CN110068270A - A kind of monocular vision box volume measurement method based on multi-line structured light image recognition - Google Patents
A kind of monocular vision box volume measurement method based on multi-line structured light image recognition Download PDFInfo
- Publication number
- CN110068270A CN110068270A CN201910312179.7A CN201910312179A CN110068270A CN 110068270 A CN110068270 A CN 110068270A CN 201910312179 A CN201910312179 A CN 201910312179A CN 110068270 A CN110068270 A CN 110068270A
- Authority
- CN
- China
- Prior art keywords
- cabinet
- line
- structured light
- coordinate
- camera
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
- G01B11/00—Measuring arrangements characterised by the use of optical techniques
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Length Measuring Devices By Optical Means (AREA)
Abstract
The monocular vision box volume measurement method based on multi-line structured light image recognition that the present invention relates to a kind of, including following operating procedure: 1) camera calibration and multi-thread laser plane parameter obtain.2) calculating of the two dimensional coordinate map to three dimensional space coordinate.3) cabinet edge detection and laser center line drawing are carried out with deep learning method.4) different cabinets is measured using groined type, double flat line and cross type structure light respectively, obtains the volume of regular cabinet.The present invention by measurement cabinet planar projective multi-line structured light (groined type, cross, it is parallel linear), by the intersection point of image procossing detection line-structured light and chest edge, it will be put from two-dimensional coordinate system according to structure light triangulation theory and be converted to computation rule cabinet length, width and height under three-dimensional system of coordinate, and then calculate its volume.This method algorithm complexity is low, can be advantageously applied to the end PC and mobile end equipment, and energy real-time online measuring meets portability requirements.
Description
Technical field
The invention belongs to a kind of methods of box volume measurement, and in particular to a kind of based on multi-line structured light image recognition
Monocular vision box volume measurement method.
Background technique
Box volume measurement is mainly used in the fields such as logistics, electric business, warehouse storage, aviation warehouse, provide calculate freight charges,
The various services such as cargo is put, cabinet is recommended.Its application scenarios is complex, so how quick, quasi- under complex environment
Really, the volume for efficiently measuring cabinet is these industry urgent problems, furthermore how to be sent out on the basis of superperformance
Open up the urgent need that portable volume measuring instrument is also these industries.It is current there are many measurement method: artificial detection,
It is measured either manually or by ruler, not only low efficiency but also is easy to appear mistake；Grating measuring method, equipment automatic measurement, measurement accuracy
Height, but manpower is needed to move cabinet to fixed position, extremely difficult, labor intensive and instrument are measured to the cabinet higher than 1 meter
It is at high cost；Plethysmometry based on binocular stereo vision, algorithm complexity is high, and Stereo matching is difficult, measures large scale cabinet
Precision is low, is not suitable for the measurement scene of the complexity such as logistics.
Summary of the invention
The present invention proposes a kind of based on multi-thread in view of the demand of the industries such as logistics transportation, electric business, supermarket measurement box volume
The monocular vision box volume measurement method of structure light image identification, before guaranteeing to the high-acruracy survey of large scale box volume
It puts, realizes practical online portable box volume measurement method.The present invention is more by projecting to measurement tank surface
Line-structured light (groined type, cross are parallel linear), as shown in Figure 1, then by deep learning algorithm detect line-structured light with
Point is finally converted to computational length and width under three-dimensional system of coordinate according to structure light triangulation theory by the intersection point at cabinet edge
Degree repeats an adjacent surface for shooting cabinet, calculates the height of cabinet, and then calculate volume.The measurement method has non-connect
The plurality of advantages such as touching, speed are fastly, wide range, stability are good, algorithm complexity is low.
In order to achieve the above objectives, insight of the invention is that
Multiple scaling board pictures for having multi-line structured light are shot first, are calibrated in camera using Zhang Zhengyou calibration method
Ginseng, outer ginseng, distortion factor carry out distortion correction to the scaling board picture of shooting.Then a picture is taken, is extracted in its laser rays
Laser rays central point on two-dimension picture is carried out straight line fitting with least square method by heart point.Swash again in the same of plurality of pictures
Multiple points are equidistantly taken on light, constitute the point set of laser plane, carry out plane fitting with least square method.In measurement process,
Two adjacent surfaces for acquiring cabinet first, pre-process the collected cabinet picture containing multi-line structured light, are swashed
The pixel coordinate for the characteristic point that light intersects with cabinet edge.According to the inside and outside parameter of camera, distortion factor and laser plane equation
By the three-dimensional coordinate under the coordinate transformation of characteristic point to camera coordinates system, length, width and the height of cabinet is finally calculated.
According to the above invention thinking, the specific technical solution of the present invention is as follows:
A kind of monocular vision box volume measurement method based on multi-line structured light image recognition, will be multi-thread on two dimensional image
The intersection point at structure light and cabinet edge is converted under three-dimensional coordinate, calculates its length, and then calculate its volume, concrete operation step
It is as follows:
1) camera calibration and multi-thread laser plane parameter obtain: groined type structure light, double parallel line-structured light and cross are handed over
The scaling method of forked type structure light shoots multiple scaling board pictures, is obtained in the system in camera using Zhang Zhengyou calibration method
Join matrix K, the distortion factor k of camera1,k2,k3, distortion correction is carried out to the scaling board picture of shooting；Multiple are shot to be respectively provided with
The scaling board picture of groined type structure light, double parallel line-structured light and cross type structure light, is swashed by image procossing
The corresponding phase of laser optical losses each point is calculated according to camera parameter and target plane plane equation in the coordinate of light optical losses
Machine coordinate finally carries out plane fitting using least square method, obtains multiple line structure optic plane equations；
2) calculating of the two dimensional coordinate map to three dimensional space coordinate: flat using the Intrinsic Matrix K and line-structured light of camera
The intersecting point coordinate of line-structured light and cabinet edge on picture is converted to camera coordinates by pixel coordinate system by face equation；
3) cabinet edge detection and laser center line drawing are carried out using deep learning method: by the way that containing for shooting is added
Data set with laser rays and cabinet, convolutional layer and the pond number of plies to deep learning method are trained, finally optimization output
Marginal probability figure obtains the conspicuousness fringe region containing cabinet edge and laser rays；Two are carried out to the marginal probability figure of image
Value operation, Hough transformation is then carried out on binary image, finds whole straight lines, carries out straight line cluster and endpoint later
Detection；
4) box volume measures: obtaining cabinet length and width with first cabinet face of groined type structural light measurement cabinet
Degree obtains box height using double parallel line-structured light measurement adjacent surface later；For elongated cabinet, cross structure light pair is used
Cabinet measures, and obtains the length and width of cabinet；After obtaining the intersecting point coordinate at multi-line structured light and cabinet edge, according to knot
Point is converted to computational length and width under three-dimensional system of coordinate by structure light triangulation theory, and then calculates its volume.
In the step 1), using improved Zhang Zhengyou camera calibration method calibration for cameras, is shot and distinguished using video camera
Image with groined type structure light, the different angle of double parallel line-structured light and cross type structure light, depth obtains phase
Join outside the internal reference of machine and distortion factor, world coordinate system and camera coordinates system, camera coordinates system and pixel coordinate are obtained with this
Relationship between system carries out distortion correction to picture；The picture of laser optical losses in figure acceptance of the bid targeting is extracted by image procossing
It is corresponding that laser optical losses each point is calculated according to camera parameter and target plane plane equation in the coordinate of each point in plain coordinate system
Camera coordinates system under coordinate, carry out plane fitting respectively, obtain multiple line structure optic plane equations.
In the step 2), by two dimensional image on laser rays for a point P, it is assumed that P point is laser plane 1 and cabinet
On intersection a bit, P point coordinate in pixel coordinate system is set as (up,vp), coordinate of the P under camera coordinates system is set as PC(Xpc,
Ypc,Zpc), following relationship is obtained on laser plane 1 according to camera imaging principle and P point:
Wherein, K is the Intrinsic Matrix of camera, A1、B1、C1For the parameter of line-structured light plane equation；Therefore, according to solution
Equation group obtains coordinate P of the P point under camera coordinates systemC(Xpc,Ypc,Zpc)。
In the step 3), shooting has groined type structure light, double parallel line-structured light and cross type structure respectively
The data set of light and cabinet, convolutional layer and the pond number of plies to deep learning method are trained, and finally optimization output edge is general
Rate figure obtains the conspicuousness fringe region containing cabinet edge and laser rays；Binaryzation behaviour is carried out to the marginal probability figure of image
Make, Hough transformation is carried out on the binary image of cabinet image, finds whole straight lines, finds out the seat under its polar coordinate system
Mark, is set as (ρi,θi), the straight line in threshold range is classified as one kind by given threshold, therefore obtains under pixel coordinate system this
The equation of straight line:
Y=k1*x+b1
Wherein k1For straight slope, the b acquired1For the Linear intercept acquired；Find what line-structured light intersected with cabinet edge
Intersection point obtains the coordinate of cabinet measurement characteristic point.
In the step 4), for different types of chest using various ways to the length of chest, width and height into
Row measurement:
Mode 1: the length and width of groined type structure light measuring tank body: straight line D1D3 and straight line D2D4 in space most short is obtained
(D1, D2, D3, D4 are groined type structure light wherein one group of Structured Light and case by distance Dis-length, the as length of cabinet
Four intersection points at body edge):
Dis-length=f (D1D3, D2D4)
Obtain the shortest distance Dis-width of straight line D5D7 and straight line D6D8 in space, as cabinet width (D5,
D6, D7, D8 are four intersection points of groined type structure light another set Structured Light and cabinet edge):
Dis-width=f (D5D7, D6D8)
Mode 2: double parallel line-structured light measures the length in adjacent face as height, obtains the height Dis-height of cabinet
(four intersection points that D1, D2, D3, D4 are double parallel line-structured light and cabinet edge):
Dis-height=f (D1D2, D3D4)
Mode 3: cross-shaped configuration light measurement slender type cabinet controls the sides aligned parallel of cross laser and cabinet as far as possible, this
When (X1, Y1, Z1, X2, Y2, Z2, X3, Y3, Z3, X4, Y4, Z4 are that cross type structure light intersects with cabinet edge respectively
Characteristic point coordinate):
Dis-length=((X1-X2)2+(Y1-Y2)2)+(Z1-Z2)2)1/2
Dis-width=((X3-X4)2+(Y3-Y4)2)+(Z3-Z4)2)1/2
The volume that cabinet is obtained finally by the volume calculation formula of cabinet is V:
V=(Dis-length) * (Dis-width) * (Dis-height)
Compared with prior art, the present invention has the advantage that:
The method of the present invention complexity is low, can be used in the end PC and mobile end equipment well, and energy real-time online measuring meets
Portability requirements, and a variety of line-structured light measurement methods of use meet the demand of the measurement to various cabinets.
Detailed description of the invention
Fig. 1 is groined type structure light, double parallel structure light and cross structure light schematic diagram.
Fig. 2 is regular cabinet measuring system flow diagram.
Fig. 3 is regular cabinet Measuring System Models structure chart.
Fig. 4 is scaling board and the scaling board schematic diagram with multi-line structured light.
Fig. 5 is the original graph of cabinet measurement.
Fig. 6 is by deep learning method treated marginal probability figure.
Fig. 7 is the flow chart that deep learning method carries out cabinet edge detection and laser center line drawing.
Fig. 8 is the final effect figure of cabinet edge detection and laser center line drawing.
Fig. 9 is groined type line-structured light instrumentation plan.
Figure 10 is double parallel structural light measurement schematic diagram.
Figure 11 is cross type structural light measurement schematic diagram.
Specific embodiment
Details are as follows for the preferred embodiment of the present invention combination attached drawing:
The flow chart of the method for the present invention whole process is as shown in Fig. 2, the system model structure chart is as shown in Figure 3.A kind of base
In the monocular vision box volume measurement method of multi-line structured light image recognition, specific steps are as follows:
1) camera calibration and multi-thread laser plane parameter obtain:
Improved Zhang Zhengyou camera calibration method calibration for cameras is used first, obtains the geometric parameter of camera.By camera
World coordinate system and camera coordinates system known to geometric parameter, the relationship between camera coordinates system and pixel coordinate system.The present invention makes
With scaling board as shown in figure 4, being respectively provided with groined type structure light, double parallel line-structured light and right-angled intersection using video camera shooting
The image of the different angle of type structure light, depth.This calibration mode is avoided using expensive calibration object and high-precision shifting
Moving platform, and stated accuracy is high.Obtaining Intrinsic Matrix by camera calibration is K.It is theoretical according to camera imaging transitting probability
Obtain having following relationship (coordinate under (u, v) expression pixel coordinate system, (X under pixel coordinate system and in camera coordinates systemc,Yc,
Zc) indicate camera coordinates system under coordinate):
In camera imaging model, world coordinate system is that the absolute of objective world three-dimensional world is coordinate, is sat by the world
Mark system can describe the position of camera, with (Xw,Yw,Zw) indicate.Camera coordinates system and world coordinate system have following relationship
(R, T are Camera extrinsic):
Ideally, the imaging model between the corresponding points in three-dimensional geometric information and image that certain in space is put is line
Property, it is to be determined by the above camera imaging model.But in practice since the construction of camera and manufacture mounting process etc. are each
The factor of aspect will lead to camera and generate distortion.The radial distortion of image is only considered in the present invention.Radial distortion is by lens shaped
It distorts caused by shape, the fractional distortion in image closer to edge is more obvious.This kind of distortion can with following polynomial function come
It is corrected (for the point in normalization plane):
Wherein, x, y are the coordinate for the point that do not correct, xrct、yrctBe correct after point coordinate, r be away from lens centre away from
From k1、k2、k3For the distortion factor of camera.
According to external parameter matrix R, T of target under camera internal parameter matrix K and different positions and pose, by under world coordinate system
Coordinate obtain target angle point coordinate under camera coordinate system, then carry out the plane fitting of least square method, it is flat to obtain target
The plane equation (A, B, C are target plane plane equation parameter, and x, y, z is the coordinate under camera coordinate system) in face.
A*x+B*y+C*z=1 (4)
The coordinate that laser optical losses each point under pixel coordinate system on calibration target is extracted by image procossing, is set as (ui,
vi), it is calculated under the corresponding camera coordinates system of laser optical losses each point by camera parameter and target plane plane equation
Coordinate is set as (XCi,YCi,ZCi).The figure for taking N (N=10) bracing cable calibration under different positions and pose, by corresponding line-structured light
Point be placed on point and concentrate, carry out plane fitting using least square method respectively, obtain multiple line structure optic plane equations (An、
Bn、CnFor the plane equation parameter of the one of structure light of multi-line structured light).
Therefore, by taking groined type laser as an example, this step can obtain the internal reference matrix K of camera in the system, the distortion of camera
Coefficient k1,k2,k3And the optic plane equations parameter A of four laser planes1,B1,C1,A2,B2,C2,A3,B3,C3,A4,B4,C4。
2) calculating of the two dimensional coordinate map to three dimensional space coordinate:
According to the optic plane equations for obtaining the internal reference of camera in system, distortion factor and laser plane in step 1), at this
In transforming relationship between pixels illustrated coordinate system and camera coordinates system.By in two dimensional image on laser rays for a point P, it is assumed that
P point is a bit on the intersection of laser plane 1 and cabinet, and P point coordinate in pixel coordinate system is (up,vp), it is assumed that P is in camera
Coordinate under coordinate system is PC(Xpc,Ypc,Zpc).It is available as follows on laser plane 1 according to camera imaging principle and P point
Relationship:
Wherein, K is the Intrinsic Matrix of camera, A1、B1、C1For the parameter of line-structured light plane equation.Therefore, it solves equation
Coordinate P of the available P point of group (6) under camera coordinates systemC(Xpc,Ypc,Zpc)。
3) cabinet edge detection and laser center line drawing are carried out using deep learning method:
Deep learning method has the ability of multi-level Analysis On Multi-scale Features in stronger extraction image relative to traditional approach,
All there is outstanding performance in image classification, the semantics recognition of image and image segmentation.Deep learning method can effectively inhibit
Background area projecting edge feature in image.The present invention contains the data set pair with laser rays and cabinet by addition shooting
The convolutional layer and the pond number of plies of deep learning method are trained, and finally optimization output marginal probability figure is obtained containing cabinet side
The conspicuousness fringe region of edge and laser rays.The original graph of shooting is as shown in figure 5, by deep learning method treated edge
Probability graph is as shown in Figure 6.
The present invention finally carries out binarization operation to the marginal probability figure of image.It is enterprising in the binary image of cabinet image
Row Hough transformation finds whole straight lines.The coordinate under its polar coordinate system is found out, (ρ is set asi,θi), given threshold will be in threshold value
Straight line in range is classified as one kind, therefore the equation of available this straight line under pixel coordinate system:
Y=k1*x+b1 (7)
It, similarly can be in the hope of other linear equations of 7 edge lines under pixel coordinate system by taking groined type laser as an example.Root
According to two straight line intersections, the coordinate of 8 characteristic points: (u may finally be respectively obtained1,v1),(u2,v2),(u3,v3),(u4,v4),
(u5,v5),(u6,v6),(u7,v7),(u8,v8)。
The operation specific flow chart is as shown in fig. 7, final effect picture is as shown in Figure 8.This step completes in the present invention
Positioning to the characteristic point that cabinet edge and laser rays intersect.
4) box volume measures part:
The length of cabinet, width and height are surveyed using various ways in the different types of cabinet present invention
Amount:
Mode 1: when measuring first cabinet face of cabinet, projection line laser is as shown in figure 9, by step 1), and 2), 3),
4) available D1, D2, D3, D4, D5, D6, D7, the coordinate under D8 camera coordinate system are as follows: (X1, Y1, Z1), (X2, Y2,
Z2), (X3, Y3, Z3), (X4, Y4, Z4), (X5, Y5, Z5), (X6, Y6, Z6), (X7, Y7, Z7), (X8, Y8, Z8), at this point, can
It is respectively as follows: with obtaining the linear equation of straight line D1D3 and D2D4
Therefore can be in the hope of the shortest distance Dis-length of straight line D1D3 and straight line D2D4 in space, as cabinet
Length:
Dis-length=f (D1D3, D2D4) (9)
The similarly shortest distance Dis-width of available straight line D5D7 and straight line D6D8 in space, as cabinet
Width:
Dis-width=f (D5D7, D6D8) (10)
Mode 2: when measuring second face of cabinet, can generally choose the face adjacent with first measuring surface is second
Measuring surface is shot in addition to that can use with line-structured light mode identical in mode 1 in operation, can also be using such as Figure 10 institute
The mode of parallel wire structure light in showing is shot, at this time can be in the hope of the height Dis-height of cabinet:
Dis-height=f (D1D2, D3D4) (11)
Mode 3: such as encountering some leptosomatic cabinets, at this time mode 1 during measurement, and mode 2 can not normal work
Make, takes cross type line-structured light as shown in figure 11 to carry out the linear measure longimetry of cabinet at this time, in operation, as far as possible
The sides aligned parallel of cross laser and cabinet is controlled, is operated in this way, measurement result can be more accurate, at this time:
Dis-length=((X1-X2)2+(Y1-Y2)2)+(Z1-Z2)2)1/2
Dis-width=((X3-X4)2+(Y3-Y4)2)+(Z3-Z4)2)1/2 (12)
The volume that cabinet is obtained finally by the volume calculation formula of cabinet is V:
V=(Dis-length) * (Dis-width) * (Dis-height) (13)
Embodiment one:
With first face of groined type structure light rule cabinet, for the measurement process of double parallel structural light measurement adjacent surface,
Actual measurement experiment is carried out.The project structured light launched by laser forms straight line striation to regular tank surface.By
In the influence of regular tank surface pattern, around background and illumination condition, the structural light strip shot by video camera
Pattern shows as incomplete striation, and optical losses are highlighted and are partly white, and marginal portion takes on a red color.Fig. 5 is well of the present invention
To the instrumentation plan of cabinet, the resolution ratio of image is 2592 × 1944 pixels for font and double parallel structure light.According to the present invention
The monocular vision box volume measurement method based on multi-line structured light proposed, measures box volume, using c++ and
Java programming is realized.Fig. 8 is the resulting measurement result of image shown in fig. 5, and the laser striation parameter of extraction is as shown in table 1, is surveyed
The results are shown in Table 2 for amount.From experimental result as can be seen that the present invention completes the precise measurement to box volume, miss measurement
Difference can be used in the end PC and mobile end equipment within 10mm well, and energy real-time online measuring meets portability requirements.
1 groined type of table and double parallel line laser striation testing result
| Endpoint 1 | Endpoint 2 | Slope | Intercept | |
| Line segment 1 | (754,138) | (1827,1250) | 1.03564 | -642.039 |
| Line segment 2 | (429,458) | (1527,1588) | 1.02985 | 16.349 |
| Line segment 3 | (1513,519) | (543,1482) | -0.993253 | 2021.97 |
| Line segment 4 | (1855,835) | (867,1785) | -0.961475 | 2619.39 |
| Line segment 5 | (1415,602) | (2013,1197) | 0.994476 | -804.419 |
| Line segment 6 | (1069,948) | (1701,1555) | 0.96032 | -77.9783 |
The regular cabinet measurement result of table 2 and error analysis
| Measured value/mm | Actual value/mm | Error/mm | |
| Length/mm | 546.3 | 541.1 | 5.2 |
| Width/mm | 470.6 | 468.3 | 2.3 |
| Height/mm | 219.1 | 213.5 | 5.6 |
Embodiment two:
Some leptosomatic cabinets are such as encountered during measurement, at this time mode 1, mode 2 can not work normally, at this time
Cross type line-structured light is taken to carry out the linear measure longimetry of cabinet.The cross structure light projection launched by laser is to rule
Tank surface forms straight line striation.In operation, the sides aligned parallel of cross laser and cabinet, shooting gained are controlled as far as possible
Measurement picture as shown in figure 5, photo resolution be 2592 × 1944 pixels.It is proposed according to the present invention based on multi-line structured light
Monocular vision box volume measurement method in measurement method three elongated cabinet is measured, compiled using c++ and java
Cheng Shixian.
3 cruciform shape laser striation testing result of table
| Endpoint 1 | Endpoint 2 | Slope | Intercept | |
| Line segment 1 | (1103,996) | (1707,1574) | 0.956954 | -59.5199 |
| Line segment 2 | (2477,411) | (1219,1788) | -1.09459 | 3122.31 |
The regular cabinet measurement result of table 4 and error analysis
| Measured value/mm | Actual value/mm | Error/mm | |
| Length/mm | 547.0 | 541.1 | 5.9 |
| Width/mm | 216.8 | 213.5 | 3.3 |
Fig. 8 is the resulting measurement result of image shown in fig. 5, and the laser striation parameter of extraction is as shown in table 3, measurement result
As shown in table 4.From experimental result as can be seen that proposed by the present invention survey elongated cabinet with cross type line-structured light
The method of amount can to large scale rule cabinet carry out precise measurement, compared with artificial detection, grating measuring method, be based on binocular
The Plethysmometry of stereoscopic vision is compared, and there have that wide range, precision are high, non-contact, speed is fast, algorithm complexity is low etc. to be many excellent
Point manufactures portable device for industries such as logistics and lays a good foundation.
Claims (5)
1. a kind of monocular vision box volume measurement method based on multi-line structured light image recognition, which is characterized in that will be two-dimentional
The intersection point at multi-line structured light and cabinet edge is converted under three-dimensional coordinate on image, calculates its length, and then calculate its volume, is had
Steps are as follows for gymnastics work:
1) camera calibration and multi-thread laser plane parameter obtain: groined type structure light, double parallel line-structured light and cross type
The scaling method of structure light shoots multiple scaling board pictures, obtains the internal reference square of camera in the system using Zhang Zhengyou calibration method
Battle array K, the distortion factor k of camera1,k2,k3, distortion correction is carried out to the scaling board picture of shooting；It shoots multiple and is respectively provided with well word
The scaling board picture of shape structure light, double parallel line-structured light and cross type structure light, obtains laser light by image procossing
The coordinate at center is calculated the corresponding camera of laser optical losses each point according to camera parameter and target plane plane equation and sits
Mark finally carries out plane fitting using least square method, obtains multiple line structure optic plane equations；
2) calculating of the two dimensional coordinate map to three dimensional space coordinate: Intrinsic Matrix K and line-structured light plane side using camera
The intersecting point coordinate of line-structured light and cabinet edge on picture is converted to camera coordinates system by pixel coordinate system by journey；
3) cabinet edge detection and laser center line drawing are carried out using deep learning method: by the way that swashing containing band for shooting is added
The data set of light and cabinet, convolutional layer and the pond number of plies to deep learning method are trained, finally optimization output edge
Probability graph obtains the conspicuousness fringe region containing cabinet edge and laser rays；Binaryzation is carried out to the marginal probability figure of image
Operation, Hough transformation is then carried out on binary image, finds whole straight lines, carries out straight line cluster and endpoint inspection later
It surveys；
4) box volume measures: cabinet length and width is obtained with first cabinet face of groined type structural light measurement cabinet, it
Box height is obtained using double parallel line-structured light measurement adjacent surface afterwards；For elongated cabinet, using cross structure light to cabinet
It measures, obtains the length and width of cabinet；After obtaining the intersecting point coordinate at multi-line structured light and cabinet edge, according to structure light
Point is converted to computational length and width under three-dimensional system of coordinate by triangulation theory, and then calculates its volume.
2. a kind of monocular vision box volume measurement side based on multi-line structured light image recognition according to claim 1
Method, which is characterized in that in the step 1), using improved Zhang Zhengyou camera calibration method calibration for cameras, clapped using video camera
The image for being respectively provided with groined type structure light, the different angle of double parallel line-structured light and cross type structure light, depth is taken the photograph,
It obtains and joins outside the internal reference of camera and distortion factor, world coordinate system and camera coordinates system, camera coordinates system and picture are obtained with this
Relationship between plain coordinate system carries out distortion correction to picture；It is extracted in figure acceptance of the bid targeting in laser striation by image procossing
It is each that laser optical losses are calculated according to camera parameter and target plane plane equation in the coordinate of each point in the pixel coordinate system of the heart
Coordinate under the corresponding camera coordinates system of point, carries out plane fitting respectively, obtains multiple line structure optic plane equations.
3. a kind of monocular vision box volume measurement side based on multi-line structured light image recognition according to claim 1
Method, which is characterized in that in the step 2), by two dimensional image on laser rays for a point P, it is assumed that P point is 1 He of laser plane
On the intersection of cabinet a bit, P point coordinate in pixel coordinate system is set as (up,vp), coordinate of the P under camera coordinates system is set as
PC(Xpc,Ypc,Zpc), following relationship is obtained on laser plane 1 according to camera imaging principle and P point:
Wherein, K is the Intrinsic Matrix of camera, A1、B1、C1For the parameter of line-structured light plane equation；Therefore, according to solving equation
Group obtains coordinate P of the P point under camera coordinates systemC(Xpc,Ypc,Zpc)。
4. a kind of monocular vision box volume measurement side based on multi-line structured light image recognition according to claim 1
Method, which is characterized in that in the step 3), shooting has groined type structure light, double parallel line-structured light and right-angled intersection respectively
The data set of type structure light and cabinet, convolutional layer and the pond number of plies to deep learning method are trained, finally optimization output
Marginal probability figure obtains the conspicuousness fringe region containing cabinet edge and laser rays；Two are carried out to the marginal probability figure of image
Value operation, Hough transformation is carried out on the binary image of cabinet image, finds whole straight lines, is found out under its polar coordinate system
Coordinate, be set as (ρi,θi), the straight line in threshold range is classified as one kind, therefore obtained in pixel coordinate system by given threshold
The equation of this lower straight line:
Y=k1*x+b1 (2)
Wherein k1For straight slope, the b acquired1For the Linear intercept acquired；Find the friendship that line-structured light intersects with cabinet edge
Point obtains the coordinate of cabinet measurement characteristic point.
5. a kind of monocular vision box volume measurement side based on multi-line structured light image recognition according to claim 1
Method, which is characterized in that in the step 4), for different types of chest using various ways to the length of chest, width and
Height measures:
Mode 1: the length and width of groined type structure light measuring tank body: straight line D1D3 and straight line D2D4 in space most short is obtained
Distance Dis-length, the as length of cabinet, wherein D1, D2, D3, D4 are groined type structure light wherein one group of Structured Light
With four intersection points at cabinet edge:
Dis-length=f (D1D3, D2D4) (3)
Obtain the shortest distance Dis-width of straight line D5D7 and straight line D6D8 in space, the as width of cabinet, wherein D5,
D6, D7, D8 are four intersection points of groined type structure light another set Structured Light and cabinet edge:
Dis-width=f (D5D7, D6D8) (4)
Mode 2: double parallel line-structured light measures the length in adjacent face as high, obtain cabinet height Dis-height, D1,
D2, D3, D4 are four intersection points of double parallel line-structured light and cabinet edge:
Dis-height=f (D1D2, D3D4) (5)
Mode 3: cross-shaped configuration light measurement slender type cabinet controls the sides aligned parallel of cross laser and cabinet as far as possible, at this point,
X1, Y1, Z1, X2, Y2, Z2, X3, Y3, Z3, X4, Y4, Z4 are the feature that cross type structure light intersects with cabinet edge respectively
Point coordinate:
The volume that cabinet is obtained finally by the volume calculation formula of cabinet is V:
V=(Dis-length) * (Dis-width) * (Dis-height) (7).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910312179.7A CN110068270B (en) | 2019-04-18 | 2019-04-18 | Monocular vision box volume measuring method based on multi-line structured light image recognition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910312179.7A CN110068270B (en) | 2019-04-18 | 2019-04-18 | Monocular vision box volume measuring method based on multi-line structured light image recognition |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110068270A true CN110068270A (en) | 2019-07-30 |
| CN110068270B CN110068270B (en) | 2021-04-02 |
Family
ID=67368022
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910312179.7A Active CN110068270B (en) | 2019-04-18 | 2019-04-18 | Monocular vision box volume measuring method based on multi-line structured light image recognition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110068270B (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111402330A (en) * | 2020-04-03 | 2020-07-10 | 山东省科学院激光研究所 | Laser line key point extraction method based on plane target |
| CN111426280A (en) * | 2020-05-08 | 2020-07-17 | 中国科学技术大学 | Two-dimensional DIC optical extensometer out-of-plane compensation device and method based on structured light |
| CN111598939A (en) * | 2020-05-22 | 2020-08-28 | 中原工学院 | Human body circumference measuring method based on multi-vision system |
| CN111610173A (en) * | 2020-05-27 | 2020-09-01 | 中国水利水电科学研究院 | Three-dimensional fluid concentration field calibration device and calibration method |
| CN111932494A (en) * | 2020-07-01 | 2020-11-13 | 湖南省鹰眼在线电子科技有限公司 | Tire wear degree evaluation method and device |
| CN112781470A (en) * | 2020-12-24 | 2021-05-11 | 中广核核电运营有限公司 | U-shaped groove multi-dimensional gap measuring method and system |
| CN112964185A (en) * | 2021-01-20 | 2021-06-15 | 广西电网有限责任公司电力科学研究院 | Power transmission line crossing point on-line monitoring system based on binocular stereo vision |
| CN113137920A (en) * | 2021-05-19 | 2021-07-20 | 重庆大学 | Underwater measurement equipment and underwater measurement method |
| CN113222986A (en) * | 2021-06-04 | 2021-08-06 | 中冶宝钢技术服务有限公司 | Continuous casting billet angular point and edge contour point set positioning method, system, medium and device |
| CN113256738A (en) * | 2021-06-23 | 2021-08-13 | 苏州大学 | Binocular-based parcel volume measurement method |
| CN113358052A (en) * | 2021-04-09 | 2021-09-07 | 宿迁学院 | Express size measuring device and method |
| CN113610929A (en) * | 2021-08-09 | 2021-11-05 | 西安外事学院 | Camera and multi-line laser combined calibration method |
| CN114111576A (en) * | 2021-11-24 | 2022-03-01 | 易思维（杭州）科技有限公司 | Aircraft skin clearance surface difference detection method and sensor |
| CN114234808A (en) * | 2021-12-17 | 2022-03-25 | 湖南大学 | Size measuring method and device for deformation area of rotary magnetic pulse crimping part |
| WO2024069513A1 (en) * | 2022-09-30 | 2024-04-04 | Salvagnini Italia S.P.A. | Machine and method for working and/or moving metal plates or sheets comprising edge recognition means |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| FR2570489A1 (en) * | 1984-09-17 | 1986-03-21 | Tech Nles Ste Gle | Method and device for monitoring the volumes of glass after cutting |
| JP2006131168A (en) * | 2004-11-09 | 2006-05-25 | Mitsubishi Heavy Ind Ltd | Wear volume measuring device |
| CN101571373A (en) * | 2009-06-11 | 2009-11-04 | 天津大学 | Method for measuring geometric parameters of profile steel based on multi-linear structured light vision sensors |
| CN107578464A (en) * | 2017-06-30 | 2018-01-12 | 长沙湘计海盾科技有限公司 | A kind of conveyor belt workpieces measuring three-dimensional profile method based on line laser structured light |
| CN107655536A (en) * | 2017-10-24 | 2018-02-02 | 浙江华睿科技有限公司 | A kind of object volume measuring system and method |
| CN108180825A (en) * | 2016-12-08 | 2018-06-19 | 中国科学院沈阳自动化研究所 | A kind of identification of cuboid object dimensional and localization method based on line-structured light |
| CN109000559A (en) * | 2018-06-11 | 2018-12-14 | 广东工业大学 | A kind of measurement method of object volume, device, system and readable storage medium storing program for executing |
| CN109029618A (en) * | 2018-07-11 | 2018-12-18 | 苏州科技大学 | Monocular vision packing case volume measuring method |
-
2019
- 2019-04-18 CN CN201910312179.7A patent/CN110068270B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| FR2570489A1 (en) * | 1984-09-17 | 1986-03-21 | Tech Nles Ste Gle | Method and device for monitoring the volumes of glass after cutting |
| JP2006131168A (en) * | 2004-11-09 | 2006-05-25 | Mitsubishi Heavy Ind Ltd | Wear volume measuring device |
| CN101571373A (en) * | 2009-06-11 | 2009-11-04 | 天津大学 | Method for measuring geometric parameters of profile steel based on multi-linear structured light vision sensors |
| CN108180825A (en) * | 2016-12-08 | 2018-06-19 | 中国科学院沈阳自动化研究所 | A kind of identification of cuboid object dimensional and localization method based on line-structured light |
| CN107578464A (en) * | 2017-06-30 | 2018-01-12 | 长沙湘计海盾科技有限公司 | A kind of conveyor belt workpieces measuring three-dimensional profile method based on line laser structured light |
| CN107655536A (en) * | 2017-10-24 | 2018-02-02 | 浙江华睿科技有限公司 | A kind of object volume measuring system and method |
| CN109000559A (en) * | 2018-06-11 | 2018-12-14 | 广东工业大学 | A kind of measurement method of object volume, device, system and readable storage medium storing program for executing |
| CN109029618A (en) * | 2018-07-11 | 2018-12-18 | 苏州科技大学 | Monocular vision packing case volume measuring method |
Non-Patent Citations (1)
| Title |
|---|
| 张森等: "圆形煤场煤堆体积快速测量方法", 《中国计量大学学报》 * |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111402330A (en) * | 2020-04-03 | 2020-07-10 | 山东省科学院激光研究所 | Laser line key point extraction method based on plane target |
| CN111402330B (en) * | 2020-04-03 | 2024-02-27 | 山东省科学院激光研究所 | Laser line key point extraction method based on planar target |
| CN111426280A (en) * | 2020-05-08 | 2020-07-17 | 中国科学技术大学 | Two-dimensional DIC optical extensometer out-of-plane compensation device and method based on structured light |
| CN111598939A (en) * | 2020-05-22 | 2020-08-28 | 中原工学院 | Human body circumference measuring method based on multi-vision system |
| CN111598939B (en) * | 2020-05-22 | 2021-01-26 | 中原工学院 | Human body circumference measuring method based on multi-vision system |
| CN111610173A (en) * | 2020-05-27 | 2020-09-01 | 中国水利水电科学研究院 | Three-dimensional fluid concentration field calibration device and calibration method |
| CN111932494B (en) * | 2020-07-01 | 2022-06-17 | 湖南省鹰眼在线电子科技有限公司 | Tire wear degree evaluation method and device |
| CN111932494A (en) * | 2020-07-01 | 2020-11-13 | 湖南省鹰眼在线电子科技有限公司 | Tire wear degree evaluation method and device |
| CN112781470A (en) * | 2020-12-24 | 2021-05-11 | 中广核核电运营有限公司 | U-shaped groove multi-dimensional gap measuring method and system |
| CN112964185A (en) * | 2021-01-20 | 2021-06-15 | 广西电网有限责任公司电力科学研究院 | Power transmission line crossing point on-line monitoring system based on binocular stereo vision |
| CN113358052A (en) * | 2021-04-09 | 2021-09-07 | 宿迁学院 | Express size measuring device and method |
| CN113137920B (en) * | 2021-05-19 | 2022-09-23 | 重庆大学 | Underwater measurement equipment and underwater measurement method |
| CN113137920A (en) * | 2021-05-19 | 2021-07-20 | 重庆大学 | Underwater measurement equipment and underwater measurement method |
| CN113222986A (en) * | 2021-06-04 | 2021-08-06 | 中冶宝钢技术服务有限公司 | Continuous casting billet angular point and edge contour point set positioning method, system, medium and device |
| CN113222986B (en) * | 2021-06-04 | 2022-08-09 | 中冶宝钢技术服务有限公司 | Continuous casting billet angular point and edge contour point set positioning method, system, medium and device |
| CN113256738A (en) * | 2021-06-23 | 2021-08-13 | 苏州大学 | Binocular-based parcel volume measurement method |
| CN113256738B (en) * | 2021-06-23 | 2023-01-17 | 苏州大学 | Binocular-based parcel volume measurement method |
| CN113610929A (en) * | 2021-08-09 | 2021-11-05 | 西安外事学院 | Camera and multi-line laser combined calibration method |
| CN113610929B (en) * | 2021-08-09 | 2023-08-18 | 西安外事学院 | Combined calibration method of camera and multi-line laser |
| CN114111576A (en) * | 2021-11-24 | 2022-03-01 | 易思维（杭州）科技有限公司 | Aircraft skin clearance surface difference detection method and sensor |
| CN114234808A (en) * | 2021-12-17 | 2022-03-25 | 湖南大学 | Size measuring method and device for deformation area of rotary magnetic pulse crimping part |
| CN114234808B (en) * | 2021-12-17 | 2022-10-28 | 湖南大学 | Size measuring method and device for deformation area of rotary magnetic pulse crimping part |
| WO2024069513A1 (en) * | 2022-09-30 | 2024-04-04 | Salvagnini Italia S.P.A. | Machine and method for working and/or moving metal plates or sheets comprising edge recognition means |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110068270B (en) | 2021-04-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110068270A (en) | A kind of monocular vision box volume measurement method based on multi-line structured light image recognition | |
| CN110506297B (en) | High accuracy calibration system and method | |
| CN106599897B (en) | Readings of pointer type meters recognition methods and device based on machine vision | |
| CN110118528B (en) | Line structure light calibration method based on chessboard target | |
| CN110956660B (en) | Positioning method, robot, and computer storage medium | |
| CN104376558B (en) | Cuboid-based intrinsic parameter calibration method for Kinect depth camera | |
| CN103196370B (en) | Measuring method and measuring device of conduit connector space pose parameters | |
| CN108305286B (en) | Color coding-based multi-view stereoscopic vision foot type three-dimensional measurement method, system and medium | |
| CN103411553A (en) | Fast calibration method of multiple line structured light visual sensor | |
| CN108734744A (en) | A kind of remote big field-of-view binocular scaling method based on total powerstation | |
| CN107025663A (en) | It is used for clutter points-scoring system and method that 3D point cloud is matched in vision system | |
| CN110443879B (en) | Perspective error compensation method based on neural network | |
| CN110163918A (en) | A kind of line-structured light scaling method based on projective geometry | |
| CN104766309A (en) | Plane feature point navigation and positioning method and device | |
| CN104036542A (en) | Spatial light clustering-based image surface feature point matching method | |
| EP3706076A1 (en) | Method and device to determine the dimensions and distance of a number of objects in an environment | |
| CN109255818A (en) | A kind of extracting method of novel target and its sub-pixel angle point | |
| CN112001973B (en) | Quick three-dimensional human head measuring method based on digital speckle correlation | |
| CN114219855A (en) | Point cloud normal vector estimation method and device, computer equipment and storage medium | |
| CN101561269B (en) | Method for automatically matching characteristic lines of close-range photogrammetry | |
| CN103198481A (en) | Camera calibration method and achieving system of same | |
| Tushev et al. | Architecture of industrial close-range photogrammetric system with multi-functional coded targets | |
| CN102930551A (en) | Camera intrinsic parameters determined by utilizing projected coordinate and epipolar line of centres of circles | |
| CN110838146A (en) | Homonymy point matching method, system, device and medium for coplanar cross-ratio constraint | |
| CN108898629B (en) | Projection coding method for enhancing aerial luggage surface texture in three-dimensional modeling |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |