CN104573692B - License plate binarization method based on fuzzy degradation model - Google Patents
License plate binarization method based on fuzzy degradation model Download PDFInfo
- Publication number
- CN104573692B CN104573692B CN201410794680.9A CN201410794680A CN104573692B CN 104573692 B CN104573692 B CN 104573692B CN 201410794680 A CN201410794680 A CN 201410794680A CN 104573692 B CN104573692 B CN 104573692B
- Authority
- CN
- China
- Prior art keywords
- license plate
- point
- image
- skeleton
- character
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/28—Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/34—Smoothing or thinning of the pattern; Morphological operations; Skeletonisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
Abstract
The invention discloses a license plate binarization method based on a fuzzy degradation model, which comprises the following steps: preliminary binaryzation; thinning a framework; extracting a character center skeleton; extracting the color of the center of the character; acquiring a license plate image binarization threshold; and finally carrying out binarization. The invention can effectively separate the license plate characters, thereby facilitating the subsequent identification of the license plate characters; meanwhile, a degradation relation between central pixels and edge pixels of the license plate characters is estimated by adopting a fuzzy degradation model, so that correct thresholds are calculated to distinguish the license plate characters from the background of the license plate, and the problem that the thresholds cannot be self-adapted under different conditions is effectively solved.
Description
Technical Field
The invention relates to the field of image processing, in particular to a license plate binarization method based on a fuzzy degradation model.
Background
License plate recognition is a very successful application of image processing and pattern recognition technology in modern intelligent transportation. The method comprises the steps of positioning a license plate region from an image by using an image processing and analyzing technology through the image or video collected by a camera, dividing the license plate image into individual license plate character images, and finally identifying each individual character image by using a pattern recognition technology so as to obtain a final license plate result by splicing together.
The binaryzation is a key step in the image preprocessing step in the license plate recognition system, and the binaryzation effect directly influences the accuracy of license plate character segmentation and the recognition rate of character recognition. However, due to the fact that the shooting environment of the license plate image is variable, the actually acquired license plate image has differences in definition, brightness, contrast and the like. This brings difficulty to the traditional method of using fixed threshold binarization to preprocess the license plate image.
Most of the existing license plate binarization algorithms are based on a threshold value method to binarize the license plate, aiming at the calculation of the threshold value, some adopt a local threshold value, some adopt a global threshold value, and a local binarization algorithm and a global binarization algorithm are correspondingly generated, but due to the particularity of the license plate image, some illumination is uneven, the exposure is excessive, and the effect is not ideal when the two methods are applied to the contaminated license plate.
In practical application, the adhesion problem among characters is also a difficult problem, and in an image sample acquired by a low-pixel camera, the resolution of the image is low, so that many details in the image are generated by simulating interpolation among pixels, and the situation shows that connection among vehicle license characters can be formed on a vehicle license image, the characters cannot be further well separated from the characters, and the difficulty is caused in recognition of the later characters.
Disclosure of Invention
The invention aims to provide a license plate binarization method based on a fuzzy degradation model, which can effectively separate license plate characters and is convenient for subsequent license plate character recognition.
In order to achieve the purpose, the invention adopts the following technical scheme:
a license plate binarization method based on a fuzzy degradation model comprises the following steps:
s1, preliminary binarization, namely, carrying out preliminary binarization on the original license plate image to obtain a binary license plate image;
s2, skeleton refinement, namely, refining the connected region of the binary license plate image obtained in the step S1 to obtain a license plate skeleton image;
s3, extracting a character center skeleton, judging whether each skeleton point on the license plate skeleton image obtained in the step S2 belongs to a point on a license plate character, if so, retaining the skeleton point, and if not, deleting the skeleton point so as to obtain the license plate character skeleton image;
s4, extracting the color of the character center, and extracting the color value of the corresponding point on the original license plate image according to each skeleton point on the license plate character skeleton image obtained in the step S3, so as to obtain the color value of the character center;
s5, acquiring a license plate image binarization threshold, establishing a fuzzy degradation model according to the character center color value acquired in the step S4, and calculating to acquire a license plate image binarization threshold;
and S6, performing final binarization, and performing binarization on the original license plate image based on the license plate image binarization threshold value obtained in the step S5.
Preferably, in step S1, the preliminary binarization employs an Ostu binarization algorithm.
Preferably, the step S2 includes the following substeps:
s21, for a single pixel point, defining the collection of the upper, lower, left and right 4 pixel points as the 4-neighborhood of the pixel point, defining the collection of the upper, lower, left and right 4 pixel points and the 4 pixel points in the diagonal direction as the 8-neighborhood of the pixel point, defining the number of intersection points as the total number of the pixel points with the gray scale value meeting the following conditions in the 4-neighborhood of the point p for each pixel point p in the foreground,
g(Pk)-g(Pk-1)＝1，
wherein, PkRepresents the k-th pixel point, g (P)k) K is greater than or equal to 0 and less than or equal to 7, and a subscript module 8 of the pixel point p is calculated;
s22, reserving or deleting each pixel point on the binary license plate image according to the gray value of the 4-neighborhood, the gray value of the 8-neighborhood and the number of crossing points, and finally obtaining the license plate skeleton image.
Preferably, the step S3 is implemented by the following sub-steps:
s31, judging the number of pixel points with the gray value of 1 in the 8-neighborhood of each skeleton point on the license plate skeleton image obtained in the step S2, if the number is equal to 1 or more than or equal to 3, reserving the pixel points, and otherwise, executing the steps S32-S34;
s32, judging the line type of the line where the pixel point is located according to the position relation between the pixel point and the surrounding pixel points, wherein the line type comprises a vertical line, a horizontal line and an inclined line;
s33, starting from the pixel point, searching to two sides of the binary license plate image along the vertical direction of the line type until a black pixel point is met, and taking the sum of the step lengths of the two-side searching as the stroke width of the original character at the pixel point;
s34, if the stroke width of the original character at the pixel point is within the normal stroke width range of the license plate character, judging that the pixel point belongs to the point on the license plate character and reserving the point, otherwise, deleting the point;
and S35, traversing all skeleton points on the license plate skeleton image to finally obtain the license plate character skeleton image.
Preferably, the step S4 further includes storing the obtained character center color value in the array C.
Preferably, the step S5 includes the following substeps:
s51, taking the color value in the array C as a random variable X, wherein the observed value of X is represented as:
X＝{X1,X2,...,Xtin which Xi∈C；
S52, establishing a mixed Gaussian distribution model of the random variable X:
wherein, g (X, μ)i,σi) Is a Gaussian distribution function, omegaiIs a weight;
s53, initializing mu1＝0、μ2＝0、σ1＝0、σ20, according to the actual value of the random variable X to the mui、σi、ωiUpdating is carried out;
s54, acquiring a Gaussian model with the largest weight, and solving parameters mu and sigma of the Gaussian model;
s55, establishing a Gaussian distribution model of the character center color values:
for any point (x, y), if its color value is g, then the condition that it satisfies the point on the license plate character is:
|g-μ|＝3σ。
preferably, in step S6, the binarizing the original license plate image is performed by binarizing the original license plate image according to the following formula:
wherein, b (x, y) is the final license plate binary image, and f (x, y) is the gray image of the original license plate.
After adopting the technical scheme, compared with the background technology, the invention has the following advantages:
the invention can effectively separate the license plate characters, thereby facilitating the subsequent identification of the license plate characters; meanwhile, a degradation relation between central pixels and edge pixels of the license plate characters is estimated by adopting a fuzzy degradation model, so that correct thresholds are calculated to distinguish the license plate characters from the background of the license plate, and the problem that the thresholds cannot be self-adapted under different conditions is effectively solved.
Drawings
Fig. 1 is a schematic view of the working process of the present invention.
Fig. 2 shows a binary license plate image result obtained through a preliminary binarization process.
Fig. 3 shows the result of the license plate skeleton image.
Fig. 4 shows the result of the license plate character skeleton image.
FIG. 5a shows a case where the line type is a vertical line; FIG. 5b shows a case where the line type is a horizontal line; fig. 5c shows a case where the line type is an inclined line.
Fig. 6 shows the license plate binary image result obtained through the final binarization processing.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
Examples
Referring to fig. 1, the invention discloses a license plate binarization method based on a fuzzy degradation model, comprising the following steps:
s1, preliminary binarization
And (3) carrying out preliminary binarization on the original license plate image by adopting an Otsu binarization algorithm to obtain a binary license plate image. In this step, the method for obtaining the Otsu threshold is as follows:
for the license plate image f (x, y), the segmentation threshold of the license plate characters and the license plate background is T, and the proportion of pixel points belonging to the license plate characters in the whole image is marked as omega0The average gray level of these points is recorded as μ0(ii) a The proportion of the license plate background pixel points to the whole image is omega1Average gray scale of ω1. The total mean gray level of the image is denoted as μ and the inter-class variance is denoted as σ. Then there are:
σ＝ω0ω1(μ0-μ1)2
traversing T from 0 to 255, calculating the value of T when σ has the maximum value, i.e. the threshold of Otsu.
The binary license plate image obtained by the binarization processing in the step is only used for roughly estimating the position of the license plate characters and is not used for extracting the license plate characters, and the result of the binary license plate image is shown in fig. 2.
S2 skeleton refinement
And (4) thinning the connected region of the binary license plate image obtained in the step (S1) to obtain a license plate skeleton image (as shown in FIG. 3). The method is realized by the following steps:
s21, for a single pixel point, defining the collection of the upper, lower, left and right 4 pixel points as the 4-neighborhood of the pixel point, defining the collection of the upper, lower, left and right 4 pixel points and the 4 pixel points in the diagonal direction as the 8-neighborhood of the pixel point, defining the cross point number as the sum of the point numbers of the pixel points with the gray scale value meeting the following conditions in the 4-neighborhood of the point p for each pixel point p in the foreground,
g(Pk)-g(Pk-1)＝1，
wherein, PkRepresents the k-th pixel point, g (P)k) K is more than or equal to 0 and less than or equal to 7, and the subscript module of the pixel point p is 8 during operation.
S22, reserving or deleting each pixel point on the binary license plate image according to the gray value of the 4-neighborhood, the gray value of the 8-neighborhood and the number of crossing points, and finally obtaining the license plate skeleton image. Step S22 is specifically implemented by the following method:
1) and determining pixel points needing to be reserved. Sequentially traversing each point in the image, setting the current investigation point as p (the gray value of the current investigation point is 1), and defining the sum of 4-neighborhood gray levels of p as sigma4Sum of gray values of 8-neighborhood of p is sigma8P has a number of crossing points Ng。
Firstly, in order to ensure that the thinned image can reflect the salient part of the target in the original image, only a single point with the gray value of 1 is stored in the 8-neighborhood of the current point p, and the pixel point is reserved in principle.
Secondly, 4-neighborhood points of the current point p are inspected, if the gray values of the 4-neighborhood points are all 1, sigma is4If 4, it can be determined that it is the inner point of the target edge that should be preserved.
Next, consider the 8-neighborhood of p, leaving the sum of the 4-neighborhoods at 1, i.e., Σ44 and the number of crossings NgAnd the point is more than 1, so as to ensure the connectivity of the refined image.
Finally, in order to keep the points which are difficult to determine whether to delete before thinning to the width of 2, the templates a and b given below are used for detection, if one of the output results of the templates a and b is true, the current point is kept to ensure that the edge is not broken when thinning to the double line.
| x | x | x | |
| 0 | p(x,y)＝1 | 1 | 0 |
| x | x | x |
Form a
| x | 0 | x | |
| x | p(x,y)＝1 | x | |
| x | 1 | x | |
| 0 |
Form b
2) And deleting redundant pixel points. In order to ensure the image connectivity and realize the single-point connectivity between the pixel points on the framework, the following templates c and d are adopted for detection, and if the output results of the templates c and d are true, the current pixel point is deleted.
Template c
| x | 1 | 0 |
| 0 | p(x,y)＝1 | 1 |
| 0 | 0 | x |
Template d
3) And removing the redundant branch lines. In order to eliminate redundant branches in the skeleton image after image thinning and obtain a smooth image skeleton, the number N of cross points of each point on the skeleton of the thinned image obtained after the processing of the steps 1) and 2) is considered againgIf N is presentgIf not greater than 1, the data is deleted. The finally obtained license plate skeleton image is shown in fig. 3.
S3, extracting character center skeleton
And (4) judging whether each skeleton point on the license plate skeleton image obtained in the step (S2) belongs to a point on the license plate character, if so, reserving the points, and if not, deleting the points, thereby obtaining the license plate character skeleton image (as shown in FIG. 4). The method is realized by the following steps:
s31, judging the number of pixel points with the gray value of 1 in the 8-neighborhood of each skeleton point on the license plate skeleton image obtained in the step S2, if the number is equal to 1 or more than or equal to 3, reserving the pixel points, and otherwise, executing the steps S32-S34.
And S32, judging the line type of the line where the pixel point is located according to the position relation between the pixel point and the surrounding pixel points, wherein the line type comprises a vertical line, a horizontal line and an inclined line. Fig. 5a shows a case where the line type is a vertical line, fig. 5b shows a case where the line type is a horizontal line, and fig. 5c shows a case where the line type is an inclined line.
S33, starting from the pixel point, searching to two sides of the binary license plate image along the vertical direction of the line type until a black pixel point is met, and at the moment, respectively recording the step length of searching to the two sides as SlAnd SrThe stroke width of the original character at the pixel point is marked as W, and W is equal to Sl+Sr。
S34, if W is in the normal stroke width range of the license plate character, the pixel point is judged to belong to the point on the license plate character and is reserved, otherwise, the pixel point is deleted. That is, W needs to satisfy:
w1≤W≤w2，
wherein, w1＝pH/20，w2pH/8, pH is the height of the license plate image.
And S35, traversing all skeleton points on the license plate skeleton image, wherein only the points on the central skeleton of the license plate characters are reserved on the license plate skeleton image, and finally obtaining the license plate character skeleton image.
S4, extracting color of character center
And according to each skeleton point on the license plate character skeleton image obtained in the step S3, extracting the color value of the corresponding point on the original license plate image, thereby obtaining the color value of the center of the character.
And (3) setting the original license plate image as I (x, y), and setting the license plate character skeleton image obtained in the step (3) as T (x, y). Traversing the image T (x, y), if T (x)i,yi) When 1, then I (x)i,yi) The gray values of (a) are stored in an array C. After traversing is completed, the gray values of the pixel points at the central axis of most license plate characters are stored in the array C.
S5, obtaining the binary threshold value of the license plate image
And establishing a fuzzy degradation model according to the character center color value obtained in the step S4, and calculating to obtain a license plate image binarization threshold value. The method is realized by the following steps:
s51, taking the color value in the array C as a random variable X, and then the observed value of X is represented as:
X＝{X1,X2,...,Xtin which Xi∈C。
S52, establishing a mixed Gaussian distribution model of the random variable X:
wherein, g (X, μ)i,σi) Is a Gaussian distribution function, omegaiAre weights.
S53, initializing mu1＝0、μ2＝0、σ1＝0、σ20, according to the actual value of the random variable X to the mui、σi、ωiAnd (6) updating.
And S54, acquiring the Gaussian model with the maximum weight, and calculating the parameters mu and sigma of the Gaussian model. Step S54 is specifically implemented by the following method:
1) weight ωkThe update at time t is performed according to the following update expression:
ωk,t＝(1-α)ωk,t-1+αMk,t
where α is the updated learning rate, and 1/α represents a time constant that characterizes the update rate Mk,tIs a binary variable when the model is associated with XiWhen matching, Mk,t1, otherwise Mk,t＝0。
2) If Gaussian distribution model g (X, μ)k,σk) And XiMatch, mean μkVariance, variance
The updating of (2) is performed according to the following update expression:
μk,t＝(1-ρ)μk,t-1+ρXt
ρ＝αg(Xt|μk,σk)
wherein α is a constant, g (X)t|μk,σk) Is a density function of the gaussian distribution. The learning rate when updating is performed by the calculated ρ is the mean and the variance. When the model does not have a sum of XiWhen matched, mean value μkVariance, variance
The original value is kept unchanged.
In the array C, more points are points at the center of the license plate character, and the interference points occupy a small number, so after the statistics of the Gaussian mixture model, the Gaussian distribution with large weight is a statistical model for describing the distribution of the pixels at the center of the license plate character.
Through the above steps we can find the parameters μ and σ of that model that describes the center pixel of the character. Where μ and σ are the parameters of the gaussian model with the highest weight in the mixture model.
S55, according to the image blurring principle, when the image is from near to far, the degradation of the image quality can be simulated by convolution with the original image only by a gaussian low-pass filter. Thus, a Gaussian distribution model of the color values of the center of the character is established:
for any point (x, y), if its color value is g, then the condition that it satisfies the point on the license plate character is:
|g-μ|＝3σ。
s6, final binarization
Based on the license plate image binarization threshold value obtained in the step S5, the original license plate image is binarized, and the finally obtained license plate binary image is as shown in fig. 6. In the step, the original license plate image is binarized through the following formula:
wherein, b (x, y) is the final license plate binary image, and f (x, y) is the gray image of the original license plate.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (7)
1. A license plate binarization method based on a fuzzy degradation model is characterized by comprising the following steps:
s1, preliminary binarization, namely, carrying out preliminary binarization on the original license plate image to obtain a binary license plate image;
s2, skeleton refinement, namely, refining the connected region of the binary license plate image obtained in the step S1 to obtain a license plate skeleton image;
s3, extracting a character center skeleton, judging whether each skeleton point on the license plate skeleton image obtained in the step S2 belongs to a point on a license plate character, if so, retaining the skeleton point, and if not, deleting the skeleton point so as to obtain the license plate character skeleton image;
s4, extracting the color of the character center, and extracting the color value of the corresponding point on the original license plate image according to each skeleton point on the license plate character skeleton image obtained in the step S3, so as to obtain the color value of the character center;
s5, acquiring a license plate image binarization threshold, establishing a fuzzy degradation model according to the character center color value acquired in the step S4, and calculating to acquire the license plate image binarization threshold;
and S6, performing final binarization, and performing binarization on the original license plate image based on the license plate image binarization threshold value obtained in the step S5.
2. The vehicle license plate binarization method based on the fuzzy degradation model as claimed in claim 1, wherein in step S1, the preliminary binarization employs an Ostu binarization algorithm.
3. The vehicle license plate binarization method based on the fuzzy degradation model as claimed in claim 1 or 2, wherein the step S2 comprises the following substeps:
s21, for a single pixel point, defining the collection of the upper, lower, left and right 4 pixel points as the 4-neighborhood of the pixel point, defining the collection of the upper, lower, left and right 4 pixel points and the 4 pixel points in the diagonal direction as the 8-neighborhood of the pixel point, defining the cross point number as the total number of the point numbers of the pixel points with the gray scale value meeting the following conditions in the 4-neighborhood of the point P for each pixel point P in the foreground,
g(Pk)-g(Pk-1)＝1，
wherein, PkRepresents the k-th pixel point, g (P)k) K is greater than or equal to 0 and less than or equal to 7, and a subscript module 8 of the pixel point P is calculated;
s22, reserving or deleting each pixel point on the binary license plate image according to the gray value of the 4-neighborhood, the gray value of the 8-neighborhood and the number of crossing points, and finally obtaining the license plate skeleton image.
4. The vehicle license plate binarization method based on the fuzzy degradation model as claimed in claim 3, wherein the step S3 is realized by the following sub-steps:
s31, judging the number of pixel points with the gray value of 1 in the 8-neighborhood of each skeleton point on the license plate skeleton image obtained in the step S2, if the number is equal to 1 or more than or equal to 3, reserving the pixel points, and otherwise, executing the steps S32-S34;
s32, judging the line type of the line where the pixel point is located according to the position relation between the pixel point and the surrounding pixel points, wherein the line type comprises a vertical line, a horizontal line and an inclined line;
s33, starting from the pixel point, searching to two sides of the binary license plate image along the vertical direction of the line type until a black pixel point is met, and taking the sum of the step lengths of the two-side searching as the stroke width of the original character at the pixel point;
s34, if the stroke width of the original character at the pixel point is within the normal stroke width range of the license plate character, judging that the pixel point belongs to the point on the license plate character and reserving the point, otherwise, deleting the point;
and S35, traversing all skeleton points on the license plate skeleton image to finally obtain the license plate character skeleton image.
5. The vehicle license plate binarization method based on the fuzzy degradation model as claimed in claim 4, wherein the step S4 further comprises storing the obtained character center color values in an array C.
6. The vehicle license plate binarization method based on the fuzzy degradation model as claimed in claim 5, wherein the step S5 comprises the following substeps:
s51, taking the color value in the array C as a random variable X, wherein the observed value of X is represented as:
X＝{X1，X2，...，Xtin which Xi∈C；
S52, establishing a mixed Gaussian distribution model of the random variable X:
wherein, g (X, μ)i，σi) Is a Gaussian distribution function, wiIs a weight;
s53, initializing mu1＝0、μ2＝0、σ1＝0、σ20, according to the actual value of the random variable X to the mui、σi、wiUpdating is carried out;
s54, acquiring a Gaussian model with the largest weight, and solving parameters mu and sigma of the Gaussian model;
s55, establishing a Gaussian distribution model of the character center color values:
for any point (x, y), if its color value is g, then the condition that it satisfies the point on the license plate character is:
|g-μ|＝3σ。
7. the method as claimed in claim 6, wherein in step S6, the binarizing step is implemented by binarizing the original license plate image according to the following formula:
wherein, b (x, y) is the final license plate binary image, and f (x, y) is the gray image of the original license plate.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410794680.9A CN104573692B (en) | 2014-12-18 | 2014-12-18 | License plate binarization method based on fuzzy degradation model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410794680.9A CN104573692B (en) | 2014-12-18 | 2014-12-18 | License plate binarization method based on fuzzy degradation model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104573692A CN104573692A (en) | 2015-04-29 |
| CN104573692B true CN104573692B (en) | 2020-04-10 |
Family
ID=53089714
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410794680.9A Active CN104573692B (en) | 2014-12-18 | 2014-12-18 | License plate binarization method based on fuzzy degradation model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104573692B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110134071B (en) * | 2019-06-05 | 2021-09-28 | 南京航空航天大学 | Image processing-based circular cutting tool path generation method under constraint condition of skin mirror milling process |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102096821A (en) * | 2010-04-15 | 2011-06-15 | 西安理工大学 | Number plate identification method under strong interference environment on basis of complex network theory |
| CN102945360A (en) * | 2012-09-20 | 2013-02-27 | 南京信息工程大学 | Method for detecting headwind zone of Doppler radar image based on morphological and logic operations |
| CN104050472A (en) * | 2014-06-12 | 2014-09-17 | 浙江工业大学 | Self-adaptation global threshold method for gray level image binaryzation |
-
2014
- 2014-12-18 CN CN201410794680.9A patent/CN104573692B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102096821A (en) * | 2010-04-15 | 2011-06-15 | 西安理工大学 | Number plate identification method under strong interference environment on basis of complex network theory |
| CN102945360A (en) * | 2012-09-20 | 2013-02-27 | 南京信息工程大学 | Method for detecting headwind zone of Doppler radar image based on morphological and logic operations |
| CN104050472A (en) * | 2014-06-12 | 2014-09-17 | 浙江工业大学 | Self-adaptation global threshold method for gray level image binaryzation |
Non-Patent Citations (4)
| Title |
|---|
| License plate binarization method based on fuzzy classification;YAN WO et al;《International Conference on Machine Learning and Cybernetics》;20131231;第194-198页 * |
| License Plate Recognition Using Complex Network Feature;Hai-Peng Ren et al;《11th World Congress on Intelligent Control and Automation》;20140704;第5426-5431页 * |
| 基于修正思想的车牌图像定位与二值化处理;洪健 等;《计算机工程》;20070430;第33卷(第8期);第175-177页 * |
| 车牌识别中二值化方法的研究;朱浩悦 等;《计算机应用与软件》;20070228;第24卷(第2期);第33-34，90页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104573692A (en) | 2015-04-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109859171B (en) | Automatic floor defect detection method based on computer vision and deep learning | |
| CN109784333B (en) | Three-dimensional target detection method and system based on point cloud weighted channel characteristics | |
| CN113160192B (en) | Visual sense-based snow pressing vehicle appearance defect detection method and device under complex background | |
| CN109409263B (en) | Method for detecting urban ground feature change of remote sensing image based on Siamese convolutional network | |
| CN107545239B (en) | Fake plate detection method based on license plate recognition and vehicle characteristic matching | |
| CN107480649B (en) | Fingerprint sweat pore extraction method based on full convolution neural network | |
| CN108122236B (en) | Iterative fundus image blood vessel segmentation method based on distance modulation loss | |
| Zhou et al. | On detecting road regions in a single UAV image | |
| CN111611643B (en) | Household vectorization data acquisition method and device, electronic equipment and storage medium | |
| CN110838126B (en) | Cell image segmentation method, cell image segmentation device, computer equipment and storage medium | |
| CN107038416B (en) | Pedestrian detection method based on binary image improved HOG characteristics | |
| JP6811217B2 (en) | Crack identification method, crack identification device, crack identification system and program on concrete surface | |
| CN110555382A (en) | Finger vein identification method based on deep learning and Wasserstein distance measurement | |
| CN113689445B (en) | High-resolution remote sensing building extraction method combining semantic segmentation and edge detection | |
| CN110245600B (en) | Unmanned aerial vehicle road detection method for self-adaptive initial quick stroke width | |
| CN113780110A (en) | Method and device for detecting weak and small targets in image sequence in real time | |
| CN115641327A (en) | Building engineering quality supervision and early warning system based on big data | |
| CN113313107A (en) | Intelligent detection and identification method for multiple types of diseases on cable surface of cable-stayed bridge | |
| CN110598581B (en) | Optical music score recognition method based on convolutional neural network | |
| CN108765463B (en) | Moving target detection method combining region extraction and improved textural features | |
| Fang et al. | Towards real-time crack detection using a deep neural network with a Bayesian fusion algorithm | |
| Babu et al. | An efficient image dahazing using Googlenet based convolution neural networks | |
| CN106204596B (en) | Panchromatic waveband remote sensing image cloud detection method based on Gaussian fitting function and fuzzy mixed estimation | |
| CN107704864B (en) | Salient object detection method based on image object semantic detection | |
| CN104573692B (en) | License plate binarization method based on fuzzy degradation model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | 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 |