Texture map deduplication method and device and storage medium

Abstract

The invention discloses a texture map deduplication method and device and a storage medium, and the method comprises the steps: obtaining a plurality of groups of multi-source data collected by a map collection device in a motion process, and each group of multi-source data comprises a texture image and laser cloud data collected at a collection moment; at least utilizing the texture images corresponding to the acquisition moments to respectively obtain map frames corresponding to the acquisition moments; comparing the laser cloud features and the texture features of different map frames so as to find out repeated frames meeting the similarity requirement from the map frames; and forming the texture map for positioning and / or repositioning the mobile equipment by using the map frames except the repeated frames. Through the above mode, the method can achieve the de-duplication of the texture map, and can improve the accuracy of using the texture map to carry out navigation and positioning.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of visual navigation positioning, in particular to a texture map deduplication method, device and storage medium. BACKGROUND

[0002] The indoor navigation technology of ground mobile robots mainly includes active navigation such as laser navigation and visual navigation, and passive navigation such as WIFI (Wireless Fidelity), UWB (Ultra Wideband) and Bluetooth base station networking.

[0003] The laser navigation technology has been developed for a long time and is widely used, but the single-line laser has a small amount of information and is highly dependent on the environment, and usually needs to be combined with a reflector, a reflective sticker and the like for navigation and positioning. The multi-line laser has a large amount of data, and a large amount of point cloud data needs to be processed, which requires a high performance of the robot platform and is relatively expensive. The navigation methods such as WIFI, UWB and Bluetooth have low cost and strong scalability, but are easily affected by obstacles and signal interference, and perform poorly in high-precision scenarios. The visual navigation has developed rapidly in recent years, and typical cases such as two-dimensional code navigation are widely used in intelligent warehousing and logistics, but the two-dimensional code is easily contaminated and damaged, and affects the aesthetics of the site after construction. In the prior art, the ground texture information can be collected for navigation and positioning without the need for additional markers, which can to some extent avoid affecting the aesthetics of the site.

[0004] However, in actual application scenarios, in a large scene texture navigation environment, the route length can reach thousands of kilometers or even ten thousand kilometers, which leads to a rapid growth of the texture feature database and increases the possibility of similar visual texture features in the global map. At the same time, in specific scenarios (such as artificial ceramic tiles and artificial floors), the ground texture has repeatability and similarity due to the use of similar printing templates, which may cause technical problems such as positioning result jumping, repositioning result error and the like. SUMMARY

[0005] To solve the above technical problems, the present application provides a texture map deduplication method, device and storage medium to at least solve the problem that the ground texture has a certain repeatability and similarity in the related art, which may cause technical problems such as positioning result jumping and repositioning result error.

[0006] According to an embodiment of the present application, a texture map deduplication method is provided, comprising: obtaining a plurality of groups of multi-source data collected by a map collection device during movement, each group of multi-source data comprising a texture image and laser cloud data collected at a collection time; obtaining a map frame corresponding to each collection time by using at least the texture image corresponding to each collection time, respectively; comparing the laser cloud features and the texture features of different map frames to find out repeated frames from the map frames which meet the similarity requirement, wherein the laser cloud features are determined based on laser source data corresponding to the same acquisition time as the map frames, and the texture features are determined based on texture images corresponding to the same acquisition time as the map frames; using each of the map frames except the repeated frames to form a texture map used for positioning and / or repositioning of a mobile device.

[0007] To solve the above technical problems, one technical solution of the present application is to provide an electronic device, comprising a memory and a processor, wherein the memory is used to store a computer program, and the computer program is used to implement the texture map deduplication method in the above technical solution when executed by the processor.

[0008] To solve the above technical problems, one technical solution of the present application is to provide a computer readable storage medium, which is used to store a computer program, and the computer program is used to implement the texture map deduplication method in the above technical solution when executed by a processor.

[0009] Through the above scheme, the texture map deduplication method provided by the present application obtains a plurality of sets of multi-source data collected by a map collection device during movement, at least uses the texture images corresponding to each acquisition time to obtain the map frames corresponding to each acquisition time respectively, compares the laser cloud features and the texture features of different map frames to find out repeated frames from the map frames which meet the similarity requirement, and uses each of the map frames except the repeated frames to form a texture map used for positioning and / or repositioning of a mobile device. In this way, the present application combines multi-source data to perform global similarity judgment and local deduplication screening on texture map data, thereby reducing the possibility of global similarity of texture features, and being able to complete the screening of texture map data at the end of collecting texture data, and further being able to improve the accuracy of navigation positioning using the texture map. BRIEF DESCRIPTION OF DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor. Among them: Figure 1 is a flowchart of an embodiment of the texture map deduplication method provided by the present application; Figure 2 is a flowchart of another embodiment of the texture map deduplication method provided by the present application; Figure 3is a structural schematic diagram of an embodiment of an electronic device provided in the present application. Figure 4 is a structural schematic diagram of an embodiment of a computer readable storage medium provided in the present application. DETAILED DESCRIPTION

[0011] The present application will be further described below in conjunction with the accompanying drawings and embodiments. It is particularly pointed out that the following embodiments are only for illustrating the present application, but do not limit the scope of the present application. Similarly, the following embodiments are only part of the embodiments of the present application, but not all the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art without making creative efforts fall within the scope of protection of the present application.

[0012] In the present application, the phrase "embodiment" means that the specific features, structures or characteristics described in conjunction with the embodiment can be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily mean the same embodiment, nor is it an independent or alternative embodiment to other embodiments. Those skilled in the art explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0013] It should be noted that in the description of the present application, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly and specifically limited. In addition, the terms "comprise" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but can optionally include steps or units not listed, or can optionally include other steps or units inherent to these processes, methods, products or devices.

[0014] Please refer to Figure 1 , Figure 1 is a flowchart of an embodiment of a texture map deduplication method provided in the present application. It should be noted that the present embodiment is not limited to the flow order shown in Figure 1 . As shown in Figure 1 , the present embodiment includes: S110: Obtain a plurality of sets of multi-source data collected by a map collection device during movement.

[0015] The map collection device refers to a special device integrating at least an image collection module, a laser radar module, an odometer module and a data processing module, which is used to obtain geographic spatial data. The multi-source data refers to a set of geographic data collected by the map collection device at the same collection time, which is different in source but associated. Each set of multi-source data includes texture images, laser cloud data, odometer data, etc., and each type of data is associated through a time stamp.

[0016] In the plurality of sets of multi-source data collected by the map collection device during the movement, each set of multi-source data includes a texture image and laser cloud data collected at a collection time. The texture image refers to two-dimensional image data reflecting the texture characteristics of the surface of the geographical scene, which is captured by the image collection module of the map collection device, and contains the color, contour, and details of the objects in the scene. The laser cloud data refers to three-dimensional point cloud data generated after the laser radar module of the map collection device emits a laser beam and receives a reflected signal, which can be used to construct the three-dimensional geometric structure of the geographical scene.

[0017] Each set of multi-source data can also include odometer data collected at a collection time. The odometer data refers to parameter data reflecting the motion state of the device collected by the odometer module of the map collection device, which at least includes the real-time position, motion speed, and heading angle of the device, and can be used to associate multi-source data at different collection times and provide key position reference information for accurate mapping and data fusion of the map.

[0018] The plurality of sets of multi-source data collected by the map collection device during the movement are obtained. In an embodiment, each set of source data collected at different collection times is obtained, and each set of source data collected at different collection times is time-aligned to obtain a set of multi-source data collected at the same collection time.

[0019] By obtaining multi-source data, the limitations of traditional single data in map making can be overcome. For example, the visual details of the scene can be supplemented by the texture image to avoid the problem of no texture and difficult identification of laser cloud data; the three-dimensional geometric structure can be provided by the laser cloud data to make up for the defect of no depth and difficult modeling of the texture image; and the spatial splicing of multiple sets of multi-source data is realized by the odometer data to ensure the continuity of the map.

[0020] S120: At least the texture image corresponding to each collection time is used to obtain a map frame corresponding to each collection time, respectively.

[0021] The map frame refers to a two-dimensional map unit with geographical coordinate attributes generated by spatial positioning and geometric correction based on the texture image of a single collection time and the corresponding odometer data. Each map frame contains texture information and accurate spatial position, and can be used as a basic component unit of a high-precision map. Adjacent map frames are spliced to form a complete map through spatial coordinates.

[0022] In an embodiment, for each acquisition time, a map frame of the acquisition time is obtained by fusing the texture image of the acquisition time and at least one other source data of the acquisition time. In an implementation, the preprocessed multi-source data of the same acquisition time is extracted first, such as the environment texture image after distortion correction, the laser cloud data after noise reduction, the odometer data containing position and attitude, and the landmark point identification data, etc. The laser cloud data and the landmark point coordinates are converted into geographic coordinates by a specified formula to realize unification, taking the coordinate system of the odometer as the reference. Then, the feature points of the texture image and the landmark points are detected and descriptors are generated, the edge feature points of the laser cloud data are projected and descriptors are extracted, and the coordinate mapping relationship of the three types of data is established. Subsequently, the texture color and the navigation attribute of the landmark points are assigned to the three-dimensional points of the laser cloud according to the mapping, a three-dimensional point cloud with navigation attributes is generated, the geographic boundary is determined in combination with the odometer and the landmark points, and the supplementary fusion metadata is obtained, to obtain the fused map frame, so as to complete the acquisition of the map frame of the acquisition time.

[0023] S130: Comparing the laser cloud features and the texture features of different map frames to find out the repeated frames meeting the similarity requirement from each map frame.

[0024] The laser cloud features of the map frame are determined based on the laser source data of the same acquisition time corresponding to the map frame, and the texture features are determined based on the texture image of the same acquisition time corresponding to the map frame. By comparing the laser cloud features and the texture features of different map frames, the repeated frames meeting the similarity requirement are found out from each map frame.

[0025] In an embodiment, each map frame is taken as a current frame, and a plurality of to-be-processed frames located within a preset distance range of the current frame are searched from the remaining map frames except the current frame. The laser cloud features and the texture features of the current frame and the to-be-processed frames are compared to obtain the feature similarity between the current frame and the to-be-processed frames. The to-be-processed frame and the current frame with the feature similarity greater than a similarity threshold are determined as the repeated frames meeting the similarity requirement.

[0026] In an example, the nearest neighbor search of a preset distance range is performed on the current frame by using a spatial separation tree (KD-Tree) structure to obtain a plurality of to-be-processed frames, wherein the spatial separation tree structure is constructed based on the odometer data corresponding to each map frame. The laser cloud feature matching degree between the laser cloud features of the current frame and the to-be-processed frames is obtained, and the texture feature matching degree between the texture features of the current frame and the to-be-processed frames is obtained, for the plurality of to-be-processed frames obtained by the nearest neighbor search of the current frame. The laser cloud feature matching degree and the texture feature matching degree are weighted and fused to obtain the feature similarity between the current frame and the to-be-processed frames.

[0027] If the feature similarity between the current frame and the to-be-processed frame is greater than the similarity threshold, it is considered that the to-be-processed frame is similar to the current frame, and the two frames can be marked as repeated frames, which are not used in subsequent relocation to achieve the effect of positioning data deduplication. If the feature similarity between the current frame and the to-be-processed frame is not greater than the similarity threshold, it is considered that the to-be-processed frame is not similar to the current frame, and the feature similarity between the next to-be-processed frame and the current frame is calculated in turn.

[0028] In response to the existence of the to-be-processed frame with the feature similarity greater than the similarity threshold, the calculation of the feature similarity of the remaining to-be-processed frames of the current frame is stopped, and one of the remaining map frames that is not the current frame is selected as a new current frame, and the subsequent steps of searching for a plurality of to-be-processed frames within a preset distance range of the current frame from the remaining map frames except the current frame are re-executed.

[0029] If there is no to-be-processed frame with the feature similarity greater than the similarity threshold in the plurality of to-be-processed frames corresponding to the current frame, the subsequent steps of searching for a plurality of to-be-processed frames within a preset distance range of the current frame from the remaining map frames except the current frame are sequentially executed until all the map frames are traversed to obtain all the repeated frames in the map frames fused by the current map collection device in the movement process.

[0030] In an embodiment, for the map frames that have been marked as repeated frames, a repeated frame queue can be constructed, and in subsequent nearest neighbor search based on the current frame to obtain the current plurality of to-be-processed frames, the to-be-processed frames that have been determined to be repeated frames are removed from the plurality of to-be-processed frames to improve the calculation efficiency. In other embodiments, for the two-by-two map frames for which the feature similarity has been calculated, the calculation results can also be retained to avoid repeated calculation.

[0031] S140: Use each map frame except the repeated frame to form a texture map for positioning and / or relocating the mobile device.

[0032] When the mobile device moves in the target environment, it continuously collects image frames of the surrounding environment, wherein the image frames contain texture details in the environment, such as the texture of the ground, etc. In the collected image sequence, there may be some repeated or highly similar image frames, which affect the positioning and / or relocation of the mobile device and also increase the burden of data processing.

[0033] Therefore, by using the method for deduplicating the texture map provided in this embodiment, the repeated frames can be identified and marked, and the map frames with unique texture information are retained, that is, each map frame except the repeated frame is used to form a texture map for positioning and / or relocating the mobile device.

[0034] In an embodiment, when the mobile device needs to determine its current position, a current image frame can be captured, its texture features are extracted, and matched with the features in the texture map. By finding the most similar map frame or feature point to the current image frame, the position of the mobile device in the texture map can be determined.

[0035] In an embodiment, during the running of the mobile device, due to various factors such as sensor errors, environmental changes, etc., the accumulation of positioning errors can occur. Re-positioning refers to when the mobile device detects that the positioning error is large or loses positioning, the accurate position information is restored by re-matching with the texture map. For example, when the mobile device enters a new area or after a period of running, its position can be re-determined by searching for features in the texture map.

[0036] Referring to Figure 2 , Figure 2 is a flowchart of another embodiment of the texture map deduplication method provided by the present application. It should be noted that the present embodiment is not limited to the flow order shown in Figure 2 . As shown in Figure 2 , the present embodiment includes: S201: Collecting texture data of a target environment.

[0037] The texture data of the target environment is collected using a map collection device. In an embodiment, the robot is guided along a route that includes first landmark points (used to indicate the direction of travel, such as L-shaped turns, T-shaped U-turn / termination identifiers) and second landmark points (used to trigger the collection of visual navigation templates, such as cross-shaped straight-ahead, L-shaped turn identifiers). The images containing the landmark points are preprocessed and input into a neural network-multilayer perceptron (ANN_MLP) to train the identifier recognition model. When the robot is moving, the travel images are captured by the camera, and the trained model is used to identify the identifiers in the images. When the first landmark point is identified, the direction of travel is determined and the image is stored. When the second landmark point is identified, the direction is maintained and the image is stored. The stored images are matched and spliced to obtain the visual navigation image template. The template is detected for feature points and descriptors, and the Keyframe structure is used to store and establish a mapping relationship with the landmark point number. The robot image template library is compressed by serializing binary encoding to complete the texture data collection.

[0038] S202: Associating laser point cloud data and odometry data.

[0039] After the texture data, the laser point cloud data and the odometry data collected by the map collection device during the movement are acquired, the three types of data are associated, and the texture data, the laser point cloud data and the odometry data are time-stamped and aligned.

[0040] S203: Construct fusion data and acquire a queue of to-be-processed map data.

[0041] For each time-related texture, laser and odometry data, the three are bound as the same frame data, so as to construct fusion data and be pressed into the queue of to-be-processed map data, so as to acquire the queue of to-be-processed map data of the target environment.

[0042] S204: Construct a spatial separation tree structure.

[0043] After the multi-source data collection and the fusion data are completed, a spatial separation tree structure (KD-Tree) is constructed according to the value of the odometry for all the acquired fusion data, so as to find the nearest neighbor of each to-be-processed map data when the texture map is de-duplicated.

[0044] S205: Traverse the to-be-processed map data in sequence.

[0045] Each to-be-processed map data in the queue of to-be-processed map data is traversed in sequence, and the steps of S206-S210 are performed for each frame of to-be-processed map data.

[0046] S206: Screen the to-be-analyzed data according to a positioning prediction range threshold.

[0047] Determine a positioning prediction range threshold r , for each frame of to-be-processed map data , perform a nearest neighbor search with a radius r , and assume that all to-be-analyzed data searched within a range of r around the current frame is , that is and satisfy the constraint of formula (1) to obtain the to-be-analyzed data corresponding to the current to-be-processed map data .

[0048] (1) , wherein odom is the odometry data of the corresponding frame, r is a deviation range size allowed based on positioning and repositioning, and preferably, the specified radius r is 2 meters.

[0049] S207: Calculate the feature similarity of the to-be-analyzed data.

[0050] For each piece of data to be analyzed Calculate the data to be analyzed separately. With the current map data to be processed Feature similarity score between The specific calculation method is shown in formula (2) below, where, (Range 0~1) represents the data to be analyzed. With the current map data to be processed The laser matching similarity score, The weighting parameter for this score (range 0~1). (Range 0~1) represents the data to be analyzed. With the current map data to be processed Texture feature matching similarity score, The weighting parameter for this score.

[0051] (2) In one embodiment, similarity judgment is achieved using visual texture as the core. The environment image is adaptively split using a quadtree algorithm, dividing it into image patches of different scales to balance local details and global structure. Then, the FAST algorithm is used to detect feature points in each image patch, and the rBRIEF algorithm is used to binary encode the feature point pairs, generating a compact texture feature representation. Finally, the similarity is measured by calculating the Hamming distance between the binary feature codes of the current environment image and pre-stored templates in the texture information database (a smaller Hamming distance indicates higher similarity). Simultaneously, the consistency of the distance between the detected feature points in the current frame and the previous frame is combined to further filter matching point pairs, improving the accuracy of the judgment.

[0052] In one embodiment, the laser matching similarity is determined using ICP (Iterative Nearest Point) registration. For example, the laser point cloud to be matched is first divided into a source point cloud (currently acquired data) and a target point cloud (reference data). Using structures such as KD-Tree, the closest Euclidean distance point in the target point cloud is found for each point in the source point cloud, forming preliminary corresponding point pairs. Then, erroneous point pairs are eliminated using criteria such as distance thresholds and the angle between normal vectors, retaining valid point pairs with matching geometric features. Subsequently, based on the valid point pairs, the optimal rigid body transformation (rotation matrix and translation vector) is calculated using methods such as Singular Value Decomposition (SVD) to minimize the distance error between point pairs. The above steps of "finding corresponding points - eliminating errors - calculating transformations" are repeated until the error change is less than the threshold or the maximum number of iterations is reached. Finally, the laser feature similarity is measured by the converged error value; the smaller the error, the higher the similarity.

[0053] It should be noted that the calculation methods for laser matching similarity scores and texture feature similarity scores are not limited to those described in the above embodiments, and are not limited here.

[0054] S208: Determine whether the feature similarity is greater than the similarity threshold.

[0055] Define a similarity threshold ScoreThreshold If the feature similarity score of the data to be analyzed Then the data to be analyzed is considered to be With the current map data to be processed If the conditions are similar, proceed to step S209; otherwise, proceed to step S210.

[0056] S209: Perform deduplication operation.

[0057] For data to be analyzed where the feature similarity is greater than the similarity threshold The current map data to be processed With the data to be analyzed All frames are marked as duplicates so that they will not be used in subsequent relocation, thus achieving the effect of deduplication of the texture map. At the same time, the operation on the current data queue to be analyzed is terminated, and the process proceeds to step S210.

[0058] S210: Determine whether all pending map data has been processed.

[0059] If there are still unprocessed map data in the queue of pending map data, return to step S205 to proceed with the next frame of pending map data. The processing.

[0060] If there is no unchecked map data in the queue of map data to be processed, it is determined that all map data to be processed has been traversed, and the texture map deduplication step ends. Then, the texture map for positioning and / or repositioning of mobile devices is constructed using the map frames other than duplicate frames.

[0061] Please see Figure 3 , Figure 3 This is a schematic diagram of an embodiment of the electronic device provided in this application. The electronic device 60 includes a memory 61 and a processor 62 that are interconnected. The memory 61 is used to store a computer program. When the computer program is executed by the processor 62, it is used to implement the texture map deduplication method in the above embodiment.

[0062] The methods described in the above embodiments can exist in the form of a computer program; therefore, this application proposes a computer-readable storage medium. Please refer to [link / reference needed]. Figure 4 , Figure 4 This is a schematic diagram of an embodiment of a computer-readable storage medium provided in this application. The computer-readable storage medium 80 is used to store a computer program 81, which can be executed to implement the texture map deduplication method in the above embodiment.

[0063] The computer-readable storage medium 80 can be any medium capable of storing program code, such as a server, USB flash drive, external hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0064] The above description is merely an embodiment of this application and does not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A method for deduplicating texture maps, characterized in that, The method includes: Acquire several sets of multi-source data collected by the map acquisition device during its movement. Each set of multi-source data includes texture images and laser cloud data acquired at a certain acquisition time. At least the texture images corresponding to each acquisition time are used to obtain the map frames corresponding to each acquisition time; The laser cloud features and texture features of different map frames are compared to find duplicate frames that meet the similarity requirements from each map frame. The laser cloud features are determined based on laser source data at the same acquisition time corresponding to the map frame, and the texture features are determined based on texture images at the same acquisition time corresponding to the map frame. Using the map frames other than the repeating frames, a texture map is constructed for locating and / or relocating the mobile device.

2. The method according to claim 1, characterized in that, The comparison of laser cloud features and texture features of different map frames to find duplicate frames that meet the similarity requirements from each map frame includes: Each of the aforementioned map frames is taken as the current frame; From the map frames other than the current frame, search for several frames to be processed that are located within a preset distance range of the current frame; The laser cloud features and texture features of the current frame and the frame to be processed are compared to obtain the feature similarity between the current frame and the frame to be processed; The frames to be processed and the current frame whose feature similarity is greater than the similarity threshold are identified as duplicate frames that meet the similarity requirements.

3. The method according to claim 2, characterized in that, The step of comparing the laser cloud features and texture features of the current frame and the frame to be processed to obtain the feature similarity between the current frame and the frame to be processed includes: The laser cloud feature matching degree between the laser cloud feature of the current frame and the laser cloud feature of the frame to be processed is obtained, and the texture feature matching degree between the texture feature of the current frame and the texture feature of the frame to be processed is obtained. The laser cloud feature matching degree and the texture feature matching degree are weighted and fused to obtain the feature similarity between the current frame and the frame to be processed.

4. The method according to claim 2, characterized in that, After determining the frames to be processed and the current frame whose feature similarity is greater than the similarity threshold as duplicate frames that meet the similarity requirement, the method further includes: In response to the existence of a frame to be processed with a feature similarity greater than a similarity threshold, the feature similarity calculation is no longer performed on the remaining frames to be processed of the current frame. Instead, a new current frame is selected from the remaining map frames that were never used as the current frame. The steps of searching for several frames to be processed within a preset distance range of the current frame from the remaining map frames other than the current frame and subsequent steps are then re-executed. And / or, after searching for a number of frames to be processed within a preset distance range of the current frame from the remaining map frames other than the current frame, and before comparing the laser cloud features and texture features of the current frame and the frames to be processed to obtain the feature similarity between the current frame and the frames to be processed, the method further includes: From the plurality of frames to be processed, remove the frames that have been identified as duplicate frames.

5. The method according to claim 2, characterized in that, The step of searching for a number of frames to be processed from the remaining map frames other than the current frame, which are located within a preset distance range of the current frame, includes: Using a spatial partition tree structure, a nearest neighbor search within a preset distance range is performed on the current frame to obtain the plurality of frames to be processed, wherein the spatial partition tree structure is constructed based on the odometry data corresponding to each map frame.

6. The method according to claim 1, characterized in that, Each set of multi-source data includes texture images, laser cloud data, and odometry data acquired at a single acquisition time. And / or, the acquisition of map data acquisition devices during movement includes several sets of multi-source data, including: Acquire data from various sources collected at different collection times; Time alignment is performed on the source data collected at different acquisition times to obtain a set of multi-source data collected at the same acquisition time.

7. The method according to claim 1, characterized in that, The step of obtaining map frames corresponding to each acquisition time by utilizing at least the texture images corresponding to each acquisition time includes: For each acquisition time, the texture image of that acquisition time and at least one other source data of that acquisition time are fused to obtain the map frame of that acquisition time.

8. The method according to claim 1, characterized in that, Before constructing a texture map for locating and / or relocating a mobile device using the map frames other than the repeating frames, the method further includes: Construct a texture map using each of the aforementioned map frames; The step of constructing a texture map for locating and / or relocating a mobile device using all the map frames other than the repeating frames includes: A marker is added to the repeating frames in the texture map, wherein the map frames to which the marker is added are not used for positioning and / or repositioning of the mobile device.

9. An electronic device, characterized in that, The electronic device includes a processor and a memory, the processor being coupled to the memory, the processor being configured to perform one or more steps of the texture map deduplication method according to any one of claims 1 to 8 based on instructions stored in the memory.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that is executed by a processor to implement the steps of the texture map deduplication method as described in any one of claims 1 to 8.

Images