CN103593664B - A kind of preprocess method of QR code fault image - Google Patents

Abstract

The invention discloses the preprocess method of a kind of QR code fault image, belong to image processing field, using the sample point of QR code normal picture as the input value during neural network learning, using the sample point with QR code fault image as the output valve during neural network learning, carry out BP neural network learning matching distortion function；Utilizing bilinear interpolation to determine target QR code image after carrying out functional transformation, the present invention can well identify that those distortion are than the more serious or QR code of image section disappearance.Particularly it is printed on surface and easily produces the lowest QR code image at all not identified in other words of the QR code discrimination above the article of fold, use the present invention to identify more accurately.Owing to improve the discrimination of in particular cases QR code image, and the range of application of QR code can be largely promoted in the raising of discrimination, thus uses the present invention can the application of larger range of popularization QR code.

Description

A kind of preprocess method of QR code fault image

Technical field

The invention belongs to image processing field, particularly relate to the preprocess method of a kind of QR code fault image.

Background technology

QR code as a kind of outstanding Quick Response Code, under normal circumstances its density be bar code tens to hundreds of Times, it means that QR code can express more information in limited space, and so we are the most permissible Product information is all stored in a QR code.Want the information checking product to need not in advance and set up data Storehouse, is truly realized with the bar code description to " article ".In view of such advantage, it is envisaged that QR code The general trend that application in trade mark necessarily develops.To image in 2 D code generally in prior art Can well identify, but those distortion are particularly printed on table than more serious or image section disappearance Face is easily generated that the Quick Response Code discrimination above the article of fold is the lowest not to be identified in other words at all.

Summary of the invention

Because the drawbacks described above of prior art, the technical problem to be solved is to provide one can Improve the preprocess method of the QR code fault image of distortion Quick Response Code discrimination.

For achieving the above object, the invention provides the preprocess method of a kind of QR code fault image, by following Step is carried out:

Step one, choose the sample point of QR code normal picture, choose and described QR code at QR code fault image The sample point that normal picture position is corresponding；

Step 2, using the sample point of QR code normal picture as the input value during neural network learning, general With the sample point of QR code fault image as the output valve during neural network learning, carry out BP nerve net Network learns；

Step 3, the sample point chosen in pending QR code fault image obtain as through step 2 study The input value of neutral net, obtain one group of output valve, this group output valve carried out Function Fitting, determines The distortion mode of image, then by all of pixel in QR code fault image to be corrected, all passes through this letter Number carries out functional transformation；

Step 4, bilinear interpolation is utilized to determine the QR code distortion figure that target QR code image correspondence is to be corrected As the pixel value of sample point, obtain target QR code image.

Further, target QR code image step 4 obtained with the inverse process of perspective projection transformation is also included Carry out the step processed.

It is also preferred that the left described in step 3 after functional transformation, in QR code fault image, all of pixel point value is filled Become white；Then the pixel value of the former QR each sample point of code fault image is assigned to the QR code distortion after filling Image.

It is also preferred that the left choose the sample point of QR code normal picture described in step one, select at QR code fault image Take the sample point corresponding with described QR code normal picture position all to sequentially include the following steps:

A1, four summits found out in QR code image are as the first sample point；

A2, four sample points obtained with step A1 form the first tetragon for summit, take cornerwise intersection point It it is the second sample point；

A3, in the tetragon that step A2 obtains, take the midpoint of any two articles of adjacent edges respectively as the 3rd sample This point and the 4th sample point, the intersection point adding these two articles of limits and the second sample point composition the two or four taken out above Limit shape, taking its diagonal intersection point is the 5th sample point；

A4, return perform step A1.

Further, the sample point of QR code normal picture is chosen described in step one, at QR code fault image Choose the sample point corresponding with described QR code normal picture position further comprising the steps of:

QR code image does Hough transformation process, find out the boundary contour of image, by the intersection point of contour line Obtain four apex coordinates of image.

It is also preferred that the left described step 2 uses Minimum Mean Square Error BP neural network learning is carried out error analysis.

It is also preferred that the left described step 2 uses the most quick gradient descent method as of BP neural network learning Practise strategy.

It is also preferred that the left described bilinear interpolation sequentially includes the following steps:

B1, set the coordinate of QR code fault image sample point as (i j), obtains it by reflection method backward At the floating-point coordinate (i+u, j+v) that QR code fault image is corresponding, wherein i, j are nonnegative integer, u, v For the floating number that [0,1] is interval；

B2, the pixel value setting described floating-point coordinate (i+u, j+v) they are f (i+u, j+v), (i+1, j), (i, j+1), (i+1, j+1) they are the coordinate in QR code fault image, calculating f (i+u, j+v)=(1-u) × (1-v) × f (i, j)+(1-u) × v × f (i, j+1)+u × (1-v) × f (i+1, j)+u × v × f(i+1,j+1)；Draw the pixel value of described floating-point coordinate (i+u, j+v)；Described f (i, j) be coordinate (i, J) pixel value, described f (i, j+1) is the pixel value of coordinate (i, j+1), and (i+1 j) is coordinate to described f (described f (i+1, j+1) is the pixel value of coordinate (i+1, j+1) for i+1, pixel value j).

The invention has the beneficial effects as follows: the present invention can well identify that those distortion are than more serious or image The QR code of excalation.Particularly it is printed on surface and easily produces the QR code discrimination above the article of fold The lowest QR code image at all not identified in other words, uses the present invention to identify more accurately.Due to Improve the discrimination of in particular cases QR code image, and the raising of discrimination can largely be promoted The range of application of QR code, thus uses the present invention can the application of larger range of popularization QR code.

Accompanying drawing explanation

Fig. 1 is the schematic flow sheet of the embodiment of the present invention one.

Detailed description of the invention

The invention will be further described with embodiment below in conjunction with the accompanying drawings, and the example of described embodiment is at accompanying drawing Shown in, the most same or similar label represent same or similar element or have identical or The element of similar functions.The embodiment described below with reference to accompanying drawing is exemplary, is only used for explaining this Invention, and be not considered as limiting the invention.

Embodiment one, as it is shown in figure 1, the preprocess method of a kind of QR code fault image, is entered according to the following steps OK:

Step one, QR code image is done Hough transformation process, find out the boundary contour of image, pass through profile The intersection point of line obtains four apex coordinates of image.

Step 2, choose the sample point of QR code normal picture, choose and described QR code at QR code fault image The sample point that normal picture position is corresponding.

The sample point of the described QR of choosing code normal picture, choose at QR code fault image normal with described QR code The corresponding sample point in picture position all sequentially includes the following steps:

A1, four summits found out in QR code image are as the first sample point；

A2, four sample points obtained with step A1 form the first tetragon for summit, take cornerwise intersection point It it is the second sample point；

A3, in the tetragon that step A2 obtains, take the midpoint of any two articles of adjacent edges respectively as the 3rd sample This point and the 4th sample point, the intersection point adding these two articles of limits and the second sample point composition the two or four taken out above Limit shape, taking its diagonal intersection point is the 5th sample point；

A4, return perform step A1.

In the present embodiment, first QR code image is done Hough transformation and processes, find out the boundary contour of image, Obtained four apex coordinates A, B, C, D of image by the intersection point of contour line, seek the straight line of some A, C AC and the intersection point E of straight line BD crossing some B, D, then seek midpoint F and G of AB and AD respectively, obtain with A, Tetragon cornerwise intersecting point coordinate H of F, E, G 4 composition；Use such method that tetragon is continuous Reduce, in little tetragon, seek sample point by same method, finally obtain 25 points as study Sample point.

Step 3, using the sample point of QR code normal picture as the input value during neural network learning, general With the sample point of QR code fault image as the output valve during neural network learning, use Minimum Mean Square Error BP neural network learning is carried out error analysis, uses the most quick gradient descent method as BP neutral net The learning strategy of study, carries out BP neural network learning, to form a learning machine.

Step 4, pending QR code fault image is taken the sample point input value as neutral net, by step Rapid three learning machines obtained learn, and obtain one group of output valve, according to this group output valve to QR code distortion figure Shape carries out Function Fitting, after obtaining fitting function, all passes through all pixels in QR code fault image This functional transformation.

Step 5, bilinear interpolation is utilized to determine the QR code distortion figure that target QR code image correspondence is to be corrected The pixel value of the sample point of picture, obtains target QR code image.

Described bilinear interpolation sequentially includes the following steps:

B1, set the coordinate of QR code fault image sample point as (i j), obtains it by reflection method backward At the floating-point coordinate (i+u, j+v) that QR code fault image is corresponding, wherein i, j are nonnegative integer, u, v For the floating number that [0,1] is interval；

B2, the pixel value setting described floating-point coordinate (i+u, j+v) they are f (i+u, j+v), (i+1, j), (i, j+1), (i+1, j+1) they are the coordinate in QR code fault image, calculating f (i+u, j+v)=(1-u) × (1-v) × f (i, j)+(1-u) × v × f (i, j+1)+u × (1-v) × f (i+1, j)+u × v × f(i+1,j+1)；Draw the pixel value of described floating-point coordinate (i+u, j+v)；Described f (i, j) be coordinate (i, J) pixel value, described f (i, j+1) is the pixel value of coordinate (i, j+1), and (i+1 j) is coordinate to described f (described f (i+1, j+1) is the pixel value of coordinate (i+1, j+1) for i+1, pixel value j).

Step 6, the target QR code image obtained step 5 with the inverse process of perspective projection transformation process.

Can obtain a QR code image through process above, this image there may exist perspective phenomenon, So the inverse process of perspective projection transformation to be carried out processes.The image that perspective occurs is replaced with tetragon ABCD, S is camera position, owing to the position of S causes the QR code image shot to be with irregular four limits Shape ABCD.Tetragon A ' B ' C ' D ' is the tetragon ABCD projection at positive screen orientation, and it is a square.

During calculating, first find out the position on tetra-summits of tetragon ABCD, ask the process on these four summits to use Hough transformation conventional in image procossing, obtains 4 boundary lines of QR code, by the intersection point of straight line two-by-two Just can obtain A, B, C, D tetra-point coordinates, A ', B ', C ', the coordinate of D ' then can be given as required Go out suitably value.

If P is any point in ABCD, with homogeneous coordinates be expressed as (x, y, 1) ', then at projection as A ' B ' C ' D ' 1 P ' of middle existence anduniquess is corresponding with P, and its homogeneous coordinates are (x_h,y_h,h)′.Then projective transformation matrix T is full Foot P '=TP；

$\left(\begin{array}{c}{x}_h\\ y_h\\ h\end{array}\right)=\left(\begin{array}{ccc}{t}_{00}& t_{01}& t_{02}\\ t_{10}& t_{11}& t_{12}\\ t_{20}& t_{21}& t_{22}\end{array}\right)\left(\begin{array}{c}x\\ y\\ 1\end{array}\right);$

Here $\left(\begin{array}{c}{x}_h\\ y_h\\ h\end{array}\right)$ Representing the coordinate of some P ', h is a scale factor, $\left(\begin{array}{ccc}{t}_{00}& t_{01}& t_{02}\\ t_{10}& t_{11}& t_{12}\\ t_{20}& t_{21}& t_{22}\end{array}\right)$ Represent matrix T, $\left(\begin{array}{c}x\\ y\\ 1\end{array}\right)$ Represent the coordinate of some P.

Above formula abbreviation, the common coordinate of invocation point P ' is:

$x^{\′}=\frac{x_h}{h}=\frac{t_{00}x+t_{01}y+t_{02}}{t_{20}x+t_{21}y+t_{22}}---\left(1\right);$

$y^{\′}=\frac{y_h}{h}=\frac{t_{10}x+t_{11}y+t_{12}}{t_{20}x+t_{21}y+t_{22}}---\left(2\right);$

Wherein t₂₂For scale factor, the present embodiment is taken as 1.

As long as it can be seen that the unknown number obtained in (1) and (2), it is possible to determine projective transformation matrix T.

A ' in the projection of tetra-apex coordinates of A, B, C, D and correspondence, B ', C ', D ' coordinate are divided Do not bring (1) and (2) two formulas into, available 8 equations, thus can solve 8 unknown numbers in matrix, Thus obtain transformation matrix T.Point all of in tetragon ABCD is all become through projective transformation matrix T Change, it is possible to obtaining the tetragon ABCD projection A ' B ' C ' D ' at positive screen orientation, the image finally obtained is exactly The target QR code image recovered.

Embodiment two: the flow process of the present embodiment is essentially identical with embodiment one, except that: in step 4 After described functional transformation, in QR code fault image, all of pixel point value fills into white；Then by former QR code The pixel value of each sample point of fault image is assigned to the QR code fault image after filling.

In the description of this specification, reference term " embodiment ", " some embodiments ", " example ", The description of " concrete example " or " some examples " etc. means to combine the concrete spy of this embodiment or example description Levy, structure, material or feature are contained at least one embodiment or the example of the present invention.In this explanation In book, the schematic representation of above-mentioned term is not necessarily referring to identical embodiment or example.And, retouch Specific features, structure, material or the feature stated can be in any one or more embodiments or example Combine in an appropriate manner.

Although an embodiment of the present invention has been shown and described, it will be understood by those skilled in the art that: These embodiments can be carried out in the case of without departing from the principle of the present invention and objective multiple change, amendment, Replacing and modification, the scope of the present invention is limited by claim and equivalent thereof.

Claims (8)

1. the preprocess method of a QR code fault image, it is characterised in that sequentially include the following steps:

Step one, choose the sample point of QR code normal picture, choose and described QR code at QR code fault image The sample point that normal picture position is corresponding；

Step 2, using the sample point of QR code normal picture as the input value during neural network learning, general With the sample point of QR code fault image as the output valve during neural network learning, carry out BP nerve net Network learns；

Step 3, the sample point chosen in pending QR code fault image obtain as through step 2 study The input value of neutral net, obtain one group of output valve, this group output valve carried out Function Fitting, determines The distortion mode of image, then by all of pixel in QR code fault image to be corrected, all passes through this letter Number carries out functional transformation；

Step 4, bilinear interpolation is utilized to determine the QR code distortion figure that target QR code image correspondence is to be corrected As the pixel value of sample point, obtain target QR code image.

The preprocess method of a kind of QR code fault image the most as claimed in claim 1, is characterized in that: also The step processed including target QR code image step 4 obtained with the inverse process of perspective projection transformation.

The preprocess method of a kind of QR code fault image the most as claimed in claim 1 or 2, is characterized in that: Described in step 3 after functional transformation, in QR code fault image, all of pixel point value fills into white；Then The pixel value of the former QR each sample point of code fault image is assigned to the QR code fault image after filling.

The preprocess method of a kind of QR code fault image the most as claimed in claim 1, is characterized in that: step Choose the sample point of QR code normal picture described in rapid one, with described QR code just choose at QR code fault image The sample point that often picture position is corresponding all sequentially includes the following steps:

A1, four summits found out in QR code image are as the first sample point；

A2, four sample points obtained with step A1 form the first tetragon for summit, take cornerwise intersection point It it is the second sample point；

A3, in the tetragon that step A2 obtains, take the midpoint of any two articles of adjacent edges respectively as the 3rd sample This point and the 4th sample point, the intersection point adding these two articles of limits and the second sample point composition the two or four taken out above Limit shape, taking its diagonal intersection point is the 5th sample point；

A4, return perform step A1.

The preprocess method of a kind of QR code fault image the most as claimed in claim 4, is characterized in that: step Choose the sample point of QR code normal picture described in rapid one, with described QR code just choose at QR code fault image The sample point that often picture position is corresponding is further comprising the steps of:

QR code image does Hough transformation process, find out the boundary contour of image, by the intersection point of contour line Obtain four apex coordinates of image.

The preprocess method of a kind of QR code fault image the most as claimed in claim 1, is characterized in that: institute State and step 2 uses Minimum Mean Square Error BP neural network learning is carried out error analysis.

7. the preprocess method of a kind of QR code fault image as described in claim 1 or 4, is characterized in that: Described step 2 use the most quick gradient descent method as the learning strategy of BP neural network learning.

The preprocess method of a kind of QR code fault image the most as claimed in claim 1, is characterized in that: institute State bilinear interpolation to sequentially include the following steps:

B1, set the coordinate of QR code fault image sample point as (i j), obtains it by reflection method backward At the floating-point coordinate (i+u, j+v) that QR code fault image is corresponding, wherein i, j are nonnegative integer, u, v For the floating number that [0,1] is interval；

B2, the pixel value setting described floating-point coordinate (i+u, j+v) they are f (i+u, j+v), (i+1, j), (i, j+1), (i+1, j+1) they are the coordinate in QR code fault image, calculating f (i+u, j+v)=(1-u) × (1-v) × f (i, j)+(1-u) × v × f (i, j+1)+u × (1-v) × f (i+1, j)+u × v × f(i+1,j+1)；Draw the pixel value of described floating-point coordinate (i+u, j+v)；Described f (i, j) be coordinate (i, J) pixel value, described f (i, j+1) is the pixel value of coordinate (i, j+1), and (i+1 j) is coordinate to described f (described f (i+1, j+1) is the pixel value of coordinate (i+1, j+1) for i+1, pixel value j).