Centroid-based visual SLAM methods, systems, terminals, and media
By using a barycentric coordinate-based visual SLAM method, similar configurations and linear matrix equations are constructed using visual information, which solves the problems of high complexity and low efficiency of visual SLAM algorithms and achieves efficient and accurate calculation of location and landmark coordinates.
Patent Information
- Application Number
- CN202311604113.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-11-27
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2043-11-27
AI Technical Summary
Existing visual SLAM methods, while ensuring algorithm accuracy, generally suffer from high algorithm complexity and low efficiency.
The barycentric coordinate-based visual SLAM method acquires visual information of the target terminal in real time, constructs similar configurations and obtains matching coordinates, uses these coordinates to create a linear target SLAM matrix equation, solves for the position coordinates of the target terminal and the coordinates of the landmarks, and constructs the motion trajectory.
Transforming the visual SLAM problem into a localization problem of a 3D spatial network reduces algorithm complexity and significantly improves the accuracy and efficiency of position coordinates.
Smart Images

Figure CN117830396B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of computer vision technology, and in particular to a visual SLAM method, system, terminal, and medium based on centroid coordinates. Background Technology
[0002] In recent years, with the rapid development of technology, Simultaneous Localization and Mapping (SLAM) methods have been increasingly widely used and have attracted more and more attention in fields such as autonomous driving, indoor navigation, robot navigation, virtual reality, and seabed exploration. SLAM requires mobile robots to build environmental maps and determine their own position in the environment while moving. To achieve this goal, robots rely on various sensors, including sonar systems, ranging lasers, cameras, and GPS, to perceive and measure landmarks.
[0003] Currently, mainstream SLAM methods are divided into two categories: laser SLAM and visual SLAM. Visual SLAM uses camera sensors to capture continuous image frames, extracting feature points, edges, or landmarks, and then using these to build maps and estimate robot motion. It mainly includes visual SLAM based on traditional image processing and visual SLAM based on deep learning. Methods based on traditional image processing rely primarily on continuous image frames to estimate camera or robot motion, thereby achieving localization and building environmental maps. Based on whether image features need to be extracted, they are further divided into feature-point front-ends and direct front-ends that do not extract features. Feature-point front-ends detect and match feature points in images using methods such as SIFT, ORB, and SURF. While less affected by environmental changes and with good stability, they have high computational complexity, low efficiency, and high implementation costs. Direct front-ends, on the other hand, estimate camera motion by performing pixel-level matching on image brightness. While avoiding the complexity of feature extraction and matching, they are sensitive to changes in image brightness and have poor accuracy. Visual SLAM based on deep learning uses deep learning methods to improve key steps such as feature extraction, feature matching, depth estimation, and map construction. Although the accuracy is improved, the algorithm has high complexity, low efficiency, and high implementation cost.
[0004] It is evident that existing visual SLAM methods, while ensuring algorithm accuracy, generally suffer from high algorithm complexity and low efficiency. Summary of the Invention
[0005] In view of the shortcomings of the prior art, the purpose of this invention is to provide a visual SLAM method, system, terminal and medium based on centroid coordinates, which aims to solve the problems of high algorithm complexity and low efficiency in the prior art.
[0006] To achieve the above objectives, a first aspect of the present invention provides a visual SLAM method based on centroid coordinates, comprising:
[0007] The system acquires visual information collected by the target terminal in real time, obtains the orientation vectors of several landmarks based on the visual information, and obtains the relative orientation vectors of the target terminal at adjacent times based on the orientation vectors of the landmarks.
[0008] Based on the preset configuration, using the azimuth vectors of all the landmarks at any two adjacent times and the relative azimuth vector of the target terminal, several similar configurations similar to the preset configuration are constructed, and based on each similar configuration, a set of matching coordinates is obtained;
[0009] Using the coherent coordinates, the barycenter coordinates of the corresponding similar configurations are solved, and a target SLAM matrix equation is created based on all the barycenter coordinates at the same time.
[0010] Solve the target SLAM matrix equation at each time point to obtain the position coordinates of the target terminal and the coordinates of the landmark. Then, use the position coordinates of the target terminal and the coordinates of the landmark to construct the motion trajectory of the target terminal and complete the mapping.
[0011] Optionally, the step of constructing the configuration includes:
[0012] Define several nodes with known coordinates to obtain anchor nodes;
[0013] Based on the position of the target terminal at any two adjacent times, and several anchor nodes and several free nodes with unknown coordinates, a communication sensing map is constructed.
[0014] Based on the communication sensing map, the configuration is constructed using the orientation vectors of the anchor node and the free node observed by the target terminal at the two locations respectively.
[0015] Optionally, obtaining a set of matching coordinates based on each of the similar configurations includes:
[0016] Based on each of the similar configurations, a triangle is constructed with each anchor node or free node observed at the corresponding time using the position of the target terminal at any two adjacent times.
[0017] Using the azimuth vectors of all the landmarks at the two adjacent times, calculate the relative azimuth vector of the target terminal at the two adjacent times and the azimuth vectors of the anchor node and the free node in each triangle in the target terminal coordinate system, and obtain the azimuth vector of each node in the similar configuration.
[0018] Based on the orientation vectors of each node in the similar configuration, a set of matching coordinates is obtained.
[0019] Optionally, the step of using the azimuth vectors of all the landmarks at the two adjacent time points to calculate the relative azimuth vector of the target terminal's movement at the two adjacent time points, and the azimuth vectors of the anchor nodes and free nodes in each triangle in the target terminal coordinate system, to obtain the azimuth vectors of each node in the similar configuration, includes:
[0020] Based on the orientation vectors of the landmarks of the target terminal at the two adjacent times, calculate the relative rotation matrix and the relative translation matrix of the target terminal between the two adjacent times;
[0021] Based on the target terminal coordinate system, the relative translation matrix is used to calculate the relative azimuth vector of the target terminal at two adjacent times; and the relative rotation matrix is used to calculate the azimuth vector of the anchor node and the free node in each triangle in the target terminal coordinate system, thereby obtaining the azimuth vector of each node in the similar configuration.
[0022] Optionally, the step of creating a target SLAM matrix equation based on all the centroid coordinates at the same time includes:
[0023] Based on the properties of the centroid coordinates, a SLAM linear equation is constructed using each set of centroid coordinates.
[0024] All the SLAM linear equations at the same time point are merged to create a target SLAM matrix equation.
[0025] Optionally, merging all the SLAM linear equations at the same time point to create a target SLAM matrix equation includes:
[0026] All SLAM linear equations representing the same node coordinates at the same time are superimposed, and each centroid coordinate is assigned a weight. The sum of the weights of each centroid coordinate under each SLAM linear equation is 1, thus obtaining a SLAM linear fusion equation.
[0027] Based on the aforementioned SLAM linear fusion equation, an initial SLAM matrix equation is constructed;
[0028] The SLAM linear fusion equations of all landmarks and target terminal positions at two adjacent time points are merged. Based on the number of nodes of each type in the initial SLAM matrix equation, the Kronecker product is used for transformation to create a target SLAM matrix equation.
[0029] Optionally, solving the target SLAM matrix equation at each time step to obtain the location coordinates of the target terminal and the coordinates of the landmark includes:
[0030] The target SLAM matrix equation is transformed to obtain a non-singular matrix equation;
[0031] The non-singular matrix equation is solved iteratively using the conjugate gradient algorithm until a preset number of iterations is reached, or the residual between the estimated coordinates of the target terminal at the corresponding time or the coordinates of some landmarks and their respective true positions is less than a preset residual threshold.
[0032] A second aspect of the present invention provides a visual SLAM system based on centroid coordinates, the system comprising:
[0033] The information acquisition module is used to acquire visual information collected by the target terminal in real time, obtain the orientation vectors of several landmarks based on the visual information, and obtain the relative orientation vector of the target terminal at adjacent time points based on the orientation vectors of the landmarks.
[0034] The coherent coordinate solving module is used to construct several similar configurations based on a preset configuration, using the azimuth vectors of all the landmarks at any two adjacent times and the relative azimuth vector of the target terminal, and to obtain a set of coherent coordinates based on each similar configuration;
[0035] The SLAM matrix equation construction module is used to solve for the centroid coordinates of the corresponding similar configuration using the coherent coordinates; and to create a target SLAM matrix equation based on all the centroid coordinates at the same time.
[0036] The motion trajectory generation module is used to solve the target SLAM matrix equation at each time point, obtain the position coordinates of the target terminal and the coordinates of the landmark, and use the position coordinates of the target terminal and the coordinates of the landmark to construct the motion trajectory of the target terminal and complete the mapping.
[0037] A third aspect of the present invention provides a smart terminal, the smart terminal including a memory, a processor, and a barycentric coordinate-based visual SLAM program stored in the memory and executable on the processor, wherein the barycentric coordinate-based visual SLAM program, when executed by the processor, implements any of the steps of the above-described barycentric coordinate-based visual SLAM method.
[0038] A fourth aspect of the present invention provides a computer-readable storage medium storing a barycentric coordinate-based visual SLAM program, wherein the barycentric coordinate-based visual SLAM program, when executed by a processor, implements any of the steps of the above-described barycentric coordinate-based visual SLAM method.
[0039] Compared with existing technologies, the beneficial effects of this solution are as follows:
[0040] This invention, based on real-time acquired visual information from the target terminal, obtains the azimuth vectors of landmarks observed by the target terminal, thereby deriving its own relative azimuth vector. Based on a preset configuration, it constructs a similar configuration using all azimuth vectors and azimuth vectors from any two adjacent time points. Then, it obtains matching coordinates based on the three-dimensional similar configurations, and uses these matching coordinates to construct a linear target SLAM matrix equation to solve for the target terminal's position coordinates at each time point, thus constructing the motion trajectory. This method relies solely on orientation measurement in computer vision, innovatively transforming the traditional visual SLAM problem into a localization problem in a three-dimensional spatial network. It converts the complex three-dimensional nonlinear solution problem into a linear matrix equation solution problem, significantly reducing algorithm complexity and thus significantly improving the accuracy and efficiency of solving for the target terminal's position coordinates at each time point. Attached Figure Description
[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0042] Figure 1 This is a simplified flowchart of the visual SLAM method based on centroid coordinates of the present invention;
[0043] Figure 2 This is a flowchart of the overall visual SLAM method based on centroid coordinates of the present invention;
[0044] Figure 3 This is a schematic diagram of the communication sensing diagram of the present invention;
[0045] Figure 4 This is a schematic diagram of the vector relationships between the nodes that form a triangle in three-dimensional space according to the present invention.
[0046] Figure 5 This is a schematic diagram of the structure of the visual SLAM system based on centroid coordinates according to the present invention;
[0047] Figure 6 This is a schematic diagram of the intelligent terminal structure of the present invention. Detailed Implementation
[0048] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of the invention. However, those skilled in the art will understand that the invention can be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of the invention with unnecessary detail.
[0049] It should be understood that, when used in this specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.
[0050] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.
[0051] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.
[0052] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0053] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.
[0054] This invention addresses the common problem of high computational complexity and low efficiency in existing visual SLAM methods while maintaining accuracy. It proposes a barycentric coordinate-based (Baryc-based) visual SLAM algorithm. This method successfully transforms the visual SLAM problem into a series of continuous 3D spatial network localization problems based on orientation measurements, providing a new perspective for solving SLAM problems. Furthermore, it establishes linear equation constraints using local orientation measurements and constructs the required matrix to iteratively solve for the positions of the target terminal and unknown landmarks, ensuring global convergence of the results. This significantly simplifies the computation of visual SLAM, enabling accurate and efficient calculation of the positions of unknown landmarks and the trajectory of the target terminal.
[0055] Exemplary methods
[0056] This invention provides a centroid-based visual SLAM method, deployable on electronic devices such as computers, servers, and robots, and applied to autonomous localization scenarios, including but not limited to autonomous driving, indoor navigation, robot navigation, virtual reality, and seabed exploration. Taking robot mapping and trajectory tracking as an example, this invention elaborates on the steps of the centroid-based visual SLAM method, specifically as follows: Figure 1 and Figure 2 As shown, the steps of the method in this embodiment include:
[0057] Step S1000: Real-time acquisition of visual information collected by the target terminal; based on the visual information, obtain the orientation vectors of several landmarks; and based on the orientation vectors of the landmarks, obtain the relative orientation vectors of the target terminal's movement at adjacent times.
[0058] Specifically, the system acquires visual information collected by the target terminal in real time. The target terminal refers to a mobile terminal capable of acquiring its own position, including but not limited to robots and autonomous vehicles. In this embodiment, the target terminal refers to a robot, which is equipped with high-definition cameras or other similar acquisition devices and corresponding sensors. The robot acquires visual information through these acquisition devices. Then, based on the acquired visual information, the orientation vectors of landmarks in the local coordinate system observed by the target terminal are obtained. Based on these landmark orientation vectors, the relative orientation vectors of the target terminal's movement at adjacent moments are obtained. However, the target terminal cannot directly acquire its own position coordinates at various moments in the global coordinate system.
[0059] Step S2000: Based on the preset configuration, using the azimuth vectors of all landmarks and the azimuth vector of the target terminal at any two adjacent times, construct several similar configurations that are similar to the original configuration, and obtain a set of matching coordinates based on each similar configuration;
[0060] Specifically, in order to extract the movement position of the target terminal from the observed visual information, a three-dimensional geometric model needs to be pre-constructed based on factors such as the type of target terminal, movement characteristics, and the field of view acquired. Several locations that conform to the characteristics of the constructed three-dimensional geometric model are then selected as research objects, facilitating the subsequent determination of the specific coordinates of unknown locations. In this embodiment, the pre-constructed three-dimensional geometric model is referred to as the Framework. This Framework is constructed based on the trajectory positions of several target terminals and observed landmark positions, according to the vector relationships between these positions. In this embodiment, the pre-constructed Framework is used as the target Framework, with the aim of ensuring that subsequently constructed similar Frameworks are identical in shape to the target Framework, differing only in size.
[0061] Then, since landmark information in multiple global coordinate systems can be observed at each moment, similar configurations of the target configuration can be constructed using the locations of the target terminal at any two adjacent moments and multiple landmarks, thus obtaining several similar configurations. The coordinates of all nodes in each similar configuration are called coincident coordinates.
[0062] Step S3000: Using the coincident coordinates, solve for the barycenter coordinates of the corresponding similar configurations; and based on all barycenter coordinates at the same time, create a target SLAM matrix equation;
[0063] Specifically, based on the coincident coordinates of each similar configuration, a set of barycentric coordinates is calculated. Then, multiple sets of barycentric coordinates can be calculated from the coincident coordinates of all similar configurations at the same time. Then, using all the barycentric coordinates at the same time, a target SLAM matrix equation is created to characterize the orientational relationship between landmarks observed by the target terminal at the corresponding time.
[0064] Step S4000: Solve the target SLAM matrix equation at each time step to obtain the position coordinates of the target terminal and the coordinates of the landmarks. Then, use the position coordinates of the target terminal and the coordinates of the landmarks to construct the motion trajectory of the target terminal and complete the mapping.
[0065] Specifically, as time progresses, the target terminal generates a target SLAM matrix equation at each moment, thus obtaining a series of target SLAM matrix equations corresponding to each moment. By solving the target SLAM matrix equation at each moment, the position coordinates of the target terminal and the coordinates of the landmarks at that corresponding moment can be obtained. Using the position coordinates of the target terminal obtained at each moment, the movement trajectory of the target terminal within the corresponding time period can be constructed.
[0066] In this embodiment, the method constructs a similar configuration by using the azimuth vector of the target terminal's movement between two consecutive adjacent moments within a certain period and the azimuth vector of the observed landmark. This converts the azimuth vectors and their information between the target terminal and the landmark into a three-dimensional spatial network, thus building a more realistic spatial model. Then, a set of matching coordinates is obtained based on the similar configuration. The centroid coordinates satisfying the similar configuration are then calculated based on these matching coordinates. A linear target SLAM matrix equation is constructed using the centroid coordinates to solve for the target terminal's position coordinates at each moment, thereby constructing the motion trajectory and completing the mapping. It is evident that this method relies solely on orientation measurement in computer vision. It generates a dynamic three-dimensional spatial network from the target terminal's movement on a plane. By utilizing a linear equation construction method based on centroid coordinates, it innovatively transforms the traditional visual SLAM problem into a three-dimensional spatial network localization problem. It converts the complex three-dimensional nonlinear solution problem into a linear matrix equation solution problem, significantly reducing algorithm complexity and thus significantly improving the accuracy and efficiency of solving for the target terminal's position coordinates at each moment. Furthermore, it ensures that the target terminal's motion trajectory is closer to reality, demonstrating broad application prospects.
[0067] In one implementation, the step of constructing the configuration in step S2000 specifically includes:
[0068] Step S2100: Define several nodes with known coordinates to obtain anchor nodes;
[0069] Step S2200: Construct a communication sensing map based on the position of the target terminal at any two adjacent times, as well as several anchor nodes and several free nodes with unknown coordinates;
[0070] Step S2300: Based on the communication sensing map, construct the configuration using the orientation vectors of the anchor node and the free node observed by the target terminal at two locations respectively.
[0071] Specifically, the set of all nodes and edges between nodes in the configuration can be viewed as a communication sensing graph. It's easy to understand that the communication sensing graph is a two-dimensional plane, while the configuration is a three-dimensional solid model; there is a one-to-one correspondence between the two. The communication sensing graph represents the observation and orientation relationship between the target terminal and the nodes in three-dimensional space, while the configuration is a geometrically meaningful communication sensing graph, containing the corner relationships and positional information between the nodes and the target terminal; it is a three-dimensional solid model, and the two have a corresponding relationship.
[0072] For example, construct a configuration (G5, p5), where G5 represents a communication sensing map including the target terminal's position at two time points and three nodes, and p5 includes the target terminal's position coordinates at two time points, the coordinates of two known anchor nodes, and the coordinates of one unknown free node. The communication sensing map is constructed using the five nodes in this configuration. in Represents a set of nodes. Represents an edge set.
[0073] In graph theory, a graph is globally rigid only when there are at least four nodes with known coordinates (called anchor nodes), meaning the location of each node can be uniquely determined. Therefore, to ensure successful localization, the processed visual information at each time step must include at least four anchor nodes in the resulting communication perception graph. In other words, in two consecutive time steps, the target terminal needs to obtain the relative orientation information of at least four anchor nodes to accurately determine its own position coordinates.
[0074] In this embodiment, a communication perception map is first constructed based on the visual information collected by the target terminal and several landmarks with known and unknown coordinates. Then, a three-dimensional configuration is constructed using the movement of the target terminal on a two-dimensional plane, so that the orientation relationship between each node can be described more accurately and vividly.
[0075] like Figure 2 As shown, in one embodiment, step S2000, which obtains a set of matching coordinates based on each similar configuration, specifically includes:
[0076] Step S2400: Based on each similar configuration, using the position of the target terminal at any two adjacent times, construct a triangle with each anchor node or free node observed at the corresponding time.
[0077] Specifically, the coordinates of the five nodes in each similar configuration constructed based on a preset configuration are called the coincidence coordinates. A communication-aware graph is constructed using these five nodes. in This represents a set of nodes, that is, a set of matching coordinates. Represents an edge set.
[0078] For example, Figure 3 The diagram shown is a schematic of a communication sensing map, which includes the position r of the target terminal at time t. t The position r at time t-1 t-1 And the {i, j, k} of any three observed landmarks, where i and j are anchor nodes with known coordinates, and k is a free node with unknown coordinates. Based on this communication-aware graph, a similar configuration (G5, q5) is constructed, where, The position r of the target terminal at time t is indicated. t Transpose of coordinates The position r of the target terminal at time t-1 is indicated. t-1 Transpose of coordinates These represent the transposes of the coordinates of landmarks i, j, and k observed by the target terminal at time t. It's easy to understand that since the target terminal's location coordinates are unknown, the coordinates used to construct a similar configuration are not necessarily the coordinates of the actual nodes. Furthermore, the set of coordinates found that satisfies the shape similar to the configuration means that the shape of the similar configuration is similar to the preset conceptual shape, and the size can be a scaling of the configuration's size.
[0079] It needs to be stated that, Figure 2 The communication awareness map shown is the smallest locatable communication awareness map. In practical applications, the number and type of landmarks used to construct the communication awareness map can be selected according to the actual situation.
[0080] Step S2500: Using the azimuth vectors of all landmarks at two adjacent time points, calculate the relative azimuth vector of the target terminal at the two adjacent time points and the azimuth vectors of the anchor nodes and free nodes in each triangle in the target terminal coordinate system, and obtain the azimuth vectors of each node in the similar configuration.
[0081] Specifically, by using the azimuth vectors of all landmarks at two adjacent times, the azimuth information of the target terminal in each similar configuration relative to the motion at two adjacent times is calculated. That is, the vector relationship information between each node in the similar configuration and the motion azimuth is calculated to obtain the azimuth vector of each node in the corresponding similar configuration.
[0082] For example, in Figure 2 In the communication sensing diagram shown, at time t, the target terminal observes any landmark l through the camera and obtains the azimuth vector of landmark l relative to the target terminal in the target terminal coordinate system at the current time t, where the azimuth vector is defined as in formula (1):
[0083]
[0084] in, This represents the azimuth vector of landmark l as observed in the target terminal coordinate system at time t. This indicates the position of landmark l in the target terminal coordinate system. This indicates the position of the target terminal in the target terminal coordinate system. The current position of the target terminal is usually set as the origin. It can be seen that the origin changes dynamically with the movement trajectory of the target terminal.
[0085] Based on the principle of formula (1), the target terminal directly obtains the azimuth vectors of landmarks i, j, and k at the location of the target terminal in the similar configuration from time t-1 to time t through observation. That is, the target terminal at position r t Observed position r t The orientation vectors between landmark i, landmark j, and landmark k respectively and the target terminal at location r t-1 Observed position r t-1 The azimuth vectors between landmark j and landmark i, landmark j and landmark k respectively.
[0086] Step S2600: Based on the orientation vectors of each node in the similar configuration, obtain a set of matching coordinates.
[0087] Specifically, by Figure 3 It can be known that the position r of the target terminal at time t and time t-1 is... t Position r t-1 These landmarks, i, j, and k, form three triangles. Therefore, according to the geometric properties of triangles, in a planar triangle, the sum of the vectors of any two sides equals the vector of the third side. Figure 4 The diagram shows the vector relationships between the nodes that form a triangle in three-dimensional space. Let... Where 03 represents the 3-dimensional zero vector, and its position r at time t and time t-1. t r t-1 In the triangle formed with landmark i, the sides satisfy the following vector relationship as shown in formula (2):
[0088]
[0089] Where ω1 and ω2 are two scalars. Since the coordinates of each point in the triangle are three-dimensional coordinates, and due to potential measurement noise, it is not easy to find node coordinates that exactly satisfy the above equation. Therefore, scalars ω1 and ω2 are introduced to satisfy the objective function shown in formula (3):
[0090]
[0091] Solving the objective function yields ω1 and ω2, which is equivalent to finding a landmark i that approximately satisfies the aforementioned triangle property. Therefore, the coordinate representation of landmark i in this similar configuration is q. i for or Similarly, the coordinate representations q of landmark j and landmark k in the corresponding similar configurations can be solved. j and q k .
[0092] The coordinates of landmark i, landmark j, and landmark k in the corresponding similar configurations are used to represent q. i q j and q k And the location r of the target terminal at time t and time t-1. t r t-1This forms a set of matching coordinates.
[0093] In this embodiment, a triangle is constructed with each anchor node or free node at two adjacent moments using the positions of the target terminal in the similar configuration. Based on the geometric properties of the triangle, the anchor nodes and free nodes that satisfy the similar configuration are determined. A set of matching coordinates is generated using the position coordinates of all target terminals constructing the similar configuration and the found anchor nodes and free nodes. It is evident that only by finding matching coordinates that meet the shape requirements of the configuration can a similar configuration be constructed. In other words, matching coordinates are the foundation for constructing a similar configuration, and finding suitable matching coordinates lays a solid foundation for accurately calculating the coordinates of the nodes subsequently.
[0094] In one implementation, step S2500, using the azimuth vectors of all landmarks at two adjacent time points, calculates the relative azimuth vector of the target terminal's movement at two adjacent time points, as well as the azimuth vectors of anchor nodes and free nodes in each triangle in the target terminal's coordinate system, to obtain the azimuth vectors of each node in the similar configuration. Specifically, this includes:
[0095] Step S2510: Based on the azimuth vectors of the landmarks observed by the target terminal at two adjacent time points, calculate the relative rotation matrix and relative translation matrix of the target terminal between the two adjacent time points;
[0096] Step S2520: Based on the target terminal coordinate system, use the relative translation matrix to calculate the relative azimuth vector of the target terminal at two adjacent moments; and use the relative rotation matrix to calculate the azimuth vector of each anchor node and free node in each triangle in the target terminal coordinate system, so as to obtain the azimuth vector of each node in the similar configuration.
[0097] Specifically, based on the principle of epipolar geometry, the relative rotation matrix of the target terminal from time t-1 to time t is calculated according to the azimuth vectors of the landmarks at time t-1 and time t. With relative translation matrix And using relative translation matrix The azimuth vector of the target terminal's motion from time t-1 to time t is calculated, i.e.
[0098] Based on the properties of the rotation matrix, calculate the position r of the target terminal at time t (as observed at time t-1) in the target terminal coordinate system. t-1 The azimuth vectors to landmarks i, j, and k are respectively and The specific calculation formula is shown in formula (4):
[0099]
[0100]
[0101]
[0102] In this embodiment, the principle of epipolar geometry is used to first calculate the relative rotation matrix and relative translation matrix of the target terminal between any two adjacent time points in each similar configuration. Then, based on the relative rotation matrix and relative translation matrix, the orientation vectors between each node in the similar configuration (including the orientation vectors of the anchor node and the free node, as well as the relative orientation vectors of the target terminal at any two adjacent time points) are calculated, which is convenient, fast, simple and efficient.
[0103] like Figure 2 As shown, in one implementation, step S3000, which involves creating a target SLAM matrix equation based on the coordinates of all centroids at the same time, specifically includes:
[0104] Step S3100: Based on the properties of the barycenter coordinates, construct a SLAM linear equation for each set of barycenter coordinates;
[0105] Specifically, by using the coherent coordinates of similar configurations, the centroid coordinates of the target terminal and the landmark can be easily solved.
[0106] For example, the formula for solving the barycenter coordinates of configuration (G5,p5) is shown in formula (5):
[0107]
[0108] Where, p5 = [p m ,p n ,p x ,p y ,p z ] T p m p n p x p y and p z This represents the coordinates of the five nodes in the configuration, a zm a zn a zx a zy These represent the barycentric coordinates of the five nodes in the configuration. The node and barycentric coordinates are not actual coordinates; they are only used to formally represent the formula for solving the barycentric coordinates and the properties of the barycentric coordinates.
[0109] Using the set of coincident coordinates satisfying the configuration (G5, p5) obtained in the above embodiments q i q j q k ,Will qi q j q k , Replace p in the formulas for solving the centroid coordinates respectively m p n p x p y p z The centroid coordinates a are obtained. zm a zn a zx a zy .
[0110] Typically, a complete communication sensing map contains multiple sub-maps, which are several landmarks collected by the target terminal. These landmarks, along with the positions of any two adjacent time points, can construct several communication sensing maps, meaning there are multiple sets of matching coordinates. Based on the centroid coordinates of each set of matching coordinates, a SLAM linear equation is established. Clearly, the number of SLAM linear equations constructed at each time point is the same as the number of sets of matching coordinates obtained at that time point.
[0111] For example, because the centroid coordinates have the following property, as shown in formula (6):
[0112]
[0113] Therefore, based on the position r of the target terminal at time t and time t-1 t Position r t-1 The linear equation constructed from the centroid coordinates of the communication perception map built from landmarks i, j, and k is shown in Equation (7):
[0114]
[0115] in, This indicates that the target terminal is at position r. t-1 Relative to the target terminal at position r t The coordinates of the center of gravity, This indicates that landmark i is at position r relative to the target terminal. t The coordinates of the center of gravity, This indicates that landmark j is at position r relative to the target terminal. t The coordinates of the center of gravity, This indicates that the landmark k is at position r relative to the target terminal. t centroid coordinates This represents the position coordinates of the target terminal at time t. p represents the position coordinates of the target terminal at time t-1. i p j and p k These represent the coordinates of landmark i, landmark j, and landmark k, respectively.
[0116] It should be noted that the nodes in the similar configurations (G5,q5) and (G5,p5) in this embodiment are the same, that is, they both use the position r of the target terminal at time t. t The position r at time t-1 t-1 The q is constructed from the {i, j, k} of any three observed landmarks, and thus the solved q i q j and q k , and p i p j and p k Each of these represents the coincident coordinates and actual coordinates of landmark i, landmark j, and landmark k, respectively. The solved coordinates are... and Both represent the coincident coordinates and actual coordinates of the target terminal at time t, and the solved coordinates are... and Both represent the coincident coordinates and actual coordinates of the target terminal at time t-1.
[0117] Step S3300: Merge all SLAM linear equations at the same time point to create a target SLAM matrix equation.
[0118] Specifically, all SLAM linear equations representing the coordinates of the same node at the same time are merged, that is, a SLAM linear fusion equation is created using all the centroid coordinates at that time. The SLAM linear fusion equations of all nodes are merged and converted into matrix form, which is called the target SLAM matrix equation.
[0119] In this embodiment, based on the properties of barycenter coordinates in three-dimensional space, the barycenter coordinates are solved using a set of coincident coordinates, and a linear equation is constructed. Then, the barycenter coordinates representing the same node coordinates at the same time are fused to create a target SLAM matrix equation. Clearly, the target SLAM matrix equation includes the position information of all nodes observed at the same time, which is beneficial for subsequently determining the coordinates of unknown nodes. Therefore, by utilizing the properties of the barycenter in three-dimensional space to construct a linear constraint equation, the nonlinear localization problem is linearized, simplifying the computational complexity.
[0120] like Figure 2 As shown, in one implementation, step S3300, which involves merging all SLAM linear equations at the same time point to create a target SLAM matrix equation, specifically includes:
[0121] Step S3310: Superimpose all SLAM linear equations representing the same node coordinates at the same time, assign weights to each centroid coordinate, and the sum of the weights of each centroid coordinate under each SLAM linear equation is 1, to obtain a SLAM linear fusion equation.
[0122] Specifically, all positions r including the target terminal at time t. t The linear equations constructed from the communication-aware graphs are added together and weighted to obtain a SLAM linear fusion equation, as shown in formula (8):
[0123]
[0124] in, In the c-th similar configuration, r t-1 Relative to r t barycentric coordinates Indicates shared ownership Each configuration contains r t and r t-1 γ c The weights are randomized, and the sum of the weights of all the centroid coordinates of the equation is ensured to be 1, as shown in formula (9), i.e.:
[0125]
[0126] in, These represent the coordinates of the barycenter. The weighted sum, where This represents the centroid coordinates of the last node in all similar configurations of the target terminal at time t. The specific number of centroid coordinates is determined by the number of nodes in all communication sensing graphs constructed by the target terminal at time t.
[0127] Therefore, the SLAM linear fusion equation becomes formula (10), that is:
[0128]
[0129] Where, p num This represents the coordinates of the last node among all similar configurations of the target terminal at time t.
[0130] Step S3320: Construct an initial SLAM matrix equation based on the SLAM linear fusion equation;
[0131] Step S3330: Merge the SLAM linear fusion equations of all landmarks and target terminal positions at two adjacent time points. Based on the number of nodes of each type in the initial SLAM matrix equation, use the Kronecker product to transform and create a target SLAM matrix equation.
[0132] Specifically, the SLAM linear fusion equation is written in matrix form, as shown in equation (11), that is:
[0133]
[0134] Where, p a p represents an anchor node with known coordinates. f p represents a free node with unknown coordinates. r This indicates the location of the target terminal at the corresponding time. I represents the Kronecker product. n Let B0 be an n-dimensional identity matrix, and its specific form is shown in formula (12), that is:
[0135]
[0136] Where row and col represent the row-th and col-th columns of the matrix, V a V represents the set of anchor nodes. z (t)=V f (t)∪V r (t), V f V represents the set of free nodes. r This represents the set of locations of the target terminal at various times during the movement period.
[0137] definition B(t) = I n -B0(t), that is:
[0138] Among them, B a B z These are the block matrices representing the anchor node coefficients and the block matrices representing the target terminal's two time-stamp positions and landmark coefficients, respectively. z This represents the number of positions at two time points, one for the free node and one for the target terminal.
[0139] By performing a matrix transformation on the matrix corresponding to the SLAM linear fusion equation, we obtain the form of formula (13), that is:
[0140]
[0141] Finally, it can be simplified to the matrix equation shown in formula (14), that is:
[0142]
[0143] Among them, B z (t), B a (t) is called the matrix based on the centroid coordinates, and the linear equation based on the centroid coordinates is now complete.
[0144] It is easy to understand that the location of the target terminal, anchor node and free node appearing in this invention are collectively referred to as nodes, and both anchor node and free node belong to landmarks in the global coordinate system.
[0145] In this embodiment, by merging all SLAM linear equations at the same time point, a SLAM linear fusion equation is obtained. The SLAM linear fusion equation is then converted into a matrix and subjected to matrix transformation to create a target SLAM matrix equation with a simpler form. This helps to reduce algorithm complexity and improve solution efficiency.
[0146] like Figure 2 As shown, in one embodiment, step S4000, solving the target SLAM matrix equation at each time point to obtain the target terminal's position coordinates and the landmark's coordinates, specifically includes:
[0147] Step S4100: Perform matrix transformation on the target SLAM matrix equation to obtain a non-singular matrix equation;
[0148] Step S4200: Use the conjugate gradient algorithm to iteratively solve the non-singular matrix equation until the preset number of iterations is reached, or the residual between the estimated coordinates of the target terminal at the corresponding time or the coordinates of some landmarks and their respective true positions is less than the preset residual threshold.
[0149] Specifically, the matrix equation based on the barycentric coordinates has one and only in B z (t) Non-singular (B) z It can only be solved when (t) is full rank, defined as follows: Multiply both sides of the linear equation by The transpose of gives the equation shown in formula (15), that is:
[0150]
[0151] Assume B z If (t) is nonsingular, then Since it is symmetric and positive definite, the equation can be solved iteratively using the conjugate gradient algorithm. The algorithm steps are as follows:
[0152] Define the iteration number k, the residual re, and the direction vector d, and set the initial values as shown in formula (16), i.e.:
[0153] k = 0, (16)
[0154]
[0155] d[0] = re[0],
[0156] p is continuously processed through the following steps. z The re and d are iteratively updated as follows:
[0157] Calculate the scalar parameter α to ensure that each residual update is mutually orthogonal, i.e., satisfying... The calculation method is shown in formula (17), that is:
[0158]
[0159] Based on the scalar parameter α, p z re is updated, and the calculation method is shown in formula (18), that is:
[0160]
[0161]
[0162] Calculate the scalar parameter β to ensure that each update of the direction vector is conjugate, i.e., satisfying... The calculation method is shown in formula (19), that is:
[0163]
[0164] The parameter β is used to update d, and the calculation method is shown in formula (20), that is:
[0165] d[k+1]=re[k+1]+β[k]d[k] (20)
[0166] Let k = k + 1, and proceed to the next iteration until re[k + 1] is less than ε, where ε represents a very small positive constant. At this point, p f p r , The relationship is as shown in formula (21), that is:
[0167]
[0168] At this point, the locations of some landmarks and the target terminal at times t-1 and t have converged to their actual locations. This is easily understood... Figure 2 The label coordinates refer to the landmark coordinates.
[0169] In this embodiment, by introducing the residual re, the direction vector d, and the scalar parameters α and β, the target SLAM matrix equation is solved iteratively to ensure the convergence of the algorithm and to accurately and efficiently calculate the true positions of each landmark and target terminal.
[0170] Furthermore, such as Figure 2As shown, landmarks whose true coordinates are determined after iterative calculation are added to the anchor node set. As long as a landmark in the anchor node set can be observed at any given time, it can be used as a known anchor node to calculate the coordinates of subsequently observed landmarks and the coordinates of the target terminal's location. As time progresses and the target terminal continues to move, the method provided in the above embodiments continuously determines the true coordinates of more tags, resulting in more and more landmarks being added to the anchor node set.
[0171] Furthermore, such as Figure 2 As shown, over time, the position of the target terminal determined by the method provided in the above embodiments will be recorded each time it moves. When the target terminal stops moving, the coordinates of all observed landmarks and the movement trajectory of the target terminal can be obtained.
[0172] Exemplary System
[0173] like Figure 5 As shown, corresponding to the above-described visual SLAM method based on centroid coordinates, this embodiment of the invention also provides a visual SLAM system based on centroid coordinates, the above-described visual SLAM system based on centroid coordinates includes:
[0174] The information acquisition module 510 is used to acquire visual information collected by the target terminal in real time, obtain the orientation vectors of several landmarks based on the visual information, and obtain the relative orientation vectors of the target terminal at adjacent times based on the orientation vectors of the landmarks.
[0175] The coincidence coordinate solving module 520 is used to construct several similar configurations based on a preset configuration, using the azimuth vectors of all landmarks at any two adjacent times and the relative azimuth vector of the target terminal movement, and to obtain a set of coincidence coordinates based on each similar configuration.
[0176] The SLAM matrix equation construction module 530 is used to solve for the barycenter coordinates of corresponding similar configurations using coincident coordinates; and to create a target SLAM matrix equation based on all barycenter coordinates at the same time.
[0177] The motion trajectory generation module 540 is used to solve the target SLAM matrix equation at each time step, obtain the position coordinates of the target terminal and the coordinates of the landmark, and use the position coordinates of the target terminal and the coordinates of the landmark to construct the motion trajectory of the target terminal and complete the mapping.
[0178] Specifically, in this embodiment, the specific functions of the above-mentioned visual SLAM system based on centroid coordinates can also be referred to the corresponding description in the above-mentioned visual SLAM method based on centroid coordinates, and will not be repeated here.
[0179] Based on the above embodiments, the present invention also provides a smart terminal, the principle block diagram of which can be as follows: Figure 6 As shown. The aforementioned smart terminal includes a processor, memory, network interface, and display screen connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and a barycentric coordinate-based visual SLAM program. The internal memory provides an environment for the operation of the operating system and the barycentric coordinate-based visual SLAM program stored in the non-volatile storage medium. The network interface of the smart terminal is used for communication with external terminals via a network connection. When the barycentric coordinate-based visual SLAM program is executed by the processor, it implements the steps of any of the aforementioned barycentric coordinate-based visual SLAM methods. The display screen of the smart terminal can be a liquid crystal display (LCD) or an e-ink display.
[0180] Those skilled in the art will understand that Figure 6 The block diagram shown is merely a partial structural diagram related to the present invention and does not constitute a limitation on the smart terminal to which the present invention is applied. A specific smart terminal may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0181] In one embodiment, a smart terminal is provided, the smart terminal including a memory, a processor, and a barycentric coordinate-based visual SLAM program stored in the memory and executable on the processor. When the barycentric coordinate-based visual SLAM program is executed by the processor, it implements the steps of any barycentric coordinate-based visual SLAM method provided in the embodiments of the present invention.
[0182] This invention also provides a computer-readable storage medium storing a barycentric coordinate-based visual SLAM program. When the barycentric coordinate-based visual SLAM program is executed by a processor, it implements the steps of any barycentric coordinate-based visual SLAM method provided in this invention.
[0183] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.
[0184] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the above device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this invention. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.
[0185] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.
[0186] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.
[0187] In the embodiments provided by this invention, it should be understood that the disclosed apparatus / terminal devices and methods can be implemented in other ways. For example, the apparatus / terminal device embodiments described above are merely illustrative. For instance, the division of the above modules or units is merely a logical functional division, and in actual implementation, it can be divided in other ways. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.
[0188] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not mean that the essence of the corresponding technical solutions deviates from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.
Claims
1. A visual SLAM method based on centroid coordinates, characterized in that, include: The system acquires visual information collected by the target terminal in real time, obtains the orientation vectors of several landmarks based on the visual information, and obtains the relative orientation vectors of the target terminal at adjacent times based on the orientation vectors of the landmarks. Define several nodes with known coordinates to obtain anchor nodes. Based on a preset configuration, use the orientation vectors of all the landmarks at any two adjacent times and the relative orientation vector of the target terminal to construct several similar configurations that are similar to the configuration. Based on each similar configuration, obtain a set of matching coordinates. Using the coherent coordinates, the barycenter coordinates of the corresponding similar configurations are solved, and a target SLAM matrix equation is created based on all the barycenter coordinates at the same time. Solve the target SLAM matrix equation at each time point to obtain the position coordinates of the target terminal and the coordinates of the landmark. Then, use the position coordinates of the target terminal and the coordinates of the landmark to construct the motion trajectory of the target terminal and complete the mapping. Add the landmarks whose coordinates are determined after the solution to the anchor node set; In subsequent moments, if a landmark that has been added to the anchor node set is observed by the target terminal, then that landmark is used as an anchor node to calculate the position coordinates of the target terminal at the corresponding subsequent moment and the coordinates of other landmarks observed at that subsequent moment; The process of creating a target SLAM matrix equation based on all the centroid coordinates at the same time includes: Based on the properties of the centroid coordinates, a SLAM linear equation is constructed using each set of centroid coordinates at the same time; all the SLAM linear equations at the same time are merged to create a target SLAM matrix equation. The step of constructing a SLAM linear equation based on the properties of the centroid coordinates and using each set of centroid coordinates at the same time includes: based on the centroid coordinates, establishing a SLAM equation for the target terminal position at the same time and its position at the previous adjacent time, as well as multiple landmarks observed at the two adjacent times. The SLAM equation is configured to represent the position coordinates of the target terminal at the current time as a linear combination of the position coordinates of the target terminal at the previous adjacent time and the centroid coordinates of the multiple landmark coordinates. The steps for constructing the configuration include: Based on the position of the target terminal at any two adjacent times, and several anchor nodes and several free nodes with unknown coordinates, a communication sensing map is constructed. Based on the communication sensing map, the configuration is constructed using the orientation vectors of the anchor node and the free node observed by the target terminal at the two locations respectively. The process of obtaining a set of matching coordinates based on each of the similar configurations includes: Based on each of the similar configurations, a triangle is constructed with each anchor node or free node observed at the corresponding time using the position of the target terminal at any two adjacent times. Using the azimuth vectors of all the landmarks at the two adjacent times, calculate the relative azimuth vector of the target terminal at the two adjacent times and the azimuth vectors of the anchor node and the free node in each triangle in the target terminal coordinate system, and obtain the azimuth vectors of each node in the similar configuration. Based on the orientation vectors of each node in the similar configuration, a set of matching coordinates is obtained.
2. The visual SLAM method based on centroid coordinates according to claim 1, characterized in that, The step of using the azimuth vectors of all the landmarks at the two adjacent time points to calculate the relative azimuth vector of the target terminal's movement at the two adjacent time points, and the azimuth vectors of the anchor nodes and free nodes in each triangle in the target terminal coordinate system, to obtain the azimuth vectors of each node in the similar configuration, includes: Based on the azimuth vectors of the landmarks observed by the target terminal at the two adjacent times, calculate the relative rotation matrix and relative translation matrix of the target terminal between the two adjacent times; Based on the target terminal coordinate system, the relative translation matrix is used to calculate the relative azimuth vector of the target terminal at two adjacent times; and the relative rotation matrix is used to calculate the azimuth vector of the anchor node and the free node in each triangle in the target terminal coordinate system, thereby obtaining the azimuth vector of each node in the similar configuration.
3. The visual SLAM method based on centroid coordinates according to claim 2, characterized in that, The step of merging all the SLAM linear equations at the same time point to create a target SLAM matrix equation includes: All SLAM linear equations representing the same node coordinates at the same time are superimposed, and each centroid coordinate is assigned a weight. The sum of the weights of each centroid coordinate under each SLAM linear equation is 1, thus obtaining a SLAM linear fusion equation. Based on the aforementioned SLAM linear fusion equation, an initial SLAM matrix equation is constructed; The SLAM linear fusion equations of all landmarks and target terminal locations at two adjacent time points are merged. Based on the number of nodes of each type in the initial SLAM matrix equation, the Kronecker product is used for transformation to create a target SLAM matrix equation.
4. The visual SLAM method based on centroid coordinates according to claim 1, characterized in that, Solving the target SLAM matrix equation at each time step to obtain the location coordinates of the target terminal and the coordinates of the landmark includes: The target SLAM matrix equation is transformed to obtain a non-singular matrix equation; The non-singular matrix equation is solved iteratively using the conjugate gradient algorithm until a preset number of iterations is reached, or the residual between the estimated coordinates of the target terminal at the corresponding time or the coordinates of some landmarks and their respective true positions is less than a preset residual threshold.
5. A barycentric coordinate-based visual SLAM system for implementing the barycentric coordinate-based visual SLAM method according to any one of claims 1-4, characterized in that, The system includes: The information acquisition module is used to acquire visual information collected by the target terminal in real time, obtain the orientation vectors of several landmarks based on the visual information, and obtain the relative orientation vector of the target terminal at adjacent time points based on the orientation vectors of the landmarks. The coherent coordinate solving module is used to construct several similar configurations based on a preset configuration, using the azimuth vectors of all the landmarks at any two adjacent times and the relative azimuth vector of the target terminal, and to obtain a set of coherent coordinates based on each similar configuration; The SLAM matrix equation construction module is used to solve for the centroid coordinates of the corresponding similar configuration using the coherent coordinates; and to create a target SLAM matrix equation based on all the centroid coordinates at the same time. The motion trajectory generation module is used to solve the target SLAM matrix equation at each time point, obtain the position coordinates of the target terminal and the coordinates of the landmark, and use the position coordinates of the target terminal and the coordinates of the landmark to construct the motion trajectory of the target terminal and complete the mapping.
6. A smart terminal, used to implement the steps of the visual SLAM method based on centroid coordinates as described in any one of claims 1-4, characterized in that, The smart terminal includes a memory, a processor, and a barycentric coordinate-based visual SLAM program stored in the memory and capable of running on the processor.
7. A computer-readable storage medium for implementing the steps of the visual SLAM method based on centroid coordinates as described in any one of claims 1-4, characterized in that, The computer-readable storage medium stores a visual SLAM program based on centroid coordinates.
