CN103207987A - Indicating value identification method of dial instrument - Google Patents

Abstract

The invention discloses an indicating value identification method of a dial instrument. The indicating value identification method of the dial instrument belongs to the field of artificial intelligence. The technical scheme of the indicating value identification method of the dial instrument is that a computer automatically acquires a dial plate image of the dial instrument, divides out needed image feature regions, obtains the feature of a number to determine the numerical value of the number according to centre-of-gravity position, perpendicular line features and horizontal projection operations, and finally computes the actual evaluation of the dial instrument through an abscissa of the number and a former pointer and a scale mark. The image processing method for reading the indicating value of the dial instrument can accurately acquires the coordinate value of the pointer and coordinate value of an intersection of the pointer and scale marks, can avoid indicating value calculation deviation problems caused by camera lens distortion, uneven scales of dial plates, or the noncoincidence of a fixed point of the pointer and the centre of a circle which passes through the scale mark of the dial plate due to the fact that the angle method is utilized in the prior art, improves the efficiency and accuracy of reading the indicating value of the dial instrument, is applicable to a plurality of hardware platforms, and has high transportability and stability.

Description

A kind of registration recognition methods of pointer instrument

Technical field

The invention belongs to artificial intelligence field, relate in particular to a kind of registration recognition methods of pointer instrument.

Background technology

Pointer instrument still has a wide range of applications in commercial production and daily life, for the registration interpretation of pointer instrument generally all by manually finishing, be the other pointer of remarkable outlook and the relative position of scale mark and the absolute position of scale mark, estimate the indicating value of reading instrument.This method inefficiency, labour intensity are big, very easily make the mistake.

For solving this type of problem, developed the interpreting system based on the pointer instrument of computer vision technique.The shortcoming that but present method exists, and operand is big, complexity is high, the hardware platform real-time is required.

Summary of the invention

Low in the identification accuracy of gauge pointer, scale mark position at the existing method of mentioning in the background technology, problems such as inefficiency the present invention proposes a kind of registration recognition methods of pointer instrument.

A kind of registration recognition methods of pointer instrument is characterized in that, described method specifically may further comprise the steps:

Step 1: with camera collection instrument dial plate image, be sent to Computer Storage and demonstration;

Step 2: the centre of gyration coordinate of determining pointer with method of difference;

Step 3: with the Region Segmentation method instrument dial plate image is cut apart, the tracing area condition of setting dial plate is that girth is the longest, thereby obtains comprising in the dial plate image image of numeral, scale mark and pointer part, and the image that obtains is carried out binaryzation;

Step 4: the image that will comprise numeral, scale mark and pointer part carries out polar coordinate transform;

Step 5: by in the image that comprises numeral, scale mark and pointer part, making vertical straight line, obtain the information of scale mark;

Step 6: adopting contour tracing method, is the filling condition in zone with girth maximum and area minimum, is only comprised the image of numerical information;

Step 7: the image that only comprises numerical information is carried out row rectilinear scanning, the zone of the definite numeral that will identify;

Step 8: according to centre of gravity place, perpendicular line feature and horizontal projection operation, draw this digital feature to determine this digital numerical value;

Step 9: at last by this numeral horizontal ordinate of prior pointer and scale mark with it, calculate the actual evaluation of this pointer instrument.

Describedly determine that with method of difference the concrete steps of the centre of gyration coordinate of pointer comprise:

Step 201: at first gather a width of cloth original image I ₁(x, y), make pointer rotation 30～45 degree then after, gather image I again ₂(x, y);

Step 202: this two width of cloth image is carried out subtraction, and the pixel of different gray areas is set to 1, and the pixel of same grayscale is set to 0;

Step 203: obtained removing the sketch figure that identifies on background and other dial plate (x y), thereby determines the centre of gyration of pointer as Δ I by step 202.

Described process with coordinate system transformation is:

Step 401: set image that step 3 obtains and be I (x, y);

Step 402: with the origin of image in the step 401 move to pointer the centre of gyration (X, Y), to the image I after the translation ' (x-X y-Y) carries out flip vertical;

Step 403: to the upset after image I ' ' ((x-X), y-Y) carry out polar coordinate transform according to the polar coordinate transform formula; Described polar coordinate transform formula coordinate is:

$\left\{\begin{array}{c}\mathrm{\θ}=\arctan \left(\frac{y-Y}{-(x-X)}\right)\\ \mathrm{\ρ}=\sqrt{{(x-X)}^2+{(y-Y)}^2}\end{array}\right.$

Described by make vertical straight line in the dial plate image of pointer instrument, the process that obtains the information of scale mark is:

Step 501: in the image that step 3 obtains, make vertical straight line from top to bottom;

Step 502: the data that scale mark overlaps in the image that obtains by vertical line and step 3 and the number of scale mark pixel obtain the information of scale mark;

Described the image that only comprises numerical information is carried out row rectilinear scanning, determines that the process in the zone of numeral to be identified is:

Step 701: to step 6 to only comprise the numeral image carry out row rectilinear scanning, calculate each digital center horizontal ordinate cx _j

Step 702: compare each cx successively _jCentre of gyration horizontal ordinate X with pointer _pApart from d, and d=|cx _j-X _p|;

Step 703: satisfy min (d) and cx _j＜X _pNumeric area, be the zone of numeral to be identified.

Described according to centre of gravity place, perpendicular line feature and horizontal projection operation, draw this digital feature and with the process of determining this digital numerical value be:

Step 801: to the normalization of carrying out in the zone of numeral to be identified, and be benchmark with the center line, the zone of numeral to be identified is divided into upper and lower two parts, setting size simultaneously is M * N, upper and lower two parts pixel number to numeric area to be identified is added up, and computing formula is:

${\mathrm{<figure-callout\; id="1"\; label="S"\; filenames="HDA00002867206700012.png"\; state="\{\{state\}\}">S</figure-callout>}}_1=\underset{x=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{y=1}{\overset{\frac{N}{2}}{\mathrm{\&Sigma;}}}g(x,y)$ ${\mathrm{<figure-callout\; id="2"\; label="S"\; filenames="HDA00002867206700012.png"\; state="\{\{state\}\}">S</figure-callout>}}_2=\underset{x=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{y=\frac{N}{2}+1}{\overset{N}{\mathrm{\&Sigma;}}}g(x,y)$

Wherein, S ₁Number for the latter half pixel; S ₂Number for working partial pixel point;

(x y) is the coordinate of pixel in the zone of numeral to be identified to g.

Step 802: according to S ₁And S ₂Magnitude relationship, with center of gravity be divided on the upper side, 3 classes between two parties and on the lower side; Work as S ₁S ₂, for center of gravity on the upper side; Work as S ₁=S ₂, then be that center of gravity is placed in the middle; Work as S ₁＜S ₂, for center of gravity on the lower side;

Step 803: be all digital packet 3 groups according to step 802, upper group comprises 5,7 and 9; The middle part group comprises 0,1,3 and 8; Lower set comprises 2,4 and 6;

Step 804: judge whether numeral to be identified belongs to middle part group, if, execution in step 805 successively then: otherwise, begin to carry out from step 807;

Step 805: for 4 numbers of middle part group, 0,1,3,8; 3 vertical scan lines are placed the N/3 of digital picture width respectively, N/2, the 2N/3 place, and calculate the intersection point number of sweep trace and numeral;

${\mathrm{<figure-callout\; id="1"\; label="S"\; filenames="HDA00002867206700012.png"\; state="\{\{state\}\}">S</figure-callout>}}_{\frac{1}{3}}=\underset{y=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{x=\frac{M}{3}}{\mathrm{\&Sigma;}}g(x,y)$ ${\mathrm{<figure-callout\; id="1"\; label="S"\; filenames="HDA00002867206700012.png"\; state="\{\{state\}\}">S</figure-callout>}}_{\frac{1}{2}}=\underset{y=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{x=\frac{M}{2}}{\mathrm{\&Sigma;}}g(x,y)$ ${\mathrm{<figure-callout\; id="2"\; label="S"\; filenames="HDA00002867206700012.png"\; state="\{\{state\}\}">S</figure-callout>}}_{\frac{2}{3}}=\underset{y=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{x=\frac{2M}{3}}{\mathrm{\&Sigma;}}g(x,y)$

Wherein,

Be respectively the intersection point number of sweep trace and numeral; (x y) is the coordinate of pixel in the zone of numeral to be identified to g;

Therefore, each numeral is for 1 proper vector:

$R\left(i\right)=[{\mathrm{<figure-callout\; id="1"\; label="S"\; filenames="HDA00002867206700012.png"\; state="\{\{state\}\}">S</figure-callout>}}_{\frac{1}{3}},{\mathrm{<figure-callout\; id="1"\; label="S"\; filenames="HDA00002867206700012.png"\; state="\{\{state\}\}">S</figure-callout>}}_{\frac{1}{2}},{\mathrm{<figure-callout\; id="2"\; label="S"\; filenames="HDA00002867206700012.png"\; state="\{\{state\}\}">S</figure-callout>}}_{\frac{2}{3}}]i=\mathrm{0,1,3,8}$

Step 806: proper vector is organized each digital corresponding proper vector with the middle part and is mated after extracting, and jumps to 809; The proper vector table is as shown in table 3:

Table 1 proper vector table

Numeral	Proper vector
		0	R(0)=[2,2,2]
1	R(1)=[1,1,N]N>1,
		3	R(3)=[2,3,3]
8	R(8)=[3,3,3]

Step 807: feature extraction and identification are carried out in the digit basis horizontal direction projection for upper group and lower set, and numeral to be identified has so just had N feature C (i);

$C\left(i\right)=\underset{1\&le;x\&le;M}{\mathrm{\&Sigma;}}\underset{y=i}{\mathrm{\&Sigma;}}g(x,y)i=\mathrm{2,4,5,6,7,9}$

Step 808: need differentiate the numerical characteristic that extracts based on the diagnostic method of minimum distance method at last:

$J\left(i\right)=\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}|Y\left(k\right)-X(i,k)|i=\mathrm{2,4,5,6,7,9}$

Wherein, J (i) is the distance between sample to be detected and i class template; Y (k) is i feature of sample to be detected, and (i k) is i feature of k class template to X.

Step 809: obtain the intersecting point coordinate of pointer and scale mark, again according to the nearest digital value in pointer left side, calculate the registration numerical value of instrument according to the registration computing formula; Described registration computing formula is:

$V=i+\frac{L_i}{L}$

Wherein, V is the registration numerical value of instrument, and i is that the pointer left side is from the nearest digital occurrence of pointer, L _iFor pointer and scale mark intersecting point coordinate and the left side from the distance between the nearest numeral of pointer, L is the distance between two numerals on the dial plate.

Useful effect of the present invention is, the image processing method of the reading pointer type instrument indication value that provides can accurately obtain the coordinate figure of pointer and pointer and scale mark intersection point, and can avoid occurring as the available technology adopting preset angle configuration and because of cam lens distortion, dial scale is inhomogeneous or the center of circle of circle by the dial scale line does not overlap the indicating value calculation deviation problem of bringing with the point of fixity of pointer, has improved efficient and the accuracy of the indicating value of reading pointer type instrument; And can be fit to the multiple hardwares platform, have very strong portability and stable.

Description of drawings

Fig. 1 is the process flow diagram of the registration recognition methods of a kind of pointer instrument provided by the invention;

Fig. 2 is coordinate system transformation result schematic diagram provided by the invention;

Fig. 3 is each digital perpendicular line scanning synoptic diagram provided by the invention.

Embodiment

Below in conjunction with accompanying drawing, preferred embodiment is elaborated.Should be emphasized that following explanation only is exemplary, rather than in order to limit the scope of the invention and to use.

Below in conjunction with accompanying drawing the present invention is described in further detail:

" obtain pointer instrument dial plate image; storage also shows ", the dial plate image of pointer instrument is by camera collection, read in computer system by USB interface, utilize the automatic discriminating digit program in the computing machine, the form of one frame data with array read in the calculator memory, deliver to simultaneously in the corresponding display memory, the concrete indicating value with instrument is presented on the screen at last.

Determine the centre of gyration coordinate of pointer with " method of difference ".After determining the position of tested table, take a width of cloth original image, gather second width of cloth image after the width of cloth pointer wide-angle deflection then.If the nearer use least square method of two width of cloth image cursor is difficult to carry out fitting of two straight lines respectively, the image corresponding pixel points that two width of cloth pointer positions are different is done and is subtracted computing.The pixel of different gray areas is set to 1, and the location of pixels of same grayscale is 0.This just comprises background in the index dial that scale mark and Digital ID deduct and only stay vicissitudinous part in the image, diverse location pointer.Obtain the sketch figure picture, thereby can detect the existence of pointer and the centre of gyration information of definite dial plate pointer.

The pointer instrument dish view data that gather to obtain is that the form with matrix exists in the internal memory; Therefore " coordinate system transformation " will namely be that image is transformed into the polar coordinate system from rectangular coordinate system.According to the dial plate characteristics of image, earlier the dial plate image is carried out coordinate translation; The centre of gyration with pointer is new true origin; It is the upwards upset of axle that image after this translation is carried out with the horizontal ordinate, carries out polar coordinate transform afterwards.Scale mark is always perpendicular to abscissa axis, and therefore the row rectilinear scanning that the dial plate image after the conversion is carried out from top to bottom can obtain scale mark information accurately.Draw vertical line (from top to bottom) in binaryzation rectangle dial plate refined image, vertical line is combined scale mark pixel number with the data of scale mark intersection, and scale mark is the characteristics of continuous straight line in the dial plate image after variation, can effectively judge the type of scale mark.

Use the profile tracing, the condition that setting regions is filled is the maximum and area minimum of girth, obtains removing the image in pointer and scale mark zone.Try to achieve from the center of a nearest numeral of cursor line centre of gyration horizontal ordinate or the coordinate in the upper left corner, determine numeric area to be identified.

The feature extraction of numeral mainly is to carry out according to centre of gravity place, perpendicular line feature and horizontal projection.

After extracting numeric area to be identified, be benchmark with the center line of image, image is divided into upper and lower two parts; The black pixel point sum of each several part successively adds up.Two results of gained are compared, if the top sum greater than the lower part, then the numeral in the image belongs to upper group; Equate, then belong to the middle part group; The top sum then belongs to lower set less than the lower part.The numeral of upper group is 5,7,9; The numeral of middle part group is 0,1,3,8; Lower set then is 2,4,6.

For four numerals of difference middle part group, must use the perpendicular line feature.The perpendicular line feature then is 1/3,1/2,2/3 to put vertical scan line respectively on the horizontal ordinate of digital picture, and calculates the intersection point number of each perpendicular line and numeral respectively.4 numbers of middle part group, the number of hits on these perpendicular line has tangible difference, therefore can finely make a distinction.

For the differentiation of each numeral in upper group and the middle part group, then must the operation of usage level direction projection.Horizontal projection, horizontal direction projection refer to along the set of the gray-scale value accumulation calculating amount in image level direction cross section, can be reduced to the set of the count value of black pixel under bianry image.Therefore after horizontal projection, digital picture is divided into several zones in the vertical; Add up the total number of black picture element in this zone then.We claim that this total number is the feature of digital picture in this zone.After feature extraction, can adopt minimum distance method that sample characteristics to be detected and template characteristic are mated.

The above; only for the preferable embodiment of the present invention, but protection scope of the present invention is not limited thereto, and anyly is familiar with those skilled in the art in the technical scope that the present invention discloses; the variation that can expect easily or replacement all should be encompassed within protection scope of the present invention.Therefore, protection scope of the present invention should be as the criterion with the protection domain of claim.

Claims (6)

1. the registration recognition methods of a pointer instrument is characterized in that, described method specifically may further comprise the steps:

Step 1: with camera collection instrument dial plate image, be sent to Computer Storage and demonstration;

Step 2: the centre of gyration coordinate of determining pointer with method of difference;

Step 3: with the Region Segmentation method instrument dial plate image is cut apart, the tracing area condition of setting dial plate is that girth is the longest, thereby obtains comprising in the dial plate image image of numeral, scale mark and pointer part, and the image that obtains is carried out binaryzation;

Step 4: the image that will comprise numeral, scale mark and pointer part carries out polar coordinate transform;

Step 5: by in the image that comprises numeral, scale mark and pointer part, making vertical straight line, obtain the information of scale mark;

Step 6: adopting contour tracing method, is the filling condition in zone with girth maximum and area minimum, is only comprised the image of numerical information;

Step 7: the image that only comprises numerical information is carried out row rectilinear scanning, the zone of the definite numeral that will identify;

Step 8: according to centre of gravity place, perpendicular line feature and horizontal projection operation, draw this digital feature to determine this digital numerical value;

Step 9: at last by this numeral horizontal ordinate of prior pointer and scale mark with it, calculate the actual evaluation of this pointer instrument.

2. the registration recognition methods of a kind of pointer instrument according to claim 1 is characterized in that, describedly determines that with method of difference the concrete steps of the centre of gyration coordinate of pointer comprise:

Step 201: at first gather a width of cloth original image I ₁(x, y), make pointer rotation 30～45 degree then after, gather image I again ₂(x, y);

Step 202: this two width of cloth image is carried out subtraction, and the pixel of different gray areas is set to 1, and the pixel of same grayscale is set to 0;

Step 203: obtained removing the sketch figure that identifies on background and other dial plate (x y), thereby determines the centre of gyration of pointer as Δ I by step 202.

3. the registration recognition methods of a kind of pointer instrument according to claim 1 is characterized in that, described process with coordinate system transformation is:

Step 401: set image that step 3 obtains and be I (x, y);

Step 402: with the origin of image in the step 401 move to pointer the centre of gyration (X, Y), to the image I after the translation ' (x-X y-Y) carries out flip vertical;

Step 403: to the upset after image I ' ' ((x-X), y-Y) carry out polar coordinate transform according to the polar coordinate transform formula; Described polar coordinate transform formula coordinate is:

$\left\{\begin{array}{c}\mathrm{\&theta;}=\arctan \left(\frac{y-Y}{-(x-X)}\right)\\ \mathrm{\&rho;}=\sqrt{{(x-X)}^2+{(y-Y)}^2}\end{array}\right..$

4. the registration recognition methods of a kind of pointer instrument according to claim 1 is characterized in that, described by make vertical straight line in the dial plate image of pointer instrument, the process that obtains the information of scale mark is:

Step 501: in the image that step 3 obtains, make vertical straight line from top to bottom;

Step 502: the data that scale mark overlaps in the image that obtains by vertical line and step 3 and the number of scale mark pixel obtain the information of scale mark.

5. the registration recognition methods of a kind of pointer instrument according to claim 1 is characterized in that, described the image that only comprises numerical information is carried out row rectilinear scanning, determines that the process in the zone of numeral to be identified is:

Step 701: to step 6 to only comprise the numeral image carry out row rectilinear scanning, calculate each digital center horizontal ordinate cx _j

Step 702: compare each cx successively _jCentre of gyration horizontal ordinate X with pointer _pApart from d, and d=|cx _j-X _p|;

Step 703: satisfy min (d) and cx _j＜X _pNumeric area, be the zone of numeral to be identified.

6. the registration recognition methods of a kind of pointer instrument according to claim 1 is characterized in that, and is described according to centre of gravity place, perpendicular line feature and horizontal projection operation, draws this digital feature and with the process of determining this digital numerical value is:

Step 801: to the normalization of carrying out in the zone of numeral to be identified, and be benchmark with the center line, the zone of numeral to be identified is divided into upper and lower two parts, setting size simultaneously is M * N, upper and lower two parts pixel number to numeric area to be identified is added up, and computing formula is:

$S_1=\underset{x=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{y=1}{\overset{\frac{N}{2}}{\mathrm{\&Sigma;}}}g(x,y)$ $S_2=\underset{x=1}{\overset{M}{\mathrm{\&Sigma;}}}\underset{y=\frac{N}{2}+1}{\overset{N}{\mathrm{\&Sigma;}}}g(x,y)$

Wherein, S ₁Number for the latter half pixel; S ₂Number for working partial pixel point;

(x y) is the coordinate of pixel in the zone of numeral to be identified to g.

Step 802: according to S ₁And S ₂Magnitude relationship, with center of gravity be divided on the upper side, 3 classes between two parties and on the lower side; Work as S ₁S ₂, for center of gravity on the upper side; Work as S ₁=S ₂, then be that center of gravity is placed in the middle; Work as S ₁＜S ₂, for center of gravity on the lower side;

Step 803: be all digital packet 3 groups according to step 802, upper group comprises 5,7 and 9; The middle part group comprises 0,1,3 and 8; Lower set comprises 2,4 and 6;

Step 804: judge whether numeral to be identified belongs to middle part group, if, execution in step 805 successively then; Otherwise, begin to carry out from step 807;

Step 805: for 4 numbers of middle part group, 0,1,3,8; 3 vertical scan lines are placed the N/3 of digital picture width respectively, N/2, the 2N/3 place, and calculate the intersection point number of sweep trace and numeral;

$S_{\frac{1}{3}}=\underset{y=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{x=\frac{M}{3}}{\mathrm{\&Sigma;}}g(x,y)$ $S_{\frac{1}{2}}=\underset{y=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{x=\frac{M}{2}}{\mathrm{\&Sigma;}}g(x,y)$ $S_{\frac{2}{3}}=\underset{y=1}{\overset{N}{\mathrm{\&Sigma;}}}\underset{x=\frac{2M}{3}}{\mathrm{\&Sigma;}}g(x,y)$

Wherein,

Be respectively the intersection point number of sweep trace and numeral; (x y) is the coordinate of pixel in the zone of numeral to be identified to g;

Therefore, each numeral is for 1 proper vector:

$R\left(i\right)=[S_{\frac{1}{3}},S_{\frac{1}{2}},S_{\frac{2}{3}}]i=\mathrm{0,1,3,8}$

Step 806: proper vector is organized each digital corresponding proper vector with the middle part and is mated after extracting, and jumps to step 809;

Step 807: feature extraction and identification are carried out in the digit basis horizontal direction projection for upper group and lower set, and numeral to be identified has so just had N feature C (i);

$C\left(i\right)=\underset{1\&le;x\&le;M}{\mathrm{\&Sigma;}}\underset{y=i}{\mathrm{\&Sigma;}}g(x,y)i=\mathrm{2,4,5,6,7,9}$

Step 808: need differentiate the numerical characteristic that extracts based on the diagnostic method of minimum distance method at last:

$J\left(i\right)=\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}|Y\left(k\right)-X(i,k)|i=\mathrm{2,4,5,6,7,9}$

Wherein, J (i) is the distance between sample to be detected and i class template; Y (k) is i feature of sample to be detected, and (i k) is i feature of k class template to X.

Step 809: obtain the intersecting point coordinate of pointer and scale mark, again according to the nearest digital value in pointer left side, calculate the registration numerical value of instrument according to the registration computing formula; Described registration computing formula is:

$V=i+\frac{L_i}{L}$

Wherein, V is the registration numerical value of instrument, and i is that the pointer left side is from the nearest digital occurrence of pointer, L _iFor pointer and scale mark intersecting point coordinate and the left side from the distance between the nearest numeral of pointer, L is the distance between two numerals on the dial plate.

Images