Indoor positioning method, terminal and storage medium

By shooting indoor images, the keyframe image and feature vector matching points are constructed, and the 2D-3D point pair is calculated, which solves the indoor positioning problem caused by GNSS signal occlusion, and realizes high-precision and low-cost indoor positioning, expands the application range and improves the User experience.

CN114445491BActive Publication Date: 2024-11-19GUANGZHOU OKAY INFORMATION TECH CO LTD
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202111552720.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-12-17
Publication Date
2024-11-19
Estimated Expiration
2041-12-17

AI Technical Summary

Technical Problem

In multi-story buildings, GNSS signals are easily blocked and difficult to achieve high-precision indoor positioning. The existing technology equipment is costly and has strict environmental requirements, which cannot meet the requirements of high accuracy and low cost.

Method used

By taking images of indoor environments, building keyframe images, obtaining map matching points of feature vectors, calculating 2D-3D point pairs, realizing the positioning of the camera in the map, reducing positioning costs, expanding application scope, and improving accuracy through global matching .

Benefits of technology

It realizes high-precision indoor positioning, reduces positioning costs, is suitable for a wide range of indoor environments, and improves user experience.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114445491B_ABST
    Figure CN114445491B_ABST
Patent Text Reader

Abstract

The present invention provides an indoor positioning method, a terminal and a storage medium, and the indoor positioning method includes: S101: constructing a state transfer matrix and a co-viewing image according to a key frame image, and constructing a feature vector matching model using the global sorting result of feature vectors in the key frame image; S102: correcting the captured image, obtaining the feature vector and camera parameters after image correction, calculating the query vector according to the feature vector and the feature vector matching model, and obtaining the matching point set corresponding to the feature vector; S103: calculating the 2D-3D point pair of each feature vector according to the matching point set, the state transfer matrix and the query vector, and calculating the position and posture of the camera through the 2D-3D point pair and the camera parameters to realize indoor positioning. The present invention does not need to deploy a large number of positioning devices, greatly reduces the positioning cost, has low requirements on the environment, expands the application scope of indoor positioning, and improves the matching accuracy through global matching, has high positioning accuracy, and improves the user experience.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of space positioning, and in particular to an indoor positioning method, a terminal and a storage medium. Background Art

[0002] According to statistics, 70%-80% of people's activities take place indoors, so it is of great significance to carry out research on indoor positioning. When a fire occurs in a multi-story building, it is not easy to evacuate quickly. At this time, indoor navigation and positioning services are needed to guide people in danger to evacuate quickly and accurately to a safe area to prevent major casualties and property losses. High-precision location information can be obtained through GNSS in outdoor areas. GNSS requires direct communication between satellites and mobile targets. In signal blind spots such as urban areas with high-rise buildings or complex and changeable indoor environments, the signal is easily blocked by objects such as building walls, glass, and indoor equipment, making it difficult to effectively play its ranging and positioning functions. Therefore, other sensors are needed to achieve high-precision indoor positioning.

[0003] Currently, there are many related indoor positioning technologies, such as infrared, radio frequency identification, WiFi, Zigbee, visual positioning, etc. However, these devices are expensive, have strict requirements on the environment, and because of limited signal coverage, they require a large number of deployments, and the overall cost is very high, which cannot meet people's requirements for high accuracy and low cost of indoor positioning perception systems. Summary of the invention

[0004] In order to overcome the shortcomings of the prior art, the present invention proposes an indoor positioning method, a terminal and a storage medium, which capture images of indoor environments, construct map data based on key frame images in the images, obtain matching points corresponding to each feature vector in the captured images in the map data, and then obtain 2D-3D point pairs of each feature vector, and use the point pairs to obtain the position of the camera in the map, thereby realizing indoor positioning. There is no need to deploy a large number of positioning devices, which greatly reduces the positioning cost, has low requirements on the environment, expands the application scope of indoor positioning, and improves the matching accuracy through global matching, with high positioning accuracy, thereby improving the user experience.

[0005] To solve the above problems, a technical solution adopted by the present invention is: an indoor positioning method, which includes: S101: constructing a state transfer matrix and a co-viewing image according to a key frame image, and constructing a feature vector matching model using the global sorting result of the feature vector in the key frame image; S102: correcting the captured image, obtaining the feature vector and camera parameters after the image correction, calculating a query vector according to the feature vector and the feature vector matching model, and obtaining a matching point set corresponding to the feature vector; S103: calculating the 2D-3D point pair of each feature vector according to the matching point set, the state transfer matrix and the query vector, and calculating the camera's posture through the 2D-3D point pair and the camera parameters to achieve indoor positioning.

[0006] Furthermore, before the step of constructing a state transfer matrix and a co-visual map according to the key frame images, the step also includes: acquiring the key frame images, extracting feature vectors in the key frame images, and sorting the feature vectors.

[0007] Furthermore, the step of obtaining the key frame image specifically includes: taking the first frame image as the key frame image, and determining whether the ratio of the common feature points of the current frame image and the previous frame image to the feature points of the current frame is less than a preset value; if so, the current frame image is the key frame image; if not, the current frame image is not a key frame image.

[0008] Furthermore, the step of sorting the feature vectors specifically includes: constructing a coordinate system of the map, superimposing the point cloud corresponding to the key frame image onto the map, extracting and filtering the feature vectors in the key frame image, and assigning key frame numbers and global numbers to the feature vectors according to the key frame images to which the feature vectors belong and the order of the key frame images.

[0009] Furthermore, the step of constructing a state transfer matrix and a co-viewing diagram based on the key frame image specifically includes: initializing the state transfer matrix, obtaining the value of each element in the state transfer matrix according to the number of key frame numbers of the feature vector, and standardizing each column of the state transfer matrix.

[0010] Furthermore, the step of calculating the query vector based on the feature vector and the feature vector matching model specifically includes: obtaining 3D points in the map that match the feature vector according to the feature vector matching model, calculating the sum of similarities of each 3D point being matched, and obtaining the query vector according to the sum of similarities.

[0011] Furthermore, the step of calculating the 2D-3D point pair of each feature vector according to the matching point set, the state transfer matrix and the query vector specifically includes: sorting the 3D points in the matching point set in descending order according to the state transfer matrix and the query vector, obtaining the distance ratio of the first two 3D points in the descending order, and obtaining the 2D-3D point pair of each feature vector according to the distance ratio.

[0012] Furthermore, the step of obtaining the distance ratio of the first two 3D points in the descending order and obtaining the 2D-3D point pair of each feature vector according to the distance ratio specifically includes: obtaining the distance between the first two 3D points and the feature vector, and determining whether the ratio of the distance corresponding to the first 3D point to the distance corresponding to the second 3D point is less than a preset value; if so, forming a 2D-3D point pair with the feature vector and the first 3D point; if not, discarding the feature vector.

[0013] Based on the same inventive concept, the present invention also proposes an intelligent terminal, which includes a processor and a memory, wherein the processor is communicatively connected to the memory, the memory stores a computer program, and the computer program is used to execute the indoor positioning method as described above.

[0014] Based on the same inventive concept, the present invention further proposes a computer-readable storage medium, wherein the computer-readable storage medium stores program data, and the program data is used to execute the indoor positioning method as described above.

[0015] Compared with the prior art, the beneficial effects of the present invention are: capturing images of indoor environments, constructing map data based on key frame images in the images, obtaining matching points corresponding to each feature vector in the captured images in the map data, and then obtaining 2D-3D point pairs of each feature vector, and using the point pairs to obtain the position of the camera in the map, thereby achieving indoor positioning, without the need for a large number of positioning devices to be deployed, greatly reducing positioning costs, having low requirements on the environment, expanding the application scope of indoor positioning, and improving matching accuracy through global matching, with high positioning accuracy and improved user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A flowchart of an indoor positioning method according to an embodiment of the present invention;

[0017] Figure 2 A flowchart of an embodiment of constructing map data in an indoor positioning method of the present invention;

[0018] Figure 3 A flowchart of an embodiment of an indoor positioning method using captured images for positioning according to the present invention;

[0019] Figure 4This is a structural diagram of an embodiment of a smart terminal of the present invention;

[0020] Figure 5 This is a structural diagram of an embodiment of a computer-readable storage medium of the present invention. DETAILED DESCRIPTION

[0021] The following describes the implementation of the present application through specific specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the content disclosed in this specification. The present application can also be implemented or applied through other different specific implementations, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that the various embodiments of the present disclosure generally described and shown in the accompanying drawings here can be combined with each other without conflict, and the structural components or functional modules therein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present disclosure provided in the accompanying drawings is not intended to limit the scope of the present disclosure claimed for protection, but merely represents the selected embodiments of the present disclosure. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present disclosure.

[0022] See also Figures 1 to 3 ,in, Figure 1 A flowchart of an indoor positioning method according to an embodiment of the present invention; Figure 2 A flowchart of an embodiment of constructing map data in an indoor positioning method of the present invention; Figure 3 FIG. 1 is a flow chart of an embodiment of an indoor positioning method of the present invention using a captured image for positioning. Figures 1 to 3 The indoor positioning method of the present invention is described in detail.

[0023] In other embodiments, the device that executes the indoor positioning method may be a laptop computer, a desktop computer, a server, a cloud platform, or other devices that can be used to construct map data and perform positioning based on the map data.

[0024] Among them, indoor positioning methods include:

[0025] S101: constructing a state transfer matrix and a co-visual map according to the key frame image, and constructing a feature vector matching model using the global sorting result of the feature vectors in the key frame image.

[0026] In this embodiment, before the step of constructing a state transfer matrix and a co-visual map according to key frame images, the step further includes: acquiring key frame images, extracting feature vectors in the key frame images, and sorting the feature vectors.

[0027] Among them, the step of obtaining the key frame image specifically includes: taking the first frame image as the key frame image, and determining whether the ratio of the common feature points of the current frame image and the previous frame image to the feature points of the current frame is less than a preset value; if so, the current frame image is the key frame image; if not, the current frame image is not a key frame image.

[0028] In this embodiment, ORBSLAM2 is used to open the kinect depth camera, which is an RGBD camera. The depth camera is used to obtain the 3D point cloud and RGB image of the area corresponding to the map, and the key frame image is obtained through the RGB image.

[0029] In a specific embodiment, the image output size of the depth camera is set to 640*480, and the first frame image after initialization is used as the origin of the coordinate system of the complete map generated later, that is, as the origin of the coordinate system of the pose (Euclidean transformation matrix) of the key frame image, and the first frame image is used as the key frame image to obtain the key frame image in the image taken by the depth camera.

[0030] In this embodiment, the feature vector is a shift feature vector, and the steps of sorting the feature vectors specifically include: constructing a coordinate system of the map, superimposing the point cloud corresponding to the key frame image onto the map, extracting and filtering the feature vectors in the key frame image, and assigning key frame numbers and global numbers to the feature vectors according to the key frame images to which the feature vectors belong and the order of the key frame images.

[0031] In a specific embodiment, the 3D point cloud currently acquired by the kinect depth camera is converted and superimposed on the map through the inverse matrix of the pose matrix of the current key frame image output by ORBSLAM2. At the same time, the sift feature vector is extracted from the key frame image, and each feature vector is filtered in the neighborhood according to the image coordinates in the key frame image (the horizontal coordinate is the width of the image, the vertical coordinate is the height of the image, and the origin is in the upper left corner) (based on the image coordinates of the feature vector in the image, from left to right, from top to bottom) (that is, with the image coordinates of the feature vector as the center, take a certain radius (here 5 pixels, if there are other feature vectors in the circle, these feature vectors are discarded), and then the key frame sequence number (in ascending order, starting from 0) is uniformly assigned to the feature vectors retained in the frame image. If there are multiple key frame numbers, the key frame numbers are in ascending order. If there are multiple key frame numbers, the key frame numbers are in ascending order, starting from 0). If all key frame images have this feature vector (judged by the Hamming distance between feature vectors, when it is less than a certain threshold, it is considered to be the same feature vector), the serial numbers of multiple key frame images are assigned to the feature vector together, and for each key frame image, a global serial number is assigned to the feature vector in order from left to right and from top to bottom. For example, if the first frame image has 10 feature vectors, they are assigned serial numbers from 0 to 9 respectively, and if the second frame image has 20 feature vectors, they are assigned serial numbers from 10 to 19, and so on. After scanning the scene, a series of feature-3D point pairs are obtained, that is, a sparse map. Indoor positioning is performed based on the sparse map.

[0032] The steps of constructing a state transfer matrix and a co-viewing diagram according to the key frame images specifically include: initializing the state transfer matrix, obtaining the value of each element in the state transfer matrix according to the number of key frame numbers of the feature vector, and normalizing each column of the state transfer matrix.

[0033] In this embodiment, the sparse matrix (state transition matrix) corresponding to the sparse map is obtained, and the sparse matrix C is initialized. The element C in the matrix C ij It represents the probability that feature vector i is visible when feature vector j is visible. After all key frame images are acquired, for each feature vector j (i and j are global serial numbers), if the number of key frame serial numbers assigned to it is equal to 1, it means that the point is only visible in the current frame image. Then the point is visible to other points in the same frame image, but not to points in other frames. Then C ij =||P i -P j ||2 / ||P j ||2(1); where P j represents the image coordinates of j in the key frame image, P iis the image coordinate of other feature points in the same frame as j (i is also a global serial number), ||*||2 represents the 2-norm of the vector, and all other rows of the j-th column of matrix C are set to 0; if the number of key frame serial numbers of j is greater than 1, it means that the point is seen by multiple key frames, and the corresponding key frames can be found and calculated according to formula (1). Finally, each column of C is standardized so that the sum of each column is equal to 1, and the final state transition matrix is ​​used as the co-view map-graph.

[0034] In this embodiment, the global sorting result of the feature vectors is obtained by arranging all the feature vectors in ascending order according to the size of the global sequence number, and a feature vector matching model for quickly finding a matching vector is obtained by using the flann method for all the feature vectors.

[0035] S102: Correct the captured image, obtain a feature vector and camera parameters after the image correction, calculate a query vector according to the feature vector and a feature vector matching model, and obtain a matching point set corresponding to the feature vector.

[0036] In this embodiment, the captured image is obtained by shooting with a mobile phone, and the captured image is corrected using the internal parameters and distortion coefficients of the mobile phone. If the internal parameters and distortion parameters are not obtained, the mobile phone lens can be calibrated to obtain the corrected image by calibration, wherein the corrected image is an image without distorted pixels.

[0037] In other embodiments, an image of the area to be located may also be obtained by means of a camera, a network query, etc., and it is sufficient to obtain an image without distorted pixels corresponding to the image.

[0038] The camera parameters include intrinsic parameters and distortion parameters. Obtaining the feature vector and camera parameters after image correction includes: extracting the shift feature vector of the corrected image, obtaining the image coordinates of the shift feature vector in the image, and obtaining the intrinsic parameters and distortion parameters of the camera.

[0039] In this embodiment, the step of obtaining camera parameters and feature vectors and the step of performing subsequent positioning using the feature vectors and camera parameters are performed on different devices. In other embodiments, they may also be performed in the same device.

[0040] The step of calculating the query vector according to the feature vector and the feature vector matching model specifically includes: obtaining 3D points in the map that match the feature vector according to the feature vector matching model formed by the flann algorithm, calculating the sum of similarities of each 3D point being matched, and obtaining the query vector according to the sum of similarities.

[0041] In this embodiment, each feature vector obtains a preset number of the most matching 3D points according to the matching degree sorting, and forms a matching point set of the feature vector based on the 3D points, where the preset number is 100. In other embodiments, the preset number may also be 150, 200 or other numbers.

[0042] In a specific embodiment, let f be a feature vector of the rectified image, i represents the global sequence number of the 3D point, and use the Gaussian kernel function Calculate the distance between the feature vector f and the 3D point i, W fi Represents the distance between the feature vector f and the 3D point i, where H(f,i) represents the distance between the feature vector f and the 3D point i calculated using the flann method, where δ = 32.

[0043] definition q i represents the sum of similarities of the i-th 3D point being matched, that is, the importance of the point relative to the query vector set, where O(i) represents the set of feature vectors matching the i-th 3D point, N i represents the number of O(i), N f It is the number of 3D points that match the feature vector f, and N represents the total number of 3D points. Q i Normalize. All q i Combined together to form the query vector q.

[0044] S103: Calculate the 2D-3D point pair of each feature vector according to the matching point set, the state transfer matrix and the query vector, and calculate the camera's position and posture through the 2D-3D point pair and camera parameters to achieve indoor positioning.

[0045] The step of calculating the 2D-3D point pair of each feature vector according to the matching point set, the state transfer matrix and the query vector specifically includes: sorting the 3D points in the matching point set in descending order according to the state transfer matrix and the query vector, obtaining the distance ratio of the first two 3D points in the descending order, and obtaining the 2D-3D point pair of each feature vector according to the distance ratio.

[0046] The step of obtaining the distance ratio of the first two 3D points in the descending order, and obtaining the 2D-3D point pair of each feature vector according to the distance ratio specifically includes: obtaining the distance between the first two 3D points and the feature vector, and determining whether the ratio of the distance corresponding to the first 3D point to the distance corresponding to the second 3D point is less than a preset value; if so, forming a 2D-3D point pair with the feature vector and the first 3D point; if not, discarding the feature vector.

[0047] In a specific embodiment, a restarted random walk model is defined to calculate the steady-state probability vector of a 3D point set, where pv (t) represents the probability of being at node ν at time t. When t→∞, p v (t) represents the probability of being stable at node ν (here it means feature matching is successful). According to the formula p(t+1)=αCp(t)+(1-α)q(5), since (5) converges, we have p=(I-αC) -1 (1-α)q(6), which can calculate the value of P(∞). Here α = 0.85, C is the created state transition matrix. Since each 3D point corresponds to multiple matching feature vectors, the ratio-test method is used to obtain the best one-to-one 2D-3D matching pair. The specific method is as follows: obtain the feature vector of each query image and the 3D point set that matches it, use P(∞) (restart walk model) to sort these 3D point sets in descending order, extract the first two 3D points in the sorting, and calculate the distance with the feature vector of the query image respectively. If the ratio of the first distance to the second distance is less than 0.5, it is considered that the feature vector matches the first 3D point, and the 3D point and feature vector are marked as matched. If the distance ratio is greater than or equal to 0.5, it is considered that the feature vector has not been matched successfully, and it is not marked, indicating that the feature vector is discarded. By analogy, after traversing all the feature vectors of the query image, a one-to-one 2D-3D point pair can be obtained.

[0048] In this embodiment, after obtaining the 2D-3D point pair, the EPnP+RANSAC algorithm is used to calculate according to the 2D-3D point pair and the camera internal parameters and distortion parameters to obtain the final pose (the position and posture of the camera in the map), and the pose is used for indoor positioning.

[0049] Beneficial effect: The indoor positioning method of the present invention captures an image of an indoor environment, constructs map data based on key frame images in the image, obtains matching points corresponding to each feature vector in the captured image in the map data, and then obtains 2D-3D point pairs of each feature vector, and uses the point pairs to obtain the position of the camera in the map, thereby achieving indoor positioning. It does not require a large number of positioning devices to be deployed, greatly reduces the positioning cost, has low requirements on the environment, expands the application scope of indoor positioning, and improves the matching accuracy through global matching, with high positioning accuracy, thereby improving the user experience.

[0050] Based on the same inventive concept, the present invention also proposes an intelligent terminal, see Figure 4 , Figure 4 This is a structural diagram of an embodiment of the intelligent terminal of the present invention, combined with Figure 4 The intelligent terminal of the present invention is described.

[0051] In this embodiment, the intelligent terminal includes a processor and a memory. The processor is communicatively connected to the memory. The memory stores a computer program. The computer program is used to execute the indoor positioning method as described in the above embodiment.

[0052] In some embodiments, the memory may include, but is not limited to, high-speed random access memory, non-volatile memory. For example, one or more disk storage devices, flash memory devices, or other non-volatile solid-state storage devices. The processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable functional devices, discrete gates or transistor functional devices, discrete hardware components.

[0053] Based on the same inventive concept, the present invention also provides a computer-readable storage medium, see Figure 5 , Figure 5 The structure diagram of an embodiment of the computer readable storage medium of the present invention is shown in FIG. Figure 5 The computer-readable storage medium of the present invention is described.

[0054] In this embodiment, the computer-readable storage medium stores program data, and the program data is used to execute the indoor positioning method as described in the above embodiment.

[0055] The computer-readable storage medium may include, but is not limited to, a floppy disk, an optical disk, a CD-ROM (compact disk-read only memory), a magneto-optical disk, a ROM (read only memory), a RAM (random access memory), an EPROM (erasable programmable read-only memory), an EEPROM (electrically erasable programmable read-only memory), a magnetic card or an optical card, a flash memory, or other types of media / machine-readable media suitable for storing machine-executable instructions. The computer-readable storage medium may be a product that is not connected to a computer device, or a component that is connected to a computer device for use.

[0056] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0057] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An indoor positioning method, characterized in that: The indoor positioning method comprises: S101: constructing a state transfer matrix and a co-visual map according to the key frame image, and constructing a feature vector matching model using a global sorting result of feature vectors in the key frame image; S102: Correcting the captured image, obtaining a feature vector and camera parameters after the image correction, calculating a query vector according to the feature vector and a feature vector matching model, and obtaining a matching point set corresponding to the feature vector; S103: Calculate the 2D-3D point pair of each feature vector according to the matching point set, the state transfer matrix and the query vector, and calculate the position and posture of the camera through the 2D-3D point pair and camera parameters to achieve indoor positioning; The state transfer matrix and the co-viewing image are constructed according to the key frame image, and the feature vector matching model is constructed using the global sorting result of the feature vectors in the key frame image, including: Constructing a coordinate system of a map, superimposing a point cloud corresponding to a key frame image onto the map, extracting and filtering feature vectors in the key frame image, assigning a key frame sequence number and a global sequence number to the feature vector according to the key frame image to which the feature vector belongs and the sequence of the key frame images, thereby obtaining a sparse map consisting of 3D point pairs of each feature vector in the map, wherein when multiple key frame images have the same feature vector, the sequence numbers of the multiple key frame images are all assigned to the feature vector, and the global sequence numbers are numbered from left to right and from top to bottom according to the image coordinate sequence of the feature vector in the key frame image; Initialize a state transfer matrix so that the value of an element thereof represents the probability that the feature vector corresponding to the row where the element is located is visible when the feature vector corresponding to the column where the element is located is visible, obtain the value of each element in the state transfer matrix according to the number of key frame numbers of the feature vector, and perform normalization on each column of the state transfer matrix so that the sum of the probabilities of each column is equal to 1 to obtain a co-viewing diagram, wherein the state transfer matrix is ​​a sparse matrix corresponding to the sparse map, and each row and column are arranged according to the global serial number; Arrange all feature vectors in ascending order according to the global sequence number to obtain the global sorting result of the feature vectors, and use the flann method to obtain the feature vector matching model for all feature vectors; The step of calculating a query vector according to the feature vector and the feature vector matching model and obtaining a matching point set corresponding to the feature vector includes: According to the feature vector matching model, 3D points in the map that match the feature vector are obtained, the sum of similarities of each 3D point being matched is calculated, and a query vector is obtained according to the sum of similarities, which includes: according to the matching sorting result, each feature vector obtains a preset number of the most matched 3D points, and a matching point set of the feature vector is formed according to the 3D points; the sum of similarities of each 3D point being matched is combined to obtain the query vector; The step of calculating a 2D-3D point pair of each feature vector according to the matching point set, the state transfer matrix, and the query vector, and calculating the position and posture of the camera through the 2D-3D point pair and camera parameters to achieve indoor positioning includes: According to the state transfer matrix and the query vector, the 3D points in the matching point set are sorted in descending order, the distance ratio of the first two 3D points in the descending order is obtained, and the 2D-3D point pair of each feature vector is obtained according to the distance ratio, which includes: obtaining the distance between the first two 3D points and the feature vector, and judging whether the ratio of the distance corresponding to the first 3D point to the distance corresponding to the second 3D point is less than a preset value; if so, forming a 2D-3D point pair with the feature vector and the first 3D point; if not, discarding the feature vector.

2. The indoor positioning method according to claim 1, characterized in that: Before the step of constructing a state transfer matrix and a co-viewing image according to the key frame image, the following steps are also included: A key frame image is acquired, feature vectors in the key frame image are extracted, and the feature vectors are sorted.

3. The indoor positioning method according to claim 2, characterized in that: The step of obtaining the key frame image specifically includes: The first frame image is used as a key frame image, and it is determined whether the ratio of the common feature points of the current frame image and the previous frame image to the feature points of the current frame is less than a preset value; If yes, the current frame image is a key frame image; If not, the current frame image is not a key frame image.

4. An intelligent terminal, characterized in that: The intelligent terminal includes a processor and a memory, the processor is communicatively connected to the memory, the memory stores a computer program, and the computer program is used to execute the indoor positioning method according to any one of claims 1 to 3.

5. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores program data, and the program data is used to execute the indoor positioning method according to any one of claims 1 to 3.

Citation Information

Patent Citations

  • Indoor movable robot real-time navigation method based on visual information correction

    CN101576384A

  • Three-dimensional human body motion tracking method based on quantum immune clone algorithm

    CN101692284A

  • Indoor autonomous navigation method for micro unmanned aerial vehicle

    CN104236548A

  • Localization and Tracking Using Location, Signal Strength, and Pseudorange Data

    US20180074209A1

  • Real-time monocular structure from motion

    US20180247429A1