CN101281208A - Cotton stream velocity on-line estimation method using video to measure speed in isomerism fibre sorting system - Google Patents

Abstract

The invention pertains to the industry testing technique field based on machine vision, which is characterized in that two linear array cameras with a setting distance are utilized to respectively extract images in which the cotton stream is partly superposed in the setting equal sampling period, according to cotton streams doing a speed changing rectilinear movement along the channel, and the image are divided into a plurality of cotton stream sub-graphs along a direction perpendicular to the cotton stream cross section direction; according to each sub-graph, the similarity between the sub-graph picked up by a linear array camera 1 and the linear array camera 2 with a different time delay delta t, thereby a similarity curve f(delta t) varying with delta t is obtained, and then calculating a time delay delta T corresponding to the maximum value f(delta T) of the curve f(delta t), namely the time delay of a same point on the cotton stream passing through two linear array cameras; and then based on the setting distance between the two linear array cameras, the cotton stream online speed distributed along the cotton stream transversal, under the condition that the cotton stream velocity will not abruptly change in the setting sampling time delay. the invention can be used to realize accurate sorting of isomerism fiber in the cotton stream.

Description

Use the cotton stream velocity degree On-line Estimation method of video frequency speed-measuring in the foreign fiber sorting system

Technical field

The invention belongs to industrial detection technical field based on machine vision.

Background technology:

So-called foreign fiber letter sorting mainly is meant from a large amount of cottons identification and sorts out the process of foreign fiber (as the hair of humans and animals, polypropylene fibre silk, plastics etc.).In the cotton spinning industry, the foreign fiber letter sorting is very important together operation.Foreign fiber sorting system based on the machine vision principle is a kind of main method of using at present.The foreign fiber sorting system uses high speed camera (being generally line-scan digital camera) that the cotton stream in the transparent channel is carried out online shooting, with take the photograph image send into computing machine, computer software by special use identifies foreign fiber and determines its position, determines delay time and drive corresponding solenoid valve according to the cotton stream velocity degree at last foreign fiber is blown out.The whether accurate rate that blows out that directly influences foreign fiber estimated of cotton stream velocity degree wherein, the estimation of inaccurate cotton stream velocity degree can cause the solenoid valve action in advance or hysteresis, detected foreign fiber is blown by leakage.In existing systems, the cotton stream velocity degree generally estimates according to ducted wind speed, wind speed can be provided by blowing device control or ducted measuring wind speed equipment, multiply by an experience factor (as 0.78) less than 1 and be the cotton stream velocity degree on the basis of wind speed.This cotton stream velocity degree method of estimation is simple and easy to use, cotton stream velocity degree applicable to wind speed relatively steadily the time is estimated, but in practical application in industry, ducted wind speed is often also unstable, and for different cotton varieties, be used to estimate that the experience factor of cotton stream velocity degree generally is different, so use classic method to be difficult to obtain the estimation of cotton stream velocity degree accurately.

Video frequency speed-measuring is a research focus (Tong Jianjun has a talk about video frequency speed-measuring, Chinese transportation information industry, 2006, the 9 phases, 36-38 page or leaf) of " intelligent transportation field ", and there are some difficulties in its advantage although video traffic has tested the speed, has limited practical application.The present invention is applied to the foreign fiber sorting system with the video frequency speed-measuring technology.With respect to the video frequency speed-measuring in the intelligent transportation field, the video frequency speed-measuring technology is applied to the foreign fiber sorting system has following supremacy clause: 1, illumination is stable; 2, double camera is gathered vision signal; 3, camera position is fixed; 4, do not need conversion between imaging space and the physical space.

Summary of the invention:

Purpose of the present invention is to propose the cotton stream velocity degree On-line Estimation method that a kind of foreign fiber accurately sorts usefulness.The invention is characterized in that this method contains following steps successively:

Step (1), flow line-scan digital camera (1) and (2) of two setpoint distance S of being separated by of direction of motion placement perpendicular to this channel width direction and cotton in cotton circulation road one side, the cotton stream of making change speed linear motion is collected two width of cloth images that overlap in the time of setting that equates, the cotton stream picture of two width of cloth that described two line-scan digital cameras (1) and (2) are gathered is along cutting apart perpendicular to the direction of cotton stream transversal, be divided into some cotton stream subgraphs, be masked as S1 (t, x) and S2 (t, x), x is that the cotton corresponding with described camera (1) or (2) flowed the locus on the transversal, x ∈ X _pAnd { X ₁... X _p... X _mConstitute complete cotton stream transversal, and m is the number of subgraph, and the m minimum is 1, and the effective observation that is described camera (1) or (2) to the maximum is counted, and t is the corresponding time that photographs described cotton stream transversal of described camera (1) or (2);

Step (2), cotton that line-scan digital camera (1) and (2) photograph respectively stream subgraph S1 (t, x) and S2 (t x) imports a computing machine;

Step (3), computing machine be calculated as follows described cotton stream subgraph S1 (t, x) and corresponding different delayed time Δ t cotton stream subgraph S2 down (t, the x) similarity between obtain a curve f (Δ t) with the similarity degree of Δ t variation;

$f\left(\mathrm{\Δt}\right)=\frac{\underset{t,x}{\mathrm{\Σ}}\left[(S1(t,x)-m1)(S2(t+\mathrm{\Δt},x)-m2)\right]}{\sqrt{\underset{t,x}{\mathrm{\Σ}}{[S1(t,x)-m1]}^2\underset{t,x}{\mathrm{\Σ}}{[S2(t+\mathrm{\Δt},x)-m2]}^2}}$

Δ t ∈ { Δ t ₁, Δ t ₂..., Δ t _n, n is the number of Δ t, is setting value, n＞0,

M1 be S1 (t, average x) are used following formulate:

$m1=\frac{1}{L}\underset{t,x}{\mathrm{\Σ}}S1(t,x)$

M2 be S2 under the different Δ t (t+ Δ t, average x) are used following formulate:

$m2=\frac{1}{L}\underset{t,x}{\mathrm{\Σ}}S2(t+\mathrm{\Δt},x)$

In the formula, L is the number of the pixel of cotton stream subgraph;

Step (4) is found out the maximal value f (Δ T) among the curve f (Δ t), and the pairing time delay Δ of this maximal value T is cotton stream the time required when going up same point respectively by described two cameras (1) and (2);

Step (5) is separated by the velocity analysis set at interval in the I, and I be arbitrarily greater than zero number, the cotton stream velocity degree v that has estimated according to previous moment ₀Estimate next speed v constantly by following formula

$v=\mathrm{\&alpha;}{\mathrm{<figure-callout\; id="0"\; label="v"\; filenames="A20071006519200111.png"\; state="\{\{state\}\}">v</figure-callout>}}_0+(1-\mathrm{\&alpha;})\frac{S}{\mathrm{\&Delta;T}},$ α is the arbitrary value between 0 to 1, the mensuration wind speed the when initial value of v is startup;

Step (6) is handled all cotton stream subgraphs according to step (2)-described method of step (5), obtains the curve that current time cotton stream velocity degree v changes with subgraph variable p, p=1 ..., m is the space distribution curve of current time cotton stream velocity degree.

It is accurate in real time to have speed Tracking with the cotton stream velocity degree On-line Estimation method of video frequency speed-measuring, does not need additionally to add advantages such as hardware device.In our system, not using the average mistake rate of blowing of online speed-measuring method is 90%, the average mistake rate of blowing is reduced to 50% after using online speed-measuring method, and the average mistake rate of blowing described here is the inferior situation that blows out of checking 50 times of per tour, and statistics is as good as fine cotton-wool number and accounts for the ratio that always blows out the cotton-wool number.

Description of drawings:

The video frequency speed-measuring techniqueflow chart that Fig. 1, the present invention provide.

Fig. 2, signal acquiring system synoptic diagram.

Fig. 3, cotton flow graph resemble, and (a) camera 1 is gathered; (b) camera 2 is gathered.

The rate curve that the On-line Estimation of Fig. 4, a period of time obtains.

Embodiment:

Adopt wind speed to determine the defective that the cotton stream velocity degree exists indirectly at existing foreign fiber sorting system, this invention utilizes the video frequency speed-measuring technology directly to determine the cotton stream velocity degree.Existing foreign fiber sorting system adopts two or four stringing array cameras, this invention only need be utilized the vision signal of the different observation stations of cotton flow path direction to carry out Computer Processing and extract cotton stream velocity degree information, so do not need to increase or change the hardware components of system.Along cotton flow path direction two stringing array cameras are installed one in front and one in back, each only obtains a transversal (the vertical or near normal of general and cotton flow path direction) that flows by cotton constantly one stringing array camera, one 1 dimensional signal that promptly obtains, a plurality of moment combine, just obtain that the cotton flow graph by the camera check point resembles in this section moment, promptly obtain one 2 dimensional signal.The cotton flow graph that two cameras that front and back are placed collect resembles the feature of extraction (or resembled by cotton flow graph) and mates contrast, estimate the current cotton stream that passes through first camera and pass through second camera in the back, just can obtain the mistiming of the same point of current cotton stream by two cameras how long.Because the check point of two cameras on cotton stream is known, promptly the distance between them is fixing and known, so just can obtain the estimation of current cotton stream velocity degree.By the cotton stream video of high speed acquisition, the estimation of real-time update cotton stream velocity degree realizes online cotton stream velocity degree estimation.Utilize the cotton stream velocity degree of estimating in real time, during foreign fiber in finding cotton stream, the execution time of online definite solenoid valve, realize accurate foreign fiber letter sorting.

Provide one embodiment of the present invention below.We apply the present invention in the foreign fiber sorting system of a reality, and cotton stream signal acquiring system synoptic diagram as shown in Figure 2.The dotted line by two cameras and the point of crossing of cotton circulation road are their observation station.Fig. 3 a and 3b have provided camera 1 respectively and the 2 cotton flow graphs of gathering in the time period of setting resemble, and acquisition time length is 0.32 second.The corresponding cotton circulation road width of x axle is 80cm.Distance is 5cm between camera.As can be seen from Figure 3, a certain the cotton image that camera 1 collects will appear at behind certain hour in the image of camera 2 collections.The cotton flow graph of two camera collections of note resemble corresponding subgraph be masked as S1 (t, x) and S2 (t, x), t is the time, x is the locus that the cotton of corresponding camera is flowed transversal.Utilize an image similarity tolerance of cross-correlation coefficient structure f as follows:

$f\left(\mathrm{\&Delta;t}\right)=\frac{\underset{t,x}{\mathrm{\&Sigma;}}\left[(S1(t,x)-m1)(S2(t+\mathrm{\&Delta;t},x)-m2)\right]}{\sqrt{\underset{t,x}{\mathrm{\&Sigma;}}{[S1(t,x)-m1]}^2\underset{t,x}{\mathrm{\&Sigma;}}{[S2(t+\mathrm{\&Delta;t},x)-m2]}^2}}$

In the formula, m1, m2 be respectively S1 (t, x) and S2 (t+ Δ t, average x), $m1=\frac{1}{L}\underset{t,x}{\mathrm{\&Sigma;}}S1(t,x),$ $m2=\frac{1}{L}\underset{t,x}{\mathrm{\&Sigma;}}S2(t+\mathrm{\&Delta;t},x),$ Δ t is a time delay.Calculate S1 (t, x) and the image S2 of corresponding different delay Δ t (t+ Δ t, the x) similarity between obtain a similarity measurement curve f (Δ t) who changes with Δ t.If the f of corresponding time delay Δ T (Δ T) is the maximal value in f (Δ t) curve, the last same point of then cotton stream is Δ T by the time delay of two cameras.The cotton flow graph that utilizes Fig. 3 to provide resembles can estimate to obtain Δ T=7ms, then obtains this point estimation speed and is $\frac{S}{\mathrm{\&Delta;T}}=7.1m/s.$

Consider influence factors such as noise, and find by cotton circulation road air-velocity measurement, can think that gas velocity can acutely not suddenly change, can think that promptly the cotton stream velocity degree can not suddenly change, setting the non-negative Velocity Updating factor is α=0.88, velocity analysis time interval I=0.032s, adopt following Velocity Updating formula:

$v=\mathrm{\&alpha;}{v}_0+(1-\mathrm{\&alpha;})\frac{S}{\mathrm{\&Delta;T}}$

In the formula, v ₀Be the previous moment estimating speed.

Fig. 4 has provided the rate curve that a period of time On-line Estimation obtains.

The present invention has following features:

1. use the two stringing array camera high speed acquisition cotton flow vision signals of one in front and one in back installing along the cotton flow direction, utilize cotton flow figure Resemble (or by the visual feature extracted of cotton flow) and mate, obtain the same point of current cotton flow by the time difference of two cameras, Utilize the observation station distance estimations cotton flow speed of two stringing array cameras on cotton flow.

2. utilize previous moment to estimate the cotton flow speed that obtains, in conjunction with the cotton flow speed of current estimation, realize adaptive online Cotton flow speed is upgraded, and has improved the robustness of velocity estimation.

3. spatial division is carried out with the direction of cotton flow perpendicular direction in the cotton flow image edge that obtains, to each image subsection after dividing Carry out respectively the cotton flow velocity estimation that provides in the 1st and 2 characteristics, obtain the spatial distribution of cotton flow speed, adapt to cotton flow speed and exist There is situation about changing on the channel space.

4. in practice, set the subgraph number according to cotton flow passage middle cotton Flow Velocity spatial variations situation, the subgraph number is minimum to be 1, mostly be effective the points of measurement of line-scan digital camera most.

5. by setting the time interval of suitable velocity analysis, between speed follow-up control and amount of calculation, trade off. One As cotton flow speed can not suddenly change, can adopt bigger time series analysis interval.

6. two cameras can be installed in the same side of cotton flow passage, also can be installed in respectively the both sides of cotton flow passage.

7. two cameras can adopt identical or different lighting source.

8. estimate that the real-time cotton flow speed that obtains is used to determine online the time of implementation of magnetic valve, realize accurate foreign fiber Sorting.

Claims (1)

1. use the cotton stream velocity degree On-line Estimation method of video frequency speed-measuring in the foreign fiber sorting system, it is characterized in that this method contains following steps successively:

Step (1), flow line-scan digital camera (1) and (2) of two setpoint distance S of being separated by of direction of motion placement perpendicular to this channel width direction and cotton in cotton circulation road one side, the cotton stream of making change speed linear motion is collected two width of cloth images that overlap in the time of setting that equates, the cotton stream picture of two width of cloth that described two line-scan digital cameras (1) and (2) are gathered is along cutting apart perpendicular to the direction of cotton stream transversal, be divided into some cotton stream subgraphs, be masked as S1 (t, x) and S2 (t, x), x is that the cotton corresponding with described camera (1) or (2) flowed the locus on the transversal, x ∈ X _pAnd { X ₁... X _p... X _mConstitute complete cotton stream transversal, and m is the number of subgraph, and the m minimum is 1, and the effective observation that is described camera (1) or (2) to the maximum is counted, and t is the corresponding time that photographs described cotton stream transversal of described camera (1) or (2);

Step (2), cotton that line-scan digital camera (1) and (2) photograph respectively stream subgraph S1 (t, x) and S2 (t x) imports a computing machine;

Step (3), computing machine be calculated as follows described cotton stream subgraph S1 (t, x) and corresponding different delayed time Δ t cotton stream subgraph S2 down (t, the x) similarity between obtain a curve f (Δ t) with the similarity degree of Δ t variation;

$f\left(\mathrm{\&Delta;t}\right)=\frac{\underset{t,x}{\mathrm{\&Sigma;}}\left[(S1(t,x)-m1)(S2(t+\mathrm{\&Delta;t},x)-m2)\right]}{\sqrt{\underset{t,x}{\mathrm{\&Sigma;}}{[S1(t,x)-m1]}^2\underset{t,x}{\mathrm{\&Sigma;}}{[S2(t+\mathrm{\&Delta;t},x)-m2]}^2}}$

Δ t ∈ { Δ t ₁, Δ t ₂..., Δ t _n, n is the number of Δ t, is setting value, n＞0,

M1 be S1 (t, average x) are used following formulate:

$m1=\frac{1}{L}\underset{t,x}{\mathrm{\&Sigma;}}S1(t,x)$

M2 be S2 under the different Δ t (t+ Δ t, average x) are used following formulate:

$m2=\frac{1}{L}\underset{t,x}{\mathrm{\&Sigma;}}S2(t+\mathrm{\&Delta;t},x)$

In the formula, L is the number of the pixel of cotton stream subgraph;

Step (4) is found out the maximal value f (Δ T) among the curve f (Δ t), and the pairing time delay Δ of this maximal value T is cotton stream the time required when going up same point respectively by described two cameras (1) and (2);

Step (5) is separated by the velocity analysis set at interval in the I, and I be arbitrarily greater than zero number, the cotton stream velocity degree v that has estimated according to previous moment ₀Estimate next speed v constantly by following formula

$v={\mathrm{\&alpha;v}}_0+(1-\mathrm{\&alpha;})\frac{S}{\mathrm{\&Delta;T}},$ α is the arbitrary value between 0 to 1, the mensuration wind speed the when initial value of v is startup;

Step (6) is handled all cotton stream subgraphs according to step (2)-described method of step (5), obtains the curve that current time cotton stream velocity degree v changes with subgraph variable p, p=1 ..., m is the space distribution curve of current time cotton stream velocity degree.

Images