CN107490346B - RFID multi-label network three-dimensional measurement modeling method based on vision - Google Patents

Abstract

Aiming at the requirements of warehouse entry and exit information acquisition and cargo inventory of cargoes in modern smart warehouse logistics, a RFID multi-tag network three-dimensional measurement modeling method based on vision is provided, images of RFID tags are obtained from multiple angles by using two cameras, the obtained images are processed by adopting an iterative threshold segmentation and morphology method, and three-dimensional coordinates of the RFID tags are obtained by utilizing a template matching method. The method effectively reduces the complexity of the algorithm and can acquire the three-dimensional coordinates of the RFID label in real time. The invention has important theoretical and application values for modern intelligent warehouse logistics.

Description

RFID multi-label network three-dimensional measurement modeling method based on vision

Technical Field

The invention relates to the field of RFID technology and Internet of things application, in particular to three-dimensional coordinate determination and network modeling of an RFID multi-label network, particularly relates to three-dimensional measurement modeling of the RFID multi-label network by introducing a three-dimensional visual analysis method, and belongs to the field of metering test.

Background

Radio Frequency Identification (RFID) is a novel non-contact automatic Identification technology, and has been widely applied in many fields such as modern logistics, intelligent transportation, and production automation, especially in intelligent logistics, because of the efficiency of system Identification, a large number of RFID tags are needed for warehouse entry and exit information acquisition and cargo inventory of bulk cargos, and at present, an effective automatic measurement and calculation means is lacking for modeling of an RFID multi-tag network and Identification, distribution, arrangement, combination, positioning and the like of the RFID multi-tag.

In the field of three-dimensional image measurement modeling, in the existing method, a single camera is used for carrying out three-dimensional measurement modeling on an image, the direction of the camera needs to be continuously adjusted, object images in the same state are obtained from different angles, the operation is complex, the real-time performance is poor, and the method is difficult to be applied to occasions with higher real-time requirements. In the existing literature, a single camera is less used for three-dimensional measurement modeling of an image, and the three-dimensional measurement modeling of an RFID multi-tag network by using an image method is more rarely reported. In order to meet the requirements of warehouse entry and exit information acquisition and cargo inventory of cargos in modern intelligent warehouse logistics, the method for carrying out three-dimensional measurement modeling on the RFID multi-tag network by utilizing the image correlation theory is low in algorithm complexity, small in calculated amount, high in efficiency and good in robustness and is significant. According to the invention, the horizontal camera and the vertical camera are used for obtaining the image information of the RFID label from multiple angles, so that the complexity of performing three-dimensional measurement modeling on the RFID label by using an image method is reduced.

The iterative threshold segmentation is an effective method in image segmentation by a threshold method, and the optimal threshold value of the segmentation is obtained by an iterative mode, so that the method has certain adaptivity. The morphological algorithm can perform morphological operations such as erosion and expansion on the image on the basis of threshold segmentation so as to retain and extract specific geometric features in the image, and the morphological algorithm is easy to realize in a parallel processing mode and has high real-time performance. The template matching algorithm has the advantages of simple calculation method, convenient template selection, easy realization and the like, and is particularly suitable for the field of the Internet of things with higher real-time requirement.

Disclosure of Invention

The invention provides a vision-based RFID multi-tag network three-dimensional measurement modeling method, which comprises the following steps:

the first step is as follows: building a test platform, wherein the test platform consists of an RFID reader-writer, a reader-writer antenna, an RFID tag support, a control computer, a servo motor, a vertical camera, a horizontal camera, a guide rail and a turntable;

the second step is as follows: the method comprises the steps of RFID label vertical overlook image acquisition, namely, using a vertical camera to acquire images of a turntable and an RFID label, processing the images by using an iterative threshold segmentation and morphology method, and numbering all RFID label mark points;

the third step: and an RFID label radial distance extraction step, namely taking two chords from the turntable vertical overlook image obtained in the second step, making a vertical bisector of the chords, taking the intersection point of the vertical bisector as the center of the turntable, selecting one RFID label mark point as a template, matching the ith RFID label mark point by using template matching, and obtaining the distance r from the ith RFID label mark point to the center of the turntable_iThe distance is the radial distance of the ith RFID label;

the fourth step: measuring the horizontal coordinate of the RFID label, namely adjusting a servo motor to drive a turntable to rotate to obtain the rotating angle theta of the ith RFID label_iThen theta_iAnd r obtained in the above third step_iNamely the horizontal coordinate parameter of the marking point of the ith RFID label, and further obtaining the horizontal coordinate of the ith RFID label as (r)_icosθ_i，r_isinθ_i)；

The fifth step: the method comprises the steps of RFID label side view image acquisition, controlling a computer to control a servo motor to drive a turntable to rotate, controlling a horizontal camera to be adjusted back and forth, enabling the horizontal camera to clearly image one label, taking the label as a template, measuring the vertical distance h from the central point of the template to the turntable, namely the vertical coordinate of the template, enabling the horizontal camera to accurately focus the ith RFID label, and obtaining the ith RFID labelMatching the template and the ith RFID label by utilizing template matching, and calculating the pixel number difference value c between the central point of the ith RFID label and the central point of the template_i；

A sixth step: the RFID label vertical coordinate measuring step, namely determining the side length a of each pixel in the image according to the triangle similarity principle, and calculating the vertical coordinate H of the ith RFID label in the vertical direction according to the vertical coordinate of the side-view image label template and the pixel number difference value of the ith RFID label and the template in the vertical direction_iWhen the center point of the ith RFID label is positioned below the center point of the template, H_i＝h-c_iX a, when the center point of the ith RFID label is above the center point of the template, H_i＝h+c_i×a；

A seventh step of: an RFID label three-dimensional coordinate determining step, namely determining the ith RFID label three-dimensional coordinate as (r)_icosθ_i，r_isinθ_i，H_i)；

An eighth step: and repeating the third, fourth, fifth, sixth and seventh steps, and measuring the horizontal coordinates and the vertical coordinates of all the RFID labels to obtain the three-dimensional coordinates of all the RFID labels.

The vision-based RFID multi-tag network three-dimensional measurement modeling method includes, in the second step, the iterative threshold segmentation, and the following steps:

the method comprises the following steps: normalizing the vertical overlook image and setting an iterative convergence parameter T₀And selecting an initial estimation threshold T₁(min g (x, y) + max g (x, y))/2, wherein (x, y) is the coordinate of a pixel point in the image, g (x, y) is the normalized image gray value, min g (x, y) is the minimum value of g (x, y), and max g (x, y) is the maximum value of g (x, y);

step two: by estimating the threshold value T₁Dividing the vertical overhead image into G₁And G₂Two parts, wherein G₁Is { (x, y) | g (x, y) > T₁}，G₂Is { (x, y) | g (x, y) < T₁}；

Step three: calculation of G₁And G₂Average gray value mu of all pixels in₁And mu₂And a new threshold value T₂＝(μ₁+μ₂)/2；

Step four: if | T₂-T₂|＜T₀Then T is₂I.e. the optimal threshold, otherwise, T is set₂Is assigned to T₁；

Step five: repeating the second, third and fourth steps until the optimal threshold is obtained, and utilizing the finally obtained optimal threshold T₂And performing threshold segmentation on the vertical overhead view image to obtain a binary image I of the vertical overhead view image.

The vision-based RFID multi-tag network three-dimensional measurement modeling method described above, wherein the morphological method in the second step includes the following steps:

the method comprises the following steps: performing expansion operation on the binary image of the vertical overlook image to obtain an expanded image

Wherein B is a structural element, I is a binary image of a vertical overlook image,

denotes swelling, P₁Is the image after expansion;

step two: carrying out corrosion operation on the image to obtain a corroded image P₂＝P₁Theta B, wherein B is a structural element, P₁For the expanded image, Θ represents corrosion, P₂Is an image after corrosion;

step three: repeating the first and second steps, and filling the image to obtain an image P₃To P₃And executing the operation of the second step to obtain the morphologically processed image.

The vision-based RFID multi-tag network three-dimensional measurement modeling method described above, wherein the template matching in the third step and the fifth step, comprises the following steps:

the method comprises the following steps: selecting a certain specific image as a reference template image, translating the known reference template image point by point in a search area of the image, and traversing each pixel point in the search area;

step two: determining normalized cross-correlation coefficients

Matching the reference template image with the search area image with the same size in the original image, wherein (m, n) is pixel point coordinates, T (m, n) is the reference template image,

is the mean value of the reference template image, M is the maximum value of the horizontal coordinate of the pixel point coordinate, N is the maximum value of the coordinate of the pixel point coordinate column, S^i，j(m, n) is a search area image,

the mean value of the search area image is obtained;

step three: and translating the reference template image to the next pixel point, and repeating the step two until all the positions are matched, wherein the area with the normalized cross-correlation coefficient NC (i, j) larger than the set threshold lambda is the matched area to be searched.

In the above method for modeling RFID multi-tag network three-dimensional measurement based on vision, in the fifth step, the control computer controls the horizontal camera to adjust back and forth, and the steps are as follows:

the method comprises the following steps: measuring the distance L from the initial position of the level camera to the center of the turntable₁；

Step two: controlling a computer to control a servo motor to drive a turntable to rotate, enabling the ith RFID label to be in a vertical state with a horizontal camera, and calculating the distance d between the ith RFID label and the horizontal camera_i＝L₁-r_i；

Step three: calculating the object space distance required by the horizontal camera to accurately focus the ith RFID label

Wherein l' is the distance from the center of the lens of the horizontal camera to the CCD sensor in the horizontal camera, and f is the focal length of the horizontal camera;

step four: distance delta L needing to be adjusted when horizontal camera accurately focuses on ith RFID label_i＝d_i-l_iIf Δ L_iGreater than zero, the horizontal camera moves Δ L in a direction approaching the ith RFID tag_iIf Δ L_iLess than zero, the horizontal camera moves Δ L in a direction away from the ith RFID tag_i。

The RFID multi-tag network three-dimensional measurement modeling method based on vision described above, wherein the triangle similarity principle described in the sixth step includes the following steps:

the method comprises the following steps: controlling a computer to control a horizontal camera to adjust back and forth, enabling the horizontal camera to clearly image one of the labels, measuring the distance from the label to the horizontal camera, recording the distance as S, the distance l' from the center of a lens of the horizontal camera to a CCD sensor in the horizontal camera, and the side length e of a CCD pixel;

step two: calculating pixel side lengths of an image

Drawings

FIG. 1: test platform structure diagram

FIG. 2: 7 RFID tag top view

FIG. 3: schematic diagram for measuring horizontal coordinate of 1 st RFID (radio frequency identification) tag

FIG. 4: iterative threshold segmentation processing result graph

FIG. 5: swelling treatment result graph

FIG. 6: graph of corrosion treatment results

FIG. 7: filling processing result graph

FIG. 8: graph of morphological processing results

FIG. 9: 7 RFID label vertical overlook image template matching processing result graph

FIG. 10: 1 st RFID label horizontal image template matching processing result graph

Detailed Description

A RFID multi-tag network three-dimensional measurement modeling method based on vision comprises the following steps:

the first step is as follows: building a test platform, wherein the test platform is composed of a 1-RFID reader-writer, a 2-reader-writer antenna, a 3-RFID label, a 4-RFID label support, a 5-control computer, a 6-servo motor, a 7-vertical camera, an 8-horizontal camera, A9-guide rail and a 10-rotary table, a marking point is pasted at the bottom of the 4-RFID label support, the 1-RFID reader-writer is respectively connected with the 2-reader-writer antenna and the 5-control computer, the 7-vertical camera and the 8-horizontal camera are respectively connected with the 5-control computer, the RFID label adopts an ultrahigh frequency electronic label-H47, the reader-writer adopts an Impinj speed Revolution R420 reader-writer, the reader-writer antenna adopts a Laird A9028 far-field antenna, the maximum RF output power is 30dBm, and the test platform is shown in figure 1;

the second step is as follows: an RFID label vertical overlook image acquisition step, wherein a vertical camera is used for carrying out image acquisition on the turntable and the RFID label, an iterative threshold segmentation and morphology method is used for processing the vertical overlook image, all 7 RFID label mark points are numbered, and the 7 RFID label top views are shown in FIG. 2;

the third step: and an RFID label radial distance extraction step, namely taking two chords from the turntable image obtained in the second step as vertical bisectors of the chords, taking the intersection point of the vertical bisectors as the center of the turntable, selecting one RFID label mark point as a template, matching the 1 st RFID label mark point by using a template matching algorithm, and obtaining the distance r from the 1 st RFID label mark point to the center of the turntable₁198.4mm, namely the radial distance of the 1 st RFID tag;

the fourth step: and an RFID label horizontal coordinate measuring step, namely adjusting a servo motor to drive a turntable to rotate, and obtaining the angle theta of the 1 st RFID label rotation which is 48.4 degrees and r obtained in the third step₁198.4mm is the horizontal coordinate parameter of the marking point of the 1 st RFID tag, the horizontal two-dimensional coordinate of the 1 st RFID tag is further calculated to be (131.7mm, 148.4mm), and the schematic diagram of measuring the horizontal coordinate of the 1 st RFID tag is shown in fig. 3;

the fifth step: the method comprises the steps of RFID label side-looking image acquisition, controlling a servo motor to drive a turntable to rotate, controlling a horizontal camera to be adjusted front and back, enabling the horizontal camera to clearly image one label, taking the label as a template, measuring the vertical distance h between a template central point and the turntable as 328.0mm, namely the vertical coordinate of the template, enabling the horizontal camera to accurately focus the 1 st RFID label, obtaining the 1 st RFID label side-looking image, matching the template and the 1 st RFID label by utilizing template matching, and calculating the pixel number difference value c between the central point of the 1 st RFID label and the template central point₁＝71；

A sixth step: an RFID label vertical coordinate measuring step, wherein according to a triangular similarity principle, the size of each pixel in an image is determined to be a being 0.75mm, wherein 0.75mm is the size of the side length of each pixel, according to the vertical coordinate of a template, the difference value of the number of pixels of the 1 st RFID label and the template in the vertical direction, and the longitudinal coordinate value of the central point of the 1 st RFID label is greater than the longitudinal coordinate value of the central point of the template, the central point of the 1 st RFID label is judged to be above the central point of the template, and therefore the vertical coordinate H of the 1 st RFID label in the vertical direction is calculated₁＝h+c₁×a＝327.85+0.75×71＝381.1mm；

A seventh step of: an RFID label three-dimensional coordinate determining step, namely determining the 1 st RFID label three-dimensional coordinate to be (131.7, 148.4, 381.1);

an eighth step: repeating the third, fourth, fifth, sixth and seventh steps, measuring the horizontal two-dimensional coordinates and the vertical coordinates of all 7 RFID tags, namely obtaining the three-dimensional coordinates of all 7 RFID tags as (131.7, 148.4, 381.1), (116.5, 179.2, 355.9), (-232.4, 200.3, 350.5), (-168.7, -58.8, 350.3), (47.5, -106.4, 328.5), (174.2, 40.7, 340.1), (174.9, -16.7, 379.9).

The second step of the vision-based RFID multi-tag network three-dimensional measurement modeling method is iterative threshold segmentation, and the steps are as follows:

the method comprises the following steps: normalizing the vertical overlook image and setting an iterative convergence parameter T₀0.01, and select a primaryInitial estimated threshold value T₁＝0.5178；

Step two: by estimating the threshold value T₁Dividing the vertical overhead image into G, 0.5178₁And G₂Two parts, wherein G₁Is { (x, y) | g (x, y) > T₁}，G₂Is { (x, y) | g (x, y) < T₁}；

Step three: calculation of G₁And G₂Average gray value mu of all pixels in₁0.6615 and μ₂0.1165, and a new threshold T₂＝(0.6615+0.1165)/2＝0.3890；

Step four: i T₂-T₁If | is less than 0.01, then T₂I.e. the optimal threshold, otherwise, T is set₂Is assigned to T₁；

Step five: repeating the second, third and fourth steps until the optimal threshold is obtained, and utilizing the finally obtained optimal threshold T₂The vertical overhead image is subjected to threshold segmentation to obtain a binarized image I of the vertical overhead image, which is shown in fig. 4.

The morphological method in the second step of the vision-based RFID multi-tag network three-dimensional measurement modeling method comprises the following steps:

the method comprises the following steps: performing expansion operation on the binary image of the vertical overlook image to obtain an expanded image

Wherein B is a structural element, I is a binary image of a vertical overlook image,

denotes swelling, P₁The dilated image is shown in fig. 5;

step two: carrying out corrosion operation on the image to obtain a corroded image P₂＝P₁Theta B, wherein B is a structural element, P₁For the expanded image, Θ represents corrosion, P₂The corroded image is shown in FIG. 6;

step three: repeating the first and second steps, and filling the image to obtain an image P₃The filled image is shown in FIG. 7 for P₃And (5) executing the operation of the step two to obtain an image after morphological processing, wherein a final effect image of the morphological processing is shown in fig. 8.

The third step and the fifth step of the visual-based RFID multi-tag network three-dimensional measurement modeling method for template matching are as follows:

the method comprises the following steps: selecting a certain specific image as a reference template image, translating the known reference template image point by point in a search area of the image, and traversing each pixel point in the search area;

step two: determining normalized cross-correlation coefficients

Matching the reference template image with the search area image with the same size in the original image, wherein (m, n) is pixel point coordinates, T (m, n) is the reference template image,

is the mean value of the reference template image, M is the maximum value of the horizontal coordinate of the pixel point coordinate, N is the maximum value of the coordinate of the pixel point coordinate column, S^i，j(m, n) is a search area image,

the mean value of the search area image is obtained;

step three: and translating the reference template image to the next pixel point, and repeating the second step until all the positions are matched, wherein the area with the normalized cross-correlation coefficient NC (i, j) larger than the set threshold value lambda being 0.75 is the matching area to be searched, the 7 RFID tags vertically overlook the image template matching result as shown in fig. 9, and the 1 st RFID tag horizontal image template matching result as shown in fig. 10.

In the fifth step of the vision-based three-dimensional measurement modeling method for the RFID multi-tag network, the control computer controls the horizontal camera to adjust back and forth, and the steps are as follows:

the method comprises the following steps: measuring the distance L from the initial position of the level camera to the center of the turntable₁＝1.512m；

Step two: controlling a computer to control a servo motor to drive a turntable to rotate, enabling the 1 st RFID label to be in a vertical state with a horizontal camera, and calculating the distance d between the 1 st RFID label and the horizontal camera₁＝1.512m-0.406m＝1.106m；

Step three: calculating the object space distance required by the horizontal camera to accurately focus the 1 st RFID label

Wherein l 'is the distance from the center of the horizontal camera lens to the CCD sensor in the horizontal camera, l' is 0.00597m, and f is 0.006 m;

step four: distance delta L needing to be adjusted when the 1 st RFID label is accurately focused by the horizontal camera₁＝d₁-|l₁|＝1.106m-(1.194m)＝-0.088m，ΔL₁Less than zero, the horizontal camera moves 0.088m away from the 1 st RFID tag.

The visual-based RFID multi-tag network three-dimensional measurement modeling method based on the sixth step adopts the triangle similarity principle, and comprises the following steps:

the method comprises the following steps: controlling a computer to control a horizontal camera to adjust back and forth, enabling the horizontal camera to clearly image one of labels, measuring the distance from the label to the horizontal camera, and recording the distance as S1.194 m, the distance from the center of a lens of the horizontal camera to a CCD sensor in the horizontal camera, i' 0.00597m, and the side length e of a CCD pixel as 3.75 mu m;

step two: calculating pixel side lengths of an image

Claims (6)

1. A RFID multi-tag network three-dimensional measurement modeling method based on vision comprises the following steps:

the first step is as follows: building a test platform, wherein the test platform consists of an RFID reader-writer, a reader-writer antenna, an RFID tag support, a control computer, a servo motor, a vertical camera, a horizontal camera, a guide rail and a turntable;

the second step is as follows: the method comprises the steps of RFID label vertical overlook image acquisition, namely, using a vertical camera to acquire images of a turntable and an RFID label, processing the images by using an iterative threshold segmentation and morphology method, and numbering all RFID label mark points;

the third step: and an RFID label radial distance extraction step, namely taking two chords from the turntable vertical overlook image obtained in the second step, making a vertical bisector of the chords, taking the intersection point of the vertical bisector as the center of the turntable, selecting one RFID label mark point as a template, matching the ith RFID label mark point by using template matching, and obtaining the distance r from the ith RFID label mark point to the center of the turntable_iThe distance is the radial distance of the ith RFID label;

the fourth step: measuring the horizontal coordinate of the RFID label, namely adjusting a servo motor to drive a turntable to rotate to obtain the rotating angle theta of the ith RFID label_iThen theta_iAnd r obtained in the above third step_iNamely the horizontal coordinate parameter of the marking point of the ith RFID label, and further obtaining the horizontal coordinate of the ith RFID label as (r)_icosθ_i，r_isinθ_i)；

The fifth step: the method comprises the steps of RFID label side-looking image acquisition, controlling a servo motor to be controlled by a control computer to drive a turntable to rotate, controlling a horizontal camera to be controlled by the control computer to be adjusted front and back, enabling the horizontal camera to clearly image one label, taking the label as a template, measuring the vertical distance h from a template central point to the turntable, namely the vertical coordinate of the template, enabling the horizontal camera to accurately focus the ith RFID label, obtaining the ith RFID label side-looking image, matching the template and the ith RFID label by utilizing template matching, and calculating the ith RFID labelDifference value c of pixel number between central point and template central point_i；

A sixth step: the RFID label vertical coordinate measuring step, according to the triangle similarity principle, determining the side length a of each pixel in the image, according to the vertical coordinate of the side-view image label template and the pixel number difference value of the ith RFID label and the template in the vertical direction, calculating to obtain the vertical coordinate H of the ith RFID label in the vertical direction_iWhen the center point of the ith RFID label is positioned below the center point of the template, H_i＝h-c_iX a, when the center point of the ith RFID label is above the center point of the template, H_i＝h+c_i×a；

A seventh step of: an RFID label three-dimensional coordinate determining step, namely determining the ith RFID label three-dimensional coordinate as (r)_icosθ_i，r_isinθ_i，H_i)；

An eighth step: and repeating the third, fourth, fifth, sixth and seventh steps, and measuring the horizontal coordinates and the vertical coordinates of all the RFID labels to obtain the three-dimensional coordinates of all the RFID labels.

2. The vision-based RFID multi-tag network three-dimensional measurement modeling method of claim 1, wherein the iterative threshold segmentation of the second step comprises the following steps:

the method comprises the following steps: normalizing the vertical overlook image and setting an iterative convergence parameter T₀And selecting an initial estimation threshold T₁(min g (x, y) + max g (x, y))/2, wherein (x, y) is the coordinate of a pixel point in the image, g (x, y) is the normalized image gray value, min g (x, y) is the minimum value of g (x, y), and max g (x, y) is the maximum value of g (x, y);

step two: by estimating the threshold value T₁Dividing the vertical overhead image into G₁And G₂Two parts, wherein G₁Is { (x, y) | g (x, y) > T₁}，G₂Is { (x, y) | g (x, y) < T₁}；

Step three: calculation of G₁And G₂Average gray scale of all pixels inValue of mu₁And mu₂And a new threshold value T₂＝(μ₁+μ₂)/2；

Step four: if | T₂-T₁|＜T₀Then T is₂I.e. the optimal threshold, otherwise, T is set₂Is assigned to T₁；

Step five: repeating the second, third and fourth steps until the optimal threshold is obtained, and utilizing the finally obtained optimal threshold T₂And performing threshold segmentation on the vertical overhead view image to obtain a binary image I of the vertical overhead view image.

3. The vision-based RFID multi-tag network three-dimensional measurement modeling method of claim 1, wherein the morphological method of the second step comprises the following steps:

the method comprises the following steps: performing expansion operation on the binary image of the vertical overlook image to obtain an expanded image

Wherein B is a structural element, I is a binary image of a vertical overlook image,

denotes swelling, P₁Is the image after expansion;

step two: carrying out corrosion operation on the image to obtain a corroded image P₂＝P₁Theta B, wherein B is a structural element, P₁For the expanded image, Θ represents corrosion, P₂Is an image after corrosion;

step three: repeating the first and second steps, and filling the image to obtain an image P₃To P₃And executing the operation of the second step to obtain the morphologically processed image.

4. The vision-based RFID multi-tag network three-dimensional measurement modeling method of claim 1, wherein the template matching in the third step and the fifth step comprises the following steps:

the method comprises the following steps: selecting a certain specific image as a reference template image, translating the known reference template image point by point in a search area of the image, and traversing each pixel point in the search area;

step two: determining normalized cross-correlation coefficients

Matching the reference template image with the search area image with the same size in the original image, wherein (m, n) is pixel point coordinates, T (m, n) is the reference template image,

is the mean value of the reference template image, M is the maximum value of the horizontal coordinate of the pixel point coordinate, N is the maximum value of the coordinate of the pixel point coordinate column, S^i，j(m, n) is a search area image,

the mean value of the search area image is obtained;

step three: and translating the reference template image to the next pixel point, and repeating the step two until all the positions are matched, wherein the area with the normalized cross-correlation coefficient NC (i, j) larger than the set threshold lambda is the matched area to be searched.

5. The vision-based RFID multi-tag network three-dimensional measurement modeling method of claim 1, wherein in the fifth step the control computer controls the horizontal camera to adjust back and forth, the steps are as follows:

the method comprises the following steps: measuring the distance L from the initial position of the level camera to the center of the turntable₁；

Step two: controlling a computer to control a servo motor to drive a turntable to rotate, enabling the ith RFID label to be in a vertical state with a horizontal camera, and calculating the distance d between the ith RFID label and the horizontal camera_i＝L₁-r_i；

Step three: computingObject space distance required by accurate focusing of ith RFID label by horizontal camera

Wherein l' is the distance from the center of the lens of the horizontal camera to the CCD sensor in the horizontal camera, and f is the focal length of the horizontal camera;

step four: distance delta L needing to be adjusted when horizontal camera accurately focuses on ith RFID label_i＝d_i-l_iIf Δ L_iGreater than zero, the horizontal camera moves Δ L in a direction approaching the ith RFID tag_iIf Δ L_iLess than zero, the horizontal camera moves Δ L in a direction away from the ith RFID tag_i。

6. The vision-based RFID multi-tag network three-dimensional measurement modeling method of claim 1, wherein the triangle similarity principle in the sixth step comprises the following steps:

the method comprises the following steps: controlling a computer to control a horizontal camera to adjust back and forth, enabling the horizontal camera to clearly image one of the labels, measuring the distance from the label to the horizontal camera, recording the distance as S, the distance l' from the center of a lens of the horizontal camera to a CCD sensor in the horizontal camera, and the side length e of a CCD pixel;

step two: calculating pixel side lengths of an image

Images