black characters such as yellow cards and new energy cards in the intercepted single character area sample images, and forming a first data set by the single character area sample images after the color reversal processing and other single character area sample images. A single character area sample image with clear characters is selected from the first data set to form a second data set. And performing capacity-increasing processing such as random superposition noise, random transformation saturation, brightness and the like on the single character area sample image in the first data set to obtain a third data set. The images in the second data set and the third data set are scaled to a 32x32 size. And extracting the characteristics in the third data set by adopting an original neural Network model (for example, a Residual Network model), and sending the characteristics into a cross entropy function to calculate loss after passing through a pooling layer and a full connection layer until the loss function value is smaller than a preset value or reaches a preset iteration cycle number, and completing training to obtain a characteristic extraction Network. And extracting the features in the second data set through the feature extraction network as standard character feature vectors.

S502, determining cosine distances between the character feature vectors to be tested and the standard character feature vectors.

The cosine distance is also called cosine similarity, and is used for measuring the difference between two vectors.

Optionally, a measuring tool is used to determine the cosine distance between each character feature vector to be measured and the standard character feature vector. The measuring tool can be a neural network model or other calculating tools.

S503, determining the interference evaluation result of each single character area image according to the magnitude relation between each cosine distance and the sixth threshold value and the magnitude relation between each character score information and the seventh threshold value.

The sixth threshold is a preset threshold for measuring whether the cosine distance reaches a certain condition. The seventh threshold value is a preset threshold value for measuring whether the character score information reaches a certain condition. The interference evaluation result is a result obtained by evaluating each single character area image based on the cosine distance and character score information.

Optionally, in this embodiment, if the cosine distance is smaller than the sixth threshold and the character score information is smaller than the seventh threshold, it is determined that the interference evaluation result of the single character area image is that interference exists; otherwise, the interference evaluation result of the single character area image is determined to be that no interference exists.

S504, determining a second quality evaluation result of each single character area image according to the contrast evaluation result, the fuzzy evaluation result, the exposure evaluation result and the interference evaluation result of each single character area image.

Optionally, in this embodiment, the index content of the evaluation index in the second quality evaluation result may be determined according to the contrast evaluation result, the fuzzy evaluation result, the exposure evaluation result and the interference evaluation result of each single character area image, so as to determine the second quality evaluation result of each single character area image. The second quality evaluation result at least comprises a contrast index, a blurring index, an exposure index and an interference index.

In the embodiment, a feature extraction network is adopted to extract the feature vectors of the characters to be detected of each single character area image; determining cosine distances between the character feature vectors to be tested and the standard character feature vectors; according to the magnitude relation between each cosine distance and the sixth threshold value and the magnitude relation between each character score information and the seventh threshold value, the interference evaluation result of each single character area image can be accurately determined; and determining a second quality evaluation result of each single character area image according to the contrast evaluation result, the fuzzy evaluation result, the exposure evaluation result and the interference evaluation result of each single character area image, so that the comprehensiveness of the second quality evaluation result of each single character area image is increased.

In one embodiment, the character information further includes character score information of each license plate character, and in order to increase the comprehensiveness of the license plate quality evaluation result, as shown in fig. 6, an alternative implementation manner in S205 includes:

s601, determining a character score mean value according to the character score information.

The character score average value is an average value obtained by determining according to the score of each character.

S602, determining a third quality evaluation result of the target license plate image according to the size relation between the character score mean value and the eighth threshold value.

The third quality evaluation result is obtained by evaluating the target license plate image according to the character score average value.

Optionally, in this embodiment, if the character score average value is smaller than the eighth threshold value, determining that the third quality evaluation result of the target license plate image is recognition rejection; otherwise, if the character score mean value is not smaller than the eighth threshold value, determining that the third quality evaluation result of the target license plate image is normal recognition.

S603, determining license plate quality evaluation results of the target license plate image according to the third quality evaluation result, the first quality evaluation result and the second quality evaluation results.

Optionally, in this embodiment, the comprehensive evaluation result is determined according to the index content and the relevant threshold value of each evaluation index in each second quality evaluation result, and the index information of the license plate quality evaluation result of the target license plate image is determined according to the comprehensive evaluation result, the third quality evaluation result and the first quality evaluation result. For example, if the evaluation indexes of the second quality evaluation results include an interference index and an exposure index, the index content of the interference index in all the second quality evaluation results of the license plate to be tested is that the number of interference is more than two (i.e. exceeds the set threshold). And determining that the comprehensive evaluation result is that the target license plate image has interference. If the index content of the exposure index is that the number of overexposure exceeds a set threshold (for example, 1), the comprehensive evaluation result indicates that the target license plate image is interfered, and the target license plate image is overexposed. If the first quality evaluation result is image blurring and the third quality evaluation result is refusal recognition, the index information of the license plate quality evaluation result of the target license plate image is that the license plate image to be detected has interference, image overexposure, image blurring and refusal recognition. It should be noted that, the final license plate quality evaluation result is generally that only one or two items have abnormality.

In this embodiment, a character score average value is determined according to the score information of each character. And according to the size relation between the character score mean value and the eighth threshold value, a third quality evaluation result of the target license plate image can be accurately determined. And determining a license plate quality evaluation result of the target license plate image according to the third quality evaluation result, the first quality evaluation result and each second quality evaluation result, so that the comprehensiveness of the license plate quality evaluation result is increased.

In one embodiment, in order to further increase the comprehensiveness of the license plate quality evaluation result, as shown in fig. 7, an alternative implementation manner of S205 includes:

and correcting the original license plate image to obtain a target license plate image and a correction angle.

The correction angle is an angle which is adjusted from an inclined state to a required state when the original license plate image is corrected.

Optionally, in this embodiment, an image correction algorithm may be used to perform correction processing on the original license plate image, so as to obtain the target license plate image and the correction angle.

S701, determining a shielding evaluation result of the target license plate image according to the size relation between the character number information and the number threshold.

The shielding evaluation result is an evaluation result obtained by evaluating the target license plate image according to the character quantity information.

Optionally, in this embodiment, if the number of characters is smaller than the number threshold, determining that the occlusion evaluation result of the target license plate image is occlusion; otherwise, if the number of the characters is not smaller than the number threshold, determining that the shielding evaluation result of the target license plate image is that shielding does not exist.

S702, determining an inclination evaluation result of the target license plate image according to the shielding evaluation result of the target license plate image and the relation between the correction angle and the angle threshold.

The inclination evaluation result is an evaluation result obtained by comprehensively evaluating the target license plate image according to the shielding evaluation result and the correction angle.

Optionally, in this embodiment, if the occlusion evaluation result indicates that there is occlusion and the correction angle is greater than the angle threshold (for example, 60 °), it is determined that the tilt evaluation result of the target license plate image is that the license plate is excessively tilted; otherwise, determining that the inclination evaluation result of the target license plate image is that the license plate inclination is normal.

S703, determining a license plate quality evaluation result of the target license plate image according to the shielding evaluation result, the inclination evaluation result, the first quality evaluation result and the second quality evaluation results.

Optionally, in this embodiment, the index information of the license plate quality evaluation result of the target license plate image is determined according to the shielding evaluation result, the inclination evaluation result, the first quality evaluation result and each second quality evaluation result.

In this embodiment, the license plate shielding evaluation result and the license plate inclination evaluation result can be accurately determined, and the license plate quality evaluation result of the target license plate image is determined according to the shielding evaluation result, the inclination evaluation result, the first quality evaluation result and the second quality evaluation results, so that the comprehensiveness of the license plate quality evaluation result is further increased.

In one embodiment, as shown in fig. 8, an alternative implementation manner of the license plate quality evaluation method includes:

s801, correcting the original license plate image to obtain a target license plate image and a correction angle.

S802, recognizing license plate characters of the target license plate image to obtain character information of the target license plate image. The character information comprises character quantity information of the target license plate image, character position information and character score information of each license plate character.

S803, according to character position information of each license plate character, an effective character area image and a single character area image of each license plate character are intercepted from the target license plate image.

S804, gray level conversion processing is carried out on the effective character area image, and a first gray level image of the effective character area image is obtained.

S805, performing HLS color conversion processing on the effective character area image to obtain a first HLS color image of the effective character area image.

S806, determining a first gray standard deviation corresponding to the first gray image, and determining a brightness average corresponding to the first HLS color image.

S807, a ratio between the first luminance average and the first gray scale standard deviation is determined.

S808, determining a fuzzy evaluation result of the effective character area image according to the magnitude relation between the first gray standard deviation and the first threshold value and the magnitude relation between the ratio and the second threshold value.

S809, determining a first quality evaluation result of the effective character area image according to the fuzzy evaluation result of the effective character area image.

And S8010, respectively carrying out gray level conversion processing and HLS color conversion processing on each single character area image to obtain a second gray level image and a second HLS color image corresponding to each single character area image.

S8011, for each single character area image, determining a second gray standard deviation and a gray mean of a second gray image corresponding to the single character area image.

And S8012, carrying out normalization processing on the gray average value to obtain a normalized value.

And S8013, determining a contrast evaluation result of the single character area image according to the magnitude relation between the second gray standard deviation and the third threshold value.

And S8014, determining a fuzzy evaluation result of the single character area image according to the magnitude relation between the normalized numerical value and the fourth threshold value.

And S8015, determining the brightness average value of the second HLS color image corresponding to each single character area image.

And S8016, determining the exposure evaluation result of each single character area image according to the magnitude relation between each brightness average value and the fifth threshold value.

And S8017, extracting character feature vectors to be detected of each single character area image by adopting a feature extraction network.

And S8018, determining cosine distances between the character feature vectors to be tested and the standard character feature vectors.

And S8019, determining the interference evaluation result of each single character area image according to the magnitude relation between each cosine distance and the sixth threshold value and the magnitude relation between each character score information and the seventh threshold value.

S8020, determining a second quality evaluation result of each single character area image according to the contrast evaluation result, the fuzzy evaluation result, the exposure evaluation result and the interference evaluation result of each single character area image.

S8021, determining a character score mean value according to the character score information.

S8022, determining a third quality evaluation result of the target license plate image according to the size relation between the character score average value and the eighth threshold value.

S8023, determining a shielding evaluation result of the target license plate image according to the size relation between the character number information and the number threshold.

S8024, determining the inclination evaluation result of the target license plate image according to the shielding evaluation result of the target license plate image and the relation between the correction angle and the angle threshold value.

S8025, determining license plate quality evaluation results of the target license plate image according to the shielding evaluation result, the inclination evaluation result, the first quality evaluation result, the third evaluation result and the second quality evaluation results.

In the embodiment, license plate characters of a target license plate image are identified, and character information of the target license plate image is obtained; the character information comprises character position information of each license plate character in the target license plate image; according to character position information of each license plate character, capturing an effective character area image and a single character area image of each license plate character from a target license plate image; performing quality evaluation on the effective character area image, and determining a first quality evaluation result of the effective character area image; performing quality evaluation on each single character area image, and determining a second quality evaluation result of each single character area image; and determining license plate quality evaluation results of the target license plate image according to the first quality evaluation results and the second quality evaluation results. The license plate quality evaluation result of the target license plate image is formed based on the first quality evaluation result of the effective character area image and the second quality evaluation result of each single character area image, the quality condition of the license plate to be tested can be reflected in whole and part, and the recognition effect of the license plate recognition device can be evaluated based on the license plate instruction evaluation result.

Based on the same inventive concept, the embodiment of the application also provides a license plate quality evaluation device for realizing the license plate quality evaluation method. The implementation scheme of the solution provided by the device is similar to the implementation scheme described in the above method, so the specific limitation in the embodiment of the license plate quality evaluation device provided below can be referred to the limitation of the license plate quality evaluation method hereinabove, and will not be repeated here.

In one embodiment, as shown in fig. 9, there is provided a license plate quality evaluation apparatus 1, comprising:

the recognition module 10 is used for recognizing license plate characters of the target license plate image to obtain character information of the target license plate image; the character information comprises character position information of each license plate character in the target license plate image;

the intercepting module 20 is used for intercepting an effective character area image and a single character area image of each license plate character from the target license plate image according to character position information of each license plate character;

a first determining module 30, configured to perform quality evaluation on the valid character area image, and determine a first quality evaluation result of the valid character area image;

A second determining module 40, configured to perform quality evaluation on each single character area image, and determine a second quality evaluation result of each single character area image;

and a third determining module 50, configured to determine a license plate quality evaluation result of the target license plate image according to the first quality evaluation result and each second quality evaluation result.

In one embodiment, the first determining module 30 is further specifically configured to:

In one embodiment, the second determining module 40 is further specifically configured to:

In one embodiment, the third determining module 50 is further specifically configured to:

determining a shielding evaluation result of the target license plate image according to the size relation between the character quantity information and the quantity threshold value;

All or part of each module in the license plate quality evaluation device can be realized by software, hardware and a combination thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.

In one embodiment, a computer device is provided, which may be a server, and the internal structure of which may be as shown in fig. 10. The computer device includes a processor, a memory, an Input/Output interface (I/O) and a communication interface. The processor, the memory and the input/output interface are connected through a system bus, and the communication interface is connected to the system bus through the input/output interface. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer equipment is used for storing the related image data of the license plate to be tested. The input/output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used for communicating with an external terminal through a network connection. The computer program, when executed by the processor, implements a license plate quality assessment method.

It will be appreciated by those skilled in the art that the structure shown in fig. 10 is merely a block diagram of some of the structures associated with the present application and is not limiting of the computer device to which the present application may be applied, and that a particular computer device may include more or fewer components than shown, or may combine certain components, or have a different arrangement of components.

In one embodiment, a computer device is provided, comprising a memory and a processor, the memory having stored therein a computer program, the processor implementing the steps of the method embodiments described above when the computer program is executed.

In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored which, when executed by a processor, implements the steps of the method embodiments described above.

In an embodiment, a computer program product is provided, comprising a computer program which, when executed by a processor, implements the steps of the method embodiments described above.

Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, database, or other medium used in the various embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, high density embedded nonvolatile Memory, resistive random access Memory (ReRAM), magnetic random access Memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric Memory (Ferroelectric Random Access Memory, FRAM), phase change Memory (Phase Change Memory, PCM), graphene Memory, and the like. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, and the like. By way of illustration, and not limitation, RAM can be in the form of a variety of forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), and the like. The databases referred to in the various embodiments provided herein may include at least one of relational databases and non-relational databases. The non-relational database may include, but is not limited to, a blockchain-based distributed database, and the like. The processors referred to in the embodiments provided herein may be general purpose processors, central processing units, graphics processors, digital signal processors, programmable logic units, quantum computing-based data processing logic units, etc., without being limited thereto.

The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.

The above examples only represent a few embodiments of the present application, which are described in more detail and are not to be construed as limiting the scope of the present application. It should be noted that it would be apparent to those skilled in the art that various modifications and improvements could be made without departing from the spirit of the present application, which would be within the scope of the present application. Accordingly, the scope of protection of the present application shall be subject to the appended claims.

Claims

1. A license plate quality evaluation method, the method comprising:

identifying license plate characters of a target license plate image to obtain character information of the target license plate image; the character information comprises character position information of each license plate character in the target license plate image;

according to character position information of each license plate character, capturing an effective character area image and a single character area image of each license plate character from the target license plate image;

2. The method of claim 1, wherein the performing quality evaluation on the valid character area image to determine a first quality evaluation result of the valid character area image comprises:

3. The method of claim 2, wherein determining the fuzzy evaluation result of the valid character area image from the first grayscale image and the first HLS color image comprises:

determining a ratio between the first luminance mean and the first gray scale standard deviation;

4. The method of claim 1, wherein the performing quality evaluation on each single character area image to determine a second quality evaluation result of each single character area image comprises:

5. The method of claim 4, wherein determining the contrast evaluation result and the blur evaluation result of each single character area image based on the second gray level image corresponding to each single character area image comprises:

normalizing the gray average value to obtain a normalized value;

determining a contrast evaluation result of the single character area image according to the magnitude relation between the second gray standard deviation and a third threshold value;

6. The method of claim 4, wherein determining the exposure evaluation result of each single character area image based on the second HLS color image corresponding to each single character area image comprises:

7. The method of claim 4, wherein the character information further includes character score information for each license plate character, and wherein determining the second quality evaluation result for each individual character area image based on the contrast evaluation result, the blur evaluation result, and the exposure evaluation result for each individual character area image comprises:

8. The method of claim 1, wherein the character information further includes character score information of each license plate character, and wherein the determining the license plate quality evaluation result of the target license plate image according to the first quality evaluation result and each second quality evaluation result includes:

determining a third quality evaluation result of the target license plate image according to the size relation between the character score average value and an eighth threshold value;

and determining a license plate quality evaluation result of the target license plate image according to the third quality evaluation result, the first quality evaluation result and each second quality evaluation result.

9. The method according to claim 1, wherein the method further comprises:

correspondingly, the character information further includes character quantity information, and the determining the license plate quality evaluation result of the target license plate image according to the first quality evaluation result and each second quality evaluation result includes:

and determining a license plate quality evaluation result of the target license plate image according to the shielding evaluation result, the inclination evaluation result, the first quality evaluation result and each second quality evaluation result.

10. A license plate quality evaluation device, comprising: