Interactive positioning method and system based on LED array

Abstract

The application discloses an interactive positioning method and system based on an LED array, comprising the following steps: acquiring reference coordinates of each light-emitting unit in the LED array and a spatial arrangement relationship thereof, and constructing a reference template; collecting a visible light image of the LED array, obtaining an observation point set through spot extraction and center positioning; performing structural matching on the observation point set based on the spatial arrangement relationship of the reference template, and generating a candidate matching point set; performing RANSAC iteration on the candidate matching point set, and screening to obtain an inlier point set; obtaining transformation parameters between the reference template and the observation point set based on the inlier point set by using an improved Umeyama algorithm; verifying the inlier point set by using the transformation parameters, performing consistency determination in combination with the spatial arrangement relationship, and updating the inlier point set; repeatedly performing transformation solving and inlier updating until a stable condition is met, and outputting a spatial pose result of a terminal in a coordinate system of the LED array. The application can realize stable positioning in an indoor environment.

Description

Technical Field

[0001] This invention relates to the field of visual inspection and optical interaction technology, and in particular to an interactive positioning method and system based on LED arrays. Background Technology

[0002] With the development of indoor positioning and human-computer interaction technologies, positioning methods based on radio signals, inertial sensing, and visual information have been widely applied in indoor navigation, intelligent interaction, and augmented reality scenarios. However, in complex indoor environments, the Global Positioning System (GPS) cannot function effectively. Radio signal-based positioning methods are susceptible to multipath effects and signal blockage, leading to unstable positioning accuracy. Inertial sensing-based positioning methods suffer from error accumulation, making it difficult to maintain accuracy over long periods. Traditional visual positioning methods typically rely on natural features or artificial markers, which suffer from unstable features, susceptibility to changes in lighting, and high computational complexity.

[0003] In recent years, the method of positioning using visible light signals emitted by LED arrays has gradually attracted attention. This type of method achieves spatial positioning by identifying the position of LED light spots. Existing LED array-based positioning methods usually rely on simple light spot detection and geometric calculation. In practical applications, they are easily affected by factors such as ambient stray light, reflection interference, partial occlusion, and light spot adhesion, resulting in mismatches or outliers in the observation points, thus affecting the positioning accuracy. Existing methods mostly use direct geometric calculation or single model fitting, lacking an effective screening mechanism for outliers. When parts of the LED array are missing or the observation is incomplete, positioning failures or unstable results are likely to occur.

[0004] Existing technologies typically use all observation points directly for calculation during similarity transformation solutions, failing to fully utilize the structural relationships between point sets for constraints. This makes the solution results susceptible to interference in the presence of noise or errors. Furthermore, the lack of a feedback verification mechanism for the solution results makes it difficult to achieve dynamic optimization of the point set, further limiting the stability and reliability of the system in complex environments.

[0005] Therefore, how to provide an interactive positioning method and system based on LED arrays is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] One objective of this invention is to propose an interactive positioning method and system based on LED arrays. This invention combines the RANSAC iterative and improved Umeyama algorithm to perform structural constraint matching, outlier screening and iterative optimization on the observation points of the LED array, thereby realizing the spatial pose solution of the terminal in the array coordinate system. It has the advantages of strong anti-interference ability, high positioning accuracy and strong adaptability to complex environments.

[0007] The interactive positioning method and system based on LED array according to embodiments of the present invention includes the following steps:

[0008] Obtain the reference coordinates of each light-emitting unit in the LED array and their relative spatial arrangement relationship, and construct a reference template with unique structural constraints;

[0009] Acquire visible light images of the LED array, extract light spots and locate the center of the visible light images to obtain the observation point set;

[0010] Based on the spatial arrangement relationship of the reference template, structural constraint matching is performed on the observation point set to generate a candidate matching point set that satisfies local arrangement consistency.

[0011] RANSAC iteration is performed on the candidate matching point set. In each iteration, a preset number of matching points are selected from the candidate matching point set as the current sampling subset, and a candidate similarity transformation is constructed based on the sampling subset. The remaining matching points in the candidate matching point set are used as the validation set. The spatial residual of each matching point is calculated according to the candidate similarity transformation, and the consistency of the matching points in the validation set is determined in combination with the spatial arrangement relationship. The matching points that satisfy the global arrangement consistency are determined as the interior point set.

[0012] The interior point set is input into the improved Umeyama algorithm to solve for the similarity transformation parameters between the reference template and the observation point set;

[0013] The interior point set is back-projected and verified using similarity transformation parameters, and the consistency of the verification results is determined based on the spatial arrangement relationship to obtain the updated interior point set.

[0014] The similarity transformation parameters are solved again based on the updated interior point set until the interior point set satisfies the stability condition, and the spatial pose of the terminal in the LED array coordinate system is output.

[0015] Optionally, a reference template with unique structural constraints is constructed, including:

[0016] Obtain the reference coordinates of each light-emitting unit in the LED array in the array coordinate system to form a reference point set;

[0017] Based on the reference point set, the corresponding set of adjacent reference coordinates is determined for each reference coordinate in the reference point set according to the preset neighborhood rules, and the connection order between each reference coordinate in the adjacent reference coordinate set is recorded to obtain the local adjacency relationship corresponding to each reference coordinate.

[0018] Based on the local adjacency relationship of each reference coordinate, the corresponding local structural information is extracted for each reference coordinate in the reference point set.

[0019] Based on the local structural information corresponding to each reference coordinate, the reference coordinates in the reference point set are classified to obtain the initial structural identifier corresponding to each reference coordinate.

[0020] For reference coordinates with the same initial structure identifier, the initial structure identifier is distinguished based on the overall distribution relationship between the reference coordinates in the reference point set, so as to obtain a unique structure identifier corresponding to each reference coordinate.

[0021] By associating the reference point set, the local adjacency relationships corresponding to each reference coordinate, the local structural information corresponding to each reference coordinate, and the unique structural identifier, a reference template with unique structural constraints is constructed.

[0022] Optionally, spot extraction and center localization are performed on the visible light image to obtain a set of observation points, including:

[0023] A continuous visible light image containing LED array light-emitting units is acquired, and the visible light image is subjected to exposure control and image cropping to obtain an image to be processed containing the LED light-emitting area;

[0024] The image to be processed is converted to grayscale, and threshold segmentation is performed on the image to be processed based on the grayscale distribution to obtain a binary spot image;

[0025] Connected region detection is performed based on binary spot images, and each connected region is extracted as a candidate spot region.

[0026] Candidate spot regions are screened based on their area range, boundary continuity, and gray level concentration. Candidate spot regions that do not meet the preset conditions are removed to obtain the target spot region.

[0027] The center position of each target spot area is calculated based on the spatial distribution of pixels within each target spot area, and the center coordinates of the spot area corresponding to each target spot area are obtained.

[0028] The coordinates of the center of each spot are recorded and associated according to their spatial distribution in the image to be processed, forming a set of observation points.

[0029] Optionally, the generation of candidate matching point sets that satisfy local permutation consistency includes:

[0030] Based on the local adjacency relationship of each reference coordinate in the reference template, extract the reference local arrangement structure corresponding to each reference coordinate;

[0031] Based on the spatial distribution relationship of the center coordinates of each spot in the observation point set in the image to be processed, the corresponding adjacent observation points are determined for the center coordinates of each spot, and the observation local arrangement structure corresponding to the center coordinates of each spot is generated.

[0032] The reference local arrangement structure is matched one by one with the observed local arrangement structure. The matching results are filtered according to the number of adjacent points, connection order and relative position relationship to obtain the initial matching point pairs.

[0033] Based on the initial matching point pairs, determine the correspondence consistency between the observed local arrangement structure and the reference local arrangement structure in each matching point pair, and remove matching point pairs that do not meet the consistency condition;

[0034] The matching point pairs that meet the consistency condition are summarized to form a candidate matching point set that meets the local permutation consistency.

[0035] Optionally, determining the interior point set includes:

[0036] Obtain each pair of matching points in the candidate matching point set, and extract the corresponding reference coordinates and spot center coordinates from each pair of matching points;

[0037] Based on the reference coordinates and spot center coordinates of each matching point pair, establish the correspondence between matching point pairs in the candidate matching point set;

[0038] Based on the correspondence of matching point pairs, the number of samples and the preset number of iterations are set for each iteration, and in each iteration, matching point pairs corresponding to the number of samples are selected from the candidate matching point set to form the current sampling subset;

[0039] Based on the correspondence between the reference coordinates of each matching point pair in the current sampling subset and the coordinates of the spot center, the current candidate similarity transformation is calculated;

[0040] Based on the inclusion relationship between the candidate matching point set and the current sampling subset, the remaining matching point pairs in the candidate matching point set excluding the current sampling subset are determined to form the current verification set;

[0041] Based on the current candidate similarity transformation, coordinate mapping is performed on the reference coordinates of each matching point pair in the current validation set to obtain the mapped coordinates corresponding to each matching point pair in the current validation set;

[0042] Based on the mapped coordinates and the corresponding spot center coordinates of each matching point pair in the current verification set, calculate the spatial deviation of each matching point pair in the current verification set;

[0043] Based on spatial bias, a consistency determination is made for each pair of matching points in the current validation set to obtain the current set of inliers;

[0044] Based on the current set of interior points obtained in each iteration, the corresponding current candidate similarity transformations are compared to determine the optimal similarity transformation and the optimal set of interior points corresponding to the optimal similarity transformation.

[0045] When the preset number of iterations is reached or the optimal interior point set satisfies the preset stability condition, the optimal similarity transformation and interior point set are output.

[0046] Optional, improved versions of the Umeyama algorithm include:

[0047] Obtain each matching point pair in the inlier set, and extract the corresponding reference coordinates and spot center coordinates from each matching point pair;

[0048] Based on the local adjacency relationship corresponding to each reference coordinate in the reference template, reference local topology information containing the number of adjacent points, connection order and relative position relationship is extracted for each reference coordinate. Based on the spatial distribution relationship of the center coordinates of each spot in the observation point set, observation local topology information containing the number of adjacent points, connection order and relative position relationship is extracted for the center coordinates of each spot.

[0049] Based on the consistency comparison between the reference local topology information and the observed local topology information corresponding to each matching point pair, the matching point pairs with the same number of adjacent points, the same connection order, and the same relative positional relationship are determined as the set of structurally stable point pairs. The remaining matching point pairs are determined as the set of point pairs to be verified.

[0050] Based on the reference coordinates and spot center coordinates in the set of stable point pairs, the reference center coordinates and observation center coordinates are calculated respectively, and the set of stable point pairs is centered based on the reference center coordinates and observation center coordinates.

[0051] The covariance matrix is ​​constructed based on the set of stable points after centering, and the singular value decomposition is performed on the covariance matrix to obtain the initial rotation matrix, initial scale factor and initial translation vector, thus forming the initial similarity transformation parameters.

[0052] Based on the initial similarity transformation parameters, the reference coordinates in the set of stable point pairs and the set of point pairs to be verified are projected forward to obtain the corresponding forward mapped coordinates, and the center coordinates of the light spots in the set of stable point pairs and the set of point pairs to be verified are projected backward to obtain the corresponding backward mapped coordinates.

[0053] Based on the positive deviation between the forward-mapped coordinates and the corresponding spot center coordinates, the negative deviation between the reverse-mapped coordinates and the corresponding reference coordinates, and the reference local topology information and observation local topology information corresponding to each matching point pair, the matching point pairs in the set of structurally stable point pairs are retained or removed, and the matching point pairs in the set of point pairs to be verified are backfilled to obtain the updated set of stable point pairs.

[0054] Based on the updated set of stable point pairs, the calculation of reference center coordinates and observation center coordinates, centering, covariance matrix construction and singular value decomposition are re-executed to obtain the updated rotation matrix, updated scale factor and updated translation vector, thus forming the updated similarity transformation parameters.

[0055] If the updated set of stable point pairs does not meet the preset stability conditions, the forward projection, backward projection, retention or removal, and backfilling determination are repeatedly performed based on the updated similarity transformation parameters.

[0056] When the updated set of stable point pairs satisfies the preset stability conditions, the updated similarity transformation parameters are output as similarity transformation parameters between the reference template and the observation point set.

[0057] Optionally, the generation of the updated interior point set includes:

[0058] Based on the similarity transformation parameters, coordinate transformation is performed on each reference coordinate in the interior point set to obtain the projected coordinates corresponding to each reference coordinate.

[0059] Based on the positional differences between each projection coordinate and the corresponding light spot center coordinate, the projection deviation results for each matching point pair are generated.

[0060] Based on the spatial distribution relationship of the center coordinates of each spot in the observation point set in the image to be processed, the corresponding adjacent observation points are determined for the center coordinates of each spot in the in-point set, and the observation arrangement relationship corresponding to the center coordinates of each spot is generated.

[0061] Based on the local adjacency relationship of each reference coordinate in the reference template, the corresponding reference arrangement relationship is determined for each reference coordinate in the interior point set;

[0062] Based on the projection deviation results, observation arrangement relationship and reference arrangement relationship of each matching point pair, the consistency of each matching point pair in the in-point set is determined, and the matching point pairs whose projection deviation results meet the preset deviation conditions and whose observation arrangement relationship is consistent with the reference arrangement relationship are determined as valid matching point pairs.

[0063] Based on valid matching point pairs, update the composition of matching point pairs in the interior point set to generate an updated interior point set.

[0064] Optionally, the output of the terminal's spatial pose in the LED array coordinate system includes:

[0065] Based on each matching point pair in the updated interior point set, the similarity transformation parameters are solved to obtain the current similarity transformation parameters;

[0066] Based on the current similarity transformation parameters, a consistency check is performed on the updated interior point set, and the matching point pairs in the updated interior point set are updated according to the check results to obtain the current interior point set;

[0067] Based on the current set of interior points, re-execute the similarity transformation parameter solution and perform consistency verification and interior point set update;

[0068] Repeat the similarity transformation parameter solution and interior point set update until the interior point set satisfies the preset stability condition;

[0069] When the inlier set satisfies the preset stability conditions, the spatial pose of the terminal in the LED array coordinate system is determined based on the corresponding similarity transformation parameters, and the spatial pose is output.

[0070] Optionally, the LED array-based interactive positioning system, performing the LED array-based interactive positioning method according to any one of claims 1 to 8, includes:

[0071] The reference template construction module is used to obtain the reference coordinates of each light-emitting unit in the LED array and their relative spatial arrangement relationship, and construct a reference template with unique structural constraints.

[0072] The image acquisition and processing module is used to acquire visible light images containing LED array light-emitting units, and to extract light spots and locate the center of the visible light images to generate a set of observation points;

[0073] The structure matching module is used to perform structural constraint matching on the observation point set based on the spatial arrangement relationship of the reference template, and generate a candidate matching point set that satisfies the local arrangement consistency.

[0074] The robust estimation module is used to perform RANSAC iterations on the candidate matching point set to obtain the interior point set.

[0075] The similarity transformation solution module is used to execute the improved Umeyama algorithm based on the interior point set to solve for the similarity transformation parameters between the reference template and the observation point set;

[0076] The interior point update module is used to perform back projection verification on the interior point set based on similarity transformation parameters, and to determine the consistency of the verification results based on spatial arrangement relationship, thereby generating an updated interior point set.

[0077] The pose output module is used to re-execute the similarity transformation parameter solution based on the updated interior point set until the interior point set satisfies the stability condition, and output the spatial pose result of the terminal in the LED array coordinate system.

[0078] The beneficial effects of this invention are:

[0079] This invention establishes a stable correspondence between the observation point set and the reference template by constructing a reference template based on the spatial arrangement of LED arrays and combining it with a structural constraint matching method. This effectively avoids matching errors caused by misidentification, occlusion, and illumination interference in the traditional spot matching process.

[0080] This invention introduces RANSAC iteration into the candidate matching point set to perform multiple rounds of sampling and verification on the matching point pairs, thereby filtering out the set of internal points that meet the consistency conditions, effectively eliminating outliers and mismatched points, and improving the reliability of the point set participating in subsequent calculations.

[0081] This invention introduces a point set selection based on local topological relationships and a closed-loop verification mechanism combining forward and backward projection into the traditional solution process through an improved Umeyama algorithm. This enables the point set involved in the solution to be dynamically updated, thereby improving the stability and accuracy of similarity transformation parameter solution.

[0082] This invention achieves the continuity of positioning results by maintaining a valid point set in the calculation even when the LED array is partially missing or the observation is incomplete, through a consistency determination based on spatial arrangement relationship and an interior point set update mechanism. By iteratively performing similar transformation solution and interior point set update until the stability condition is met, a stable terminal spatial pose result can be output, thereby achieving high-precision and interference-resistant interactive positioning effect in complex indoor environments. Attached Figure Description

[0083] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0084] Figure 1 The flowchart shows the interactive positioning method and system based on LED array proposed in this invention.

[0085] Figure 2 This is a schematic diagram of the RANSAC iterative solution process in this invention;

[0086] Figure 3 This is a schematic diagram of the improved Umeyama algorithm processing flow in this invention. Detailed Implementation

[0087] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0088] refer to Figures 1-3 An interactive positioning method and system based on LED arrays includes the following steps:

[0089] Obtain the reference coordinates of each light-emitting unit in the LED array and their relative spatial arrangement relationship, and construct a reference template with unique structural constraints;

[0090] Acquire visible light images of the LED array, extract light spots and locate the center of the visible light images to obtain the observation point set;

[0091] Based on the spatial arrangement relationship of the reference template, structural constraint matching is performed on the observation point set to generate a candidate matching point set that satisfies local arrangement consistency.

[0092] RANSAC iteration is performed on the candidate matching point set. In each iteration, a preset number of matching points are selected from the candidate matching point set as the current sampling subset, and a candidate similarity transformation is constructed based on the sampling subset. The remaining matching points in the candidate matching point set are used as the validation set. The spatial residual of each matching point is calculated according to the candidate similarity transformation, and the consistency of the matching points in the validation set is determined in combination with the spatial arrangement relationship. The matching points that satisfy the global arrangement consistency are determined as the interior point set.

[0093] The interior point set is input into the improved Umeyama algorithm to solve for the similarity transformation parameters between the reference template and the observation point set;

[0094] The interior point set is back-projected and verified using similarity transformation parameters, and the consistency of the verification results is determined based on the spatial arrangement relationship to obtain the updated interior point set.

[0095] The similarity transformation parameters are solved again based on the updated interior point set until the interior point set satisfies the stability condition, and the spatial pose of the terminal in the LED array coordinate system is output.

[0096] In this embodiment, a reference template with unique structural constraints is constructed, including:

[0097] Obtain the reference coordinates of each light-emitting unit in the LED array in the array coordinate system to form a reference point set;

[0098] Based on the reference point set, the corresponding set of adjacent reference coordinates is determined for each reference coordinate in the reference point set according to the preset neighborhood rules, and the connection order between each reference coordinate in the adjacent reference coordinate set is recorded to obtain the local adjacency relationship corresponding to each reference coordinate.

[0099] Specifically, the preset neighborhood rule takes each reference coordinate in the reference point set as the center, selects other reference coordinates whose distance is less than the threshold as adjacent reference coordinates according to a fixed distance threshold, and determines the connection order according to the polar angle of the adjacent reference coordinates relative to the reference coordinate.

[0100] Based on the local adjacency relationship of each reference coordinate, the corresponding local structural information is extracted for each reference coordinate in the reference point set.

[0101] Based on the local structural information corresponding to each reference coordinate, the reference coordinates in the reference point set are classified to obtain the initial structural identifier corresponding to each reference coordinate.

[0102] For reference coordinates with the same initial structure identifier, the initial structure identifier is distinguished based on the overall distribution relationship between the reference coordinates in the reference point set, so as to obtain a unique structure identifier corresponding to each reference coordinate.

[0103] By associating the reference point set, the local adjacency relationships corresponding to each reference coordinate, the local structural information corresponding to each reference coordinate, and the unique structural identifier, a reference template with unique structural constraints is constructed.

[0104] In this embodiment, spot extraction and center localization are performed on the visible light image to obtain a set of observation points, including:

[0105] A continuous visible light image containing LED array light-emitting units is acquired, and the visible light image is subjected to exposure control and image cropping to obtain an image to be processed containing the LED light-emitting area;

[0106] The image to be processed is converted to grayscale, and threshold segmentation is performed on the image to be processed based on the grayscale distribution to obtain a binary spot image;

[0107] Connected region detection is performed based on binary spot images, and each connected region is extracted as a candidate spot region.

[0108] Specifically, the exposure control involves adjusting the exposure time of the image acquisition device based on the pixel grayscale value distribution in the visible light image to ensure that the pixel grayscale values ​​of the LED emitting area are within a preset grayscale range; the image cropping involves determining the boundary of the emitting area based on the area with higher brightness than the background in the visible light image, and cropping this area to form the image to be processed; the grayscale processing involves converting each pixel of the image to be processed from multi-channel values ​​to single-channel grayscale values; the threshold segmentation involves determining the segmentation threshold based on the grayscale histogram of the image to be processed, classifying pixels with grayscale values ​​higher than the threshold as light spot areas, and classifying the rest as background areas; and the connected component detection involves determining the connectivity of adjacent pixels in the binary image, and classifying the connected light spot pixel set into independent regions.

[0109] Candidate spot regions are screened based on their area range, boundary continuity, and gray level concentration. Candidate spot regions that do not meet the preset conditions are removed to obtain the target spot region.

[0110] Specifically, for each candidate spot region, the number of pixels it contains is counted as its area value, and it is determined whether the area value is within a preset area range; the boundary pixel sequence of the candidate spot region is extracted, and the spatial continuity of the boundary pixels is determined, with the number of boundary break points not exceeding a preset threshold as the boundary continuity judgment condition; at the same time, the distribution of pixel grayscale values ​​in the candidate spot region is counted, the proportion of high grayscale pixels is calculated, and it is determined whether the proportion is greater than a preset proportion threshold; the candidate spot regions that simultaneously meet the area range, boundary continuity, and grayscale concentration judgment conditions are retained as target spot regions, and the remaining candidate spot regions are eliminated;

[0111] The center position of each target spot area is calculated based on the spatial distribution of pixels within each target spot area, and the center coordinates of the spot area corresponding to each target spot area are obtained.

[0112] The coordinates of the center of each spot are recorded and associated according to their spatial distribution in the image to be processed, forming a set of observation points;

[0113] Specifically, the coordinates of pixels within each target spot area are statistically analyzed, and the average value of all pixel coordinates is calculated to obtain the corresponding spot center coordinates. The spatial distribution relationship is as follows: based on the relative position of each spot center coordinate in the image to be processed, the center coordinates of each spot are sorted according to the size of the horizontal and vertical coordinates, and their adjacent relationships are recorded.

[0114] In this embodiment, the generation of a candidate matching point set that satisfies local permutation consistency includes:

[0115] Based on the local adjacency relationship of each reference coordinate in the reference template, extract the reference local arrangement structure corresponding to each reference coordinate;

[0116] Based on the spatial distribution relationship of the center coordinates of each spot in the observation point set in the image to be processed, the corresponding adjacent observation points are determined for the center coordinates of each spot, and the observation local arrangement structure corresponding to the center coordinates of each spot is generated.

[0117] The reference local arrangement structure is matched one by one with the observed local arrangement structure. The matching results are filtered according to the number of adjacent points, connection order and relative position relationship to obtain the initial matching point pairs.

[0118] Specifically, the reference coordinates in each reference local arrangement structure are used as benchmarks and compared with the center coordinates of the light spots in each observation local arrangement structure. The number of adjacent points in the corresponding structure is compared to see if they are consistent. If the number of adjacent points is consistent, the arrangement order of adjacent points is compared one by one according to the connection order. If the connection order is consistent, the relative positional relationship between adjacent points is compared to see if they are consistent. Matching results that simultaneously satisfy the consistency of the number of adjacent points, connection order, and relative positional relationship are retained, and other matching results are discarded.

[0119] Based on the initial matching point pairs, determine the correspondence consistency between the observed local arrangement structure and the reference local arrangement structure in each matching point pair, and remove matching point pairs that do not meet the consistency condition;

[0120] Specifically, for each initial matching point pair, its corresponding reference local arrangement structure and observed local arrangement structure are extracted. The relative position direction and arrangement order between each adjacent point in the two are compared one by one, and it is determined whether the relative position relationship between the corresponding adjacent points is consistent. On the basis of consistent relative position relationship, it is calculated whether the offset difference between the corresponding adjacent points in the two is less than the preset deviation range. Matching point pairs that simultaneously meet the conditions of consistent arrangement order and satisfactory offset difference are retained, and the remaining matching point pairs are removed.

[0121] The matching point pairs that meet the consistency condition are summarized to form a candidate matching point set that meets the local permutation consistency.

[0122] In this embodiment, the determination of the interior point set includes:

[0123] Obtain each pair of matching points in the candidate matching point set, and extract the corresponding reference coordinates and spot center coordinates from each pair of matching points;

[0124] Based on the reference coordinates and spot center coordinates of each matching point pair, establish the correspondence between matching point pairs in the candidate matching point set;

[0125] Based on the correspondence of matching point pairs, the number of samples and the preset number of iterations are set for each iteration, and in each iteration, matching point pairs corresponding to the number of samples are selected from the candidate matching point set to form the current sampling subset;

[0126] Specifically, the minimum number of matching point pairs that can uniquely determine the similarity transformation relationship is selected from the candidate matching point set; the preset number of iterations is: the upper limit of iteration determined by the total number of matching point pairs in the candidate matching point set and the preset proportion of internal points; the selection method is: in each iteration, matching point pairs consistent with the sampling number are randomly selected from the candidate matching point set in a non-repeating manner to form the current sampling subset;

[0127] Based on the correspondence between the reference coordinates of each matching point pair in the current sampling subset and the coordinates of the spot center, the current candidate similarity transformation is calculated;

[0128] Based on the inclusion relationship between the candidate matching point set and the current sampling subset, the remaining matching point pairs in the candidate matching point set excluding the current sampling subset are determined to form the current verification set;

[0129] Specifically, taking the candidate matching point set as the whole set, the selected matching point pairs in the current sampling subset are removed from the candidate matching point set, and the remaining unselected matching point pairs are summarized to obtain the remaining matching point pairs in the candidate matching point set excluding the current sampling subset, and the remaining matching point pairs are used as the current verification set.

[0130] Based on the current candidate similarity transformation, coordinate mapping is performed on the reference coordinates of each matching point pair in the current validation set to obtain the mapped coordinates corresponding to each matching point pair in the current validation set;

[0131] Based on the mapped coordinates and the corresponding spot center coordinates of each matching point pair in the current verification set, calculate the spatial deviation of each matching point pair in the current verification set;

[0132] Based on spatial bias, a consistency determination is made for each pair of matching points in the current validation set to obtain the current set of inliers;

[0133] Specifically, the spatial deviations of each matching point pair in the current verification set are compared. Matching point pairs with spatial deviations less than a preset deviation threshold are determined to be consistent matching point pairs, while matching point pairs with spatial deviations greater than or equal to the preset deviation threshold are determined to be inconsistent matching point pairs. All consistent matching point pairs are then summarized to form the current inlier set.

[0134] Based on the current set of interior points obtained in each iteration, the corresponding current candidate similarity transformations are compared to determine the optimal similarity transformation and the optimal set of interior points corresponding to the optimal similarity transformation.

[0135] Specifically, in each iteration, the number of matching point pairs in the corresponding current interior point set is recorded. The current interior point set with the largest number of matching point pairs is taken as the optimal interior point set, and the current candidate similarity transformation corresponding to the generation of the optimal interior point set is determined as the optimal similarity transformation. When there are multiple current interior point sets with the same number of matching point pairs, the sum of spatial deviations of the corresponding matching point pairs is further compared, and the current candidate similarity transformation corresponding to the one with the smaller sum of spatial deviations is selected as the optimal similarity transformation.

[0136] When the preset number of iterations is reached or the optimal interior point set satisfies the preset stability condition, the optimal similarity transformation and interior point set are output.

[0137] In this embodiment, the improved Umeyama algorithm includes:

[0138] Obtain each matching point pair in the inlier set, and extract the corresponding reference coordinates and spot center coordinates from each matching point pair;

[0139] Based on the local adjacency relationship corresponding to each reference coordinate in the reference template, reference local topology information containing the number of adjacent points, connection order and relative position relationship is extracted for each reference coordinate. Based on the spatial distribution relationship of the center coordinates of each spot in the observation point set, observation local topology information containing the number of adjacent points, connection order and relative position relationship is extracted for the center coordinates of each spot.

[0140] Based on the consistency comparison between the reference local topology information and the observed local topology information corresponding to each matching point pair, the matching point pairs with the same number of adjacent points, the same connection order, and the same relative positional relationship are determined as the set of structurally stable point pairs. The remaining matching point pairs are determined as the set of point pairs to be verified.

[0141] Based on the reference coordinates and spot center coordinates in the set of stable point pairs, the reference center coordinates and observation center coordinates are calculated respectively, and the set of stable point pairs is centered based on the reference center coordinates and observation center coordinates.

[0142] Specifically, the reference center coordinates are subtracted from the reference coordinates of each matching point pair in the set of structurally stable point pairs to obtain the centered reference coordinates, and the observation center coordinates are subtracted from the spot center coordinates of each matching point pair to obtain the centered observation coordinates, so as to eliminate the overall translation effect.

[0143] The covariance matrix is ​​constructed based on the set of stable points after centering, and the singular value decomposition is performed on the covariance matrix to obtain the initial rotation matrix, initial scale factor and initial translation vector, thus forming the initial similarity transformation parameters.

[0144] Specifically, based on the set of structurally stable point pairs after centralization, the centralized observation coordinates in each matching point pair are multiplied by the corresponding centralized reference coordinates and summed to form a covariance matrix; singular value decomposition is performed by decomposing the covariance matrix to obtain the corresponding orthogonal matrix and singular value matrix, and the initial rotation matrix is ​​determined based on the orthogonal matrix; the initial scale factor is determined based on the distribution relationship between the singular value matrix and the centralized reference coordinates, and the initial translation vector is calculated by combining the reference center coordinates and the observation center coordinates;

[0145] Based on the initial similarity transformation parameters, the reference coordinates in the set of stable point pairs and the set of point pairs to be verified are projected forward to obtain the corresponding forward mapped coordinates, and the center coordinates of the light spots in the set of stable point pairs and the set of point pairs to be verified are projected backward to obtain the corresponding backward mapped coordinates.

[0146] Specifically, forward projection involves substituting the reference coordinates from the set of stable point pairs and the set of point pairs to be verified into the initial similarity transformation parameters to perform coordinate transformation and obtain the corresponding forward mapping coordinates; backward projection involves substituting the center coordinates of the light spots from the set of stable point pairs and the set of point pairs to be verified into the inverse transformation of the initial similarity transformation parameters to perform coordinate transformation and obtain the corresponding backward mapping coordinates.

[0147] Based on the positive deviation between the forward-mapped coordinates and the corresponding spot center coordinates, the negative deviation between the reverse-mapped coordinates and the corresponding reference coordinates, and the reference local topology information and observation local topology information corresponding to each matching point pair, the matching point pairs in the set of structurally stable point pairs are retained or removed, and the matching point pairs in the set of point pairs to be verified are backfilled to obtain the updated set of stable point pairs.

[0148] Specifically, retention or elimination involves comparing the positive and negative deviations of each matching point pair in the set of structurally stable point pairs, retaining matching point pairs whose positive and negative deviations are both less than the corresponding preset thresholds, and eliminating the remaining matching point pairs; the replenishment judgment involves judging the positive and negative deviations of each matching point pair in the set of point pairs to be verified, and adding matching point pairs that simultaneously meet the preset threshold conditions and whose corresponding reference local topology information is consistent with the observed local topology information to the set of structurally stable point pairs.

[0149] Based on the updated set of stable point pairs, the calculation of reference center coordinates and observation center coordinates, centering, covariance matrix construction and singular value decomposition are re-executed to obtain the updated rotation matrix, updated scale factor and updated translation vector, thus forming the updated similarity transformation parameters.

[0150] If the updated set of stable point pairs does not meet the preset stability conditions, the forward projection, backward projection, retention or removal, and backfilling determination are repeatedly performed based on the updated similarity transformation parameters.

[0151] Specifically, when the change in the number of matching point pairs in the stable point pair set obtained from two consecutive iterations does not exceed a preset number threshold, and the change between the corresponding updated similarity transformation parameters is less than a preset parameter change threshold, the stable point pair set is determined to meet the stability condition.

[0152] When the updated set of stable point pairs satisfies the preset stability conditions, the updated similarity transformation parameters are output as similarity transformation parameters between the reference template and the observation point set.

[0153] In this embodiment, the generation of the updated interior point set includes:

[0154] Based on the similarity transformation parameters, coordinate transformation is performed on each reference coordinate in the interior point set to obtain the projected coordinates corresponding to each reference coordinate.

[0155] Based on the positional differences between each projection coordinate and the corresponding light spot center coordinate, the projection deviation results for each matching point pair are generated.

[0156] Specifically, the coordinate difference between each projected coordinate and the corresponding spot center coordinate is calculated to obtain the difference in each coordinate axis direction, and the spatial offset between the projected coordinate and the corresponding spot center coordinate is characterized based on the combination of the differences in each direction.

[0157] Based on the spatial distribution relationship of the center coordinates of each spot in the observation point set in the image to be processed, the corresponding adjacent observation points are determined for the center coordinates of each spot in the in-point set, and the observation arrangement relationship corresponding to the center coordinates of each spot is generated.

[0158] Based on the local adjacency relationship of each reference coordinate in the reference template, the corresponding reference arrangement relationship is determined for each reference coordinate in the interior point set;

[0159] Based on the projection deviation results, observation arrangement relationship and reference arrangement relationship of each matching point pair, the consistency of each matching point pair in the in-point set is determined, and the matching point pairs whose projection deviation results meet the preset deviation conditions and whose observation arrangement relationship is consistent with the reference arrangement relationship are determined as valid matching point pairs.

[0160] Specifically, the positional difference between each pair of matching points is less than a preset deviation threshold; the observed arrangement relationship is considered consistent with the reference arrangement relationship when the number of adjacent points in the corresponding pair of matching points is consistent, the connection order is consistent, and the relative positional relationship is consistent.

[0161] Based on the valid matching point pairs, update the composition of matching point pairs in the interior point set to generate the updated interior point set;

[0162] Specifically, the matching point pairs that are determined to be valid matching point pairs are retained from the inner point set, and the matching point pairs that do not meet the consistency determination conditions are removed from the inner point set, so that the retained valid matching point pairs can be reconstructed into the updated inner point set.

[0163] In this embodiment, the output of the terminal's spatial pose result in the LED array coordinate system includes:

[0164] Based on each matching point pair in the updated interior point set, the similarity transformation parameters are solved to obtain the current similarity transformation parameters;

[0165] Based on the current similarity transformation parameters, a consistency check is performed on the updated interior point set, and the matching point pairs in the updated interior point set are updated according to the check results to obtain the current interior point set;

[0166] Based on the current set of interior points, re-execute the similarity transformation parameter solution and perform consistency verification and interior point set update;

[0167] Repeat the similarity transformation parameter solution and interior point set update until the interior point set satisfies the preset stability condition;

[0168] When the inlier set satisfies the preset stability conditions, the spatial pose of the terminal in the LED array coordinate system is determined based on the corresponding similarity transformation parameters, and the spatial pose is output.

[0169] Specifically, the rotation, translation, and scale relationships in the similarity transformation parameters are used as the spatial transformation relationships of the terminal relative to the LED array. The rotation relationship represents the orientation of the terminal, the translation relationship represents the position coordinates of the terminal, and the combination of the two is expressed as the spatial pose result of the terminal in the LED array coordinate system.

[0170] In this embodiment, the interactive positioning system based on an LED array executes the interactive positioning method based on an LED array as described in any one of claims 1 to 8, including:

[0171] The reference template construction module is used to obtain the reference coordinates of each light-emitting unit in the LED array and their relative spatial arrangement relationship, and construct a reference template with unique structural constraints.

[0172] The image acquisition and processing module is used to acquire visible light images containing LED array light-emitting units, and to extract light spots and locate the center of the visible light images to generate a set of observation points;

[0173] The structure matching module is used to perform structural constraint matching on the observation point set based on the spatial arrangement relationship of the reference template, and generate a candidate matching point set that satisfies the local arrangement consistency.

[0174] The robust estimation module is used to perform RANSAC iterations on the candidate matching point set to obtain the interior point set.

[0175] The similarity transformation solution module is used to execute the improved Umeyama algorithm based on the interior point set to solve for the similarity transformation parameters between the reference template and the observation point set;

[0176] The interior point update module is used to perform back projection verification on the interior point set based on similarity transformation parameters, and to determine the consistency of the verification results based on spatial arrangement relationship, thereby generating an updated interior point set.

[0177] The pose output module is used to re-execute the similarity transformation parameter solution based on the updated interior point set until the interior point set satisfies the stability condition, and output the spatial pose result of the terminal in the LED array coordinate system.

[0178] Example 1:

[0179] To verify the feasibility of this invention in practice, it was applied to a large indoor exhibition space. To achieve real-time interaction between visitors and the exhibits, high-precision positioning of handheld devices was required. Multiple LED light-emitting units were evenly arranged in a regular array on the ceiling of the space. Each LED unit was installed at a predetermined position and possessed stable light-emitting characteristics. Due to the complex ambient light sources in this scene, including spotlights, reflective surfaces, and pedestrian obstructions, traditional positioning methods based on visual features or simple light spot recognition struggled to obtain stable and reliable positioning results. Especially in cases of localized light spot loss or misidentification, positioning drift or even failure was prone to occur.

[0180] In this scenario, the interactive positioning method based on LED arrays described in this invention is applied to the positioning process of a terminal device. The terminal device continuously acquires visible light images of the LED array area through its built-in image acquisition module and processes the images to extract the center position of the light spots. Due to strong stray light and local reflections in the environment, the initially extracted light spots contain a certain number of misidentified points. For example, in actual testing, approximately 120 candidate light spots are extracted per frame on average, of which approximately 96 are real LED light spots, and the rest are reflected light spots or noise points.

[0181] After spot extraction, the system performs structural constraint matching on the observation point set based on a pre-constructed reference template. By analyzing the spatial arrangement of each emitting unit in the reference template, the spots in the observation points are matched according to local adjacency relationships. In this process, by filtering the number of adjacent points, connection order, and relative positional relationships, approximately 85 matching point pairs are initially obtained, while most obvious outliers are effectively eliminated. Compared to the case without structural constraint matching, the matching accuracy is improved from approximately 78% to approximately 91%.

[0182] After obtaining the candidate matching point set, the system performs RANSAC iterative processing. In each iteration, the minimum number of matching point pairs from the candidate matching point set is selected as a sampling subset. A similarity transformation model is constructed using this subset, and the remaining points are used as a validation set for consistency determination. During the experiment, the number of iterations was set to 200, with an average of 3 sets of matching point pairs selected as the sampling subset per iteration. Even with approximately 15% of outlier matching points, after RANSAC processing, the average size of the interior point set stabilized at around 78 sets, achieving an outlier removal rate of approximately 93%. Without this method, the positioning error caused by outlier points in the calculation could reach over 0.25 meters, while the error significantly decreased after introducing this step.

[0183] After obtaining the interior point set, the system uses an improved Umeyama algorithm to process the point set. Based on local topological relationships, the interior point set is further filtered, with matching point pairs satisfying structural consistency classified as stable point pairs, and the remaining point pairs used as the set to be verified. In this test, of the approximately 78 point pairs in the initial interior point set, about 65 were classified as structurally stable point pairs, and the remaining point pairs entered the set to be verified. Based on the stable point pair set, an initial similarity transformation is performed to obtain the initial rotation, translation, and scale parameters.

[0184] The similarity transformation parameters are used to perform forward and backward projection processing on the point set. In forward projection, the reference coordinates are mapped to the observation coordinate space; in backward projection, the observation coordinates are mapped backward to the reference space. By calculating the deviation between the projection results and the actual coordinates, matching point pairs are further filtered. In this process, about 5 point pairs in the stable point pair set are removed because the deviation exceeds the threshold, while about 8 point pairs in the set to be verified meet the consistency condition and are reinstated into the stable point pair set, ultimately maintaining the size of the stable point pair set at approximately 68 pairs.

[0185] Based on the updated set of stable point pairs, the similarity transformation is performed again, and the projection verification and point set update process described above is repeated. In two consecutive iterations, the size of the stable point pair set changes by no more than two sets of point pairs, and the change in the similarity transformation parameters is less than a preset threshold, satisfying the stability condition, at which point the iteration process stops. The final similarity transformation parameters are used to determine the spatial pose of the terminal device in the LED array coordinate system.

[0186] In this scenario, by continuously testing 500 sets of positioning data and statistically analyzing the positioning error, the results showed that under complex lighting and occlusion conditions, the system's average positioning error was 0.028 meters, with a standard deviation of 0.009 meters; while without the method of this invention, the average error was 0.162 meters, with a standard deviation of 0.047 meters. Regarding the positioning success rate, the method of this invention achieved 98.6% in the test, while the comparative method only achieved 82.3%. Even with approximately 20% of the LEDs obscured, the method of this invention still maintained a positioning success rate of approximately 95%, while the comparative method dropped to approximately 68%.

[0187] In terms of processing efficiency, under the condition that the terminal device has the processing power of a conventional mobile processor, the average processing time per frame is 18 milliseconds, of which spot extraction takes about 6 milliseconds, structure matching about 4 milliseconds, RANSAC processing about 5 milliseconds, and improved similarity transformation solution and update about 3 milliseconds, which generally meets the requirements of real-time positioning. In the comparative experiment, the algorithm without structural constraints and closed-loop optimization had an average processing time of 15 milliseconds, but due to more mismatches, additional correction steps were required, and the overall response latency actually increased.

[0188] As demonstrated by the above applications, the method of this invention can effectively solve the positioning instability problems caused by spot misidentification, occlusion, and environmental interference in complex indoor environments. Through structural constraint matching, robust point set selection, and improved similarity transformation solution and closed-loop optimization mechanisms, the point set participating in the positioning calculation is made more reliable, significantly improving positioning accuracy and stability. While ensuring controllable computational complexity, a balance between high accuracy and real-time performance is achieved, meeting the application requirements of practical interactive positioning scenarios.

[0189] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An interactive positioning method based on LED arrays, characterized in that, include: Obtain the reference coordinates of each light-emitting unit in the LED array and their relative spatial arrangement relationship, and construct a reference template with unique structural constraints; Acquire visible light images of the LED array, extract light spots and locate the center of the visible light images to obtain the observation point set; Based on the spatial arrangement relationship of the reference template, structural constraint matching is performed on the observation point set to generate a candidate matching point set that satisfies local arrangement consistency. RANSAC iteration is performed on the candidate matching point set. In each iteration, a preset number of matching points are selected from the candidate matching point set as the current sampling subset, and a candidate similarity transformation is constructed based on the sampling subset. The remaining matching points in the candidate matching point set are used as the validation set. The spatial residual of each matching point is calculated according to the candidate similarity transformation, and the consistency of the matching points in the validation set is determined in combination with the spatial arrangement relationship. The matching points that satisfy the global arrangement consistency are determined as the interior point set. The interior point set is input into the improved Umeyama algorithm to solve for the similarity transformation parameters between the reference template and the observation point set; The interior point set is back-projected and verified using similarity transformation parameters, and the consistency of the verification results is determined based on the spatial arrangement relationship to obtain the updated interior point set. The similarity transformation parameters are solved again based on the updated interior point set until the interior point set satisfies the stability condition, and the spatial pose of the terminal in the LED array coordinate system is output.

2. The interactive positioning method based on LED array according to claim 1, characterized in that, Construct a reference template with unique structural constraints, including: Obtain the reference coordinates of each light-emitting unit in the LED array in the array coordinate system to form a reference point set; Based on the reference point set, the corresponding set of adjacent reference coordinates is determined for each reference coordinate in the reference point set according to the preset neighborhood rules, and the connection order between each reference coordinate in the adjacent reference coordinate set is recorded to obtain the local adjacency relationship corresponding to each reference coordinate. Based on the local adjacency relationship of each reference coordinate, the corresponding local structural information is extracted for each reference coordinate in the reference point set. Based on the local structure information corresponding to each reference coordinate, the reference coordinates in the reference point set are classified to obtain the initial structure identifier corresponding to each reference coordinate. For reference coordinates with the same initial structure identifier, the initial structure identifier is distinguished based on the overall distribution relationship between the reference coordinates in the reference point set, so as to obtain a unique structure identifier corresponding to each reference coordinate. By associating the reference point set, the local adjacency relationships corresponding to each reference coordinate, the local structural information corresponding to each reference coordinate, and the unique structural identifier, a reference template with unique structural constraints is constructed.

3. The interactive positioning method based on LED array according to claim 1, characterized in that, Spot extraction and center localization are performed on visible light images to obtain a set of observation points, including: A continuous visible light image containing LED array light-emitting units is acquired, and the visible light image is subjected to exposure control and image cropping to obtain an image to be processed containing the LED light-emitting area; The image to be processed is converted to grayscale, and threshold segmentation is performed on the image to be processed based on the grayscale distribution to obtain a binary spot image; Connected region detection is performed based on binary spot images, and each connected region is extracted as a candidate spot region. Candidate spot regions are screened based on their area range, boundary continuity, and gray level concentration. Candidate spot regions that do not meet the preset conditions are removed to obtain the target spot region. The center position of each target spot area is calculated based on the spatial distribution of pixels within each target spot area, and the center coordinates of the spot area corresponding to each target spot area are obtained. The coordinates of the center of each spot are recorded and associated according to their spatial distribution in the image to be processed, forming a set of observation points.

4. The interactive positioning method based on LED array according to claim 1, characterized in that, The generation of candidate matching point sets that satisfy local permutation consistency includes: Based on the local adjacency relationship of each reference coordinate in the reference template, extract the reference local arrangement structure corresponding to each reference coordinate; Based on the spatial distribution relationship of the center coordinates of each spot in the observation point set in the image to be processed, the corresponding adjacent observation points are determined for the center coordinates of each spot, and the observation local arrangement structure corresponding to the center coordinates of each spot is generated. The reference local arrangement structure is matched one by one with the observed local arrangement structure. The matching results are filtered according to the number of adjacent points, connection order and relative position relationship to obtain the initial matching point pairs. Based on the initial matching point pairs, determine the correspondence consistency between the observed local arrangement structure and the reference local arrangement structure in each matching point pair, and remove matching point pairs that do not meet the consistency condition; The matching point pairs that meet the consistency condition are summarized to form a candidate matching point set that meets the local permutation consistency.

5. The interactive positioning method based on LED array according to claim 1, characterized in that, Determining the interior point set includes: Obtain each pair of matching points in the candidate matching point set, and extract the corresponding reference coordinates and spot center coordinates from each pair of matching points; Based on the reference coordinates and spot center coordinates of each matching point pair, establish the correspondence between matching point pairs in the candidate matching point set; Based on the correspondence of matching point pairs, the number of samples and the preset number of iterations are set for each iteration, and in each iteration, matching point pairs corresponding to the number of samples are selected from the candidate matching point set to form the current sampling subset; Based on the correspondence between the reference coordinates of each matching point pair in the current sampling subset and the coordinates of the spot center, the current candidate similarity transformation is calculated; Based on the inclusion relationship between the candidate matching point set and the current sampling subset, the remaining matching point pairs in the candidate matching point set excluding the current sampling subset are determined to form the current verification set; Based on the current candidate similarity transformation, coordinate mapping is performed on the reference coordinates of each matching point pair in the current validation set to obtain the mapped coordinates corresponding to each matching point pair in the current validation set; Based on the mapped coordinates and the corresponding spot center coordinates of each matching point pair in the current verification set, calculate the spatial deviation of each matching point pair in the current verification set; Based on spatial bias, a consistency determination is made for each pair of matching points in the current validation set to obtain the current set of inliers; Based on the current set of interior points obtained in each iteration, the corresponding current candidate similarity transformations are compared to determine the optimal similarity transformation and the optimal set of interior points corresponding to the optimal similarity transformation. When the preset number of iterations is reached or the optimal interior point set satisfies the preset stability condition, the optimal similarity transformation and interior point set are output.

6. The interactive positioning method based on LED array according to claim 1, characterized in that, The improved Umeyama algorithm includes: Obtain each matching point pair in the inlier set, and extract the corresponding reference coordinates and spot center coordinates from each matching point pair; Based on the local adjacency relationship corresponding to each reference coordinate in the reference template, reference local topology information containing the number of adjacent points, connection order and relative position relationship is extracted for each reference coordinate. Based on the spatial distribution relationship of the center coordinates of each spot in the observation point set, observation local topology information containing the number of adjacent points, connection order and relative position relationship is extracted for the center coordinates of each spot. Based on the consistency comparison between the reference local topology information and the observed local topology information corresponding to each matching point pair, the matching point pairs with the same number of adjacent points, the same connection order, and the same relative positional relationship are determined as the set of structurally stable point pairs. The remaining matching point pairs are determined as the set of point pairs to be verified. Based on the reference coordinates and spot center coordinates in the set of stable point pairs, the reference center coordinates and observation center coordinates are calculated respectively, and the set of stable point pairs is centered based on the reference center coordinates and observation center coordinates. The covariance matrix is ​​constructed based on the set of stable points after centering, and the singular value decomposition is performed on the covariance matrix to obtain the initial rotation matrix, initial scale factor and initial translation vector, thus forming the initial similarity transformation parameters. Based on the initial similarity transformation parameters, the reference coordinates in the set of stable point pairs and the set of point pairs to be verified are projected forward to obtain the corresponding forward mapped coordinates, and the center coordinates of the light spots in the set of stable point pairs and the set of point pairs to be verified are projected backward to obtain the corresponding backward mapped coordinates. Based on the positive deviation between the forward-mapped coordinates and the corresponding spot center coordinates, the negative deviation between the reverse-mapped coordinates and the corresponding reference coordinates, and the reference local topology information and observation local topology information corresponding to each matching point pair, the matching point pairs in the set of structurally stable point pairs are retained or removed, and the matching point pairs in the set of point pairs to be verified are replenished to obtain the updated set of stable point pairs. Based on the updated set of stable point pairs, the calculation of reference center coordinates and observation center coordinates, centering, covariance matrix construction and singular value decomposition are re-executed to obtain the updated rotation matrix, updated scale factor and updated translation vector, thus forming the updated similarity transformation parameters. If the updated set of stable point pairs does not meet the preset stability conditions, the forward projection, backward projection, retention or removal, and backfilling determination are repeatedly performed based on the updated similarity transformation parameters. When the updated set of stable point pairs satisfies the preset stability conditions, the updated similarity transformation parameters are output as similarity transformation parameters between the reference template and the observation point set.

7. The interactive positioning method based on LED array according to claim 1, characterized in that, The updated generation of the interior point set includes: Based on the similarity transformation parameters, coordinate transformation is performed on each reference coordinate in the interior point set to obtain the projected coordinates corresponding to each reference coordinate. Based on the positional differences between each projection coordinate and the corresponding light spot center coordinate, the projection deviation results for each matching point pair are generated. Based on the spatial distribution relationship of the center coordinates of each spot in the observation point set in the image to be processed, the corresponding adjacent observation points are determined for the center coordinates of each spot in the in-point set, and the observation arrangement relationship corresponding to the center coordinates of each spot is generated. Based on the local adjacency relationship of each reference coordinate in the reference template, the corresponding reference arrangement relationship is determined for each reference coordinate in the interior point set; Based on the projection deviation results, observation arrangement relationship and reference arrangement relationship of each matching point pair, the consistency of each matching point pair in the in-point set is determined, and the matching point pairs whose projection deviation results meet the preset deviation conditions and whose observation arrangement relationship is consistent with the reference arrangement relationship are determined as valid matching point pairs. Based on valid matching point pairs, update the composition of matching point pairs in the interior point set to generate an updated interior point set.

8. The interactive positioning method based on LED array according to claim 1, characterized in that, The output of the terminal's spatial pose in the LED array coordinate system includes: Based on each matching point pair in the updated interior point set, the similarity transformation parameters are solved to obtain the current similarity transformation parameters; Based on the current similarity transformation parameters, a consistency check is performed on the updated interior point set, and the matching point pairs in the updated interior point set are updated according to the check results to obtain the current interior point set; Based on the current set of interior points, re-execute the similarity transformation parameter solution and perform consistency verification and interior point set update; Repeat the similarity transformation parameter solution and interior point set update until the interior point set satisfies the preset stability condition; When the inlier set satisfies the preset stability conditions, the spatial pose of the terminal in the LED array coordinate system is determined based on the corresponding similarity transformation parameters, and the spatial pose is output.

9. An interactive positioning system based on an LED array, comprising the interactive positioning method based on an LED array as described in any one of claims 1 to 8, characterized in that, include: The reference template construction module is used to obtain the reference coordinates of each light-emitting unit in the LED array and their relative spatial arrangement relationship, and construct a reference template with unique structural constraints. The image acquisition and processing module is used to acquire visible light images containing LED array light-emitting units, and to extract light spots and locate the center of the visible light images to generate a set of observation points; The structure matching module is used to perform structural constraint matching on the observation point set based on the spatial arrangement relationship of the reference template, and generate a candidate matching point set that satisfies the local arrangement consistency. The robust estimation module is used to perform RANSAC iterations on the candidate matching point set to obtain the interior point set. The similarity transformation solution module is used to execute the improved Umeyama algorithm based on the interior point set to solve for the similarity transformation parameters between the reference template and the observation point set; The interior point update module is used to perform back projection verification on the interior point set based on similarity transformation parameters, and to determine the consistency of the verification results based on spatial arrangement relationship, thereby generating an updated interior point set. The pose output module is used to re-execute the similarity transformation parameter solution based on the updated interior point set until the interior point set satisfies the stability condition, and output the spatial pose result of the terminal in the LED array coordinate system.

Images