CN109492642B - License plate recognition method, license plate recognition device, computer equipment and storage medium - Google Patents
License plate recognition method, license plate recognition device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN109492642B CN109492642B CN201811113599.4A CN201811113599A CN109492642B CN 109492642 B CN109492642 B CN 109492642B CN 201811113599 A CN201811113599 A CN 201811113599A CN 109492642 B CN109492642 B CN 109492642B
- Authority
- CN
- China
- Prior art keywords
- license plate
- character
- target
- image
- characters
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 46
- 238000003860 storage Methods 0.000 title claims abstract description 14
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 16
- 230000011218 segmentation Effects 0.000 claims description 37
- 238000003708 edge detection Methods 0.000 claims description 21
- 238000004590 computer program Methods 0.000 claims description 17
- 238000007781 pre-processing Methods 0.000 claims description 15
- 238000012937 correction Methods 0.000 claims description 10
- 230000005484 gravity Effects 0.000 claims description 8
- 229910052704 radon Inorganic materials 0.000 claims description 7
- SYUHGPGVQRZVTB-UHFFFAOYSA-N radon atom Chemical compound [Rn] SYUHGPGVQRZVTB-UHFFFAOYSA-N 0.000 claims description 7
- 230000009466 transformation Effects 0.000 claims description 6
- 238000005520 cutting process Methods 0.000 claims description 5
- 238000001514 detection method Methods 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 230000005764 inhibitory process Effects 0.000 claims 1
- 238000000605 extraction Methods 0.000 abstract 1
- 230000006870 function Effects 0.000 description 9
- 238000012545 processing Methods 0.000 description 8
- 230000001629 suppression Effects 0.000 description 7
- 230000008859 change Effects 0.000 description 6
- 238000009826 distribution Methods 0.000 description 4
- 238000010606 normalization Methods 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 238000005260 corrosion Methods 0.000 description 2
- 230000007797 corrosion Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- 238000009795 derivation Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012806 monitoring device Methods 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/14—Image acquisition
- G06V30/148—Segmentation of character regions
- G06V30/153—Segmentation of character regions using recognition of characters or words
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Character Discrimination (AREA)
- Character Input (AREA)
Abstract
The invention discloses a license plate recognition method, a license plate recognition device, computer equipment and a storage medium, wherein the license plate recognition method comprises the following steps: the initial license plate image is obtained, the initial license plate image is preprocessed to obtain the target license plate image, a preset sliding window is adopted, sliding operation is carried out on the target license plate image from left to right, each sliding operation is carried out, the target license plate image in the range of the sliding window is obtained to serve as a sliding window image, the license plate image is divided into a plurality of partially overlapped sliding window images, characters in the license plate number repeatedly appear, the situation that part of characters in the license plate number cannot be identified due to the fact that the target license plate image is manually divided is avoided, universality of license plate identification is improved, the obtained sliding window image is input into a convolutional neural network for identification, identification results are obtained, a plurality of repeated characters exist in adjacent identification results, correct extraction of the characters in the license plate number is facilitated, and accuracy of license plate identification is improved.
Description
Technical Field
The present invention relates to the field of image processing technologies, and in particular, to a license plate recognition method, device, computer equipment, and storage medium.
Background
With the development of social economy, more and more automobiles are in road traffic or parking places, which brings many convenience to people's life, but the management of automobiles is becoming more and more complex. Such as vehicle charging and management, traffic flow detection, parking lot charging management, illegal vehicle monitoring, fake license plate vehicle identification, and the like.
To solve the problems, the currently adopted main method is to manage the vehicle by recognizing license plates, the currently adopted license plate recognition technology is to divide license plate images by directly and manually dividing the license plate images, and then to recognize the license plate images after division.
Disclosure of Invention
The embodiment of the invention provides a license plate recognition method, a license plate recognition device, computer equipment and a storage medium, which are used for solving the problems of low accuracy and weak universality of the current license plate recognition.
A license plate recognition method comprising:
acquiring an initial license plate image;
preprocessing the initial license plate image to obtain a target license plate image with a size of a multiplied by b pixels, wherein a and b are positive integers;
sliding operation is carried out on the target license plate image from left to right by adopting a sliding window with preset size of c multiplied by b pixels and taking 1 pixel as a step length, and each sliding operation is carried out to obtain the target license plate image in the range of the sliding window as a sliding window image, wherein c is a positive integer;
after sliding for a-c times, stopping sliding operation to obtain a-c sliding window images;
inputting a-c sliding window images into a convolutional neural network for recognition to obtain a-c recognition results, wherein each recognition result comprises a plurality of characters;
and forming a character string by all the characters in the a-c recognition results, dividing the character string according to a preset dividing mode, acquiring a target character from each obtained sub-character string, and determining a target license plate number according to the target character.
A license plate recognition device comprising:
the acquisition module is used for acquiring an initial license plate image;
the preprocessing module is used for preprocessing the initial license plate image to obtain a target license plate image with a size of a multiplied by b pixels, wherein a and b are positive integers;
The sliding module is used for adopting a sliding window with preset size of c multiplied by b pixels, performing sliding operation from left to right on the target license plate image by taking 1 pixel as a step length, and acquiring the target license plate image within the range of the sliding window as a sliding window image every time, wherein c < a and c are positive integers;
the termination module is used for terminating the sliding operation after sliding a-c times to obtain a-c sliding window images;
the identification module is used for inputting the a-c sliding window images into a convolutional neural network for identification to obtain a-c identification results, wherein each identification result comprises a plurality of characters;
the segmentation module is used for forming character strings from all the characters in the a-c recognition results, segmenting the character strings according to a preset segmentation mode, acquiring target characters from each obtained sub-character string, and determining a target license plate number according to the target characters.
A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the steps of the license plate recognition method described above when the computer program is executed.
A computer readable storage medium storing a computer program which when executed by a processor performs the steps of the license plate recognition method described above.
According to the license plate recognition method, the device, the computer equipment and the storage medium, on one hand, the initial license plate image is obtained, the initial license plate image is preprocessed to obtain the target license plate image with the size of a multiplied by b pixels, then a sliding window with the preset size of c multiplied by b pixels is adopted, sliding operation is carried out on the target license plate image from left to right by taking 1 pixel as a step length, each sliding time, the target license plate image in the range of the sliding window is obtained and is used as one sliding window image, a-c sliding window images are obtained after sliding for a-c times, the sliding operation is stopped, the license plate image is divided into a plurality of partially overlapped sliding window images, characters in license plate numbers repeatedly appear, the situation that part of characters in the license plate number cannot be recognized due to a manually divided character cutting template is avoided, and the universality of license plate recognition is improved; on the other hand, the obtained a-c sliding window images are input into a convolutional neural network for recognition, a-c recognition results are obtained, and a plurality of repeated characters exist in the adjacent recognition results, so that the characters in the license plate number can be extracted correctly, meanwhile, the target characters are extracted from the characters in the a-c recognition results, the target license plate number is further determined, and the accuracy of license plate recognition is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the description of the embodiments of the present invention will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic view of an application environment of a license plate recognition method according to an embodiment of the present invention;
fig. 2 is a flowchart of an implementation of a license plate recognition method according to an embodiment of the present invention;
fig. 3 is a flowchart illustrating implementation of step S20 in the license plate recognition method according to the embodiment of the present invention;
fig. 4 is a flowchart illustrating implementation of step S21 in the license plate recognition method according to the embodiment of the present invention;
fig. 5 is a flowchart illustrating implementation of step S60 in the license plate recognition method according to the embodiment of the present invention;
fig. 6 is a flowchart of implementation of step S64 in the license plate recognition method according to the embodiment of the present invention;
fig. 7 is a schematic diagram of a license plate recognition device according to an embodiment of the present invention;
fig. 8 is a schematic diagram of a computer device according to an embodiment of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are some, but not all embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 1, fig. 1 illustrates an application environment of a license plate recognition method according to an embodiment of the present invention. The license plate recognition method is applied to a license plate recognition scene aiming at the photographed license plate of the vehicle. The identification scene comprises a server and a client, wherein the server and the client are connected through a network, the client sends a photographed license plate image to the server, the server receives the license plate image sent by the client and identifies the license plate image, the client can be particularly but not exclusively an overspeed camera, a space network monitor, an electronic police, various personal computers, a notebook computer, a smart phone, a tablet personal computer and portable wearable equipment, and the server can be particularly realized by an independent server or a server cluster formed by a plurality of servers.
Referring to fig. 2, fig. 2 shows a license plate recognition method according to an embodiment of the present invention, and the method is applied to the server in fig. 1 for illustration, and is described in detail as follows:
s10: and acquiring an initial license plate image.
Specifically, after the client obtains the initial license plate image, the initial license plate image is sent to the server through a network transmission protocol, and the server receives the initial license plate image through the network transmission protocol.
The client can be monitoring equipment with shooting functions, such as an overspeed camera, a space net monitoring device, an electronic police device and the like, can directly shoot to obtain an initial license plate image, can also be intelligent terminal equipment, such as various personal computers, notebook computers, intelligent mobile phones or tablet computers and the like, and has the functions of storing the initial license plate image and performing network interaction with the server.
Among them, network transport protocols include, but are not limited to: internet control message protocol (Internet Control Message Protocol, ICMP), address resolution protocol (ARP Address Resolution Protocol, ARP), and file transfer protocol (File Transfer Protocol, FTP), among others.
S20: and preprocessing the initial license plate image to obtain a target license plate image with a size of a multiplied by b pixels, wherein a and b are positive integers.
Specifically, due to the influence of factors such as shooting angle, distance, high-speed running of an automobile and the like, the acquired initial license plate image is low in quality, and the accuracy of direct identification is low, so that the initial license plate image needs to be preprocessed so as to reduce the influence caused by the factors, the accuracy of subsequent identification is improved, and a target license plate image with a size of a multiplied by b is obtained after the initial license plate image is preprocessed.
Wherein the pretreatment includes, but is not limited to: image cropping, normalization, tilt correction, and the like.
The basic unit of a and b is a preset value, the basic unit of a and b is a width of 1 pixel, a is a horizontal side length, that is, a width of a pixel is a horizontal side length, b is a vertical side length, that is, b width of b pixel is a vertical side length, a and b are positive integers, and specific values thereof can be set according to needs, and are not particularly limited herein.
Preferably, in the embodiment of the present invention, the value of a takes 140, and the value of b takes 28, that is, the target license plate image with 140×28 pixels is finally obtained.
S30: and sliding operation is carried out on the target license plate image from left to right by adopting a sliding window with preset size of c multiplied by b pixels and taking 1 pixel as a step length, and one target license plate image in the range of the sliding window is obtained as one sliding window image in each sliding, wherein c is a positive integer.
Specifically, through presetting a sliding window with the size of c×b pixels, sliding operation is performed on the target license plate image from left to right by taking 1 pixel as a step length, and each sliding operation is performed, the target license plate image in the range of the sliding window is obtained and is used as one sliding window image, so that after multiple sliding operations, a plurality of sliding windows have the same overlapped image parts, and the follow-up optimization of target characters is facilitated.
The width b of the sliding window is the same as the width of the target license plate image, and the length c can be set according to practical situations, which is not particularly limited herein.
Preferably, the value of c in the embodiment of the present invention is set to 28, that is, a square of 28×28 pixels is used as a sliding window for sliding.
S40: after sliding a-c times, the sliding operation is terminated, and a-c sliding window images are obtained.
Specifically, the length of the target license plate image is a, the length of the sliding window is b, the sliding step length is 1, the right end of the sliding window is overlapped with the right end of the target license plate image after sliding for a-c times, namely the sliding window moves the rightmost end of the target license plate image, and a-c sliding window images are obtained in total at the moment.
S50: inputting the a-c sliding window images into a convolutional neural network for recognition to obtain a-c recognition results, wherein each recognition result comprises a plurality of characters.
Specifically, the a-c sliding window images obtained in the step S40 are used as input images, input into a trained convolutional neural network for recognition, 71 classifiers are preset in a fully connected layer of the convolutional neural network, each classifier corresponds to one preset character, and 71 preset characters are respectively: the method comprises the steps of inputting preset 35 Chinese characters, preset 26 capital English letters and preset 10 Arabic numerals into a full-connection layer after convolution operation of a sliding window image, and classifying and identifying by using preset 71 classifiers to obtain an identification result.
It should be noted that each recognition result is one character or a plurality of characters of the 71 preset characters.
Wherein, the preset 35 Chinese characters comprise: beijing, jin, ji, jin, mongolian, liao, ji, black, hu, su, zhe, wan, min, gan, lu, yu, hu, xiang, yue, gui, qong, yu, chuan, qian, yun, tibetan, shan, gan, qing, tibetan, ning, xin, tai, gang and Australia.
Wherein, the preset 26 capital English letters comprise: A. b, C, D, E, F, G, H, I, J, K, L, M, N, O, P, Q, R, S, T, U, V, W, X, Y, Z.
Wherein, 10 Arabic numerals of presetting include: 0. 1, 2, 3, 4, 5, 6, 7, 8, 9.
The convolutional neural network (Convolutional Neural Network, CNN) is a feedforward neural network, and an artificial neuron of the convolutional neural network can respond to surrounding units in a part of coverage range, so that image processing can be performed quickly and efficiently.
S60: and forming character strings by all the characters in the a-c recognition results, dividing the character strings according to a preset dividing mode, acquiring target characters from each obtained sub-character string, and determining a target license plate number according to the target characters.
Specifically, all the characters in the a-c recognition results are formed into character strings, the character strings are segmented according to a preset segmentation mode to obtain at least two sub-character strings, one target character is optimized from each sub-character string, and the target license plate number is obtained by combining each target character.
The preset dividing mode may be to divide the character strings equally according to the number of sub-character strings to be divided, or divide the sub-character strings according to a preset range or a preset dividing point, or set according to actual conditions, which is not limited specifically herein.
It should be noted that, because the license plate number is generally 7 characters, as a preferred mode, the embodiment of the invention divides the character string composed of a-c recognition results into 7 sub-character strings according to a preset division mode.
In the embodiment, an initial license plate image is obtained, the initial license plate image is preprocessed to obtain a target license plate image with the size of a multiplied by b, a sliding window with the preset size of c multiplied by b pixels is adopted, sliding operation is carried out on the target license plate image from left to right by taking 1 pixel as a step length, each sliding time, the target license plate image in the range of the sliding window is obtained, the target license plate image is taken as one sliding window image, a-c sliding window images are obtained after sliding a-c times, sliding operation is stopped, the license plate image is divided into a plurality of partially overlapped sliding window images, characters in license plate numbers repeatedly appear, the situation that part of characters in license plate numbers cannot be identified due to a manually divided character cutting template is avoided, the universality of license plate identification is improved, the obtained a-c sliding window images are input into a convolutional neural network for identification, a-c identification results exist in adjacent identification results, a plurality of repeated characters are favorable for correctly extracting characters in the license plate numbers, meanwhile, the characters in the target license plate number are extracted from the a-c identification results, the target license plate number is extracted, and the license plate number recognition accuracy is improved.
Based on the corresponding embodiment of fig. 2, a specific implementation method for preprocessing the initial license plate image mentioned in step S20 to obtain the target license plate image with a size of a×b pixels is described in detail below by using a specific embodiment.
Referring to fig. 3, fig. 3 shows a specific implementation flow of step S20 provided in the embodiment of the present invention, which is described in detail below:
s21: and acquiring the upper boundary of the license plate and the lower boundary of the license plate in the initial license plate image through an edge detection algorithm.
Specifically, due to the influence of factors such as shooting angles and distances, the shot license plate images generally comprise a plurality of non-license plate images outside the license plate areas, and the non-license plate images can interfere with subsequent license plate recognition, so that in order to improve the accuracy of subsequent license plate recognition, the upper boundary of the license plate in the initial license plate image and the lower boundary of the license plate are required to be found out through an edge checking algorithm, and the license plate range in the initial license plate image is determined.
The edge detection algorithm is an algorithm for performing edge detection, and edge detection is a fundamental problem in image processing and computer vision, and in the embodiment of the present invention, the purpose of edge detection is to identify points with obvious brightness change in license plate images, namely points of license plate boundaries, and significant changes in image attributes generally reflect important events and changes of the attributes, and the image attributes include but are not limited to: discontinuities in depth, surface direction discontinuities, material property changes, scene lighting changes, etc.
The edge of the image refers to the region where the gray level of the image changes sharply, and the change of the gray level of the image can be reflected by the gradient of the gray level distribution.
Common edge detection algorithms include, but are not limited to: sobel operator edge detection algorithm, laplace operator edge detection algorithm, lobez cross edge detection (Roberts Cross operator) algorithm, canny multi-level edge detection algorithm, and the like.
Preferably, the edge detection algorithm adopted by the embodiment of the invention is a Canny multi-stage edge detection algorithm.
It is worth to say that, according to the edge detection algorithm, the obtained upper boundary of the license plate and the lower boundary of the license plate are two segments. That is, the upper boundary of the obtained license plate comprises a left vertex and a right vertex, and the lower boundary of the obtained license plate comprises a left vertex and a right vertex.
S22: and determining a range image of the license plate according to the upper boundary of the license plate and the lower boundary of the license plate.
Specifically, the upper boundary of the license plate obtained in step S21 is connected with the lower boundary of the license plate, that is, the left vertex of the upper boundary of the license plate is connected with the left vertex of the lower boundary of the license plate, and the right vertex of the upper boundary of the license plate is connected with the right vertex of the lower boundary of the license plate, so that a quadrilateral is obtained, and an image within the range of the quadrilateral is used as a range image of the license plate.
S23: and performing inclination correction on the range image by using the radon transformation to obtain a corrected base image.
Specifically, due to the influence of shooting angles and distances, the acquired initial license plate image is inclined, and further the acquired range image is inclined, so that in order to improve the accuracy of subsequent license plate recognition, the range image is subjected to inclination correction through radon transformation, and a corrected basic image is obtained.
The Radon transform (Radon transform) is a method for obtaining a corrected image by determining an image inclination angle by determining an angle when a maximum projection value is found through directional projection superposition.
S24: and cutting the basic image by taking the center of gravity of the basic image as the center to obtain a target license plate image with the size of a multiplied by b pixels.
Specifically, the base image obtained in step S23 is cut with the center of gravity of the base image as the center, a as the horizontal side length and b as the vertical side length, to obtain a rectangular target license plate image with a×b pixels.
Preferably, in the embodiment of the present invention, a is preset to 140, and b is preset to 28, i.e. the final target license plate image with 140×28 pixels is obtained.
For example, in one embodiment, the center of gravity of the base image is (82, 21), the horizontal side length is 140, and the vertical side length is 28, so that the upper left corner vertex coordinates (12, 7), the upper right corner vertex coordinates (152,7), the lower left corner vertex coordinates (12, 35), the lower right corner vertex coordinates (152, 35), a 140×28 pixel rectangle composed of the four vertices, and the base image is cut along the rectangle, so that the image within the rectangle range, that is, the target license plate image is obtained.
Preferably, after the target license plate image is obtained, the embodiment of the invention also carries out mean value removal and normalization on the target license plate image, and eliminates the difference between different dimension image data in the target license plate image.
The normalization is to normalize the image feature amplitude in the target license plate image to the same range, namely dividing the standard deviation of all the image features by each image feature, and taking the obtained result as the image feature after the normalization of the image features.
The mean value removing refers to that each dimension of the image features in the target license plate image is centered to 0, namely, the center point of the target license plate image is pulled back to the origin of the coordinate system.
In this embodiment, the upper boundary of the license plate and the lower boundary of the license plate in the initial license plate image are obtained through an edge detection algorithm, and then the range image of the license plate is determined according to the upper boundary of the license plate and the lower boundary of the license plate, and then the range image is subjected to inclination correction by using radon transformation to obtain a corrected basic image, and then the basic image is cut by taking the center of gravity of the basic image as the center to obtain a target license plate image with a size of a×b pixels, so that when the basic image is used for identification subsequently, identification errors caused by interference and inclination factors of images outside the range of the license plate are avoided, the quality of the target license plate image is enhanced, and the accuracy of identification of subsequent license plates is improved.
Based on the corresponding embodiment of fig. 3, a specific implementation method for acquiring the upper boundary of the license plate and the lower boundary of the license plate in the initial license plate image by the edge detection algorithm mentioned in step S21 is described in detail below by using a specific embodiment.
Referring to fig. 4, fig. 4 shows a specific implementation flow of step S21 provided in the embodiment of the present invention, which is described in detail below:
s211: and removing noise from the initial license plate image through Gaussian blur to obtain a denoising license plate image.
Specifically, the edge of the license plate image is a high-frequency signal, but the noise of the image is also concentrated in the high-frequency signal and is easily recognized as the edge by mistake, so that the noise of the image needs to be removed, and the interference of the noise of the image on the determined edge is avoided. In the embodiment of the invention, the Gaussian blur is adopted to remove noise of the initial license plate image, and the denoised license plate image is obtained.
The edge of the license plate image refers to an area with sharp change of gray level in the image at the juncture of the license plate area and the non-license plate area. The change in gray scale of the image can be reflected by the gradient of the gray scale distribution.
The noise of the image, that is, the noise of the image, refers to unnecessary or redundant interference information existing in the image data, and the noise seriously affects the quality of the image, so that the noise must be corrected before the image enhancement processing and the classification processing.
Among these, gaussian Blur (Gaussian blue), also known as Gaussian smoothing, is an image Blur filter that uses a normal distribution to calculate the transform for each pixel in an image, typically to reduce image noise and to reduce the level of detail.
It should be noted that, both the image edge and the noise are high-frequency signals, so that the radius selection of the gaussian blur is important, and too large a radius can easily prevent some weak edge points from being detected, and the specific setting of the radius can be adjusted according to the actual situation, which is not limited herein.
S212: and calculating gradient values in the horizontal direction and the vertical direction of the denoising license plate image by using a preset gradient operator to obtain an initial gradient value set.
Specifically, the edges of the image can point to different directions, and gradient values in the horizontal direction and the vertical direction of the denoising license plate image are calculated by using a preset operator, so that an initial gradient value set is obtained.
The digital image is a discrete point value spectrum, and can also be called a two-dimensional discrete function, the gradient of the image is the result of the derivation of the two-dimensional discrete function, and a gradient operator, namely a method for calculating the gradient.
Wherein, the preset gradient operator includes but is not limited to: sobel operator (Sobel operator), prewitt operator, roberts operator (Roberts operator), and Canny operator.
Preferably, the gradient operator adopted in the embodiment of the invention is a Canny operator.
S213: and carrying out edge refinement treatment on the initial gradient value set by adopting a non-maximum value suppression mode to obtain a gradient edge with a pixel width.
Specifically, edge refinement processing is performed on the initial gradient value set by adopting a non-maximum value suppression mode, so as to obtain a gradient edge with a pixel width.
Where non-maximum suppression (Non Maximum Suppression, NMS) is an element that suppresses non-maxima, it is understood that a maximum search is locally performed, helping to preserve the local maximum gradient while suppressing all other gradient values, meaning that only the sharpest locations in the gradient change are preserved.
For example, in a specific embodiment, in the vertical direction, a gradient value with a width of 4 pixels forms a part, and a non-maximum suppression mode is adopted in the part of the cover, and a pixel point with the maximum gradient value in the gradient value of the part is searched out and used as a gradient edge, so that edge refinement is realized.
S214: and filtering weak edge points in the gradient edge by using a preset double threshold value to obtain strong edge points in the gradient edge.
Specifically, edge pixels are distinguished by setting a double threshold, i.e., a high threshold and a low threshold. If the edge pixel point gradient value is greater than the high threshold, it is considered a strong edge point, if the edge gradient value is less than the high threshold, it is greater than the low threshold, it is marked as a weak edge point, and points less than the low threshold are suppressed, the strong edge point may be confirmed as a true edge, but the weak edge point may be a true edge, or may be caused by noise or color change, and in order to obtain an accurate result, the weak edge point is also suppressed in the embodiment of the present invention.
S215: and determining the upper boundary of the license plate and the lower boundary of the license plate according to the strong edge points.
Specifically, according to the strong edge points, the upper boundary of the license plate and the lower boundary of the license plate are obtained through image processing such as corrosion extension.
Wherein corrosion may eliminate individual ones of the strong edge points and extension may connect adjacent but unconnected strong edge points.
In this embodiment, noise is removed from an initial license plate image through gaussian blur, a denoised license plate image is obtained, gradient values in the horizontal direction and the vertical direction of the denoised license plate image are calculated by using a preset gradient operator, an initial gradient value set is obtained, further, edge refinement is performed on the initial gradient value set in a non-maximum value suppression mode, a gradient edge with a pixel width is obtained, weak edge points in the gradient edge are filtered by using a preset double threshold value, a strong edge point in the gradient edge is obtained, and then the upper boundary of a license plate and the lower boundary of the license plate are determined according to the strong edge point, so that the accuracy of license plate edge detection is improved, and the subsequent determination of a license plate range is facilitated.
Based on the corresponding embodiment of fig. 2, the specific implementation method for forming the character strings from all the characters in the a-c recognition results mentioned in step S60 by using a specific embodiment, dividing the character strings according to a preset dividing manner, obtaining the target characters from each obtained sub-character string, and determining the target license plate number according to the target characters is described in detail below.
Referring to fig. 5, fig. 5 shows a specific implementation flow of step S60 provided in the embodiment of the present invention, which is described in detail below:
s61: sequentially placing characters in the a-c recognition results into a character set X to obtain X= { X 1 ，x 2 ，...，x q Q is the total number of characters contained in a-c recognition results, x i For the ith character in character set X, i ε [1, q]Q is a positive integer.
Specifically, sequentially placing characters contained in each recognition result into a character set X with an initial value of empty to obtain X= { X 1 ，x 2 ，...，x q Q is the total number of characters contained in a-c recognition results, x i Is the ith character in character set X.
For example, in one embodiment, 112 recognition results are obtained, wherein the first 20 recognition results are: "," "I", "L", "O", "Shanghai" Shanghai "," Sha wherein the content in each double-quotation mark is a recognition result, each recognition result can be 1 character, or may be empty, or may be a plurality of characters, the 20 recognition results are sequentially stored into a character set X to obtain X= { "I", "L", "O", "Shanghai", shanghai ",.
S62: and determining k dividing points of the character set X according to k+1 preset ranges, wherein k is smaller than q, and k is a positive integer.
Specifically, after the character set X is obtained in step S61, the character set X needs to be segmented into a plurality of sub-character sets according to the number of target characters, and then a target character is determined according to each sub-character set.
In the embodiment of the invention, provided that k+1 ranges are preset according to the number k+1 of characters of the license plate to be recognized, k segmentation points in the character set X are determined according to the k+1 preset ranges.
The preset range is used to define a preferred range of each target character, for example, a first preset range is (0, 15), that is, the first recognition result to the 15 th recognition result are the first preset range.
Taking the example in step S61 as an example, if the first preset range is (0, 15), the obtained character set X= { "I", "L", "O", "Shanghai" Shanghai "," Shaw "," Sha then the 15 th recognition result is selected as a segmentation point, i.e. the position between the 15 th element "I" and the 16 th element "B" of the character set X is the first segmentation point.
It is noted that the preset range may include a fraction, for example, the second preset range is (15, 31.5), that is, the second preset range includes all characters from the 16 th recognition result to the 31 st recognition result, and the first 50% of characters of the second recognition result.
The target characters refer to characters in license plate number characters, and most license plate numbers are 7 target characters or 8 target characters currently, so that the number of target characters is usually 7 or 8, and the target characters can be set to other numerical values according to actual needs, and the number is not particularly limited.
S63: and performing character segmentation on q characters in the character set X by using k segmentation points to obtain k+1 sub-character sets.
Specifically, character segmentation is performed on characters in the character set X by using the obtained k segmentation points to obtain k+1 sub-character sets, namely k+1 preferred ranges.
The segmentation refers to taking two adjacent segmentation points as boundaries, taking all characters in the range of the two adjacent segmentation points as a sub-character set, taking all characters from a first character set to the segmentation point range as a first sub-character set for a first segmentation point, and taking all characters from a last segmentation point to a last character range as a last sub-character set for a last segmentation point.
S64: and obtaining the character with the largest occurrence number in each sub-character set as a target character to obtain k+1 target characters.
Specifically, counting the occurrence times of each character in each sub-character set aiming at each sub-character set, and taking the character with the largest occurrence times as the target character of the sub-character set to obtain k+1 target characters.
For example, in one embodiment, the first and second embodiments, one of the sub-character sets is { "B", "L", "B") "B", "O", "L", "B" }, the result is that the result contains 12 characters 'B', two characters 'L', and one character 'O', and the character 'B' is used as the target character of the sub-character set.
S65: and combining the k+1 target characters according to the sequence of the k+1 preset ranges to obtain the target license plate number.
Specifically, the obtained k+1 target characters are ordered according to the order of the k+1 preset ranges, and the target license plate number is obtained.
For example, in a specific embodiment, there are 7 preset ranges, the target character of the sub-word character set corresponding to the first preset range is "Shanghai", the target character of the sub-word character set corresponding to the second preset range is "Shanghai", the target character of the sub-word character set corresponding to the third preset range is "2", the target character of the sub-word character set corresponding to the fourth preset range is "6", the target character of the sub-word character set corresponding to the fifth preset range is "A", the target character of the sub-word character set corresponding to the sixth preset range is "6", the target character of the sub-word character set corresponding to the seventh preset range is "Shanghai", and the target license plate number is "Shanghai B26A63" obtained by combining.
In this embodiment, the characters in the a-c recognition results are sequentially placed in the character set X, k segmentation points of the character set X are determined according to k+1 preset ranges, and then the q characters in the character set X are segmented by using the k segmentation points to obtain k+1 sub-character sets, the character with the largest occurrence number in each sub-character set is obtained as a target character, k+1 target characters are obtained, then the k+1 target characters are combined according to the order of the k+1 preset ranges to obtain a target license number, so that the characters in the target license image can be obtained without segmenting the target license image, the problem that part of characters cannot be recognized due to the fact that the target license image is subjected to preset templates is avoided, the accuracy of the recognized characters is not high, and the accuracy and stability of license recognition are improved.
Based on the corresponding embodiment of fig. 5, a detailed description will be given below of a specific implementation method for obtaining the k+1 target characters by using, as the target characters, the characters with the largest occurrence number in each sub-character set mentioned in step S64.
Referring to fig. 6, fig. 6 shows a specific implementation flow of step S64 provided in the embodiment of the present invention, which is described in detail below:
S641: and acquiring the character with the largest occurrence number in the first sub-character set as the first character.
Specifically, the number of occurrences of each character in the first sub-character set is counted, and the character with the largest number of occurrences is taken as the first character.
For example, in one embodiment, the first sub-character set is { "I", "L", "shanghu", "U", "shanghu", "I" }, and the statistics result is that the character "L" appears 6 times, the character "shanghu" appears 5 times, the character "U" appears two times, the character "I" appears 2 times, and the character "L" with the largest number of occurrences is taken as the first character.
S642: if the first character is a non-Chinese character, the Chinese character with the largest occurrence number in the first sub-character set is obtained as the first target character.
Specifically, in actual use of the license plate, the first target character of the license plate number is a kanji, but in recognition of the neural network, due to insufficiency of the kanji in the partial sliding image, the kanji may be recognized as a capital english letter, and if the first character obtained in step S641 is a non-chinese character, the chinese character with the largest occurrence number in the first sub-character set is obtained as the first target character.
Taking the example in step 641 as an example, the statistics result is that the character "L" appears 6 times, the character "Shanghai" appears 5 times, the character "U" appears twice, the character "I" appears 2 times, the character "L" with the largest number of appearance is taken as the first character, and the first character is detected, and the "L" is a non-Chinese character, so that the Chinese character "Shanghai" with the largest number of appearance is taken as the first target character.
S643: and if the first character is a Chinese character, taking the first character as a first target character.
Specifically, if the first character obtained in step S641 is a chinese character, the first character is taken as the first target character.
It should be noted that, the step S642 and the step S643 are not necessarily executed sequentially, and may be executed in parallel, which is not limited herein.
S644: and acquiring the character with the largest occurrence number in the jth character set as the jth target character, wherein j is greater than 1 and less than or equal to k+1.
Specifically, for the jth sub-character set, counting the number of times of the characters appearing in the sub-character set, and obtaining the character with the largest number of times as the target character of the sub-character set.
For example, in a specific embodiment, the number of target characters is 7, that is, k=6, and for the 2 nd to 7 th sub-character sets, the character with the largest occurrence number in each sub-character set is obtained as the target character, and the 6 target characters are respectively: "B", "2", "3", "D", "6" and "3".
In this embodiment, the character with the largest occurrence number in the first sub-character set is obtained and used as the first character, and whether the first character is a chinese character is determined, if the first character is a non-chinese character, the chinese character with the largest occurrence number in the first sub-character set is obtained and used as the first target character, and if the first character is a chinese character, the first character is used as the first target character, and for other characters than the first character, the character with the largest occurrence number in the character set is obtained and used as the target character of the character set, so that each target character of the license plate number is obtained quickly and accurately, and the accuracy of license plate recognition is improved.
It should be understood that the sequence number of each step in the foregoing embodiment does not mean that the execution sequence of each process should be determined by the function and the internal logic, and should not limit the implementation process of the embodiment of the present invention.
In an embodiment, a license plate recognition device is provided, where the license plate recognition device corresponds to the license plate recognition method in the above embodiment one by one. As shown in fig. 7, the license plate recognition device includes an acquisition module 10, a preprocessing module 20, a sliding module 30, a termination module 40, a recognition module 50, and a segmentation module 60. The functional modules are described in detail as follows:
The acquisition module 10 is used for acquiring an initial license plate image;
the preprocessing module 20 is used for preprocessing the initial license plate image to obtain a target license plate image with a size of a multiplied by b pixels, wherein a and b are positive integers;
the sliding module 30 is configured to perform sliding operation from left to right on the target license plate image with 1 pixel as a step size by adopting a sliding window with a preset size of c×b pixels, and obtain the target license plate image within the range of the sliding window as a sliding window image each time, where c < a and c are positive integers;
a termination module 40, configured to terminate the sliding operation after sliding a-c times, so as to obtain a-c sliding window images;
the recognition module 50 is used for inputting the a-c sliding window images into the convolutional neural network for recognition to obtain a-c recognition results, wherein each recognition result comprises a plurality of characters;
the segmentation module 60 is configured to segment the character string formed by all the characters in the a-c recognition results according to a preset segmentation method, obtain a target character from each obtained sub-character string, and determine the target license plate number according to the target character.
Further, the preprocessing module 20 includes:
The detecting unit 21 is configured to obtain an upper boundary of a license plate and a lower boundary of the license plate in the initial license plate image through an edge detection algorithm;
a determining unit 22, configured to determine a range image of the license plate according to an upper boundary of the license plate and a lower boundary of the license plate;
a correction unit 23 for performing tilt correction on the range image using radon transform to obtain a corrected base image;
and the clipping unit 24 is used for clipping the basic image by taking the gravity center of the basic image as the center to obtain a target license plate image with a size of a multiplied by b pixels.
Further, the detection unit 21 includes:
the denoising subunit 211 is configured to perform noise removal on the initial license plate image through gaussian blur, so as to obtain a denoised license plate image;
a calculating subunit 212, configured to calculate gradient values in the horizontal direction and the vertical direction of the denoised license plate image by using a preset gradient operator, so as to obtain an initial gradient value set;
a refinement subunit 213, configured to perform edge refinement processing on the initial gradient value set by using a non-maximum suppression manner, so as to obtain a gradient edge with a pixel width;
a filtering subunit 214, configured to filter weak edge points in the gradient edge by using a preset dual threshold value, so as to obtain strong edge points in the gradient edge;
The delimiting subunit 215 is configured to determine an upper boundary of the license plate and a lower boundary of the license plate according to the strong edge points.
Further, the segmentation module 60 includes:
a generating unit 61 for sequentially placing the characters in the a-c recognition results into the character set X to obtain x= { X 1 ，x 2 ，...，x q Q is the total number of characters contained in a-c recognition results, x i For the ith character in character set X, i ε [1, q]Q is positiveAn integer;
a selecting unit 62, configured to determine k segmentation points of the character set X according to k+1 preset ranges, where k is a positive integer and k is < q;
a segmentation unit 63, configured to perform character segmentation on q characters in the character set X by using k segmentation points, to obtain k+1 sub-character sets;
a optimizing unit 64, configured to obtain, as target characters, characters with the largest occurrence number in each sub-character set, and obtain k+1 target characters;
and the combining unit 65 is configured to combine the k+1 target characters according to the order of the k+1 preset ranges, so as to obtain the target license plate number.
Further, the preferred unit 64 includes:
a statistics subunit 641, configured to obtain, as a first character, a character having the largest number of occurrences in the first sub-character set;
a first screening subunit 642, configured to obtain, if the first character is a non-chinese character, a chinese character having the largest number of occurrences in the first sub-character set as the first target character;
A second deleting subunit 643, configured to take the first character as the first target character if the first character is a chinese character;
a determining subunit 644, configured to obtain, as the jth target character, a character that occurs most frequently in the jth character set, where j is greater than 1 and less than or equal to k+1.
For specific limitations of the license plate recognition device, reference may be made to the above limitations of the license plate recognition method, and no further description is given here. The above-mentioned various modules in the license plate recognition device can be implemented in whole or in part 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. 8. The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. 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 device is used for storing the trained convolutional neural network model and the input initial license plate image. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program when executed by a processor implements a license plate recognition method.
In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor executes the computer program to implement the steps of the license plate recognition method of the above embodiment, such as steps S10 to S60 shown in fig. 2. Alternatively, the processor may implement the functions of each module/unit of the license plate recognition device in the above embodiment, such as the functions of the modules 10 to 60 shown in fig. 7, when executing the computer program. In order to avoid repetition, a description thereof is omitted.
In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, where the computer program when executed by a processor implements the license plate recognition method in the above method embodiment, or where the computer program when executed by a processor implements the functions of each module/unit in the license plate recognition device in the above device embodiment. In order to avoid repetition, a description thereof is omitted.
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, storage, database, or other medium used in embodiments provided herein may include non-volatile and/or volatile memory. The nonvolatile memory can include Read Only Memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link DRAM (SLDRAM), memory bus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), among others.
It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-described division of the functional units and modules is illustrated, and in practical application, the above-described functional distribution may be performed by different functional units and modules according to needs, i.e. the internal structure of the apparatus is divided into different functional units or modules to perform all or part of the above-described functions.
The above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention, and are intended to be included in the scope of the present invention.
Claims (8)
1. The license plate recognition method is characterized by comprising the following steps of:
acquiring an initial license plate image;
preprocessing the initial license plate image to obtain a target license plate image with a size of a multiplied by b pixels, wherein a and b are positive integers;
Sliding operation is carried out on the target license plate image from left to right by adopting a sliding window with preset size of c multiplied by b, and taking 1 pixel as a step length, and one target license plate image in the range of the sliding window is obtained as one sliding window image in each sliding, wherein c is a positive integer;
after sliding for a-c times, stopping sliding operation to obtain a-c sliding window images;
inputting a-c sliding window images into a convolutional neural network for recognition to obtain a-c recognition results, wherein each recognition result comprises a plurality of characters;
forming character strings by all the characters in the a-c recognition results, dividing the character strings according to a preset dividing mode, acquiring target characters from each obtained sub-character string, and determining a target license plate number according to the target characters;
the preprocessing of the initial license plate image to obtain a target license plate image with a size of a multiplied by b pixels comprises the following steps:
acquiring an upper boundary of a license plate and a lower boundary of the license plate in the initial license plate image through an edge detection algorithm;
determining a range image of the license plate according to the upper boundary of the license plate and the lower boundary of the license plate;
Performing inclination correction on the range image by using Lato transformation to obtain a corrected basic image;
cutting the basic image by taking the gravity center of the basic image as the center to obtain the target license plate image with the size of a multiplied by b pixels;
the steps of forming a character string from all the characters in the a-c recognition results, obtaining target characters from the character string according to a preset mode, and determining a target license plate number according to the target characters include:
all the characters in the a-c recognition results are sequentially put into a character set X to obtainX= Wherein，qFor the total number of characters contained in a-c recognition results, < ->For the i-th character in character set X, -, is->，qIs a positive integer;
according tok+1 preset ranges, determining the character set XkA plurality of dividing points, wherein,k<qand (2) andkis a positive integer;
usingkEach of the segmentation points is corresponding to the character set XqCharacter segmentation is carried out on each character to obtaink+1 sub-character sets;
acquiring the character with the largest occurrence number in each sub-character set as a target character to obtaink+1 of the target characters;
according tok+1 of said predetermined ranges, forkAnd combining +1 target characters to obtain the target license plate number.
2. The license plate recognition method according to claim 1, wherein the acquiring, by an edge detection algorithm, the upper boundary of the license plate and the lower boundary of the license plate in the initial license plate image includes:
noise removal is carried out on the initial license plate image through Gaussian blur, and a denoising license plate image is obtained;
calculating gradient values of the denoising license plate image in the horizontal direction and the vertical direction by using a preset gradient operator to obtain an initial gradient value set;
performing edge refinement treatment on the initial gradient value set in a non-maximum value inhibition mode to obtain a gradient edge with a pixel width;
filtering weak edge points in the gradient edge by using a preset double threshold value to obtain strong edge points in the gradient edge;
and determining the upper boundary of the license plate and the lower boundary of the license plate according to the strong edge points.
3. The license plate recognition method of claim 1, wherein the obtaining the character having the largest number of occurrences in each of the sub-character sets as the target character comprises:
acquiring the character with the largest occurrence number in the first sub-character set as a first character;
if the first character is a non-Chinese character, acquiring a Chinese character with the largest occurrence frequency in the first sub-character set as a first target character;
If the first character is a Chinese character, the first character is used as a first target character;
and acquiring the character with the largest occurrence number in the jth character set as the jth target character, wherein j is greater than 1 and less than or equal to k+1.
4. A license plate recognition device, characterized in that the license plate recognition device comprises:
the acquisition module is used for acquiring an initial license plate image;
the preprocessing module is used for preprocessing the initial license plate image to obtain a target license plate image with a size of a multiplied by b pixels, wherein a and b are positive integers;
the sliding module is used for adopting a sliding window with preset size of c multiplied by b pixels, performing sliding operation from left to right on the target license plate image by taking 1 pixel as a step length, and acquiring the target license plate image within the range of the sliding window as a sliding window image every time, wherein c < a and c are positive integers;
the termination module is used for terminating the sliding operation after sliding a-c times to obtain a-c sliding window images;
the identification module is used for inputting the a-c sliding window images into a convolutional neural network for identification to obtain a-c identification results, wherein each identification result comprises a plurality of characters;
The segmentation module is used for forming a character string from the characters in the a-c recognition results, segmenting the character string according to a preset segmentation mode, acquiring target characters from each obtained sub-character string, and determining a target license plate number according to the target characters;
the preprocessing of the initial license plate image to obtain a target license plate image with a size of a multiplied by b pixels comprises the following steps:
acquiring an upper boundary of a license plate and a lower boundary of the license plate in the initial license plate image through an edge detection algorithm;
determining a range image of the license plate according to the upper boundary of the license plate and the lower boundary of the license plate;
performing inclination correction on the range image by using Lato transformation to obtain a corrected basic image;
cutting the basic image by taking the gravity center of the basic image as the center to obtain the target license plate image with the size of a multiplied by b pixels;
the steps of forming a character string from all the characters in the a-c recognition results, obtaining target characters from the character string according to a preset mode, and determining a target license plate number according to the target characters include:
all the characters in the a-c recognition results are sequentially put into a character set X to obtain X= Wherein, the method comprises the steps of, wherein,qfor the total number of characters contained in a-c recognition results, < ->For the i-th character in character set X, -, is->，qIs a positive integer;
according tok+1 preset ranges, determining the character set XkA plurality of dividing points, wherein,k<qand (2) andkis a positive integer;
usingkEach of the segmentation points is corresponding to the character set XqThe character is subjected to character segmentation of the individual characters,obtainingk+1 sub-character sets;
acquiring the character with the largest occurrence number in each sub-character set as a target character to obtaink+1 of the target characters;
according tok+1 of said predetermined ranges, forkAnd combining +1 target characters to obtain the target license plate number.
5. The license plate recognition device of claim 4, wherein the preprocessing module comprises:
the detection unit is used for acquiring the upper boundary of the license plate and the lower boundary of the license plate in the initial license plate image through an edge detection algorithm;
the determining unit is used for determining a range image of the license plate according to the upper boundary of the license plate and the lower boundary of the license plate;
the correction unit is used for performing inclination correction on the range image by using the radon transformation to obtain a corrected basic image;
and the clipping unit is used for clipping the basic image by taking the gravity center of the basic image as the center to obtain the target license plate image with the size of a multiplied by b pixels.
6. The license plate recognition device of claim 4, wherein the segmentation module comprises:
a generating unit for sequentially placing the characters in the a-c recognition results into the character set X to obtainX= Wherein, the method comprises the steps of, wherein,qfor the total number of characters contained in a-c recognition results, < ->For the i-th character in character set X, -, is->，qIs a positive integer;
a selecting unit for according tok+1 preset ranges, determining the character set XkA plurality of dividing points, wherein,k<qand (2) andkis a positive integer;
a segmentation unit for usingkEach of the segmentation points is corresponding to the character setXIn (a) and (b)qCharacter segmentation is carried out on each character to obtaink+1 sub-character sets;
a preferred unit for obtaining the character with the largest occurrence number in each sub-character set as a target character to obtaink+1 of the target characters;
a combination unit for followingk+1 of said predetermined ranges, forkAnd combining +1 target characters to obtain the target license plate number.
7. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the license plate recognition method of any one of claims 1 to 3.
8. A computer-readable storage medium storing a computer program, wherein the computer program when executed by a processor implements the steps of the license plate recognition method of any one of claims 1 to 3.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811113599.4A CN109492642B (en) | 2018-09-25 | 2018-09-25 | License plate recognition method, license plate recognition device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811113599.4A CN109492642B (en) | 2018-09-25 | 2018-09-25 | License plate recognition method, license plate recognition device, computer equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109492642A CN109492642A (en) | 2019-03-19 |
| CN109492642B true CN109492642B (en) | 2023-11-24 |
Family
ID=65689336
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811113599.4A Active CN109492642B (en) | 2018-09-25 | 2018-09-25 | License plate recognition method, license plate recognition device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109492642B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110232381B (en) * | 2019-06-19 | 2023-06-20 | 梧州学院 | License plate segmentation method, license plate segmentation device, computer equipment and computer readable storage medium |
| CN110264489B (en) * | 2019-06-24 | 2022-07-05 | 北京奇艺世纪科技有限公司 | Image boundary detection method and device and terminal |
| CN112686246B (en) * | 2019-10-18 | 2024-01-02 | 深圳市优必选科技股份有限公司 | License plate character segmentation method and device, storage medium and terminal equipment |
| CN111144390B (en) * | 2019-12-20 | 2023-10-27 | 武汉瑞纳捷半导体有限公司 | Missing character searching method and system in license plate recognition and electronic equipment |
| CN111681229B (en) * | 2020-06-10 | 2023-04-18 | 创新奇智(上海)科技有限公司 | Deep learning model training method, wearable clothes flaw identification method and wearable clothes flaw identification device |
| CN112434700A (en) * | 2020-11-25 | 2021-03-02 | 创新奇智(上海)科技有限公司 | License plate recognition method, device, equipment and storage medium |
| CN112906643A (en) * | 2021-03-22 | 2021-06-04 | 高新兴科技集团股份有限公司 | License plate number identification method and device |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5542006A (en) * | 1994-06-21 | 1996-07-30 | Eastman Kodak Company | Neural network based character position detector for use in optical character recognition |
| CN1801181A (en) * | 2006-01-06 | 2006-07-12 | 华南理工大学 | Robot capable of automatically recognizing face and vehicle license plate |
| US8761498B1 (en) * | 2012-01-26 | 2014-06-24 | Google Inc. | Face and license plate detection in street level images with 3-D road width features estimated from laser data |
| CN104715252A (en) * | 2015-03-12 | 2015-06-17 | 电子科技大学 | License plate character segmentation method with combination of dynamic template and pixel points |
| CN105373794A (en) * | 2015-12-14 | 2016-03-02 | 河北工业大学 | Vehicle license plate recognition method |
| CN107016366A (en) * | 2017-03-29 | 2017-08-04 | 浙江师范大学 | A kind of guideboard detection method based on Adaptive windowing mouthful and convolutional neural networks |
| CN108416348A (en) * | 2018-01-29 | 2018-08-17 | 重庆邮电大学 | Plate location recognition method based on support vector machines and convolutional neural networks |
-
2018
- 2018-09-25 CN CN201811113599.4A patent/CN109492642B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5542006A (en) * | 1994-06-21 | 1996-07-30 | Eastman Kodak Company | Neural network based character position detector for use in optical character recognition |
| CN1801181A (en) * | 2006-01-06 | 2006-07-12 | 华南理工大学 | Robot capable of automatically recognizing face and vehicle license plate |
| US8761498B1 (en) * | 2012-01-26 | 2014-06-24 | Google Inc. | Face and license plate detection in street level images with 3-D road width features estimated from laser data |
| CN104715252A (en) * | 2015-03-12 | 2015-06-17 | 电子科技大学 | License plate character segmentation method with combination of dynamic template and pixel points |
| CN105373794A (en) * | 2015-12-14 | 2016-03-02 | 河北工业大学 | Vehicle license plate recognition method |
| CN107016366A (en) * | 2017-03-29 | 2017-08-04 | 浙江师范大学 | A kind of guideboard detection method based on Adaptive windowing mouthful and convolutional neural networks |
| CN108416348A (en) * | 2018-01-29 | 2018-08-17 | 重庆邮电大学 | Plate location recognition method based on support vector machines and convolutional neural networks |
Non-Patent Citations (1)
| Title |
|---|
| 基于边缘特征和神经网络的汽车牌照定位算法;邱刚 等;微机发展;第15卷(第04期);第30-32页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109492642A (en) | 2019-03-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109492642B (en) | License plate recognition method, license plate recognition device, computer equipment and storage medium | |
| US10789465B2 (en) | Feature extraction and matching for biometric authentication | |
| CN110414507B (en) | License plate recognition method and device, computer equipment and storage medium | |
| CN109859227B (en) | Method and device for detecting flip image, computer equipment and storage medium | |
| CN109447117B (en) | Double-layer license plate recognition method and device, computer equipment and storage medium | |
| US8977072B1 (en) | Method and system for detecting and recognizing text in images | |
| CN110569721A (en) | Recognition model training method, image recognition method, device, equipment and medium | |
| CN111461170A (en) | Vehicle image detection method and device, computer equipment and storage medium | |
| US20220058422A1 (en) | Character recognition method and terminal device | |
| CN113781406B (en) | Scratch detection method and device for electronic component and computer equipment | |
| CN109255802B (en) | Pedestrian tracking method, device, computer equipment and storage medium | |
| CN114049499A (en) | Target object detection method, apparatus and storage medium for continuous contour | |
| CN114387591A (en) | License plate recognition method, system, equipment and storage medium | |
| CN108665459B (en) | Image blur detection method, computing device and readable storage medium | |
| CN113723309A (en) | Identity recognition method, identity recognition device, equipment and storage medium | |
| CN114049641B (en) | Character recognition method and system based on deep learning | |
| CN109712134B (en) | Iris image quality evaluation method and device and electronic equipment | |
| CN116434071B (en) | Determination method, determination device, equipment and medium for normalized building mask | |
| CN113159037B (en) | Picture correction method, device, computer equipment and storage medium | |
| CN114529570A (en) | Image segmentation method, image identification method, user certificate subsidizing method and system | |
| CN115082326A (en) | Processing method for deblurring video, edge computing equipment and central processor | |
| CN113762244A (en) | Document information extraction method and device | |
| CN116681654A (en) | Image detection method, device, equipment and medium | |
| Kryachko et al. | The Algorithm for Cars License Plates Segmentation | |
| CN116416147A (en) | Image processing method, electronic device, and computer-readable medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |