CN103337067A - Visual sense detection method for single needle scanning type screw thread measuring instrument probe X-axis rotation deviation - Google Patents

Abstract

The invention provides a visual sense detection method for single needle scanning type screw thread measuring instrument probe X-axis rotation deviation and relates to the image processing technology field. The visual sense detection method comprises steps of: step1, initializing calibration; step2, detecting a probe tip in real time; and step3, calculating a deviation angle. The visual sense detection method can realize measurement and adjustment on X-axis rotation positioning deviation through an image processing means, and improves efficiency while mounting precision is guaranteed.

Description

The visible detection method of single needle scan-type screw measurement instrument probe X-axis rotating deviation

Technical field

The invention provides a kind of visible detection method of single needle scan-type screw measurement instrument probe X-axis rotating deviation, relate to technical field of image processing.

Background technology

Threaded connector is the mechanical component of mechanical industry widespread use, its manufacturing accuracy directly affects reliability, assembly precision and the interchangeability that parts connect, particularly in the manufacturing and designing of aerospacecraft, the a large amount of employing is threaded, connection between nearly part more than 50% will realize by threaded engagement, connects reliability and life-span that quality has determined industrial products such as aircraft.

For STRTHD, often adopt ring standard gauge (feeler gauge) to screw differentiation, yet for the higher screw thread of accuracy requirement, as the measurement of ring gauge (feeler gauge) itself, just needing more accurately, instrument detect.At present, on high precision screw surveying instrument equipment, ripe both at home and abroad product is not a lot.The MSXP screw measurement instrument that has Dutch IAC company to produce abroad, the contact pin type scanning profile method of employing.Domestic also do not have a special manufacturer production contact screw measurement instrument.Breathing out quantity set group has the surface profile instrument of contact pin type, but is not to detect at screw thread specially, and the calculating of screw thread comprehensive parameters and error compensation is imperfection also.Harbin Institute of Technology, Changchun University of Science and Technology, Zhejiang University etc. utilize laser scanner technique to obtain the thread contour data and measure, but on the data of present announcement, its measuring accuracy is still not as good as contact scanning.University Of Tianjin, Nanjing Aero-Space University etc. detect the thread parameter based on image vision and did research, and its method has significantly raising to detection efficiency, but the also too late contact pin type scanning of image vision measuring accuracy, and the very difficult accurate measurement that solves internal thread.

In the contact pilotage scan-type screw measurement, in order to overcome the measuring error that probe wearing and tearing and some other environmental change cause, adopt the relative measurement method usually.Namely earlier carry out calibration measurements with standard thread, again whorl work piece to be measured is measured.And this just requires instrument that high bearing accuracy is arranged, otherwise when standard thread and thread size to be measured differ big, the error of measurement result will be exaggerated.

Utilize the homogeneous transformation method of volume coordinate, can judge, the positioning error that centering footpath parameter measurements has the greatest impact mainly contains deviation and the upper and lower needle point line of probe and the rotating deviation (under the coordinate system shown in Figure 1) of X-axis of Y direction.Yet the existence of actual mismachining tolerance and rigging error makes these two deviations be difficult to control in allowed band.The axial deviation of Y-axis can calculate by compensation, and the rotating deviation of X-axis then needs to control by effective detection method.Do not find as yet at present that related data shows and to carry out Automatic Detection and Control to this deviation.

Summary of the invention

The object of the present invention is to provide a kind of visible detection method that can realize the single needle scan-type screw measurement instrument probes probes X-axis rotating deviation of robotization detection control.

Step 1, initialization are demarcated:

Step 1-1, collection the 1st frame comprise the RGB image in probe motion zone, according to following formula the RGB image are converted into gray level image;

Gray=(R*19595+G*38469+B*7472)>>16

Wherein, R, G, B represent three color components of red, green, blue of the correspondence in the RGB image respectively, and Gray represents gray-scale value;

Step 1-2, utilize the SUSAN Corner Detection Algorithm, gray level image is carried out Corner Detection, obtain comprising all angle point informations of upper and lower probe cusp;

Two angle points of upper and lower probe cusp correspondence are chosen in step 1-3, manually-operated;

Step 2, real-time detector probe cusp:

Step 2-1, collection k+1 frame comprise the RGB image in probe motion zone, are translated into gray level image, wherein k 〉=1;

Step 2-2, when k=1, serve as prediction cusp, i.e. P ' with the position of last two field picture middle probe cusp _K+1=P _kWhen k 〉=2, according to the position of the gray level image middle probe cusp of k-1 frame and k frame, by the probe cusp position in the following predictor formula prediction k+1 two field picture;

P′ _k+1＝P _k+(P _k-P _k-1)＝2P _k-P _k-1；

Wherein, P is the coordinate vector of probe cusp, and P ' is the prediction coordinate vector of probe cusp, and k is the frame number of image;

Step 2-3, based on probe pinpoint point prediction coordinate vector P ', make up local region of interest at k+1 frame gray level image;

Step 2-4, by local region of interest is carried out the SUSAN Corner Detection, filter out upper and lower probe cusp position;

Step 2-5, the upper and lower probe cusp that filters out is carried out apart from verification;

Apart from verification condition 1 be

$|d(P_{up},P_{\mathrm{down}})-d_0(P_{up},P_{\mathrm{down}})|<{\mathrm{\&epsiv;}}_1$

Wherein, d (P _Up, P _Down) distance between the upper and lower probe cusp in the expression k+1 two field picture; d ₀(P _Up, P _Down) distance between the upper and lower probe cusp in the expression k two field picture; ε ₁Be error threshold;

Step 2-6, determining step 2-5 check results, as satisfy condition 1, then change step 2-10 over to, otherwise change step 2-7;

Step 2-7, detect image edge information in the local region of interest with the Canny operator, utilize the Hough conversion to extract outline of straight line, further obtain the element of cone of cusp and following cusp correspondence according to following rule respectively:

A) in the straight line that extracts, the X-axis corner dimension is between 60 °～80 ° in two probe element of cones and the image;

B) angle between two element of cones satisfies

θ ₁, θ ₂Be respectively two

Bar probe element of cone and image X-axis angle;

C) element of cone intersection point Q is the peak of all successful matching straight-line intersections, and in image;

Step 2-8, according to the element of cone of cusp on the gained and following cusp correspondence, calculate upper and lower profile intersection point, the row distance verification of going forward side by side;

Apart from verification condition 2 be

$|d\left(Q_{\mathrm{up}}{,Q}_{\mathrm{down}}\right)-d_0(Q_{\mathrm{up}},Q_{\mathrm{down}})|<{\mathrm{\&epsiv;}}_2$

Wherein, d (Q _Up, Q _Down) distance between the upper and lower probe cusp of expression in the current frame image; d ₀(Q _Up, Q _Down) distance between the upper and lower probe cusp in the expression former frame image; ε ₂Be error threshold;

Step 2-9, determining step 2-8 check results, as satisfy condition 2, then according to desirable probe cusp Q in the previous frame image ₀With P ₀Distance and current frame image in desirable probe cusp Q, be calculated as follows actual probes cusp P in the present frame.

P=Q+P ₀-Q ₀, change step step 2-10 over to;

If condition 2 does not also satisfy, think that then current frame image detects failure, measurement data points does not deposit in the database, changes step 2-1;

Step 2-10 records position and the line slope k thereof of upper and lower probe cusp;

Step 3, calculating rotating deviation angle

The probe motion straight line is obtained in step 3-1, the upper and lower cusp of difference match probe position, and detailed process is as follows;

Probe once in the video sequence of upper and lower capturing movement, if detected effective probe pinpoint is counted and is N, utilizes equation of linear regression

$k_M=\frac{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}(x_i-\stackrel{\&OverBar;}{x})(y_i-\stackrel{\&OverBar;}{y})}{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{(x_i-\stackrel{\&OverBar;}{x})}^2}=\frac{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{x}_i{y}_i-N\stackrel{\&OverBar;}{\mathrm{xy}}}{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{x}_i^2-N{\stackrel{\&OverBar;}{x}}^2}$

Obtain the probe motion straight line; Wherein, k _MSlope for line of motion; x _i, y _iBe respectively horizontal stroke, the ordinate of i two field picture middle probe cusp; $\stackrel{\&OverBar;}{x}=\frac{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{x}_i}{N},\stackrel{\&OverBar;}{y}=\frac{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{y}_i}{N};$

Step 3-2, according to the upper and lower probe cusp position in every two field picture, obtain each two field picture middle probe place rectilinear direction; Namely

$k_i=\frac{y_{\mathrm{up}}-y_{\mathrm{down}}}{x_{\mathrm{up}}-x_{\mathrm{down}}},(i=\mathrm{1,2}...,N)$

Wherein, k _iBe in the i two field picture, the line slope of the upper and lower cusp of probe; (x _Up, y _Up) be last probe pinpoint point coordinate, (x _Down, y _Down) be following probe pinpoint point coordinate;

The average of the upper and lower cusp line of probe slope k is

$k_L=\frac{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{k}_i}{N}$

Utilize the upper and lower cusp line of probe average gradient k _LObtain the fitting a straight line of a upper and lower cusp;

The angle theta of step 3-3, calculating probe motion straight line and upper and lower cusp fitting a straight line

$\mathrm{\&theta;}=a\tan \left|\frac{k_L-{\mathrm{<figure-callout\; id="1"\; label="k\; M"\; filenames="130528163938.png"\; state="\{\{state\}\}">k</figure-callout>}}_M}{1+k_L{k}_M}\right|$

θ is the rotating deviation angle.

The present invention can measure, regulate the X-axis rotation deviations size of probe by image processing means, realizes automatic measurement, under the prerequisite that guarantees installation accuracy, raises the efficiency.

Description of drawings

Fig. 1 is the definition of workpiece coordinate system;

Fig. 2 is camera and probe relative position relation;

Fig. 3 is the Corner Detection result of entire image;

Fig. 4 is local region of interest Corner Detection result;

Fig. 5 is three kinds of Corner Detection Algorithm Corner Detection results near the area-of-interest the probe cusp when the shade of having powerful connections disturbs.Wherein, (a) be the result of Harris Corner Detection Algorithm; (b) be the result of SUSAN Corner Detection Algorithm; (c) be the result of FAST Corner Detection Algorithm;

Fig. 6 is that the position of desirable probe cusp and actual probes cusp concerns synoptic diagram.Wherein Q represents desirable probe cusp, and P is the actual probes cusp;

Fig. 7 is that three kinds of edge detection algorithms are to the edge detection results of probe topography.Wherein, (a) being the testing result of Sobel operator, (b) is the testing result of Prewitt operator, (c) is the testing result of Canny operator;

Fig. 8 is the synoptic diagram of non-maximum value inhibition method;

Fig. 9 is Hough straight-line detection design sketch;

Figure 10 is probe element of cone testing result;

Figure 11 is the synoptic diagram of probe cusp when overflowing area-of-interest;

Figure 12 is the video sequence sectional drawing that needle point point is followed the tracks of;

Number in the figure title: 1, the illusion cylinder of workpiece, 2, probe, 3, articulate panel, 4 cameras

Embodiment

In order to guarantee the accuracy rate of probe cusp identification, the present invention adopts Corner Detection, profile identification and three aspects of motion prediction comprehensively to identify, improve the accuracy of probe cusp to a great extent, effectively suppressed the probe cusp flase drop that background interference causes.By the fitting a straight line to the cusp position, calculate the angle of line of motion (measuring instrument Z axle) and upper and lower cusp line then.

Shown in Figure 2, camera is placed in the X-axis positive dirction of probe, parallel in Z-direction with probe.By the upper and lower motion of camera collection probe, line and motion by the upper and lower cusp of the automatic capturing probe of image recognition calculate the rotating deviation angle.Regulate the probe rotation by rearmounted device again, the control rotating deviation.Because image is 2-D data, so in the back in the derivation of equation that image is handled, replace the z axle to explain with the y axle.The software program step is as follows:

Step 1, initialization are demarcated:

Step 1-1, collection one width of cloth comprise the RGB image in probe motion zone, according to following formula the RGB image are converted into gray level image;

Gray=(R*19595+G*38469+B*7472)>>16

Wherein, R, G, B represent three color components of red, green, blue of the correspondence in the RGB image respectively, and Gray represents gray-scale value;

Step 1-2, utilize the SUSAN Corner Detection Algorithm, gray level image is carried out Corner Detection, obtain comprising all angle point informations of upper and lower probe cusp;

Angle point does not have the explicit mathematical definition, but people generally believe that angle point is the point that two dimensional image brightness changes curvature maximum value on violent point or the image border curve.The detection of probe cusp can be used the method for image Corner Detection, and the detection method of image angle point mainly contains three major types at present: 1. based on the angular-point detection method of boundary curve.These class methods mainly are the curvature local maximum points of finding out on the curve of image border.The performance of edge detection algorithm will directly influence the quality that detects angle point, and the situation that the edge line interruption takes place during as rim detection will cause detecting false angle point.2. based on the angular-point detection method of template.At first set up a series of angle point templates with different angles, the similarity degree between testing image and the standard form relatively in certain window then comes angle point in the detected image with this.Because the complicacy of angle point structure can not design the template that covers all directions and angle point, the big and more complicated of this class angular-point detection method calculated amount.3. based on the angular-point detection method of gradation of image, mainly be to carry out Corner Detection by the differential character of calculating pixel, as Harris algorithm, SUSAN algorithm, FAST Corner Detection Algorithm etc.

Two angle points of upper and lower probe cusp correspondence are chosen in step 1-3, manually-operated;

Step 2, real-time detector probe cusp:

Step 2-1, collection k+1 frame comprise the RGB image in probe motion zone, are translated into gray level image, wherein k 〉=1;

When step 2-2, structure local region of interest, only centered by the position of an above two field picture middle probe cusp, make up a rectangular area as area-of-interest.In the topography that makes up, owing to the time that the different images detection is required is inconsistent, can make that the position of probe can be fluctuated, when probe motion speed is very fast, might make the probe cusp overflow area-of-interest, as shown in figure 11.

Yet, if increase the setting of local interested, can make calculated amount increase again, be unfavorable for the real-time processing of image.Therefore, need and to the position of next frame image middle probe cusp, to do a simple prediction according to the rectilinear motion of probe and detected probe pinpoint dot information, guarantee that substantially the cusp of probe is in the center of area-of-interest always.

Though the movement velocity of probe changes,, can carry out uniform motion and handle in the time at very short consecutive image.Then the cusp position of next frame can be calculated by following formula

When k=1, serve as prediction cusp, i.e. P ' with the position of last two field picture middle probe cusp _K+1=P _kWhen k 〉=2, according to the position of the gray level image middle probe cusp of k-1 frame and k frame, by the probe cusp position in the following predictor formula prediction k+1 two field picture

P′ _k+1＝P _k+(P _k-P _k-1)＝2P _k-P _k-1；

Wherein, P is the coordinate vector of probe cusp, and P ' is the prediction coordinate vector of probe cusp, and k is the frame number of image;

Step 2-3, based on probe pinpoint point prediction coordinate vector P ', make up local region of interest at k+1 frame gray level image;

Step 2-4, by local region of interest is carried out the SUSAN Corner Detection, filter out upper and lower probe cusp position;

Entire image is carried out Corner Detection, as shown in Figure 3, because background is not solid color, can produce a lot of detection angle points, plain and screening has brought difficulty to searching of probe cusp P for this.Consider that image middle probe cusp only has two, make up local region of interest, as shown in Figure 4.

Fig. 4 specifies correct probe cusp P by artificial, and carries out Corner Detection in it faces the territory, can effectively reduce the interference of background color.But in the process that moves up and down of probe, because reflective characteristic and the background of material have significant change, make conventional Corner Detection Algorithm can not find the cusp P of probe often.

Three images of Fig. 5 are respectively Harris, SUSAN and three kinds of Corner Detection Algorithm of FAST to certain result of probe image constantly.

As can be seen, three kinds of algorithms all do not detect real probe cusp P.Analysis image finds that there is color change in background object, just with near the probe cusp P overlaps, and therefore, for the detection of probe cusp, just can't obtain correct result based on the angular-point detection method of gradation of image.For this reason, we consider profile information again.

Step 2-5, the upper and lower probe cusp that filters out is carried out apart from verification;

Apart from verification condition 1 be

$|d(P_{\mathrm{up}},P_{\mathrm{down}})-d_0(P_{\mathrm{up}},P_{\mathrm{down}})|<{\mathrm{\&epsiv;}}_1$

Wherein, d (P _Up, P _Down) distance between the upper and lower probe cusp in the expression k+1 two field picture; d ₀(P _Up, P _Down) distance between the upper and lower probe cusp in the expression k two field picture; ε ₁Be error threshold;

Step 2-6, determining step 2-5 check results, as satisfy condition 1, then change step 2-10 over to, otherwise change step 2-7;

Step 2-7, detect image edge information in the local region of interest with the Canny operator, utilize the Hough conversion to extract outline of straight line, further obtain the element of cone of cusp and following cusp correspondence according to following rule respectively:

A) in the straight line that extracts, the X-axis corner dimension is between 60 °～80 ° in two probe element of cones and the image;

B) angle between two element of cones satisfies θ ₁, θ ₂Be respectively two

Bar probe element of cone and image X-axis angle;

C) element of cone intersection point Q is the peak of all successful matching straight-line intersections, and in image;

The ideal model of end of probe is a conical structure (actual head cusp is that a radius is the spherical of 50um), and the shooting direction of visual pattern and its center line parallel.If can detect two element of cones, calculate its joining, can think desirable probe cusp Q.In the probe motion process, the distance between desirable cusp Q and the actual cusp P remains unchanged.Fig. 6 is the difference between P, the Q point as can be seen.

In order to extract the probe outline line, at first need image is carried out rim detection.The method of gray-scale Image Edge Detection mainly is divided into two big classes: the single order differential map is as edge detection operator and second-order differential Image Edge-Detection operator.Wherein single order differential edge detection operator comprises: Roberts operator, Sobel operator, Krisch operator, Prewitt operator etc., and the second-order differential edge detection operator mainly contains: Laplacian operator, LOG operator; In addition also have detection methods such as Canny, SUSAN, statistics differentiation.

Fig. 7 is that several method is to the result of probe topography rim detection.

On the effect, the effect of three operators is all good, but many pixel wide can appear in Sobel operator and the detected edge of Prewitt operator, the profile with the back is detected ask the element of cone intersection point can produce certain error.And the edge of Canny operator is more clear, accurately the position probe profile.The basic thought of Canny operator is: at first select certain Gauss wave filter to carry out smothing filtering to image, adopt non-extreme value inhibition technology to handle then and obtain last edge image.Its step is;

A) with Gauss filter smoothing image.

Here with a Gaussian function H(x who omits coefficient, y):

$H(x,y)=\exp (-\frac{x^2+{\mathrm{<figure-callout\; id="2"\; label="y"\; filenames="130528163900.png,130528163938.png"\; state="\{\{state\}\}">y</figure-callout>}}^2}{{2\mathrm{\&sigma;}}^2})$

G(x,y)＝f(x,y)*H(x,y)

Wherein (x y) is view data to f.

B) with the finite difference of single order local derviation assign to amplitude and the direction of compute gradient.

Utilize first order difference convolution masterplate:

${\mathrm{<figure-callout\; id="1"\; label="H"\; filenames="130528163938.png"\; state="\{\{state\}\}">H</figure-callout>}}_1=\left|\begin{array}{cc}-1& -1\\ 1& 1\end{array}\right|$

${\mathrm{<figure-callout\; id="2"\; label="H"\; filenames="130528163900.png,130528163938.png"\; state="\{\{state\}\}">H</figure-callout>}}_2=|\begin{array}{cc}1& -1\\ 1& -1\end{array}|$

Can obtain:

Amplitude:

Direction:

C) gradient magnitude being carried out non-maximum value suppresses.

The gradient that only obtains the overall situation is not sufficient to determine the edge.For determining the edge, must keep the point of partial gradient maximum, and suppress non-maximum value, be about to non local maximum point zero setting to obtain the edge of refinement.

As shown in Figure 8, the label of 4 sectors is 0 to 3, and 4 kinds of corresponding 3 * 3 neighborhoods may be made up.To a point, the center pixel M of neighborhood compares with two pixels along gradient line.If the Grad of M is big unlike two neighbor Grad along gradient line, then make M=0.

D) usefulness dual threshold algorithm detects and is connected the edge.

Use two threshold value T ₁And T ₂(T ₁＜T ₂), thereby can obtain two threshold value edge image N ₁[i, j] and N ₂[i, j].Because N ₂[i, j] uses high threshold to obtain, thereby contains false edge seldom, but interruption (not closed) is arranged.The dual threshold method will be at N ₂In [i, j] edge is connected into profile, when arriving the end points of profile, this algorithm is just at N ₁The edge that can be connected on the profile is sought in 8 adjoint point positions of [i, j], and like this, algorithm is constantly at N ₁Collect edge in [i, j], up to N ₂Till [i, j] couples together.T ₂Be used for finding every line segment, T ₁Be used for extending fracture place of seeking the edge at the both direction of these line segments, and connect these edges.

After utilizing the Canny operator to obtain edge of image, need to extract the probe profile that we are concerned about, i.e. two element of cones.The method that general outline of straight line extracts is transformed to the master with Hough.The thought of straight line Hough conversion employing " ballot " detects straight line or the line segment in the digital picture, and it is the classic algorithm of an image processing and straight-line detection.Any straight line in the plane can be determined with ρ and two parameters of θ.Wherein ρ has determined the distance of straight line to initial point, and θ has determined the orientation of straight line.Its funtcional relationship is

ρ＝xcosθ+ysinθ x∈[0,π]

Every bit (x in the image space _i, y _i) be mapped to one group of totalizer C in the Hough space (ρ, θ), i.e. so-called voting process, (ρ θ) meets the pixel count of formula (2) to C in the presentation video space.Behind the poll closing, and C (ρ, each local maximum θ) is with regard to a corresponding straight-line segment, and namely Dui Ying ρ and θ can determine this straight line uniquely.Fig. 9 is the result after the Hough conversion.

Analyze this topography's characteristics: a) in the straight line that extracts, X-axis angle basically identical in two probe element of cones and the image, and size is between 60 °～80 °, and this just means θ ₁, θ ₂∈ [π/3,4 π/9]; B) angle between two element of cones is according to dispatching from the factory report as can be known about 45 °, and namely its slope is approximately 1.Set testing conditions according to the angle formula:

$0.9<\left|\frac{\tan {\mathrm{\&theta;}}_1-\tan {\mathrm{\&theta;}}_2}{1+\tan {\mathrm{\&theta;}}_1\tan {\mathrm{\&theta;}}_2}\right|<1.5$

Can filter and obtain two element of cones of asking.Element of cone to extracting extends again, finds desirable probe cusp Q, as shown in figure 10.Certainly also needing to add a verification condition c) element of cone intersection point Q is the peak of all successful matching straight-line intersections, and in image.

So far, just detected desirable probe cusp Q.Ideally, the upper and lower needle point line of probe both can be used L (P _Up, P _Down) expression, also can use L (Q _Up, Q _Down) expression, two straight lines overlap.But because actual mismachining tolerance, two straight lines can not overlap fully.And in measuring process, contact point is actual cusp, and this reliability that just makes P order will be higher than the Q point.Therefore, when cusp detects, also take L (P _Up, P _Down) be main, L (Q _Up, Q _Down) be the strategy of assisting.

Step 2-8, according to the element of cone of cusp on the gained and following cusp correspondence, calculate upper and lower profile intersection point, the row distance verification of going forward side by side;

Apart from verification condition 2 be

$|d(Q_{up},Q_{\mathrm{down}})-d_0(Q_{up},Q_{\mathrm{down}})|<{\mathrm{\&epsiv;}}_2$

Wherein, d (Q _Up, Q _Down) distance between the upper and lower probe cusp of expression in the current frame image; d ₀(Q _Up, Q _dOwn) distance between the upper and lower probe cusp in the expression former frame image; ε ₂Be error threshold;

Step 2-9, determining step 2-8 check results, as satisfy condition 2, then according to desirable probe cusp Q in the previous frame image ₀With P ₀Distance and current frame image in desirable probe cusp Q, be calculated as follows actual probes cusp P in the present frame.

P=Q+P ₀-Q ₀, change step step 2-10 over to;

Change step step 2-10 over to;

If condition 2 does not also satisfy, think that then current frame image detects failure, measurement data points does not deposit in the database, changes step 2-1;

Step 2-10 records position and the line slope k thereof of upper and lower probe cusp;

Step 3, calculating rotating deviation angle

The probe motion straight line is obtained in step 3-1, the upper and lower cusp of difference match probe position, and detailed process is as follows;

Probe once in the video sequence of upper and lower capturing movement, if detected effective probe pinpoint is counted and is N, utilizes equation of linear regression

$k_M=\frac{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}(x_i-\stackrel{\&OverBar;}{x})(y_i-\stackrel{\&OverBar;}{y})}{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{(x_i-\stackrel{\&OverBar;}{x})}^2}=\frac{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{x}_i{y}_i-N\stackrel{\&OverBar;}{\mathrm{xy}}}{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{x}_i^2-N{\stackrel{\&OverBar;}{x}}^2}$

Obtain the probe motion straight line; Wherein, k _MSlope for line of motion; x _i, y _iBe respectively horizontal stroke, the ordinate of i two field picture middle probe cusp; $\stackrel{\&OverBar;}{x}=\frac{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{x}_i}{N},$ $\stackrel{\&OverBar;}{y}=\frac{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{y}_i}{N};$

Step 3-2, according to the upper and lower probe cusp position in every two field picture, obtain each two field picture middle probe place rectilinear direction; Namely

$k_i=\frac{y_{\mathrm{up}}-y_{\mathrm{down}}}{x_{\mathrm{up}}-x_{\mathrm{down}}},(i=\mathrm{1,2}...,N)$

Wherein, k _iBe in the i two field picture, the line slope of the upper and lower cusp of probe; (x _Up, y _Up) be last probe pinpoint point coordinate, (x _Down, y _Down) be following probe pinpoint point coordinate;

The average of the upper and lower cusp line of probe slope k is

$k_L=\frac{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{k}_i}{N}$

Utilize the upper and lower cusp line of probe average gradient k _LObtain the fitting a straight line of a upper and lower cusp;

The angle theta of step 3-3, calculating probe motion straight line and upper and lower cusp fitting a straight line

$\mathrm{\&theta;}=a\tan \left|\frac{k_L-{\mathrm{<figure-callout\; id="1"\; label="k\; M"\; filenames="130528163938.png"\; state="\{\{state\}\}">k</figure-callout>}}_M}{1+k_L{k}_M}\right|$

θ is the rotating deviation angle.

Claims (1)

1. the visible detection method to the X-axis rotating deviation of single needle scan-type screw measurement instrument measuring probe is characterized in that, comprises following process:

Step 1, initialization are demarcated:

Step 1-1, gather the RGB image that the 1st frame comprises the probe motion zone, according to following formula with RGB

Image is converted into gray level image;

Gray=(R*19595+G*38469+B*7472)>>16

Wherein, R, G, B represent three color components of red, green, blue of the correspondence in the RGB image respectively,

Gray represents gray-scale value;

Step 1-2, utilize the SUSAN Corner Detection Algorithm, gray level image is carried out Corner Detection, obtain bag

Draw together upper and lower probe cusp at all interior angle point informations;

Two angle points of upper and lower probe cusp correspondence are chosen in step 1-3, manually-operated;

Step 2, real-time detector probe cusp:

Step 2-1, collection k+1 frame comprise the RGB image in probe motion zone, are translated into gray level image, wherein k 〉=1;

Step 2-2, when k=1, serve as prediction cusp, i.e. P ' with the position of last two field picture middle probe cusp _K+1=P _kWhen k 〉=2, according to the position of the gray level image middle probe cusp of k-1 frame and k frame, by the probe cusp position in the following predictor formula prediction k+1 two field picture;

P′ _k+1＝P _k+(P _k-P _k-1)＝2P _k-P _k-1；

Wherein, P is the coordinate vector of probe cusp, and P ' is the prediction coordinate vector of probe cusp, and k is the frame number of image;

Step 2-3, based on probe pinpoint point prediction coordinate vector P ', make up local region of interest at k+1 frame gray level image;

Step 2-4, by local region of interest is carried out the SUSAN Corner Detection, filter out upper and lower probe cusp position;

Step 2-5, the upper and lower probe cusp that filters out is carried out apart from verification;

Apart from verification condition 1 be

$|d(P_{up},P_{\mathrm{down}})-d_0(P_{up},P_{\mathrm{down}})|<{\mathrm{\&epsiv;}}_1$

Wherein, d (P _Up, P _Down) distance between the upper and lower probe cusp in the expression K+1 two field picture; d ₀(P _Up, P _Down) expression kDistance between the upper and lower probe cusp in the two field picture; ε ₁Be error threshold;

Step 2-6, determining step 2-5 check results, as satisfy condition 1, then change step 2-10 over to, otherwise change step 2-7;

Step 2-7, detect image edge information in the local region of interest with the Canny operator, utilize the Hough conversion to extract outline of straight line, further obtain the element of cone of cusp and following cusp correspondence according to following rule respectively:

A) in the straight line that extracts, the X-axis corner dimension is between 60 °～80 ° in two probe element of cones and the image;

B) angle between two element of cones satisfies

θ ₁, θ ₂Be respectively two

Bar probe element of cone and image X-axis angle;

C) element of cone intersection point Q is the peak of all successful matching straight-line intersections, and in image;

Step 2-8, according to the element of cone of cusp on the gained and following cusp correspondence, calculate upper and lower profile intersection point, the row distance verification of going forward side by side;

Apart from verification condition 2 be

$|d(Q_{\mathrm{up}},Q_{\mathrm{down}})-d_0(Q_{\mathrm{up}},Q_{\mathrm{down}})|<{\mathrm{\&epsiv;}}_2$

Wherein, d (Q _Up, Q _Down) distance between the upper and lower probe cusp of expression in the current frame image; d ₀(Q _Up, Q _Down) distance between the upper and lower probe cusp in the expression former frame image; ε ₂Be error threshold;

Step 2-9, determining step 2-8 check results, as satisfy condition 2, then according to desirable probe cusp Q in the previous frame image ₀With P ₀Distance and current frame image in desirable probe cusp Q, be calculated as follows actual probes cusp P in the present frame;

P=Q+P ₀-Q ₀, change step step 2-10 over to;

If condition 2 does not also satisfy, think that then current frame image detects failure, measurement data points does not deposit in the database, changes step 2-1;

Step 2-10 records position and the line slope k thereof of upper and lower probe cusp;

Step 3, calculating rotating deviation angle

The probe motion straight line is obtained in step 3-1, the upper and lower cusp of difference match probe position, and detailed process is as follows;

Probe once in the video sequence of upper and lower capturing movement, if detected effective probe pinpoint is counted and is N, utilizes equation of linear regression

$k_M=\frac{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}(x_i-\stackrel{\&OverBar;}{x})(y_i-\stackrel{\&OverBar;}{y})}{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{(x_i-\stackrel{\&OverBar;}{x})}^2}=\frac{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{x}_i{y}_i-N\stackrel{\&OverBar;}{\mathrm{xy}}}{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{x}_i^2-N{\stackrel{\&OverBar;}{x}}^2}$

Obtain the probe motion straight line; Wherein, k _MSlope for line of motion; x _i, y _iBe respectively horizontal stroke, the ordinate of i two field picture middle probe cusp; $\stackrel{\&OverBar;}{x}=\frac{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{x}_i}{N},$ $\stackrel{\&OverBar;}{y}=\frac{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{y}_i}{N};$

Step 3-2, according to the upper and lower probe cusp position in every two field picture, obtain each two field picture middle probe place rectilinear direction; Namely

$k_i=\frac{y_{\mathrm{up}}-y_{\mathrm{down}}}{x_{\mathrm{up}}-x_{\mathrm{down}}},(i=\mathrm{1,2}...,N)$

Wherein, k _iBe in the i two field picture, the line slope of the upper and lower cusp of probe; (x _Up, y _Up) be last probe pinpoint point coordinate, (x _Down, y _Down) be following probe pinpoint point coordinate;

The average of the upper and lower cusp line of probe slope k is

$k_L\text{=}\frac{\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{k}_i}{N}$

Utilize the upper and lower cusp line of probe average gradient k _LObtain the fitting a straight line of a upper and lower cusp;

The angle theta of step 3-3, calculating probe motion straight line and upper and lower cusp fitting a straight line

$\mathrm{\&theta;}=a\tan \left|\frac{k_L-k_M}{1+k_L{k}_M}\right|$

θ is the rotating deviation angle.

Images