CN107239748A - Robot target identification and localization method based on gridiron pattern calibration technique - Google Patents
Robot target identification and localization method based on gridiron pattern calibration technique Download PDFInfo
- Publication number
- CN107239748A CN107239748A CN201710342229.7A CN201710342229A CN107239748A CN 107239748 A CN107239748 A CN 107239748A CN 201710342229 A CN201710342229 A CN 201710342229A CN 107239748 A CN107239748 A CN 107239748A
- Authority
- CN
- China
- Prior art keywords
- mtd
- mtr
- image
- msub
- mrow
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/20—Scenes; Scene-specific elements in augmented reality scenes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/80—Analysis of captured images to determine intrinsic or extrinsic camera parameters, i.e. camera calibration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
Abstract
The invention discloses the identification of the robot target based on gridiron pattern calibration technique and localization method, the localization method is divided into target identification and target positions two parts, wherein, target identification extracts effective target by the way that robot acquired image is analyzed and handled；Gridiron pattern calibration technique is applied in target positioning, and precise positioning goes out position of the effective target on court.Target identification of the present invention and localization method, robot can accurately recognize target under environment complicated and changeable and carry out precise positioning to it, do not influenceed by factors such as intensity of illumination, noise and the dressings of outside audience at scene, drastically increase the target identification and setting accuracy of robot.
Description
Technical field
The present invention relates to robot target identification and localization method, and in particular to the robot based on gridiron pattern calibration technique
Target identification and localization method, belong to object recognition and detection technical field.
Background technology
The research of Soccer robot is developed rapidly in recent years so that the problem of this multi-crossed disciplines is by more
Carry out more concerns.There is provided more simplify as the prediction project that RoboCup SPL compete for the match of NAO robotic golfs
But highly important research platform.In NAO robots golf project, the real time environment of golf course is complicated and changeable,
Intensity of illumination, noise and the dressing of outside audience at scene etc. all can cause significant impact to the normal work of NAO robots.
How robot, which can accurately recognize target under environment complicated and changeable and carry out precise positioning to it, seems extremely important.
The content of the invention
The technical problems to be solved by the invention are：The robot target identification based on gridiron pattern calibration technique is provided with determining
Position method, greatly improves the target identification and positioning precision of robot.
The present invention uses following technical scheme to solve above-mentioned technical problem：
Robot target identification and localization method based on gridiron pattern calibration technique, comprise the following steps：
Step 1, two camera cameras above and below robot head setting, two camera cameras are located at same straight line
On, gather target object image scene using any one camera camera；
Step 2, in target object institute at the scene, target object color characteristic is obtained by the way of scene obtains rgb value
Threshold value, and determine by off-line training the threshold range of target object color characteristic；
Step 3, target object image scene is split using the partitioning algorithm based on threshold value, extracts effective target special
Levy area image；
Step 4, preliminary treatment is carried out to effective target feature regional images using median filtering algorithm；
Step 5, the image after step 4 preliminary treatment is divided into several size identical block of pixels, to each picture
Pixel in plain block is scanned from left to right, from top to bottom, judges whether the rgb value of the pixel obtains in step 2
In threshold range, if, then it is assumed that the color of the pixel is target object color characteristic, obtains whole pixels of target object
Point information；
Step 6, smooth, the image after obtaining smoothly is carried out to the image obtained through step 5 using gaussian filtering method；
Step 7, the image after smooth under RGB color is converted to the image under hsv color space；
Step 8, rim detection is carried out to the image under hsv color space using Canny edge detection algorithms, be identified
Effective target image；
Step 9, using the upper left corner of effective target image as the origin of coordinates, on the upper left corner of the image two intersecting sides point
Not as u axles and v axles, the pixel coordinate system of effective target image is set up；
Step 10, the pixel coordinate system of effective target image is changed to world coordinate system, according to pixel coordinate system and generation
Transformational relation between boundary's coordinate system calculates the intrinsic parameter and outer parameter of camera, the final positional information for obtaining target object.
As a preferred embodiment of the present invention, in Canny edge detection algorithms described in step 8, the gradient width of pixel
Value is with angle calculation formula：
θ (x, y)=tan-1(GY(x,y)/GX(x, y)),
Wherein, G (x, y) is the gradient magnitude of pixel (x, y), and θ (x, y) is the angle of pixel (x, y), GX(x,y)
For the gradient of pixel (x, y) in the X direction, GY(x, y) is the gradient of pixel (x, y) in the Y direction.
As a preferred embodiment of the present invention, the pixel coordinate system of effective target image is changed to generation described in step 10
Boundary's coordinate system detailed process is as follows：
1) pixel coordinate system of effective target image is changed to image coordinate system；
2) image coordinate system is changed to camera coordinates system；
3) camera coordinates system is changed to world coordinate system.
As a preferred embodiment of the present invention, the pixel coordinate system, which is changed to the conversion formula of image coordinate system, is：
Wherein, x, y are respectively that image coordinate fastens coordinate points in x-axis, the value of y-axis, and u, v are respectively that pixel coordinate fastens seat
Punctuate is in the value of u axles, v axles, u0、v0Respectively value of the pixel coordinate system origin in u axles, v axles.
As a preferred embodiment of the present invention, described image coordinate system, which is changed to the conversion formula of camera coordinates system, is：
Wherein, x, y are respectively that image coordinate fastens coordinate points in x-axis, the value of y-axis, and f is camera coordinates system origin and image
The distance between coordinate origin, xc、yc、zcRespectively camera coordinates fasten coordinate points in xcAxle, ycAxle, zcThe value of axle.
As a preferred embodiment of the present invention, the camera coordinates system, which is changed to the conversion formula of world coordinate system, is：
Wherein, xc、yc、zcRespectively camera coordinates fasten coordinate points in xcAxle, ycAxle, zcThe value of axle, R is spin matrix, t
For translation matrix, xw、yw、zwCoordinate points are fastened in x for world coordinateswAxle, ywAxle, zwThe value of axle.
The present invention uses above technical scheme compared with prior art, with following technique effect：
Target identification of the present invention and localization method, robot can accurately recognize target and right under environment complicated and changeable
It carries out precise positioning, is not influenceed, greatly carried by factors such as intensity of illumination, noise and the dressings of outside audience at scene
The target identification and setting accuracy of Gao Liao robots.
Brief description of the drawings
Fig. 1 is NAO robots camera schematic diagram of the present invention, wherein, (a) is side view, and (b) is rearview.
Fig. 2 is that picture is taken the photograph by NAO robots of the present invention.
Fig. 3 is RGB color illustraton of model.
Fig. 4 is hsv color spatial model figure.
Fig. 5 is the image under hsv color space.
Fig. 6 is the effective target that Canny edge detection algorithms are obtained.
Fig. 7 is the effective target image that target identification is obtained.
Fig. 8 is the geometrical relationship figure of 4 coordinate system imagings.
Fig. 9 is the mapping between image coordinate system and camera coordinates system.
Figure 10 (a), (b) is the demarcation picture of different angle shots respectively.
Embodiment
Embodiments of the present invention are described below in detail, the example of the embodiment is shown in the drawings.Below by
The embodiment being described with reference to the drawings is exemplary, is only used for explaining the present invention, and is not construed as limiting the claims.
The present invention mainly introduces a kind of target identification and precise positioning method used in game of golf, is discussed in detail
Target identification module and target locating module in this method, target identification that this method greatly improves robot are accurate
Degree.Localization method proposed by the present invention mainly has target identification module to be cooperated with target locating module completion.Target identification module
Effective target is extracted, the final precise positioning of combining target locating module goes out the accurate location of effective target.
NAO robots mainly use its vision system to recognize target object while perceiving ball in game of golf
Field environment.Two cameras are arranged vertically above and below robot head, it is possible to provide resolution ratio is 640*480 YUV422 images,
And the frame speed of 30 frame per second may insure the real-time of image, shown in such as Fig. 1 (a) and (b).In the localization method, target
Identification module extracts effective target by the way that robot acquired image is analyzed and handled, and is positioned in conjunction with target
Module, using gridiron pattern calibration technique, precise positioning goes out position of the effective target on court.
1st, target identification
The picture that robot is gathered to itself carries out appropriate analysis and processing, and the process for obtaining effective target is target
Being completed needed for identification module for task.Its detailed operation flow is as follows：
The pretreatment of 1.1 images
The camera of NAO robots has certain limitation, during the games, and acquired image information exists certain
Noise and distortion.The accuracy of these uncertain factor strong influences effective target identification, it is to obtain higher-quality
Image information to these noises and distortion, it is necessary to carry out denoising and correcting process, image preprocessing is mainly used in suppressing useless letter
Breath, strengthens useful information and improves picture quality.Preconditioning technique used in the present invention is described below：
1) determination of color threshold
Under the conditions of different intensities of illumination and different image-forming ranges, the color of object can have certain deviation, obtain
Metastable color threshold is obtained, is the key for obtaining effective target.In order to save robot pre-games debug time, the present invention is adopted
The mode that rgb value is obtained with scene determines target object color to obtain the color threshold of target object, then by off-line training
Feature is with respect to stable threshold.Simplify the calculation process of threshold value, improve processing speed, greatly promote match performance.
2) image is split
Robot, which was collected, utilizes image Segmentation Technology extraction effective target characteristic area after image information.In NAO robots
In vision system, characteristic area is effective target in court, and the present invention is carried effective target using the partitioning algorithm based on threshold value
Take out, and realize follow-up detailed identification work on this basis.
3) picture noise is handled
A large amount of salt-pepper noises occur in image information after splitting through image, and salt-pepper noise is brought to the accurate processing of image
It is many difficult, directly affect feature extraction and image recognition.What the present invention was produced after being split using median filtering algorithm to image
Spiced salt noise carries out preliminary treatment.
The identification of 1.2NAO effective targets
In pre-games debug time, NAO robots have obtained the color threshold of court effective target, with golf yellow flag
Exemplified by bar, determine to meet the yellow pixel point in the threshold range obtained in pretreatment in the pictorial information that robot is obtained.
NAO recognizes that the main thought of flagpole is：Yellow target area is found on court, and by a series of image analysis processing,
If the region highlighted is rectangle, and this rectangle is identical with the flagpole ratio of artwork, then it is assumed that have identified target flagpole.
For 640 × 480 original image information (as shown in Figure 2) acquired in NAO robots camera, following image point is carried out
Analysis is handled, to recognize effective target.
1) picture pixels point information is read
It is the square pixels block that image is divided into 32 × 24 by 20 pixel grids with width.In RGB color, by background
Pixel syntax values are set to (0,0,0), and yellow pixel point grammer codomain is set to (15,120,120) to (30,255,255).
The pictorial information of acquisition is reconstructed, composition standard two-dimensional counts array count [32] [24], 32*24 two-dimemsional number
Group just represents a block of pixels, and array element just represents pixel.After image preprocessing, to every a line, each pixel is clicked through
Row scanning, when meeting the yellow threshold value of real racetrack place acquisition, it is a yellow pixel point to be considered as the pixel, works as figure
After scanned, that is, whole pixel information of effective target are obtained, judge the region as yellow flag by NAO robots
Bar.
2) gaussian filtering smoothed image
Gaussian filtering is a kind of linear smoothing filtering, is widely used in the noise abatement process in image procossing.In simple terms, Gauss
Filtering is exactly that average process is weighted to entire image, highly effective to the noise of suppression Normal Distribution.Specific behaviour
Work is that the weighted average gray value of pixel in each pixel with a convolution scan image, the field determined with convolution goes to replace
For the value of convolution central pixel point.
3) color space conversion
RGB color and hsv color space are all the widely used color systems of current image procossing.RGB is by red
The letter abbreviations of color, green and blue three primary colours, by the different degrees of superposition of three primary colours, to produce various different face
Color, can cover all colours that human eyesight can perceive, RGB color illustraton of model as shown in figure 3, the model is image
In the RGB of each pixel distribute 0-255 gray value, therefore, the full color of RGB image has 16581375 kinds of face
Color.The pixel color value of RGB color can be expressed as (Red, Green, Blue), and wherein white is (255,255,255),
Black is (0,0,0), and yellow is (111,111,111).
HSV (Hue, Saturation, Value) color space is a kind of color sky created according to the intuitive nature of color
Between, illustraton of model is as shown in Figure 4.H parameters represent the position of color information, i.e. residing spectral color, one angular metric of the parameter
To represent, red, green, blue is separated by 120 degree respectively, and complementary colours differs 180 degree respectively.Purity S be a ratio value, scope from 0 to 1,
It is expressed as the ratio between the purity of selected color and the maximum purity of the color, there was only gray scale during S=0.V represents color
Light levels, scope is from 0 to 1.Conical tip is meaningless corresponding to V=0, H and S value, and this point represents black, circular conical surface center
Place, S=0, V=1, H values are meaningless, represent white.
In real racetrack, intensity of illumination is inevitable interference factor, and RGB color is influenceed by intensity of illumination
It is inadequate to the resolution of object than larger, it could even be possible to causing robot " blindness ".Abundant experimental results show, for same
The object of one color, under the irradiation of different illumination intensity or different light source, its RGB color Distribution value it is very discrete, this
So that RGB color threshold is difficult to determine, it is easy to which noise is included, or recognized object is lost, eventually led
Cause during application match, the target Loss brought by the uneven illumination in place.And hsv color space can't be with
The change of intensity of illumination and change, influence of the illumination condition to robot vision is reduced to a certain extent, machine is enhanced
The adaptive ability of the vision system of device people.The present invention carries out colour recognition and processing using hsv color space, and Fig. 5 is conversion
For the image behind hsv color space.
4) Canny rim detections
Canny edge detection algorithms are to be calculated by John F.Canny one kind developed based on image gradient for 1986
Edge detection algorithm, be one of method for detecting image edge classic algorithm.Classical Canny edge detection algorithms are generally all
Since Gaussian smoothing, realize that edge connection terminates to based on dual threshold.Gaussian smoothing is mainly for reducing picture noise, favorably
In the gradient and edge amplitude that more accurately calculate image.The present invention uses 2*2 Sobel operators, and its mathematic(al) representation is as follows：
Gx(x,y)≈[S(x,y+1)-S(x+1,y+1)-S(x+1,y)]/2 (1)
Gy(x,y)≈[S(x,y)-S(x+1,y)-S(x+1,y+1)]/2 (2)
Wherein Gx(x, y) is the gradient in X-direction, Gy(x, y) is the gradient in Y-direction.Can according to the gradient of X and Y-direction
To calculate gradient magnitude and angle of the image in the pixel：
θ (x, y)=tan-1(Gy(x,y)/Gx(x,y)) (4)
Wherein G (x, y) is the size of gradient magnitude, the i.e. gray scale, and θ (x, y) is the angle of the point.According to anti-triangle letter
Number angle values scope beFor the ease of calculating, added on angle valueSo that angle value scope is at 0 °
Between~180 °.
After the edge amplitude and angle of each pixel of image is obtained, non-peak signal compacting is carried out.Main purpose
It is to realize edge thinning, is handled by the step, edge pixel is further reduced.After non-peak signal compacting, still have a small amount of
Non-edge pixels be incorporated into result, so will be by taking threshold value to be accepted or rejected.Canny proposes to be based on dual threshold (Fuzzy
Threshold) method realizes edge selection well, in actual applications, and dual threshold also has the effect that edge is connected.Dual threashold
Value selection with edge connection method by assuming that two threshold values one of them be high threshold TH, another is Low threshold TL, then
Have：
A. it is less than TL then discarding for any edge pixel；
B. it is higher than TH then reservation for any edge pixel；
C. for any edge pixel values between TL and TH, if a pixel can be connected to by edge more than TH
And edge all pixels are more than minimum threshold TL then reservation, otherwise abandon.
As a result display such as Fig. 6.The effective target image finally recognized is shown in Fig. 7.
2nd, target is positioned
NAO robots are after target object is recognized accurately, the method precise positioning target object demarcated using grid, greatly
Improve robot accuracy greatly.From after coming out one after another Tsai and Zhang classic paper, camera calibration be considered as one into
Ripe technology, the object in space is reduced using the image of shot by camera.Based on this, NAO robots are by using head
The corresponding tessellated situation that portion's video camera is shot is accurately positioned out the position data of target object.The algorithm of gridiron pattern standardization
It is described as：
1) print a template and be affixed on ground；2) shoot from different perspectives template photo several；3) detect in image
Characteristic point；4) intrinsic parameter and outer parameter of camera are obtained.
By such process, obtain and join outside high-precision 4 internal references and 6, using these information, final realization is three-dimensional
Information recovering, reaches the purpose of precise positioning.
The geometrical model of 2.1 camera imagings
Image shot by camera is actually the process of an optical imagery.This process can be divided into 3 steps, be under the jurisdiction of 4
Individual coordinate system, this 3 steps connect the picture pixels point coordinates of shooting and actual locus coordinate.Four coordinates
System is respectively：
1) pixel coordinate system：The origin of coordinates is located at the upper left corner for taking the photograph image, and u axles and v axles are respectively parallel to the plane of delineation
Two vertical edges.Coordinate value is represented with (u, v), is discrete integer value.
2) image coordinate system：The origin of coordinates is located at the centre for taking the photograph image, and x-axis and y-axis are respectively parallel to pixel coordinate system
U axles and v axles, coordinate value with (x, y) represent.
3) camera coordinates system (photocentre coordinate system)：The origin of coordinates is the photocentre of camera, xcAxle, ycAxle is respectively parallel to image
The x-axis and y-axis of coordinate system, the optical axis of camera is zcAxle, coordinate value (xc,yc,zc) represent.
4) world coordinate system：The coordinate system that robot is selected according to natural environment, coordinate value (xw,yw,zw) represent.
3 steps are：
1) information in pixel coordinate system is changed to image coordinate system respectively；
2) image coordinate system is changed to camera coordinates system；
3) camera coordinates system is changed to world coordinate system.
The demarcation of camera first has to be chosen to the geometrical model of picture, so that it is determined that inside and outside parameter, finally obtains target location
Coordinate information.In order to describe imaging geometry model that this patent used, it is necessary to by above-mentioned 4 coordinate systems.What is be imaged is several
What relation is shown in Fig. 8.
4 coordinate system transformational relations are：
1) pixel coordinate system (u, v) is changed to image coordinate system (x, y)
The central point for choosing the plane of delineation is O1, O1Coordinate value in pixel coordinate system is (u0,v0), according to coordinate system
Conversion understands O1Coordinate value of the point in image coordinate system beWherein dx, dy represent that each pixel exists respectively
Physical size in x-axis and y-axis.Then the homogeneous coordinates transformational relation for obtaining pixel planes and the plane of delineation is：
2) image coordinate system (x, y) is changed to camera coordinates system (xc,yc,zc)
Known according to the property of camera, camera coordinates system origin O and image coordinate system origin O1Line OO1As camera
Focal length f, the Intrinsic Matrix of cameraPoint Pc(xc,yc,zc) project in image coordinate system be P points, by
Similar triangle theory is obtained：
Mapping relations between image coordinate system and camera coordinates system are as shown in figure 9, event transformational relation is：
3) camera coordinates system (xc,yc,zc) change to world coordinate system (xw,yw,zw)
Therebetween conversion this to be related to two kinds of conversion：Translation and rotation.Specific transformational relation is as follows, wherein R
=[r1,r2,r3] be 3 × 3 spin matrix, t represents the translation matrix of 3-dimensional column vector.
Joint (1) (2) (3) obtains the transformational relation between pixel coordinate system and world coordinate system, has：
Demarcation thing is plane again, world coordinates series structure in zwIt is i.e. available in=0 plane：
The change is the singly change of reflecting property, i.e., one plane to the mapping of another plane, square defined in computer vision
Battle array H=[h1 h2 h3]=A [r1 r2T] it is referred to as singly reflecting property matrix.
The calibration algorithm of 2.2 inside and outside parameters
(xw,yw) for the coordinate of demarcation thing, can be known quantity by designer's manual control, (u, v) is pixel coordinate, can be with
Directly obtained by video camera.For one group of corresponding (xw,yw) → (u, v), by equation [h1 h2 h3]=A [r1 r2T], knot
The property of spin matrix is closed, two constraintss can be obtained.Therefore 4 characteristic points of detection, 8 equations can be obtained.
Internal reference matrix A containsu0, v0Spin matrix R in four unknown quantitys, outer ginseng matrix containsThree
Individual unknown quantity, translation matrix t contains xw,yw,zwThree unknown quantitys.By changing the relative position between video camera and demarcation thing
Two photos are obtained, and internal reference matrix immobilizes, outer ginseng matrix changes with the change of positional information, therefore can produce
16 unknown quantitys.Equally, two photos can obtain 8 characteristic points, 16 equations.6*2+4=8*2, is thus solved all unknown
Amount, completes the demarcation of inside and outside parameter, the final positional information for obtaining target object.
Figure 10 (a) and (b) is respectively the difference captured by present invention application camera calibration technology completion precise positioning
The picture of angle.Global calibration parametric solution the results are shown in Table 1 and table 2.
The inner parameter of the global calibration of table 1
The external parameter of table 2 (a) global calibration
The external parameter of table 2 (b) global calibration
The technological thought of above example only to illustrate the invention, it is impossible to which protection scope of the present invention is limited with this, it is every
According to technological thought proposed by the present invention, any change done on the basis of technical scheme each falls within the scope of the present invention
Within.
Claims (6)
1. robot target identification and localization method based on gridiron pattern calibration technique, it is characterised in that comprise the following steps：
Step 1, two camera cameras above and below robot head setting, two camera cameras are located on the same line,
Target object image scene is gathered using any one camera camera；
Step 2, in target object institute at the scene, the threshold of target object color characteristic is obtained by the way of scene obtains rgb value
Value, and determine by off-line training the threshold range of target object color characteristic；
Step 3, target object image scene is split using the partitioning algorithm based on threshold value, extracts effective target characteristic area
Area image；
Step 4, preliminary treatment is carried out to effective target feature regional images using median filtering algorithm；
Step 5, the image after step 4 preliminary treatment is divided into several size identical block of pixels, to each block of pixels
In pixel be scanned from left to right, from top to bottom, judge the pixel rgb value whether the threshold value obtained in step 2
In the range of, if, then it is assumed that the color of the pixel is target object color characteristic, obtains whole pixels letter of target object
Breath；
Step 6, smooth, the image after obtaining smoothly is carried out to the image obtained through step 5 using gaussian filtering method；
Step 7, the image after smooth under RGB color is converted to the image under hsv color space；
Step 8, rim detection is carried out to the image under hsv color space using Canny edge detection algorithms, what is be identified has
Imitate target image；
Step 9, using the upper left corner of effective target image as the origin of coordinates, in the upper left corner of the image, two intersecting sides are made respectively
For u axles and v axles, the pixel coordinate system of effective target image is set up；
Step 10, the pixel coordinate system of effective target image is changed to world coordinate system, sat according to pixel coordinate system and the world
Transformational relation between mark system calculates the intrinsic parameter and outer parameter of camera, the final positional information for obtaining target object.
2. the robot target based on gridiron pattern calibration technique is recognized and localization method according to claim 1, its feature exists
In in Canny edge detection algorithms described in step 8, gradient magnitude and the angle calculation formula of pixel are：
<mrow>
<mi>G</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<msqrt>
<mrow>
<msubsup>
<mi>G</mi>
<mi>X</mi>
<mn>2</mn>
</msubsup>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
<mo>)</mo>
</mrow>
<mo>+</mo>
<msubsup>
<mi>G</mi>
<mi>Y</mi>
<mn>2</mn>
</msubsup>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
<mo>)</mo>
</mrow>
</mrow>
</msqrt>
<mo>,</mo>
</mrow>
θ (x, y)=tan-1(GY(x,y)/GX(x, y)),
Wherein, G (x, y) is the gradient magnitude of pixel (x, y), and θ (x, y) is the angle of pixel (x, y), GX(x, y) is pixel
The gradient of point (x, y) in the X direction, GY(x, y) is the gradient of pixel (x, y) in the Y direction.
3. the robot target based on gridiron pattern calibration technique is recognized and localization method according to claim 1, its feature exists
In the pixel coordinate system of effective target image being changed described in step 10 as follows to world coordinate system detailed process：
1) pixel coordinate system of effective target image is changed to image coordinate system；
2) image coordinate system is changed to camera coordinates system；
3) camera coordinates system is changed to world coordinate system.
4. the robot target based on gridiron pattern calibration technique is recognized and localization method according to claim 3, its feature exists
In the pixel coordinate system, which is changed to the conversion formula of image coordinate system, is：
<mrow>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<mi>x</mi>
</mtd>
</mtr>
<mtr>
<mtd>
<mi>y</mi>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>=</mo>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<mrow>
<mi>d</mi>
<mi>x</mi>
</mrow>
</mtd>
<mtd>
<mn>0</mn>
</mtd>
<mtd>
<mrow>
<mo>-</mo>
<msub>
<mi>u</mi>
<mn>0</mn>
</msub>
<mi>d</mi>
<mi>x</mi>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>0</mn>
</mtd>
<mtd>
<mrow>
<mi>d</mi>
<mi>y</mi>
</mrow>
</mtd>
<mtd>
<mrow>
<mo>-</mo>
<msub>
<mi>v</mi>
<mn>0</mn>
</msub>
<mi>d</mi>
<mi>y</mi>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>0</mn>
</mtd>
<mtd>
<mn>0</mn>
</mtd>
<mtd>
<mn>1</mn>
</mtd>
</mtr>
</mtable>
</mfenced>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<mi>u</mi>
</mtd>
</mtr>
<mtr>
<mtd>
<mi>v</mi>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
Wherein, x, y are respectively that image coordinate fastens coordinate points in x-axis, the value of y-axis, and u, v are respectively that pixel coordinate fastens coordinate points
In the value of u axles, v axles, u0、v0Respectively value of the pixel coordinate system origin in u axles, v axles.
5. the robot target based on gridiron pattern calibration technique is recognized and localization method according to claim 3, its feature exists
In described image coordinate system, which is changed to the conversion formula of camera coordinates system, is：
<mrow>
<msub>
<mi>z</mi>
<mi>c</mi>
</msub>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<mi>x</mi>
</mtd>
</mtr>
<mtr>
<mtd>
<mi>y</mi>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>=</mo>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<mi>f</mi>
</mtd>
<mtd>
<mn>0</mn>
</mtd>
<mtd>
<mn>0</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>0</mn>
</mtd>
<mtd>
<mi>f</mi>
</mtd>
<mtd>
<mn>0</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>0</mn>
</mtd>
<mtd>
<mn>0</mn>
</mtd>
<mtd>
<mn>1</mn>
</mtd>
</mtr>
</mtable>
</mfenced>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<msub>
<mi>x</mi>
<mi>c</mi>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>y</mi>
<mi>c</mi>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>z</mi>
<mi>c</mi>
</msub>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
Wherein, x, y are respectively that image coordinate fastens coordinate points in x-axis, the value of y-axis, and f is camera coordinates system origin and image coordinate
It is the distance between origin, xc、yc、zcRespectively camera coordinates fasten coordinate points in xcAxle, ycAxle, zcThe value of axle.
6. the robot target based on gridiron pattern calibration technique is recognized and localization method according to claim 3, its feature exists
In the camera coordinates system, which is changed to the conversion formula of world coordinate system, is：
<mrow>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<msub>
<mi>x</mi>
<mi>c</mi>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>y</mi>
<mi>c</mi>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>z</mi>
<mi>c</mi>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>=</mo>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<mi>R</mi>
</mtd>
<mtd>
<mi>t</mi>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>0</mn>
</mtd>
<mtd>
<mn>1</mn>
</mtd>
</mtr>
</mtable>
</mfenced>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<msub>
<mi>x</mi>
<mi>w</mi>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>y</mi>
<mi>w</mi>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>z</mi>
<mi>w</mi>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
Wherein, xc、yc、zcRespectively camera coordinates fasten coordinate points in xcAxle, ycAxle, zcThe value of axle, R is spin matrix, and t is flat
Move matrix, xw、yw、zwCoordinate points are fastened in x for world coordinateswAxle, ywAxle, zwThe value of axle.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710342229.7A CN107239748A (en) | 2017-05-16 | 2017-05-16 | Robot target identification and localization method based on gridiron pattern calibration technique |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710342229.7A CN107239748A (en) | 2017-05-16 | 2017-05-16 | Robot target identification and localization method based on gridiron pattern calibration technique |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107239748A true CN107239748A (en) | 2017-10-10 |
Family
ID=59985138
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710342229.7A Pending CN107239748A (en) | 2017-05-16 | 2017-05-16 | Robot target identification and localization method based on gridiron pattern calibration technique |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107239748A (en) |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108230401A (en) * | 2018-01-12 | 2018-06-29 | 上海鼎盛汽车检测设备有限公司 | 3D four-wheel position finder automatic camera calibration method and system |
| CN108563220A (en) * | 2018-01-29 | 2018-09-21 | 南京邮电大学 | The motion planning of apery Soccer robot |
| CN108596942A (en) * | 2018-03-21 | 2018-09-28 | 黄启萌 | A kind of system and method precisely judging ball drop point using single camera |
| CN108628808A (en) * | 2018-04-04 | 2018-10-09 | 华南农业大学 | The coordinate transformation method of camera sampled point |
| CN109087341A (en) * | 2018-06-07 | 2018-12-25 | 华南农业大学 | A kind of fusion method of short distance EO-1 hyperion camera and distance measuring sensor |
| CN109397294A (en) * | 2018-12-05 | 2019-03-01 | 南京邮电大学 | A kind of robot cooperated localization method based on BA-ABC converged communication algorithm |
| CN109472829A (en) * | 2018-09-04 | 2019-03-15 | 顺丰科技有限公司 | A kind of object positioning method, device, equipment and storage medium |
| CN109895238A (en) * | 2018-12-20 | 2019-06-18 | 中铁十四局集团房桥有限公司 | The automatic installation method and system of III type track plate pre-buried sleeve of CRTS- |
| CN110186459A (en) * | 2019-05-27 | 2019-08-30 | 深圳市海柔创新科技有限公司 | Air navigation aid, mobile vehicle and navigation system |
| CN110378970A (en) * | 2019-07-08 | 2019-10-25 | 武汉理工大学 | A kind of monocular vision deviation detecting method and device for AGV |
| WO2019206247A1 (en) * | 2018-04-27 | 2019-10-31 | Shanghai Truthvision Information Technology Co., Ltd | System and method for camera calibration |
| CN110666811A (en) * | 2019-09-26 | 2020-01-10 | 同济大学 | RoboCup standard platform group-based ball position prediction method |
| CN111652069A (en) * | 2020-05-06 | 2020-09-11 | 天津博诺智创机器人技术有限公司 | Target identification and positioning method of mobile robot |
| CN111798524A (en) * | 2020-07-14 | 2020-10-20 | 华侨大学 | Calibration system and method based on inverted low-resolution camera |
| CN111932627A (en) * | 2020-09-15 | 2020-11-13 | 蘑菇车联信息科技有限公司 | Marker drawing method and system |
| CN112033408A (en) * | 2020-08-27 | 2020-12-04 | 河海大学 | Paper-pasted object space positioning system and positioning method |
| CN112184807A (en) * | 2020-09-22 | 2021-01-05 | 深圳市衡泰信科技有限公司 | Floor type detection method and system for golf balls and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102175261A (en) * | 2011-01-10 | 2011-09-07 | 深圳大学 | Visual measuring system based on self-adapting targets and calibrating method thereof |
| US20160035079A1 (en) * | 2011-07-08 | 2016-02-04 | Restoration Robotics, Inc. | Calibration and Transformation of a Camera System's Coordinate System |
| CN106127737A (en) * | 2016-06-15 | 2016-11-16 | 王向东 | A kind of flat board calibration system in sports tournament is measured |
| CN106251337A (en) * | 2016-07-21 | 2016-12-21 | 中国人民解放军空军工程大学 | A kind of drogue space-location method and system |
-
2017
- 2017-05-16 CN CN201710342229.7A patent/CN107239748A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102175261A (en) * | 2011-01-10 | 2011-09-07 | 深圳大学 | Visual measuring system based on self-adapting targets and calibrating method thereof |
| US20160035079A1 (en) * | 2011-07-08 | 2016-02-04 | Restoration Robotics, Inc. | Calibration and Transformation of a Camera System's Coordinate System |
| CN106127737A (en) * | 2016-06-15 | 2016-11-16 | 王向东 | A kind of flat board calibration system in sports tournament is measured |
| CN106251337A (en) * | 2016-07-21 | 2016-12-21 | 中国人民解放军空军工程大学 | A kind of drogue space-location method and system |
Non-Patent Citations (2)
| Title |
|---|
| 刘琳 等: "轮廓特征与神经网络相结合的行人检测", 《光电工程》 * |
| 刘禾: "《数字图像处理及应用》", 31 December 2005, 北京：中国电力出版社 * |
Cited By (30)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108230401A (en) * | 2018-01-12 | 2018-06-29 | 上海鼎盛汽车检测设备有限公司 | 3D four-wheel position finder automatic camera calibration method and system |
| CN108563220A (en) * | 2018-01-29 | 2018-09-21 | 南京邮电大学 | The motion planning of apery Soccer robot |
| CN108596942A (en) * | 2018-03-21 | 2018-09-28 | 黄启萌 | A kind of system and method precisely judging ball drop point using single camera |
| CN108628808A (en) * | 2018-04-04 | 2018-10-09 | 华南农业大学 | The coordinate transformation method of camera sampled point |
| CN112106110A (en) * | 2018-04-27 | 2020-12-18 | 上海趋视信息科技有限公司 | System and method for calibrating camera |
| US11468598B2 (en) | 2018-04-27 | 2022-10-11 | Shanghai Truthvision Information Technology Co., Ltd. | System and method for camera calibration |
| WO2019206247A1 (en) * | 2018-04-27 | 2019-10-31 | Shanghai Truthvision Information Technology Co., Ltd | System and method for camera calibration |
| CN109087341A (en) * | 2018-06-07 | 2018-12-25 | 华南农业大学 | A kind of fusion method of short distance EO-1 hyperion camera and distance measuring sensor |
| CN109087341B (en) * | 2018-06-07 | 2022-07-05 | 华南农业大学 | Fusion method of close-range hyperspectral camera and ranging sensor |
| CN109472829A (en) * | 2018-09-04 | 2019-03-15 | 顺丰科技有限公司 | A kind of object positioning method, device, equipment and storage medium |
| CN109472829B (en) * | 2018-09-04 | 2022-10-21 | 顺丰科技有限公司 | Object positioning method, device, equipment and storage medium |
| CN109397294A (en) * | 2018-12-05 | 2019-03-01 | 南京邮电大学 | A kind of robot cooperated localization method based on BA-ABC converged communication algorithm |
| CN109895238A (en) * | 2018-12-20 | 2019-06-18 | 中铁十四局集团房桥有限公司 | The automatic installation method and system of III type track plate pre-buried sleeve of CRTS- |
| CN110186459B (en) * | 2019-05-27 | 2021-06-29 | 深圳市海柔创新科技有限公司 | Navigation method, mobile carrier and navigation system |
| CN110186459A (en) * | 2019-05-27 | 2019-08-30 | 深圳市海柔创新科技有限公司 | Air navigation aid, mobile vehicle and navigation system |
| CN110378970B (en) * | 2019-07-08 | 2023-03-10 | 武汉理工大学 | Monocular vision deviation detection method and device for AGV |
| CN110378970A (en) * | 2019-07-08 | 2019-10-25 | 武汉理工大学 | A kind of monocular vision deviation detecting method and device for AGV |
| CN110666811A (en) * | 2019-09-26 | 2020-01-10 | 同济大学 | RoboCup standard platform group-based ball position prediction method |
| CN110666811B (en) * | 2019-09-26 | 2022-08-05 | 同济大学 | RoboCup standard platform group-based ball position prediction method |
| CN111652069A (en) * | 2020-05-06 | 2020-09-11 | 天津博诺智创机器人技术有限公司 | Target identification and positioning method of mobile robot |
| CN111652069B (en) * | 2020-05-06 | 2024-02-09 | 天津博诺智创机器人技术有限公司 | Target identification and positioning method for mobile robot |
| CN111798524A (en) * | 2020-07-14 | 2020-10-20 | 华侨大学 | Calibration system and method based on inverted low-resolution camera |
| CN111798524B (en) * | 2020-07-14 | 2023-07-21 | 华侨大学 | Calibration system and method based on inverted low-resolution camera |
| CN112033408B (en) * | 2020-08-27 | 2022-09-30 | 河海大学 | Paper-pasted object space positioning system and positioning method |
| CN112033408A (en) * | 2020-08-27 | 2020-12-04 | 河海大学 | Paper-pasted object space positioning system and positioning method |
| CN111932627B (en) * | 2020-09-15 | 2021-01-05 | 蘑菇车联信息科技有限公司 | Marker drawing method and system |
| CN111932627A (en) * | 2020-09-15 | 2020-11-13 | 蘑菇车联信息科技有限公司 | Marker drawing method and system |
| CN112184807A (en) * | 2020-09-22 | 2021-01-05 | 深圳市衡泰信科技有限公司 | Floor type detection method and system for golf balls and storage medium |
| WO2022062153A1 (en) * | 2020-09-22 | 2022-03-31 | 深圳市衡泰信科技有限公司 | Golf ball floor type detection method, system, and storage medium |
| CN112184807B (en) * | 2020-09-22 | 2023-10-03 | 深圳市衡泰信科技有限公司 | Golf ball floor type detection method, system and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107239748A (en) | Robot target identification and localization method based on gridiron pattern calibration technique | |
| CN105894499B (en) | A kind of space object three-dimensional information rapid detection method based on binocular vision | |
| CN107203973B (en) | Sub-pixel positioning method for center line laser of three-dimensional laser scanning system | |
| CN104091324B (en) | Quick checkerboard image feature matching algorithm based on connected domain segmentation | |
| CN110232389B (en) | Stereoscopic vision navigation method based on invariance of green crop feature extraction | |
| CN104835164B (en) | A kind of processing method and processing device of binocular camera depth image | |
| CN107907048A (en) | A kind of binocular stereo vision method for three-dimensional measurement based on line-structured light scanning | |
| CN106123772B (en) | A kind of nuclear fuel rod pose automatic identification equipment and method | |
| CN109961485A (en) | A method of target positioning is carried out based on monocular vision | |
| CN108731587A (en) | A kind of the unmanned plane dynamic target tracking and localization method of view-based access control model | |
| CN110033407B (en) | Shield tunnel surface image calibration method, splicing method and splicing system | |
| CN109523551B (en) | Method and system for acquiring walking posture of robot | |
| CN104760812B (en) | Product real-time positioning system and method on conveyer belt based on monocular vision | |
| CN109685913A (en) | Augmented reality implementation method based on computer vision positioning | |
| CN109767473A (en) | A kind of panorama parking apparatus scaling method and device | |
| CN111739031A (en) | Crop canopy segmentation method based on depth information | |
| CN109961399A (en) | Optimal stitching line method for searching based on Image distance transform | |
| CN107977996A (en) | Space target positioning method based on target calibrating and positioning model | |
| CN108074265A (en) | A kind of tennis alignment system, the method and device of view-based access control model identification | |
| CN109308702A (en) | A kind of real-time recognition positioning method of target | |
| CN108171753A (en) | Stereoscopic vision localization method based on centroid feature point Yu neighborhood gray scale cross correlation | |
| CN110648362A (en) | Binocular stereo vision badminton positioning identification and posture calculation method | |
| CN110210292A (en) | A kind of target identification method based on deep learning | |
| CN114820817A (en) | Calibration method and three-dimensional reconstruction method based on high-precision line laser 3D camera | |
| CN110909571B (en) | High-precision face recognition space positioning method |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171010 |
|
| RJ01 | Rejection of invention patent application after publication |