CN105678757A - Object displacement measurement method - Google Patents

Abstract

An object displacement measurement method disclosed by the present invention comprises the following steps of 1 selecting a first frame of image P1 and any one frame of image M1 after the first frame; 2 carrying out the gray processing on the images; 3 obtaining images P2 and M2; 4 defining a specific area in the image P1 as a template T1; 5, obtaining a matching area similar with the template T1; 6 realizing the pixel-level matching; 7 calculating the correlation coefficients of each pixel point in an area of n*n, selecting the coordinate of the maximum value of a fitted curve as a matching point; 8 obtaining a corresponding coordinate position in the image M2; 9 utilizing the template T1 to select a corresponding template T2 in the image P2; 10 calculating the correlation coefficients of the template T2 at each point; 11 comparing the coordinate value corresponding to an optimal matching point with the coordinate of the template in the image P1; 12 repeating the steps 4 to 11 to calculate the statistics average value of the displacement, and taking the statistics average value as a final displacement result. The object displacement measurement method of the present invention has the advantages of high calculation efficiency, etc.

Description

A kind of ohject displacement measuring method

Technical field

The present invention relates to the sub-pix displacement measuring technology of a kind of digital picture, in particular to a kind of ohject displacement measuring method, this ohject displacement measuring method can be widely used in the measurement of the displacement of object, distortion and strain, it is also possible to object is carried out vibration analysis.

Background technology

To test as means, adopt the method for opticmeasurement, to study displacement, stress and strain as the flash ranging mechanics of main task, it is combine the cross discipline that multiple subject technology is integrated, it is widely used in a lot of test and detection field, and plays very important effect.

Digital picture correlation method (DigitalImageCorrelationMethod, DICM), as a kind of flash ranging mechanical technology, has measurement of full field, can't harm, measurement environment is required the advantages such as lower. Obtain and develop fast. In the last few years, the theoretical and engineer applied for digital image correlation technique, a lot of scholar had carried out a large amount of research work, achieved certain achievement in research. Along with constantly perfect, its Application Areas is more and more broader, will play an increasingly important role in engineering field.

Related operation is the key issue in digital image correlation technique, it is to increase the measuring accuracy of Digital Image Correlation Method is the urgent requirement of construction quality nondestructive testing. The cost improving measuring accuracy by promoting hardware is expensive and unrealistic, and from optimizing the angle of algorithm, it is to increase the thought of image sub-pixel position accuracy is economically viable. The algorithm of Integer Pel search is very ripe and perfect at present, comparatively speaking, calculates consuming time fewer, and sub-pixel location is the key calculating precision, is also wherein compare time-consuming link, and it directly affects the efficiency of relevant search, calculates precision and stability.

The Integer Pel positioning precision of the relevant search algorithm of great majority is consistent, just some difference in calculated amount, counting yield, noise robustness, stability etc., therefore the principal element determining the calculating precision of Digital Image Correlation Method is sub-pixel position accuracy, common method has sub-pix gray-level interpolation method, surface fitting method, coordinate rotate method, newton La Pusenfa, quasi-Newton method, gradient method, frequency domain correlation method, genetic algorithm, neural network algorithm etc., and the positioning precision that these algorithms can reach is from 0.005 to 0.1 pixel not etc.

Gray-level interpolation method requires by the method for interpolation, discrete gray scale field is carried out sub-pixel reconstruct, and the simplest gray-level interpolation method is that nearest-neighbor method of interpolation and these two kinds of interpolation method precision of bilinear interpolation are very low, and interpolation reconstruction can produce fuzzy. The interpolation method that precision comparison is high has Lagrange's interpolation, cube interpolation, bicubic spline interpolation, Fifth system. By to discrete gray scale field by the method for interpolation so that digital picture becomes approximate continuous image, then carry out essence search, choose the maximum position of relation conefficient as Optimum Matching position. This kind of method calculated amount is huge, and efficiency is lower.

Summary of the invention

The shortcoming that it is an object of the invention to overcome prior art is with not enough, it is provided that a kind of ohject displacement measuring method, this ohject displacement measuring method meets the requirement measuring high precision in actual engineering, ensure that the reasonableness of counting yield.

The object of the present invention can be achieved through the following technical solutions: a kind of ohject displacement measuring method, mainly comprises the following steps:

S1: still image gathers the position of equipment, then utilizes image capture device to gather the continuous moving image of target to be measured; The position of image capture device can not change, to obtain the image that reaction target moves; Choose any two field picture (M1) after the first two field picture (P1) and the first frame;

S2: if image is coloured image, then first it is carried out gray processing process;

S3: utilize interpolation algorithm P1 and M1 to carry out interpolation k doubly, obtain P2 and M2 after interpolation respectively;

S4: frame goes out a specific region comprising obvious characteristic as template (T1) in P1, the coordinate (x of the upper left angle point writing down Prototype drawing T1 in P1₀,y₀);

S5: utilize climbing method to mate in M1, obtains the matching area roughly similar to T1;

S6: in region obtained in the previous step, utilize SDA SSD algorithm accurately mate the coordinate (x obtaining template image₁,y₁), it is achieved Pixel-level is mated;

S7: with (x₁,y₁) centered by, calculating the relation conefficient of each pixel in n*n region, utilize the relation conefficient obtained to carry out surface fitting, the coordinate of maximum value choosing matching rear curved surface is as matching point (x₂,y₂);

S8: the coupling coordinate position (x in the M1 that step S7 is obtained₂,y₂) map to, in M2, obtaining coordinate position (x corresponding in M2₃,y₃);

S9: utilize the masterplate figure (T1) chosen in P1, chooses the Prototype drawing (T2) that it is corresponding in P2;

S10: in M2, with (x₃,y₃) it is starting point, repeating step S6, finds the coordinate of optimal match point, centered by this coordinate, chooses the rectangular area of m*m, calculates the relation conefficient of T2 on each point;

S11: utilize the relation conefficient obtained in previous step to carry out surface fitting, choose the coordinate (x of the maximum value after matching₄,y₄) as optimal match point, and this coordinate figure is mapped back coordinate figure (x corresponding in M1₅,y₅), and by (x₅,y₅) with the coordinate (x in P1 of template₀,y₀) compare, measure (Δ x, Δ y) with the precise displacement of realize target;

S12: above step S4 to step S11 is circulated and performs p time, choose different Prototype drawing every time, calculate the statistical average value of displacement, using this value as final mean annual increment movement result.

The calculation formula that the NCC algorithm calculating relation conefficient in above step S5, S6, S10 adopts is as follows:

$r_{xy}=\frac{\underset{i=1}{\overset{n}{\Σ}}\·\underset{j=1}{\overset{m}{\Σ}}\[\left(S_{x,y}\right(i,j)-\stackrel{\‾}{S_{x,y}})\left(T_{x,y}\right(i,j)-\stackrel{\‾}{T_{x,y}})\]}{\sqrt{\underset{i=1}{\overset{n}{\Σ}}\·\underset{j=1}{\overset{m}{\Σ}}{\[S_{x,y}(i,j)-\stackrel{\‾}{S_{x,y}}\]}^2}\sqrt{\underset{i=1}{\overset{n}{\Σ}}\·\underset{j=1}{\overset{m}{\Σ}}{\[T_{x,y}(i,j)-\stackrel{\‾}{T_{x,y}}\]}^2}},$

In above formula, r_xyIt is that the m*n subregion being initial point with point (x, y) follows the relation conefficient between template image, S_x,yWhat represent is the m*n subregion intercepted for initial point with point (x, y) in image to be matched, S_x,y(i, j) refers to the gray-scale value that on this subregion, coordinate (i, j) is put,Refer to the mean value of gray scale on this subregion.T_x,yThe gray-scale value that in (i, j) finger print version, coordinate (i, j) is put,The mean value of the gray scale on finger print plate. M, n represent row number and the line number of template respectively.

In above step S7, S11, the method carrying out surface fitting has conicoid fitting, three surface fitting methods, Gauss curved fitting process, two dimension Lagrange method surface fittings; Generally get conicoid fitting. The Binary quadratic functions that conicoid fitting adopts is as follows:

r(x_i,y_i)=a₀+a₁x_i+a₂y_i+a₃x²+a₄x_iy_i+a₅y_i ²,

In formula, r (x_i,y_i) represent that Prototype drawing is at coordinate (x_i,y_i) relation conefficient that calculates of place, coefficient a₀～a₅For this quadric coefficient. The coordinate formula of quadric maximum value is as follows:

$x=\frac{2a_1{a}_5-a_2{a}_4}{{a_4}^2-4a_3{a}_5},y=\frac{2a_2{a}_3-a_1{a}_3}{{a_4}^2-4a_3{a}_5},$

In formula, x, y represent X-coordinate and the ordinate zou of quadric maximum value respectively, coefficient a₀～a₅For this quadric coefficient.

In above step S8, the formula of virtual borderlines is as follows:

x₃'=(x₂-1)*n+1

y₃'=(y₂-1) * n+1,

x₃=the most contiguous x₃' integer,

y₃=the most contiguous y₃' integer

In formula, coordinate (x₂,y₂) represent the coupling coordinate position of the template obtained in step S7 in M1, coordinate (x'₃,y'₃) denotation coordination map result, coordinate position (x₃,y₃) represent the coordinate position that template is corresponding in M2.

The concrete grammar of above step S9 is as follows: first by the coordinate (x of the upper left angle point of Prototype drawing T1 at P1₀,y₀) map to the coordinate in P2, it is designated as (x₀',y₀'), choose the coordinate of Prototype drawing T1 bottom right angle point in P1 simultaneously, and mapped the coordinate in P2, be designated as (x₀",y₀"), so template T2 is taken as coordinate (x₀',y₀') and coordinate (x₀",y₀Rectangular area between "). Virtual borderlines formula is as follows:

x₂=(x₁-1)*k+1

y₂=(y₁-1) * k+1,

Wherein, coordinate (x₁,y₁) and (x₂,y₂) coordinate that to be respectively in P1 and P2 corresponding, k represents the multiple of interpolation in step s3.

In above step S11, virtual borderlines formula is as follows:

x₅=(x₄-1)/k+1,

y₅=(y₄-1)/k+1

In formula, coordinate (x₄,y₄) represent the optimum matching coordinate that M2 is total, coordinate (x₅,y₅) representing the optimum matching coordinate in M1, it is obtained by above-mentioned mapping equation, and k represents the multiple of interpolation in step s3.

In above step S11, displacement calculation formula is as follows:

Δ x=x₅-x₀,

Δ y=y₅-y₀

In formula, coordinate (x₅,y₅) represent the optimum matching coordinate in M1, coordinate (x₀,y₀) representing the coordinate of the upper left angle point of Prototype drawing T1 in step S4 in P1, Δ x represents the precise displacement of target on X-coordinate direction, and Δ y represents the precise displacement of target on ordinate zou direction.

The value of k, n, m and p in step S3, S7, S10 and S12 can be any positive integer, and the more big computing amount of value is more big, and span is preferably [1,10].

In step S3, S7, S8, S9, S10 and S11, the sub-pixel interpolation algorithm of the relevant search of classics and surface fitting method are combined, drastically increases the precision of sub-pixel displacement measurement; Simultaneously by the image before interpolation and after interpolation mates, significantly reduce computation complexity.

The object of the present invention can also be achieved through the following technical solutions: a kind of ohject displacement measuring method, comprises the following steps: (1) utilizes image capture device to gather the continuous displacement image of target compound to be measured; (2) utilize interpolation algorithm that the image gathered is carried out interpolation; (3) in first frame, choose Prototype drawing, then utilize climbing method and SDA the SSD algorithm rough coordinates position of finding template in successive image; (4) it is carried out matching, obtain similarity curved surface, ask for curved surface peak value coordinate, compare with the coordinate in former figure, calculate displacement; (5) repeating (3)～(4) step several times, the mean value of displacement calculating is as net result.The inventive method is when engineer applied, repeatedly choose Prototype drawing to measure, simultaneously by image calculating before interpolation and after interpolation, and combine ingenious to interpolation algorithm and Algorithm for Surface Fitting, improve the efficiency of calculating, it is achieved that accurate sub-pixel displacement measurement.

The present invention has following advantage and effect relative to prior art:

Surface fitting method assumes that the relation conefficient matrix of Integer Pel displacement relevant search result and consecutive point thereof can fit to continuous curve surface, then using the extreme point position of this curved surface as the central position of the rear image subsection of distortion. Quadric surface, three curved surfaces, Gauss curved and two dimension Lagrange curved surface etc. are had for the curved surface type of matching. The method calculating precision height of surface fitting, noise resisting ability are stronger. The present invention combines ingenious to the gray-level interpolation method in sub-pix displacement measurement and surface fitting method so that the precision of calculating is greatly improved; Simultaneously by calculation template on first picture before interpolation matched position, and then transfer to the exact position of calculation template figure on the picture after interpolation, the advantage of this mode is that the picture before movement mates, the size of Prototype drawing and former figure is all less, counting yield is higher, comparing calculation template figure on direct picture after interpolation, counting yield is greatly improved; And choose multiple coupling figure, finally calculate the average of statistics, to obtain more accurate result.

Accompanying drawing explanation

Fig. 1 is the concrete implementing procedure figure of algorithm of climbing the mountain.

Fig. 2 is specific embodiment of the invention schema.

Embodiment

Below in conjunction with embodiment and accompanying drawing, the present invention is described in further detail, but embodiments of the present invention are not limited to this.

Embodiment

As shown in Figure 2, a kind of ohject displacement measuring method, specifically comprises the following steps:

S1: still image gathers the position of equipment, then utilizes image capture device to gather the continuous moving image of target to be measured; The position of image capture device can not change, to obtain the image that reaction target moves; Choose any two field picture (M1) after the first two field picture (P1) and the first frame; And in order to obtain higher measuring accuracy, the photo obtained should have sufficiently high quality.

S2: if image is coloured image, then first it is carried out gray processing process;

S3: utilize interpolation algorithm P1 and M1 to carry out interpolation k doubly, obtain P2 and M2 after interpolation respectively; The interpolation algorithm that can select is a lot, generally utilizes cube sum algorithm can obtain the high picture of precision comparison, and the multiple of interpolation should be determined according to the actual needs of counting yield, generally gets 10 times.

S4: the specific region that frame goes out to comprise obvious characteristic in P1 as template (T1), the coordinate (x of the upper left angle point writing down Prototype drawing T1 in P1₀,y₀); The big I of template is determined according to the actual needs of counting yield, is generally taken as 41*41 pixel or 51*51 pixel.

S5: utilize climbing method to mate in M1, obtains the matching area roughly similar to T1; ;

S6: in region obtained in the previous step, utilize SDA SSD algorithm accurately mate the coordinate (x obtaining template image₁,y₁), it is achieved Pixel-level is mated;

S7: with (x₁,y₁) centered by, calculating the relation conefficient of each pixel in n*n region, utilize the relation conefficient obtained to carry out surface fitting, the coordinate of maximum value choosing matching rear curved surface is as matching point (x₂,y₂);Choosing of curved surface fitting method can require according to concrete counting yield and determine, and generally, chooses conicoid fitting and just can meet requirement.

S8: the coupling coordinate position (x in the M1 that step S7 is obtained₂,y₂) map to, in M2, obtaining coordinate position (x corresponding in M2₃,y₃);

S9: utilize the masterplate figure (T1) chosen in P1, chooses the Prototype drawing (T2) that it is corresponding in P2;

S10: in M2 is Prototype drawing taking T2, with (x₃,y₃) it is starting point, repeating step S6, finds the coordinate of optimal match point, centered by this coordinate, chooses the rectangular area of m*m, calculates the relation conefficient of T2 on each point; Due to the matched position (x that step S8 obtains₂,y₂) very close to the matched position of reality, so this step is quickly at the coordinate speed utilizing step S6 to search for Integer Pel point Optimum Matching point.

S11: utilize the relation conefficient obtained to carry out surface fitting, choose the coordinate (x of the maximum value after matching₄,y₄) as optimal match point, and this coordinate figure is mapped back coordinate figure (x corresponding in M1₅,y₅), and by (x₅,y₅) with the coordinate (x in P1 of template₀,y₀) compare, to obtain the precise displacement (Δ x, Δ y) of target;

S12: repeat above S4～S11 step p time, choose different Prototype drawing every time, calculate the statistical average value of displacement, using this value as final mean annual increment movement result; Specifically need repetition how many times can determine according to the counting yield that reality is wanted.

The performance of the search procedure of climbing the mountain in above-mentioned steps S5 is by very multifactor impacts such as masterplate and image size to be matched, the intensity profile of template image, the algorithms of calculating relation conefficient. As shown in Figure 1, step S5 can Further Division be:

S5.1: according to the relative magnitude relationship of subgraph to be matched with masterplate subgraph, chooses all horizontal and vertical spacing between starting point of initially climbing the mountain in climbing method, generates all starting points of climbing the mountain in image to be matched. If the actual target travel window measured is less, it is possible to all starting points of initially climbing the mountain are set in coordinate (x₀,y₀) near, to improve the speed climbed the mountain. In order to react former figure more accurately with the degree of correlation between masterplate subgraph, selected starting point of climbing the mountain should be uniformly distributed as far as possible, and distance between starting point is moderate.

S5.2: the relation conefficient calculating all starting points of climbing the mountain, and they reverse. Through the Rational choice starting point of S5.1, it is possible to think that the relation conefficient size of starting point exists certain positive correlation with the distance between itself and the maximum value of relation conefficient, check these starting points can find target quickly according to relation conefficient backward.

S5.3: according to the requirement to coupling tolerance range, select a suitable relation conefficient upper limit, is used for terminating search procedure. In addition, at S5.1 step Rational choice on the basis of starting point of climbing the mountain, it is believed that the excessively little point of relation conefficient with object point apart from too far away, it is possible to directly give up. Choosing with choosing of starting point in S5.1 of this lower value is closely related, has got 0.25 in embodiment.

S5.4: after determining current point, calculates the relation conefficient of each point in surrounding 3*3 matrix successively. If there being the point that relation conefficient is bigger, then it is continued to search as new starting point, if not, enter next step.

S5.5: judge whether the maximum correlation coefficient that S5.4 finds exceedes the upper limit of step S5.3 setting, if not, the calculating inspection next one is climbed the mountain starting point, otherwise thinks and have found satisfactory point, directly terminates climbing the mountain process.

Above-described embodiment is that the present invention preferably implements mode; but embodiments of the present invention are not restricted to the described embodiments; the change done under the spirit of other any the present invention of not deviating from and principle, modification, replacement, combination, simplification; all should be the substitute mode of equivalence, it is included within protection scope of the present invention.

Claims (4)

1. an ohject displacement measuring method, it is characterised in that, comprise the following steps:

S1, still image gather the position of equipment, then utilize image capture device to gather the continuous moving image of target to be measured; Choose any two field picture M1 after the first two field picture P1 and the first frame;

If S2 image is coloured image, then first it is carried out gray processing process;

S3, utilize interpolation algorithm that P1 and M1 is carried out interpolation k times of computing, after interpolation, obtain P2 and M2 respectively;

S4, in P1, frame goes out a specific region comprising obvious characteristic as template T1;

S5, utilize climbing method to mate in M1, obtain the matching area roughly similar to T1;

S6, in region obtained in the previous step, utilize SDA SSD algorithm accurately mate the coordinate (x obtaining template image₁,y₁), it is achieved Pixel-level is mated;

S7, with (x₁,y₁) centered by, calculating the relation conefficient of each pixel in n*n region, utilize the relation conefficient obtained to carry out surface fitting, the coordinate of maximum value choosing the curved surface that matching obtains is as matching point (x₂,y₂);

Coupling coordinate position (x in S8, the M1 that step S7 is obtained₂,y₂) map to, in M2, obtaining coordinate position (x corresponding in M2₃,y₃);

The masterplate T1 that S9, utilization are chosen in P1, chooses the template T2 that it is corresponding in P2;

S10, in M2, be Prototype drawing taking T2, with (x₃,y₃) it is starting point, repeating step S6, finds the coordinate of optimal match point, centered by this coordinate, chooses the rectangular area of m*m, calculates the relation conefficient of T2 on each point;

The relation conefficient that S11, utilization obtain carries out surface fitting, chooses the coordinate (x of the maximum value of the curved surface that matching obtains₄,y₄) as optimal match point, and this coordinate figure is mapped back coordinate figure (x corresponding in M1₅,y₅), and by (x₅,y₅) compare with the coordinate in P1 of template;

S12, described step S4～step S11 circulation is performed p time, calculate the statistical average value of displacement, using statistical average value as final mean annual increment movement result.

2. ohject displacement measuring method as claimed in claim 1, it is characterised in that, the value of k, n, m and p in step S3, S7, S10 and S12 is any positive integer.

3. ohject displacement measuring method as claimed in claim 2, it is characterised in that, the span of k, n, m and p is [1,10].

4. ohject displacement measuring method as claimed in claim 1, it is characterised in that, in step S3, S7, S8, S9, S10 and S11, sub-pixel interpolation algorithm and surface fitting method are combined, simultaneously by the image before interpolation and after interpolation mates.