CN105184792A - Circular saw web wear extent online measuring method - Google Patents

Abstract

The invention discloses a circular saw web wear extent online measuring method. The method is realized through the following steps: an industrial control computer triggering an image acquisition card, and obtaining circular saw web images from an industrial camera; preprocessing the images; based on a self-adaptive threshold, finding out candidate angle points of circular saw webs by use of an eight-neighborhood gray scale similarity screening method; determining whether the candidate angle points are unique angle points in a neighborhood or angle point response function values are the maximum so as to rejecting pseudo angle points; determining whether the angle points are maximum points, and otherwise rejecting the angle points so as to determining whole pixel coordinates of circular saw web tip points; carrying out sub-pixel positioning on the tip points by use of a cubic surface fitting method; and solving a radius value of a circle where the tip points are disposed by use of a least square method, and obtaining an actual wear extent of the circular saw webs through a camera calibration relation. According to the invention, the circular saw web wear extent online measuring method is good in real-time performance, high in detection precision and high in effectiveness. Besides, a detection result provides a basis for a compensation mechanism to execute compensation.

Description

A kind of saw blade wear extent On-line Measuring Method

Technical field

The invention belongs to technical field of image processing, relate to a kind of cutting-tool wear state visible detection method, more particularly, relate to a kind of saw blade wear extent On-line Measuring Method.The method is by identifying image border angle point, solves the method for saw blade wear extent according to the position of angle point again.

Background technology

Machine vision is as a kind of emerging detection technique, and feature that is quick with it, real-time, intelligent and low cost obtains widespread use.Measurement based on machine vision belongs to non-contact measurement mode, not only can measure workpiece features in real time, improve the efficiency measured, and can according to the size of workpiece, the parameters such as the focal length of adjustment industrial camera, realize wider dimensional measurement, the measuring error because the change of survey crew self psychology produces can be avoided simultaneously.

The detection of tool abrasion, mainly contains following several method at present.1. monitoring vibration signal and motor current signal, builds the relation of vibration signal and motor current signal and tool abrasion, thus detects the state of wear of cutter; 2. monitor acoustic emission signal in process, set up the relation of acoustic emission signal and tool abrasion, detect the state of wear of cutter; 3. along with the develop rapidly of ccd sensor and application technology thereof, the non-contact detection technology based on machine vision is widely used in the fields such as size, displacement, surface shape detection.The method of applied for machines vision-based detection tool wear has three kinds: 1. detect tool surface image; 2. workpiece surface texture image is detected; 3. chip image is detected.

The monitoring of saw blade wear extent is conventional above method also.The people such as Electronic University Of Science & Technology Of Hangzhou Zhao Ling, construct saw blade geometric parameter measurement system based on machine vision.The method is based on saw blade contour optimization, circular hole in saw blade is proposed to the quadratic polynomial interpolation sub-pixel positioning method of improvement, the least square method improved is adopted to carry out matching to crown two sections of straight lines, improve accuracy of detection, but the method must obtain saw blade entire image, the on-line measurement of saw blade wear extent thus cannot be realized.The people such as Sweden Ekevad construct in sawing beech process, relation between saw blade wear extent and its saw blade vibration signal, although the method achieves the on-line measurement of saw blade wear extent, but the exact magnitude relation of saw blade wear extent and its saw blade vibration signal, be difficult to find, qualitative detection can only be carried out to saw blade wear extent.

Because measured object varies, its architectural feature has very large difference, and the measurement based on machine vision does not also have a kind of general method, for different objects, needs to adopt diverse ways.At present based on its poor real of saw blade abrasion amount measuring method of classical Harris method, be usually point of a knife point some non-point of a knife point erroneous judgements, although improve to some extent in real-time based on the saw blade abrasion amount measuring method of the Harris method improved, but still be usually point of a knife point some non-point of a knife point erroneous judgements, cannot realize accurately locating different saw blade point of a knife point, the bigger error measured, cannot be applied in actual measurement process.

Summary of the invention

The object of the invention is to the deficiency overcoming existing saw blade abrasion amount measuring method, propose a kind of On-line Measuring Method of the saw blade wear extent based on machine vision, to improve the degree of accuracy of saw blade wear extent on-line measurement.

For achieving the above object, the present invention realizes the technical scheme that object adopts and is:

A kind of saw blade wear extent On-line Measuring Method, comprises the steps:

1) obtain saw blade image: by rack-mount industrial camera, by manual shift, make industrial camera face measured circle saw blade, industrial computer trigger image capture card, obtain saw blade image;

2) medium filtering is adopted to carry out noise reduction pre-service to image;

3) based on adaptive threshold, use 8 neighborhood gray scale similarity screening techniques, find out the candidate angular of saw blade;

The mobile grey scale change E (u, v) caused of windowsill translation vector (u, v) specifically centered by (x, y) is:

$E(u,v)=\underset{x,y}{\mathrm{\Σ}}\mathrm{\ω}\left(x,y\right){\[I\left(x+u,y+v\right)-I\left(x,y\right)\]}^2$

I (x+u, y+v) is the gray-scale value after translation, and I (x, y) is the gray-scale value before translation, and ω (x, y) is Gauss's window function,

Its differential form is

$E(u,v)=\underset{x,y}{\Σ}\mathrm{\ω}(x,y)(I_x^2{u}^2+2I_x{I}_{y}u\mathrm{\ν}+I_y^2{\mathrm{\ν}}^2)$

Wherein $I_x=\frac{\∂I}{\∂x},I_y=\frac{\∂I}{\∂y}$

Its matrix form is

$E(u,v)=\[\begin{array}{cc}u& v\end{array}\]M\left[\begin{array}{c}u\\ v\end{array}\right]$

Wherein, M is the autocorrelation function matrix of target pixel points (x, y)

$M=\underset{x,y}{\Σ}\mathrm{\ω}(x,y)\left[\begin{array}{cc}{I}_x^2& I_x{I}_y\\ I_x{I}_y& I_y^2\end{array}\right]$

Gauss's window function $\mathrm{\ω}(x,y)=e^{(x^2+y^2)/{\mathrm{\σ}}^2}$

The angle point response function value of target pixel points (x, y):

CRF(x,y)＝det(M)-k(trace(M)) ²

The wherein determinant of det (M) representing matrix M, the mark of trace (M) representing matrix, k gets 0.04 ~ 0.06;

Using the standard deviation of impact point (x, y) and each pixel image intensity value in its 8 contiguous range as the similar decision threshold t of 8 neighborhood gray scale, maximum angle point response function value CRF _maxone of percentage as angle point response detection threshold T,

Impact point (x, y) with the difference △ I of gray-scale value of each pixel in its 8 contiguous range, and add up the pixel number n of △ I in [-t, t] scope, meet 2≤n≤6 and its angle point response function value and be greater than T and the maximum impact point in local is candidate angular.

4) for candidate angular, judge whether it is that the interior unique angle point of neighborhood or its angle point response function value are maximum, weeds out pseudo-angle point;

Retaining those is candidate angle maximum for the CRF value in the candidate angle of unique angle point or its 5 × 5 neighborhood in its 5 × 5 neighborhood.

5) judge that whether angle point is that on curve, maximum point, to reject tooth root point, thus determines the Integer Pel coordinate of saw blade point of a knife point

The slope k of angle point and previous angle point line _i1

$k_{i1}=\frac{y_i-y_{i-1}}{x_i-x_{i-1}}$

The slope k of angle point and a rear angle point line _i2

$k_{i2}=\frac{y_i-y_{i+1}}{x_i-x_{i+1}}$

(x _i, y _i) be the pixel coordinate of i-th angle point, (x _i-1, y _i-1) be the previous corner pixels coordinate of i-th angle point, (x _i+1, y _i+1) be the rear corner pixels coordinate of i-th angle point,

If k _i1<0 and k _i2>0, judges this point not as point of a knife point, is rejected,

6) cubic surface fitting process tool setting cusp is utilized to carry out sub-pixel positioning

The binary cubic function form of the CRF of each point in Integer Pel point of a knife point (x, y) and certain neighborhood thereof:

$CRF(x_i,y_i)=a_{00}+a_{01}{y}_i+a_{02}{y}_i^2+a_{03}{y}_i^3+a_{10}{x}_i+a_{20}{x}_i^2+a_{30}{x}_i^3+a_{11}{x}_i{y}_i+a_{21}{x}_i^2{y}_i+a_{12}{x}_i{y}_i^2$

The error sum of squares of matching

$\mathrm{\&epsiv;}=\underset{i=1}{\overset{25}{\mathrm{\&Sigma;}}}{\left(CRF\right(x,y)-CRF(x_i,y_i\left)\right)}^2$

try to achieve a ₀₀, a ₀₁, a ₀₂, a ₀₃, a ₁₀, a ₂₀, a ₃₀, a ₁₁, a ₂₁, a ₁₂

Utilize the cubic surface expression formula determined to solve the CRF that Integer Pel point of a knife point is subdivided into 3 × 3 sub-pix points, get the coordinate of the subpixel coordinates in 9 sub-pix points corresponding to CRF maximal value as this point of a knife point.

7) use least square method to solve point of a knife point place radius of a circle value, drawn the actual wear amount of saw blade by camera calibration relation

Definition

$a=n\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i^2-\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i$

$b=n\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i{y}_i-\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_i$

$c=n\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i^3+n\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i{y}_i^2-\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}(x_i+y_i)\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i$

$d=n\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_i^2-\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_i\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_i$

$e=N\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i^2{y}_i+N\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_i^3-\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}(x_i^2+y_i^2)\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_i$

$A=\frac{be-cd}{ad-b^2}$

$B=\frac{ae-bc}{b^2-ad}$

$C=-\frac{1}{n}\left(\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\right(x_i^2+y_i^2)+A\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i+B\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_i)$

r ²＝(A ²+B ²-4C)/4

Again according to camera calibration relation, obtain the radius value of cutter reality, the difference of twice testing result is the wear extent of saw blade.

The present invention compared with prior art has the following advantages and beneficial effect:

1, by twice screening of described step 3 and step 4, eliminate pseudo-angle point, avoid now methodical angle point and to cluster phenomenon.

2, by described step 5, reject tooth root point, thus determine the Integer Pel coordinate of saw blade point of a knife point, avoid existing method and the erroneous judgement of tooth root point is decided to be the deviation even mistake that point of a knife point causes round matching.

Accompanying drawing explanation

Fig. 1 is that the measurement mechanism of the embodiment of the present invention forms schematic block diagram,

Fig. 2 is the industrial camera scheme of installation of the embodiment of the present invention,

Fig. 3 is the realization flow figure of the embodiment of the present invention,

Fig. 4 the present invention and existing method to the point of a knife point extraction comparison figure of the saw blade of smooth transition between adjacent saw-tooth,

Fig. 5 the present invention and existing method are to the point of a knife point extraction comparison figure of the saw blade of transition rapid between adjacent saw-tooth.

Embodiment

In order to make object of the present invention, technical scheme and advantage clearly understand, below in conjunction with drawings and Examples, the present invention is described in further detail.Should be appreciated that specific embodiment described herein only for explaining the present invention, being not intended to limit the present invention.

A kind of saw blade wear extent On-line Measuring Method of the present invention, before on-line measurement saw blade wear extent, industrial camera is demarcated, its method is, is installing the position of saw blade, the demarcation thing that installation dimension is known, industrial computer trigger image capture card obtains the image demarcating thing, according to the image obtained, calculate the pixel value of known dimensions, known dimensions obtains the physical size value of each pixel representative divided by pixel value.As shown in Figure 3, concrete steps are as follows:

Step 1, industrial camera is rack-mount, by manual shift, make industrial camera face measured circle saw blade;

As shown in Figure 1, whole measurement mechanism comprises industrial camera, image pick-up card, industrial computer and Survey Software.As shown in Figure 2, before measurement saw blade wear extent, by on industrial camera 3 mounting bracket 4, make industrial camera 3 face saw blade 2 on lathe 1 by manual shift, industrial camera netting twine 5 is connected with the image pick-up card (not shown in FIG.) in industrial computer 6.

Step 2, industrial computer trigger image capture card, obtain saw blade image by industrial camera;

Industrial computer is as master controller, image pick-up card is communicated with industrial computer by PCI-e bus, industrial camera faces measured circle saw blade, measured circle saw blade images on industrial camera after transmitted light source irradiates, the Digital Image Transmission that collects to industrial computer, thus is obtained the image of saw blade by image pick-up card.

Step 3, pre-service is carried out to image;

For restraint speckle impact, industrial camera is collected original image and carry out noise reduction process.Adopt medium filtering to carry out noise reduction process, make object and background evenly single separately, contrast is large, without other lines and the details being difficult to differentiation.

Step 4, based on adaptive threshold, use 8 neighborhood gray scale similarity screening techniques, find out the candidate angular of saw blade;

In order to improve the overall adaptability of the method, reducing because threshold value arranges the unreasonable and angle point false retrieval that causes and undetected, in saw blade point of a knife point detecting method, using adaptive threshold to choose way.Using the standard deviation of image intensity value as the similar decision threshold t of 8 neighborhood gray scale; By the maximum angle point response function value CRF of image _maxone of percentage as angle point response detection threshold T.

The mobile grey scale change E (u, v) caused of windowsill translation vector (u, v) centered by (x, y) is:

$E(u,v)=\underset{x,y}{\mathrm{\&Sigma;}}\mathrm{\&omega;}\left(x,y\right){\&lsqb;I\left(x+u,y+v\right)-I\left(x,y\right)\&rsqb;}^2$

I (x+u, y+v) is the gray-scale value after translation, and I (x, y) is the gray-scale value before translation, and ω (x, y) is Gauss's window function,

Its differential form is

$E(u,v)=\underset{x,y}{\&Sigma;}\mathrm{\&omega;}(x,y)(I_x^2{u}^2+2I_x{I}_{y}u\mathrm{\&nu;}+I_y^2{\mathrm{\&nu;}}^2)$

Wherein $I_x=\frac{\&part;I}{\&part;x},I_y=\frac{\&part;I}{\&part;y}$

Its matrix form is

$E(u,v)=\&lsqb;\begin{array}{cc}u& v\end{array}\&rsqb;M\left[\begin{array}{c}u\\ v\end{array}\right]$

Wherein, M is the autocorrelation function matrix of target pixel points (x, y)

$M=\underset{x,y}{\&Sigma;}\mathrm{\&omega;}(x,y)\left[\begin{array}{cc}{I}_x^2& I_x{I}_y\\ I_x{I}_y& I_y^2\end{array}\right]$

Gauss's window function $\mathrm{\&omega;}(x,y)=e^{(x^2+y^2)/{\mathrm{\&sigma;}}^2}$

The angle point response function value of target pixel points (x, y):

CRF(x,y)＝det(M)-k(trace(M)) ²

The wherein determinant of det (M) representing matrix M, the mark of trace (M) representing matrix, k gets 0.04 ~ 0.06.

Using the standard deviation of impact point (x, y) and each pixel image intensity value in its 8 contiguous range as the similar decision threshold t of 8 neighborhood gray scale, maximum angle point response function value CRF _maxone of percentage as angle point response detection threshold T,

Impact point (x, y) with the difference △ I of gray-scale value of each pixel in its 8 contiguous range, and add up the pixel number n of △ I in [-t, t] scope, meet 2≤n≤6 and its angle point response function value and be greater than T and the maximum impact point in local is candidate angular.

Step 5, for candidate angular, judge its be whether in neighborhood unique angle point or its angle point response function value maximum, weed out pseudo-angle point;

The judgement of true and false angle point is carried out to candidate angular.Using 5 × 5 neighborhoods of each impact point as area-of-interest, judge whether impact point is unique angle point, if so, then this impact point is considered as real angle point and preserves; If not, then again judge.Search out other angle points in its 5 × 5 neighborhood and pixel maximum for CRF in these angle points is preserved as real angle point.

Step 6, judge that whether angle point is that on curve, maximum point, to reject tooth root point, thus determines the Integer Pel coordinate of saw blade point of a knife point;

Extract all angle points on saw blade, but these angle points not only comprise the point of a knife point of saw blade, also may comprise the tooth root point between adjacent saw-tooth, and these non-point of a knife points can cause the deviation even mistake of round matching, therefore will again screen to reject non-point of a knife point.The edge contour of saw blade is regarded as a continuous print curve, then point of a knife point is the maximum point on this curve, and the tooth root point between adjacent saw-tooth is the minimum point of curve.

The slope k of angle point and previous angle point line _i1

$k_{i1}=\frac{y_i-y_{i-1}}{x_i-x_{i-1}}$

The slope k of angle point and a rear angle point line _i2

$k_{i2}=\frac{y_i-y_{i+1}}{x_i-x_{i+1}}$

(x _i, y _i) be the pixel coordinate of i-th angle point, (x _i-1, y _i-1) be the previous corner pixels coordinate of i-th angle point, (x _i+1, y _i+1) be the rear corner pixels coordinate of i-th angle point,

If k _i1<0 and k _i2>0, judges this point not as point of a knife point, is rejected,

Step 7, cubic surface fitting process tool setting cusp is utilized to carry out sub-pixel positioning;

In order to improve accuracy of detection under the prerequisite not changing hardware device, tool setting cusp carries out sub-pixel positioning.Sub-pix carries out k segmentation to physical picture element exactly, if original image is n, capable m arranges, and after k segmentation, becomes the capable km row of kn, this means that each pixel is divided into less unit, thus implement interpolation arithmetic to improve the method precision to these less unit.

The binary cubic function representation of the CRF of each point in Integer Pel point of a knife point (x, y) and certain neighborhood thereof:

$CRF(x_i,y_i)=a_{00}+a_{01}{y}_i+a_{02}{y}_i^2+a_{03}{y}_i^3+a_{10}{x}_i+a_{20}{x}_i^2+a_{30}{x}_i^3+a_{11}{x}_i{y}_i+a_{21}{x}_i^2{y}_i+a_{12}{x}_i{y}_i^2$

The error sum of squares of matching

$\mathrm{\&epsiv;}=\underset{i=1}{\overset{25}{\mathrm{\&Sigma;}}}{\left(CRF\right(x,y)-CRF(x_i,y_i\left)\right)}^2$

try to achieve a ₀₀, a ₀₁, a ₀₂, a ₀₃, a ₁₀, a ₂₀, a ₃₀, a ₁₁, a ₂₁, a ₁₂

Utilize the cubic surface expression formula determined to solve the CRF that Integer Pel point of a knife point is subdivided into 3 × 3 sub-pix points, get the coordinate of the subpixel coordinates in 9 sub-pix points corresponding to CRF maximal value as this point of a knife point.

Step 8, use least square method solve point of a knife point place radius of a circle value, are drawn the actual wear amount of saw blade by camera calibration relation

Definition

$a=n\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i^2-\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i$

$b=n\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i{y}_i-\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_i$

$c=n\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i^3+n\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i{y}_i^2-\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}(x_i+y_i)\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i$

$d=n\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_i^2-\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_i\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_i$

$e=N\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i^2{y}_i+N\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_i^3-\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}(x_i^2+y_i^2)\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_i$

$A=\frac{be-cd}{ad-b^2}$

$B=\frac{ae-bc}{b^2-ad}$

$C=-\frac{1}{n}\left(\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\right(x_i^2+y_i^2)+A\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i+B\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_i)$

According to

r ²＝(A ²+B ²-4C)/4

Obtain the radius value of cutter reality, the difference of twice testing result is the pixel value of the wear extent of saw blade.

Beneficial effect of the present invention can be further illustrated by following experiment

1, experiment content

Test with 2 kinds of saw blades, between the adjacent saw-tooth of Fig. 4 a saw blade, transition is mild, and between the adjacent saw-tooth of Fig. 5 a saw blade, transition is rapid, with the Harris method of classical Harris method, improvement and the point of a knife point of method of the present invention extraction saw blade.

2, experimental provision

It is E5300 that industrial computer adopts Taiwan to grind magnificent industrial computer IPC-610H, CPU, internal memory 2G; Image pick-up card adopts Taiwan Ling Hua company PCIe-GIE64+ image pick-up card; Industrial camera adopts 5,000,000 pixel acA2500-14gm industrial cameras of German BASLER company, and camera lens adopts the M5018-MP2 tight shot of Japanese computar company.

3, experimental result

In order to verify validity of the present invention, devise saw blade point of a knife point Detection results control experiment.The saw blade of two kinds of different profile nature is selected in experiment, adopt classical Harris method, the Harris method of improvement and the point of a knife point of method of the present invention extraction saw blade respectively, as shown in Table 1 and Table 2, Detection results figure as shown in Figure 4 and Figure 5 in its testing result list.

Between table 1 adjacent saw-tooth transition mild saw blade detected parameters

	Working time (s)	Point of a knife is counted out	The angle point number detected
				Classical Harris method	5.25	10	10
The Harris method improved	1.078	10	14
				The inventive method	1.571	10	10

The detected parameters of the saw blade that transition is rapid between table 2 adjacent saw-tooth

	Working time (s)	Point of a knife is counted out	The angle point number detected
				Classical Harris method	4.39	12	24
The Harris method improved	1.031	12	32
				The inventive method	1.304	12	12

The Harris method improved is the shortest for working time, less than 20% of classical Harris method; The inventive method working time is also less than 35% of classical Harris method.The inventive method point of a knife point extraction effect is best; Classical Harris method to transition between adjacent sawtooth mild saw blade point of a knife point extraction effect good, but to the rapid saw blade of transition between adjacent saw-tooth, the angle point that its extracts not only comprises point of a knife point, further comprises the tooth root point between adjacent saw-tooth; There is the serious phenomenon that clusters in the angle point that the Harris method improved is extracted, Detection results is the poorest.Visible, the inventive method is more effective in the application of saw blade abrasion detection.

Claims (8)

1. a saw blade wear extent On-line Measuring Method, is characterized in that, comprises the following steps:

(1) saw blade image is obtained;

(2) noise reduction pre-service is carried out to the image obtained;

(3) based on adaptive threshold, use 8 neighborhood gray scale similarity screening techniques, find out the candidate angular of saw blade;

(4) for candidate angular, judge whether it is that the interior unique angle point of neighborhood or its angle point response function value are maximum, weeds out pseudo-angle point; Retaining those is candidate angular maximum for the angle point response function value in the candidate angular of unique angle point or its 5 × 5 neighborhood in its 5 × 5 neighborhood;

(5) judge that whether angle point is that on curve, maximum point, to reject tooth root point, thus determines the Integer Pel coordinate of saw blade point of a knife point;

(6) cubic surface fitting process tool setting cusp is utilized to carry out sub-pixel positioning;

(7) use least square method to solve point of a knife point place radius of a circle value, demarcate by industrial camera the actual wear amount that relation draws saw blade.

2. saw blade wear extent On-line Measuring Method according to claim 1, it is characterized in that, the method that described step 1 obtains saw blade image is: by rack-mount industrial camera, pass through manual shift, industrial camera is made to face measured circle saw blade, industrial computer trigger image capture card, obtains saw blade image.

3. saw blade wear extent On-line Measuring Method according to claim 1, is characterized in that, described step 2 is carried out the pretreated method of noise reduction to the image obtained and is: adopt mean filter, medium filtering or Wavelet Denoising Method.

4. saw blade wear extent On-line Measuring Method according to claim 1, is characterized in that, the defining method of described step 3 candidate angular is:

The mobile grey scale change E (u, v) caused of windowsill translation vector (u, v) centered by (x, y) is:

$E(u,v)=\underset{x,y}{\mathrm{\&Sigma;}}\mathrm{\&omega;}\left(x,y\right){\&lsqb;I\left(x+u,y+v\right)-I\left(x,y\right)\&rsqb;}^2$

In formula: I (x+u, y+v) is the gray-scale value after translation, I (x, y) is the gray-scale value before translation, and ω (x, y) is Gauss's window function,

Its differential form is

$E(u,v)=\underset{x,y}{\&Sigma;}\mathrm{\&omega;}(x,y)(I_x^2{u}^2+2I_x{I}_{y}u\mathrm{\&nu;}+I_y^2{\mathrm{\&nu;}}^2)$

Wherein $I_x=\frac{\&part;I}{\&part;x},I_y=\frac{\&part;I}{\&part;y}$

Its matrix form is

$E(u,v)=\&lsqb;\begin{array}{cc}u& v\end{array}\&rsqb;M\left[\begin{array}{c}u\\ v\end{array}\right]$

Wherein, M is the autocorrelation function matrix of target pixel points (x, y)

$M=\underset{x,y}{\&Sigma;}\mathrm{\&omega;}(x,y)\left[\begin{array}{cc}{I}_x^2& I_x{I}_y\\ I_x{I}_y& I_y^2\end{array}\right]$

Gauss's window function $\mathrm{\&omega;}(x,y)=e^{(x^2+y^2)/{\mathrm{\&sigma;}}^2}$

Angle point response function value CRF (x, y) of target pixel points (x, y):

CRF(x,y)＝det(M)-k(trace(M)) ²

The wherein determinant of det (M) representing matrix M, the mark of trace (M) representing matrix, k gets 0.04 ~ 0.06,

Using the standard deviation of impact point (x, y) and each pixel image intensity value in its 8 contiguous range as the similar decision threshold t of 8 neighborhood gray scale, maximum angle point response function value CRF _maxone of percentage as angle point response detection threshold T,

Impact point (x, y) with the difference △ I of gray-scale value of each pixel in its 8 contiguous range, the statistics pixel number n of △ I in [-t, t] scope, meets 2≤n≤6 and its angle point response function value is greater than T and the maximum impact point in local is candidate angular.

5. saw blade wear extent On-line Measuring Method according to claim 1, is characterized in that, the method that described step 5 rejects saw blade tooth root point is as follows:

The slope k of angle point and previous angle point line _i1

$k_{i1}=\frac{y_i-y_{i-1}}{x_i-x_{i-1}}$

The slope k of angle point and a rear angle point line _i2

$k_{i2}=\frac{y_i-y_{i+1}}{x_i-x_{i+1}}$

(x _i, y _i) be the pixel coordinate of i-th angle point, (x _i-1, y _i-1) be the previous corner pixels coordinate of i-th angle point, (x _i+1, y _i+1) be the rear corner pixels coordinate of i-th angle point,

Reject those k _i1<0 and k _i2the angle point of >0.

6. saw blade wear extent On-line Measuring Method according to claim 1, is characterized in that, the method that described step 6 tool setting cusp carries out sub-pixel positioning is as follows:

The CRF binary cubic function representation of each point in Integer Pel point of a knife point (x, y) and certain neighborhood thereof:

$CRF(x_i,y_i)=a_{00}+a_{01}{y}_i+a_{02}{y}_i^2+a_{03}{y}_i^3+a_{10}{x}_i+a_{20}{x}_i^2+a_{30}{x}_i^3+a_{11}{x}_i{y}_i+a_{21}{x}_i^2{y}_i+a_{12}{x}_i{y}_i^2$

The error sum of squares of matching

$\mathrm{\&epsiv;}=\underset{i=1}{\overset{25}{\mathrm{\&Sigma;}}}{\left(CRF\right(x,y)-CRF(x_i,y_i\left)\right)}^2$

try to achieve a ₀₀, a ₀₁, a ₀₂, a ₀₃, a ₁₀, a ₂₀, a ₃₀, a ₁₁, a ₂₁, a ₁₂

Utilize the cubic surface expression formula determined to solve the CRF that Integer Pel point of a knife point is subdivided into 3 × 3 sub-pix points, get the coordinate of the subpixel coordinates in 9 sub-pix points corresponding to CRF maximal value as this point of a knife point.

7. saw blade wear extent On-line Measuring Method according to claim 1, is characterized in that, the computing method that described step 7 solves point of a knife point place radius of a circle value are as follows:

Definition

$a=n\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i^2-\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i$

$b=n\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i{y}_i-\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_i$

$c=n\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i^3+n\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i{y}_i^2-\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}(x_i+y_i)\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i$

$d=n\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_i^2-\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_i\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_i$

$e=N\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i^2{y}_i+N\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_i^3-\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}(x_i^2+y_i^2)\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_i$

$A=\frac{be-cd}{ad-b^2}$

$B=\frac{ae-bc}{b^2-ad}$

$C=-\frac{1}{n}\left(\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}\right(x_i^2+y_i^2)+A\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_i+B\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{y}_i)$

According to formula r ²=(A ²+ B ²-4C)/4, calculate the sub-pixel values of point of a knife point place radius of circle.

8. saw blade wear extent On-line Measuring Method according to claim 1, is characterized in that, the computing method of the actual wear amount of the saw blade of described step 7 are as follows:

(1) camera calibration, at the demarcation thing that the position installation dimension of installing saw blade is known, industrial computer trigger image capture card obtains the image demarcating thing, according to the image obtained, calculate the pixel value of known dimensions, known dimensions obtains the physical size value of each pixel representative divided by pixel value;

(2) according to the radius value obtaining cutter reality, the difference of twice testing result is the pixel value of saw blade wear extent, then is multiplied by the physical size value of each pixel representative, obtains the size of the wearing and tearing value of saw blade.