A RFID target imaging method based on dynamic selection of initial contours

Abstract

A method for RFID target imaging based on dynamic initial contour selection specifically includes the following steps: establishing a passive target imaging system based on ultra-high frequency RFID; constructing linear and arc-shaped initial contours of targets; completing target image estimation based on a discrete tag array to obtain an estimated image of the target horizontal cross-section corresponding to the array; estimating the target horizontal cross-section based on maximizing overlap, using the horizontal coordinate set of four propagation distance starting points as optimization variables and maximizing overlap as the optimization objective, and introducing a particle swarm algorithm for optimization; and completing target imaging based on dynamic initial contour selection. The proposed imaging method pre-determines the target edge type based on the overlap and selects an appropriate initial contour based on the judgment result, significantly improving target imaging accuracy.

Description

Technical Field

[0001] The invention belongs to the technical field of mobile wireless communications and designs an RFID target imaging method based on dynamic selection of initial contours. Background Art

[0002] With the emergence of IoT concepts such as ubiquitous computing and contextually aware services, passive object imaging technology has garnered significant attention from both industry and academia over the past decade. Passive object imaging offers a novel approach to environmental perception and is poised for widespread adoption in IoT fields such as logistics and warehousing, smart spaces, and smart homes.

[0003] Existing passive target imaging systems usually use computer vision, ultra-wideband, radio frequency identification (RFID) technology, etc. to complete the construction of the system architecture and the design of the core algorithm. Among the above technologies, RFID has become a very competitive system solution due to its non-contact, non-line-of-sight, fast reading and writing, and mobile identification. A typical RFID system consists of four parts: a reader, an electronic tag, an antenna, and a background processor. The TagScan system proposed by Wang Ju et al. is a typical representative of the passive target imaging system based on RFID, with the advantages of low cost, easy implementation, and anti-multipath. It should be pointed out that since TagScan uses a straight initial contour to complete the target image estimation, its imaging accuracy for targets with arc-shaped edges is relatively low. To address this problem, the present invention proposes an RFID target imaging method based on dynamic selection of initial contours. Summary of the Invention

[0004] The purpose of the present invention is to provide an RFID target imaging method based on dynamic selection of initial contours. Specifically, the method comprises the following steps:

[0005] Step 1: Establish a passive target imaging system based on ultra-high frequency RFID. The system consists of a reader, two sets of tag arrays, and a reader antenna device. Define the first set of tag arrays as array 1, and define the second set of tag arrays as array 2. By controlling the reader antenna to move along a predetermined trajectory to collect phase information in all directions, target imaging is completed. Then, the overall framework of the target imaging method is constructed. The framework mainly consists of four parts: data acquisition and preprocessing, construction of the initial target contour, target image estimation based on discrete tag arrays, and target horizontal cross-section estimation based on maximizing overlap.

[0006] Step 2: Move the reader antenna along the predetermined trajectory, perform data acquisition and preprocessing, and when no imaging target is set between the reader antenna and the tag array, perform the image acquisition and preprocessing at N consecutive angles {θ1, θ2, ..., θ N} is tested to obtain the tag carrier phase {φ1, φ2, ..., φ N},φ i It contains the phases of multiple tags. This test only needs to be performed once. When an imaging target is placed between the reader antenna and the tag array, the reader is allowed to collect phase information at the same trajectory and speed, and the phase {φ1′, φ2′, …, φ N ′}, the phase change caused by the target at N consecutive angles can be obtained and recorded as Δφ={φ1,φ2,…,φ N}-{φ1′, φ2′, …, φ N ′}={Δφ1, Δφ2,…, Δφ N}, considering that the complex multipath environment will break the linear relationship between the propagation distance and the observed parameters, data preprocessing is performed on the above-mentioned phase changes, and the phase difference between adjacent tags in the tag array is used to suppress multipath. The multipath suppression method is divided into two stages. The first is to identify relatively "clean" channels, and then to assign weights, that is, to reduce the impact of multipath by assigning different weights to the measured signals on different channels and different tags.

[0007] Step 3: Execute the construction of the initial contour of the target. First, the two angular boundaries of the target are determined by using the cumulative phase change method based on the sliding window. Then, the lower boundary angle θ of the tag array is detected by comparing the phase change with the preset threshold. 1，l and the upper boundary angle θ 1，u Finally, the two angle boundaries of the target are measured using the label array to form the target boundary quadrilateral. For the convenience of explanation, the four vertices of the target boundary quadrilateral are defined as Q1, Q2, Q3, and Q4, and an arbitrary point is selected on the straight line segments Q1Q2, Q2Q3, Q3Q4, and Q4Q1 respectively. The four arbitrary points are defined as contour vertices V1, V2, V3, and V4. V1, V2, V3, and V4 are connected in sequence to construct the initial contour of the target. The initial contour can be regarded as composed of four curve segments, and both ends of each curve segment are contour vertices. If the line type of the curve segment is a straight line, the constructed initial contour is called a straight line initial contour. If the line type of the curve segment is an arc, the constructed initial contour is called an arc initial contour.

[0008] Step 4: Calculate the propagation distance of the signal inside the target using the phase change caused by the target, and perform target image estimation based on the discrete tag array. Based on the linear initial profile and the arc initial profile constructed in step 3, two estimated images based on the discrete tag array can be obtained for each initial profile. The target imaging estimation method based on the discrete tag array can be expressed as follows: First, use the phase change Δφ obtained in data acquisition = {Δφ1, Δφ2, ..., Δφ N} to calculate the propagation distance vector D based on the phase difference, that is, D = (Δφ + 2ξπ) / (β tar -β air ), where each element in D represents the propagation distance inside the target corresponding to the signal collected in a specific direction, λ air and λ tar Represent the signal wavelengths in air and in the target, β air =2π / λ air and β tar =2π / λ tar They represent the phase constants of the signal in the air and in the target respectively. ξ is an integer used to represent the full-cycle ambiguity feature. For objects with an average size smaller than the carrier wavelength, the phase change caused by it does not exceed 2π, so ξ is taken as 0. From the above formula, we can get D1, D2, ..., D N The value of each propagation distance, then, for array 1 and array 2, select the two curve segments closest to the label array from the initial contour as the two edges of the target estimated image, and assume that the starting points of the propagation distance are located on these two curve segments, according to {θ1, θ2, ..., θ N} and the coordinates of the array to determine the coordinates of the starting point, according to {θ1, θ2, ..., θ N} and D determine the coordinates of the ending points corresponding to each starting point, connect each ending point to obtain another edge segment of the target estimation image, and finally, connect the three edges mentioned above to obtain the target estimation images of array 1 and array 2 on the horizontal section.

[0009] Step 5: Execute target horizontal cross-section estimation based on maximizing overlap. If there is a difference between the estimated image of array 1 and the estimated image of array 2, it means that the selection of contour vertices V1, V2, V3, and V4 is not ideal. In order to correctly select the starting point of the contour vertex, it is necessary to search different combinations of V1, V2, V3, and V4 on the four edges Q1Q2, Q2Q3, Q3Q4, and Q4Q1 to construct the target initial contour. That is, the target imaging problem is converted into an optimization problem. The maximization of the overlap of the estimated images of the two tag arrays is used as the optimization goal. The horizontal coordinates x1, x2, x3, and x4 of the four points V1, V2, V3, and V4 are used to construct a set, and Ω is used as the variable to be optimized. The vertical coordinates y1, y2, y3, and y4 of the four points V1, V2, V3, and V4 can be obtained by calculating the expressions of each curve segment in the initial contour and the values ​​of x1, x2, x3, and x4. The optimization model of the UHF RFID target imaging system is constructed. The above optimization problem model can be expressed as: s (Ω)=arg(max(F)), where O s (Ω) is the optimal solution of the optimization model, F is the global optimization objective function based on overlap, O s(Ω)=arg(max(F)) means the variable value of Ω when F takes its maximum value. In this model, the overlap represents the ratio of the overlapping part of the estimated images obtained by the two label arrays to the merged image. The larger the overlap, the closer the merged image is to the actual image. The particle swarm algorithm is introduced to iteratively search for the optimal solution to the above optimization problem.

[0010] Step 6: Use deviation and overlap to analyze the impact of the selection of the initial contour on the imaging accuracy. Define deviation as the average distance between the target estimated image contour and the corresponding position points in the actual contour. According to the different target edge types, the targets are divided into two types: targets with edges composed of straight lines and targets with edges composed of arcs. Select representative quadrilateral targets and circular targets as imaging objects. Calculate the deviation and overlap of the target estimated images obtained by using different types of initial contours in the optimization process, and obtain the deviation curve and overlap curve of the arc-shaped initial contour and the straight-line initial contour. Comparing the two types of curves, it can be seen that For quadrilateral targets and circular targets, as the number of iterations gradually increases, the value of the deviation curve gradually decreases and approaches 0, and the value of the overlap curve gradually increases from 0 and approaches 1. The changes of the two types of curves are inversely proportional. Therefore, the size of the overlap is used to characterize the target imaging accuracy. At the same time, for the same imaging target, the arc-type initial profile and the straight-line initial profile are used for imaging respectively. The overlap curves of the two estimated images have obvious differences. For circular targets, the arc-type initial profile can achieve better imaging accuracy, and for quadrilateral targets, the straight-line initial profile can achieve better imaging accuracy.

[0011] Step 7: For imaging targets with unknown edge types, a straight line initial profile and an arc-type initial profile are used to image the target respectively. When imaging with an arc-type initial profile, the maximum image overlap is estimated as C1. When imaging with a straight line initial profile, the maximum image overlap is estimated as C2. Compare C1 and C2 to complete the judgment of the target edge. If C1>C2, the edge of the imaging target is considered to be an arc, and the arc-type initial profile is selected for target imaging. If C1≤C2, the straight line initial profile is selected for target imaging. The estimated images of array 1 and array 2 are further merged to obtain the final horizontal cross-sectional image of the target.

[0012] It should be pointed out that the method for constructing the straight line initial contour in the above step 3 is: define the coordinates of points V1, V2, V3 and V4 as (x1, y1), (x2, y2), (x3, y3) and (x4, y4) respectively, and use straight lines to connect the above four points in sequence to construct the straight line initial contour. Assume that the expression of the straight line segment is y=kx+b, x and y represent the horizontal coordinate and vertical coordinate of each point in the straight line segment respectively, k and b represent the slope and intercept respectively, and use the two-point method to calculate the slope and intercept, and then obtain the accurate expression of each straight line segment in the straight line initial contour, with the straight line segment For example, the slope is k = y3-y2 / x3-x2, and the intercept is b = (y2x3-y3x2) / (x3-x2). Combining x∈[x2, x3] and y∈[y2, y3], we can determine the straight line segment The specific expression and geometric characteristics of the other three straight line segments can be completed in the same way. Specific expressions and estimation of geometric characteristics.

[0013] It should be pointed out that the method for constructing the arc-shaped initial contour in the above step 3 is: define the coordinates of points V1, V2, V3 and V4 as (x1, y1), (x2, y2), (x3, y3) and (x4, y4) respectively, use arcs to connect the above four points in sequence to construct the arc-shaped initial contour, use the circumscribed circle method to calculate the curvature, and then obtain the expression of each curve segment in the arc-shaped initial contour. Take the curve segment V3V4 as an example to illustrate, select a combination of three points from V1, V2, V3 and V4, and calculate the center (a, b) of the circumscribed circle corresponding to the combination. Taking the three points in the combination as V2, V3 and V4 as an example, then a=(y2-y4)f1-(y3-y4)f2 / 2f3, b=(x2-x4)f4-(x3-x4)f5 / 2f6 f3=(x3-x4)(y2-y4)-(x2-x4)(y3-y4), f6=(y3-y4)(x2-x4)-(y2-y4)(x3-x4), further, the radius r of the circumscribed circle can be obtained by Calculated, the curve segment The curvature of the curve segment is K = 1 / r. Assume that the expression of the curve segment is y = f(x), where x and y represent the horizontal and vertical coordinates of each point in the curve segment, respectively. The first-order derivative and second-order derivative of y with respect to x are y′ and y″, respectively. Whether the arc segment meets the requirements of target contour construction is determined by whether the center of the circle is within the area of ​​the quadrilateral Q1Q2Q3Q4. If the center of the circle is within the area of ​​the quadrilateral Q1Q2Q3Q4, then K = |y″| / (1+(y′) 2 ) 1.5, x∈[x3,x4], y∈[y3,y4] can determine the specific expression of y=f(x) to describe the curve segment If the center of the circle is not within the area of ​​the quadrilateral Q1Q2Q3Q4, then select another set of three points from V1, V2, V3 and V4 to solve the center and radius to determine the arc segment. The expression and geometric characteristics of the other three arc segments can be completed in the same way. and Estimation of the expression and geometric characteristics of . BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a flowchart of the present invention;

[0015] Figure 2 Schematic diagram of the imaging deployment of the phase-based target imaging system of the present invention;

[0016] Figure 3 is a schematic diagram of boundary detection of the present invention;

[0017] Figure 4 This is a schematic diagram of the initial outline of the arc-shaped target of the present invention;

[0018] Figure 5 This is a schematic diagram of a straight line initial contour imaging estimation image of the present invention;

[0019] Figure 6 This is a schematic diagram of an arc-shaped initial contour imaging estimation image of the present invention;

[0020] Figure 7 It is the optimization algorithm program diagram of the present invention;

[0021] Figure 8 It is an optimization curve of the iterative process with a circular target as the imaging object;

[0022] Figure 9 It is an iterative process optimization curve with a quadrilateral target as the imaging object;

[0023] Figure 10 It is a quadrilateral estimated image using the initial outline of a straight target. DETAILED DESCRIPTION

[0024] The main purpose of the present invention is to propose an RFID target imaging method based on dynamic selection of initial contours. This method dynamically selects the appropriate target initial contour according to the judgment conclusion to complete target imaging, greatly improving the target imaging accuracy.

[0025] The following is combined with Figure 1 , Attachment Figure 2 , Attachment Figure 3 , Attachment Figure 4, Attachment Figure 5 , Attachment Figure 6 , Attachment Figure 7 , Attachment Figure 8 , Attachment Figure 9 , Attachment Figure 10 The embodiments of the present invention are described in further detail.

[0026] The workflow of the method proposed by the present invention is shown in the attached Figure 1 As shown in the figure, firstly, a passive target imaging system based on ultra-high frequency RFID is built. The system consists of a reader, two sets of tag arrays and a reader antenna. By controlling the reader antenna to move along a predetermined trajectory, phase information in all directions is collected to complete target imaging. Then, the overall framework of the target imaging method is constructed. The framework mainly consists of four parts: data acquisition and preprocessing, construction of the initial contour of the target, target image estimation based on discrete tag arrays, and target horizontal cross-section estimation based on maximizing overlap. Finally, target imaging based on dynamic selection of the initial contour is performed to obtain the final horizontal cross-section image of the target.

[0027] According to the attached Figure 2 After the system is built, the reader antenna is moved along the predetermined trajectory to perform data acquisition and preprocessing. When no imaging target is set between the reader antenna and the tag array, the image is captured at N consecutive angles {θ1, θ2, ..., θ N} is tested to obtain the tag carrier phase {φ1, φ2, ..., φ N},φ i It contains the phases of multiple tags. This test only needs to be performed once. When an imaging target is placed between the reader antenna and the tag array, the reader is allowed to collect phase information at the same trajectory and speed, and the phase {φ1′, φ2′, …, φ N ′}, the phase change caused by the target at N consecutive angles can be obtained and recorded as Δφ=={φ1,φ2,…,φ N}-{φ1′, φ2′, …, φ N′}={Δφ1, Δφ2,…, Δφ N Considering that the complex multipath environment will break the linear relationship between the propagation distance and the observation parameters, data preprocessing is performed on the above phase changes, and the phase difference between adjacent tags in the tag array is used to suppress multipath.

[0028] Then, the construction of the initial contour of the target is performed. First, the two angular boundaries of the target are determined by using the cumulative phase change method based on the sliding window. Then, the lower boundary angle θ of the tag array is detected by comparing the phase change with the preset threshold. 1，l and the upper boundary angle θ 1，uFinally, the two angle boundaries of the target are measured using the label array to form the target boundary quadrilateral. For the convenience of explanation, the four vertices of the target boundary quadrilateral are defined as Q1, Q2, Q3, and Q4, and an arbitrary point is selected on the straight line segments Q1Q2, Q2Q3, Q3Q4, and Q4Q1 respectively. The four arbitrary points are defined as contour vertices V1, V2, V3, and V4. V1, V2, V3, and V4 are connected in sequence to construct the initial contour of the target. The initial contour can be regarded as composed of four curve segments. The two ends of each curve segment are contour vertices. If the line type of the curve segment is a straight line as shown in the attached figure Figure 3 As shown in the figure, the constructed initial contour is called a straight line initial contour. If the line type of the curve segment is an arc type as shown in the figure, Figure 4 As shown, the constructed initial contour is called an arc-type initial contour.

[0029] The phase change caused by the target is used to calculate the propagation distance of the signal inside the target, and the target image estimation based on the discrete tag array is performed. Based on the linear initial profile and the arc initial profile constructed in step 3, two estimated images based on the discrete tag array can be obtained for each initial profile. The target imaging estimation method based on the discrete tag array can be expressed as follows: first, the phase change Δφ obtained in data acquisition is used = {Δφ1, Δφ2, ..., Δφ N} to calculate the propagation distance vector D based on the phase difference, that is, D = (Δφ + 2ξπ) / (β tar -β air ), where each element in D represents the propagation distance inside the target corresponding to the signal collected in a specific direction, λ air and λ tar Represent the signal wavelengths in air and in the target, β air =2π / λ air and β tar =2π / λ tar They represent the phase constants of the signal in the air and in the target respectively. ξ is an integer used to represent the full-cycle ambiguity feature. For objects with an average size smaller than the carrier wavelength, the phase change caused by them generally does not exceed 2π, so ξ is taken as 0. From the above formula, we can get D1, D2, ..., D N The value of each propagation distance, then, for array 1 and array 2, select the two curve segments closest to the label array from the initial contour as the two end edges of the target estimation image, and assume that the starting points of the propagation distance are located on these two curve segments, according to {θ1, θ2, ..., θ N} and determine the coordinates of the starting point, according to {θ1, θ2, ..., θ N} and D determine the coordinates of the end points corresponding to each starting point, connect each end point to obtain another edge of the target estimation image, and finally connect the three edges mentioned above. For the linear initial contour, the target estimation images of array 1 and array 2 on the horizontal section are as shown in the attached figure. Figure 5 As shown; for the arc-shaped initial profile, the target estimation images of array 1 and array 2 on the horizontal section are shown in the attached Figure 6 shown.

[0030] Perform target horizontal cross-section estimation based on maximizing overlap. If there is a difference between the estimated image of array 1 and the estimated image of array 2, it means that the selection of contour vertices V1, V2, V3, and V4 is not ideal. In order to correctly select contour vertices, it is necessary to search different combinations of V1, V2, V3, and V4 on the four edges Q1Q2, Q2Q3, Q3Q4, and Q4Q1 to construct the target initial contour. That is, the target imaging problem is converted into an optimization problem. The maximization of the overlap of the estimated images of the two tag arrays is used as the optimization goal. The horizontal coordinates x1, x2, x3, and x4 of the four points V1, V2, V3, and V4 are used to construct the set Ω, and Ω is used as the variable to be optimized. The vertical coordinates y1, y2, y3, and y4 of the four points V1, V2, V3, and V4 can be obtained by calculating the expressions of each curve segment in the initial contour and the values ​​of x1, x2, x3, and x4. The optimization model of the UHF RFID target imaging system is constructed. The above optimization problem model can be expressed as: s (Ω)=arg(max(F)), where O s (Ω) is the optimal solution of the optimization model, F is the global optimization objective function based on overlap, O s (Ω) = arg(max(F)) means that when F takes the maximum value, the variable value of Ω is taken. In this model, the overlap degree represents the ratio of the overlapping part of the estimated image obtained by the two label arrays to the merged image. The larger the overlap degree, the closer the merged image is to the actual image. The particle swarm algorithm is introduced to iteratively search for the optimal solution to the above optimization problem. The solution process is shown in the attached figure. Figure 7 shown.

[0031] Deviation and overlap are used to analyze the influence of the selection of initial contour on imaging accuracy. Deviation is defined as the average distance between the target estimated image contour and the corresponding position points in the actual contour. According to the different target edge types, the targets are divided into two types: targets with edges composed of straight lines and targets with edges composed of arcs. Representative quadrilateral targets and circular targets are selected as imaging objects. The deviation and overlap of the target estimated images obtained by using different types of initial contours in the optimization process are calculated, and the deviation curves and overlap curves of the arc-shaped initial contour and the straight-line initial contour are obtained. By comparing the two types of curves, it can be seen that for quadrilateral targets and circular targets, as the number of iterations gradually increases, the value of the deviation curve gradually decreases and approaches 0, and the value of the overlap curve gradually increases from 0 and approaches 1. The changes of the two types of curves are inversely proportional. Therefore, the size of the overlap is used to characterize the target imaging accuracy. At the same time, for the same imaging target, the arc-shaped initial contour and the straight-line initial contour are used for imaging respectively, and the overlap curves of the two estimated images are significantly different. For circular targets as shown in the attached figure, the deviation curves of the arc-shaped initial contour and the straight-line initial contour are significantly different. Figure 8 As shown in the figure, the arc overlap is close to 1, and the arc deviation is close to 0, so the arc-shaped initial contour can obtain better imaging accuracy; for quadrilateral targets such as the attached Figure 9 As shown in FIG, the overlap of the straight lines is close to 1, and the deviation of the straight lines is close to 0, so a straight line initial profile can obtain better imaging accuracy.

[0032] The following are specific embodiments: Figure 2 As shown, the imaging scene is arranged in a 3m×3m configuration. The centers of the two tag arrays are located at (150cm, 0) and (0, 150cm), respectively. Each array consists of eight tags, with a 4cm distance between adjacent tags. The target to be measured is located in the XOY plane. The reader moves along a predetermined trajectory in the XOY plane, imaging a quadrilateral. First, in a target-free scenario, the phase value of each tag is recorded during the reader's movement. This process only needs to be performed once. Because the spacing between adjacent tags in the same array is much smaller than the distance between the reader and the tag array, the spacing between tags is ignored. It is reasonable to assume that tags in the same array are positioned identically. The corresponding angles of all tags in each array are then averaged, and this average phase value is used as a reference for subsequent phase changes. Then, in a target-present scenario, the phase is acquired and imaged.

[0033] After obtaining the estimated image, the particle swarm algorithm is introduced to optimize the maximum overlap of the estimated image. Taking the quadrilateral target as an example, the population size is set to 40 and the number of iterations is set to 600. As the number of iterations increases, the straight line overlap curve continues to rise and the straight line deviation curve continues to decrease. Figure 9 is the overlap and deviation curve of the iterative process, and the image with the largest overlap is output after the number of iterations is reached. Figure 10 This is the result after optimizing the maximum overlap. In order to facilitate calculation and analysis, the errors of four vertices are used to evaluate the imaging accuracy. The coordinates of the four vertices V1, V2, V3, and V4 of the actual target are preset to be (145 cm, 200 cm), (200 cm, 140 cm), (160 cm, 120 cm), and (120 cm, 160 cm). The coordinates of the four vertices V1, V2, V3, and V4 of the estimated image are (144.9282 cm, 199.9752 cm), (199.9327 cm, 139.8116 cm), (160.0 141cm, 120.1574cm), (120.0180cm, 159.8116cm). Therefore, according to the above coordinates, it can be seen that there is a small error between the actual target image and the estimated target image. The coordinate errors of the above four vertices are (-0.0718cm, -0.0248cm), (-0.0673cm, -0.1884cm), (0.0141, 0.1574cm), (0.0180, -0.1884cm), respectively. According to the results, the maximum error of the four vertices is 0.1884cm and the minimum error is 0.0141cm. The above results show that the imaging method proposed in this patent can achieve better imaging accuracy.

Claims

1. A RFID target imaging method based on dynamic selection of initial contours, characterized in that: The specific steps are as follows: Step 1: Establish a passive target imaging system based on ultra-high frequency RFID. The system consists of a reader, two sets of tag arrays, and a reader antenna device. The first set of tag arrays is defined as array 1, and the second set of tag arrays is defined as array 2. By controlling the reader antenna to move along a predetermined trajectory to collect phase information in all directions, target imaging is completed. Then, an overall framework of the target imaging method is constructed. This framework mainly consists of four parts: data acquisition and preprocessing, construction of the initial target contour, target image estimation based on discrete tag arrays, and target horizontal cross-section estimation based on maximizing overlap. Step 2: Move the reader antenna along the predetermined trajectory, perform data acquisition and preprocessing, and when no imaging target is set between the reader antenna and the tag array, perform the image acquisition and preprocessing at N consecutive angles {θ1, θ2, ..., θ N } is tested to obtain the tag carrier phase {φ1, φ2, ..., φ N },φ i It contains the phases of multiple tags. This test only needs to be performed once. When an imaging target is placed between the reader antenna and the tag array, the reader is allowed to collect phase information at the same trajectory and speed, and the phase {φ1′, φ2′, …, φ N ′}, the phase change caused by the target at N consecutive angles can be obtained and recorded as Δφ={φ1,φ2,…,φ N }-{φ1′,φ2′,…,φ N ′}={Δφ1, Δφ2,…, Δφ N Considering that complex multipath environments can disrupt the linear relationship between propagation distance and observed parameters, data preprocessing is performed on the phase changes mentioned above. The phase difference between adjacent tags in the tag array is used to suppress multipath. The multipath suppression method is divided into two stages: first, identifying relatively "clean" channels, and then assigning weights. That is, different weights are assigned to the measured signals on different channels and different tags to reduce the impact of multipath. Step 3: Execute the construction of the initial contour of the target. First, the two angular boundaries of the target are determined by using the cumulative phase change method based on the sliding window. Then, the lower boundary angle θ of the tag array is detected by comparing the phase change with the preset threshold. 1,l and the upper boundary angle θ 1，u Finally, the two angle boundaries of the target are measured using the label array to form the target boundary quadrilateral. For the convenience of explanation, the four vertices of the target boundary quadrilateral are defined as Q1, Q2, Q3, and Q4, and an arbitrary point is selected on the straight line segments Q1Q2, Q2Q3, Q3Q4, and Q4Q1 respectively. The four arbitrary points are defined as contour vertices V1, V2, V3, and V4. V1, V2, V3, and V4 are connected in sequence to construct the initial contour of the target. The initial contour can be regarded as composed of four curve segments. The two ends of each curve segment are contour vertices. If the line type of the curve segment is a straight line, the constructed initial contour is called a straight line initial contour. If the line type of the curve segment is an arc, the constructed initial contour is called an arc initial contour. Step 4: Calculate the propagation distance of the signal inside the target using the phase change caused by the target, and perform target image estimation based on the discrete tag array. Based on the linear initial profile and the arc initial profile constructed in step 3, two estimated images based on the discrete tag array can be obtained for each initial profile. The target imaging estimation method based on the discrete tag array can be expressed as follows: First, use the phase change Δφ obtained in data acquisition = {Δφ1, Δφ2, ..., Δφ N } to calculate the propagation distance vector D based on the phase difference, that is, D = (Δφ + 2ξπ) / (β tar -β air ), where each element in D represents the propagation distance inside the target corresponding to the signal collected in a specific direction, λ air and λ tar Represent the signal wavelengths in air and in the target, β air =2π / λ air and β tar =2π / λ tar They represent the phase constants of the signal in the air and in the target respectively. ξ is an integer used to represent the full-cycle ambiguity feature. For objects with an average size smaller than the carrier wavelength, the phase change caused by it does not exceed 2π, so ξ is taken as 0. From the above formula, we can get D1, D2, ..., D N The value of each propagation distance, then, for array 1 and array 2, select the two curve segments closest to the label array from the initial contour as the two end edges of the target estimation image, and assume that the starting points of the propagation distance are located on these two curve segments, according to {θ1, θ2, ..., θ N } and the coordinates of the array to determine the coordinates of the starting point, according to {θ1, θ2, ..., θ N } and D determine the coordinates of the ending points corresponding to each starting point, connect each ending point to obtain another edge segment of the target estimation image, and finally, connect the three edges mentioned above to obtain the target estimation image of array 1 and array 2 on the horizontal section; Step 5: Execute target horizontal cross-section estimation based on maximizing overlap. If there is a difference between the estimated image of array 1 and the estimated image of array 2, it means that the selection of contour vertices V1, V2, V3, and V4 is not ideal. In order to correctly select contour vertices, it is necessary to search for different combinations of V1, V2, V3, and V4 on the four edges Q1Q2, Q2Q3, Q3Q4, and Q4Q1 to construct the target initial contour. That is, the target imaging problem is converted into an optimization problem. The maximization of the overlap of the estimated images of the two tag arrays is used as the optimization goal. The horizontal coordinates x1, x2, x3, and x4 of the four points V1, V2, V3, and V4 are used to construct the set Ω, and Ω is used as the variable to be optimized. The vertical coordinates y1, y2, y3, and y4 of the four points V1, V2, V3, and V4 can be obtained by calculating the expressions of the curve segments in the initial contour and the values ​​of x1, x2, x3, and x4. The optimization model of the UHF RFID target imaging system is constructed and expressed as: s (Ω)=arg(max(F)), where O s (Ω) is the optimal solution of the optimization model, F is the global optimization objective function based on overlap, O s (Ω) = arg(max(F)) means that when F takes its maximum value, the variable value of Ω is taken. In this model, the overlap represents the ratio of the overlapping part of the estimated image obtained by the two label arrays to the merged image. The larger the overlap, the closer the merged image is to the actual image. The particle swarm algorithm is introduced to iteratively search for the optimal solution to the above optimization problem. Step 6: Use deviation and overlap to analyze the impact of the selection of the initial contour on the imaging accuracy. Define deviation as the average distance between the target estimated image contour and the corresponding position points in the actual contour. According to the different target edge types, the targets are divided into two types: targets with edges composed of straight lines and targets with edges composed of arcs. Select representative quadrilateral targets and circular targets as imaging objects. Calculate the deviation and overlap of the target estimated images obtained by using different types of initial contours in the optimization process, and obtain the deviation curve and overlap curve of the arc-shaped initial contour and the straight-line initial contour. Comparing the two types of curves, it can be seen that For quadrilateral targets and circular targets, as the number of iterations gradually increases, the value of the deviation curve gradually decreases and approaches 0, and the value of the overlap curve gradually increases from 0 and approaches 1. The changes of the two types of curves are inversely proportional. Therefore, the size of the overlap is used to characterize the target imaging accuracy. At the same time, for the same imaging target, the overlap curves of the two estimated images obtained by imaging with an arc-shaped initial profile and a straight-line initial profile are significantly different. For circular targets, the arc-shaped initial profile can obtain better imaging accuracy, while the straight-line initial profile can obtain better imaging accuracy for quadrilateral targets. Step 7: For imaging targets with unknown edge types, use a straight line initial profile and an arc-type initial profile to image the target respectively. When imaging with an arc-type initial profile, the maximum image overlap is estimated as C1. When imaging with a straight line initial profile, the maximum image overlap is estimated as C2. Compare C1 and C2 to complete the judgment of the target edge. If C1>C2, the edge of the imaging target is considered to be an arc, and the arc-type initial profile is selected for target imaging. If C1≤C2, the straight line initial profile is selected for target imaging. Further merge the estimated images of array 1 and array 2 to obtain the final horizontal cross-sectional image of the target.

2. The method according to claim 1, characterized in that The method for constructing the straight line initial contour in step 3 is as follows: define the coordinates of points V1, V2, V3 and V4 as (x1, y1), (x2, y2), (x3, y3) and (x4, y4) respectively, and use straight lines to connect the above four points in sequence to construct the straight line initial contour. Assume that the expression of the straight line segment is y=kx+b, x and y represent the horizontal coordinate and vertical coordinate of each point in the straight line segment respectively, k and b represent the slope and intercept respectively, and use the two-point method to calculate the slope and intercept, and then obtain the accurate expression of each straight line segment in the straight line initial contour, with the straight line segment as the starting point. For example, the slope is k = y3-y2 / x3-x2, and the intercept is b = (y2x3-y3x2) / (x3-x2). Combining x∈[x2, x3] and y∈[y2, y3], we can determine the straight line segment The specific expression and geometric characteristics of the other three straight line segments can be completed in the same way. Specific expressions and estimation of geometric characteristics.

3. The method according to claim 1, characterized in that The method for constructing the arc-shaped initial contour in step 3 is as follows: define the coordinates of points V1, V2, V3 and V4 as (x1, y1), (x2, y2), (x3, y3) and (x4, y4) respectively, connect the above four points in sequence with arcs to construct the arc-shaped initial contour, use the circumscribed circle method to calculate the curvature, and then obtain the expression of each curve segment in the arc-shaped initial contour, and use the curve segment For example, select a combination of three points from V1, V2, V3 and V4, and calculate the center (a, b) of the circumscribed circle corresponding to the combination. Taking the three points in the combination as V2, V3 and V4 as an example, a=(y2-y4)f1-(y3-y4)f2 / 2f3, b=(x2-x4)f4-(x3-x4)f5 / 2f6, where f3=(x3-x4)(y2-y4)-(x2-x4)(y3-y4), f6=(y3-y4)(x2-x4)-(y2-y4)(x3-x4), further, the radius r of the circumscribed circle can be obtained by Calculated, the curve segment The curvature of the curve segment is K = 1 / r. Assume that the expression of the curve segment is y = f(x), where x and y represent the horizontal and vertical coordinates of each point in the curve segment, respectively. The first-order derivative and second-order derivative of y with respect to x are y′ and y″, respectively. Whether the arc segment meets the requirements of target contour construction is determined by whether the center of the circle is within the area of ​​the quadrilateral Q1Q2Q3Q4. If the center of the circle is within the area of ​​the quadrilateral Q1Q2Q3Q4, then K = |y″| / (1+(y′) 2 ) 1.5 , x∈[x3,x4], y∈[y3,y4] can determine the specific expression of y=f(x) to describe the curve segment If the center of the circle is not within the area of ​​the quadrilateral Q1Q2Q3Q4, then select another set of three points from V1, V2, V3 and V4 to solve the center and radius to determine the arc segment. The expression and geometric characteristics of the other three arc segments can be completed in the same way. and Estimation of the expression and geometric characteristics of .

Images