Degradation scene positioning method and device based on reflective pattern, and storage medium

Abstract

The invention discloses a degraded scene positioning method and device based on reflective patterns and a storage medium, and belongs to the technical field of laser positioning, and the method comprises the steps: setting a plurality of reflective patterns in a feature degraded scene according to a preset interval, and carrying out the unique coding of each reflective pattern; wherein each reflective pattern comprises a plurality of reflective areas which are arranged differently, and the reflective patterns are uniquely coded according to the relative position distribution of the reflective areas contained in the reflective patterns; pre-calibrating the reflective pattern with the unique code to obtain a global map position of the reflective pattern; acquiring a global map position for positioning an effective reflective pattern in the surrounding environment of the lost robot; and according to the global map position of the effective reflective pattern and the relative position of the positioning loss robot and the effective reflective pattern, obtaining the global map position of the positioning loss robot. The reflecting pattern with the unique code is used as a positioning reference, other extra hardware is not needed, and the method has the advantages of being high in reliability and wide in application range.

Description

Technical Field

[0001] This invention belongs to the field of laser positioning technology, and particularly relates to a method, device and storage medium for positioning in degraded scenes based on reflective patterns. Background Technology

[0002] In special environments such as tunnels, underground utility tunnels, and long corridors, traditional robotic solutions that rely on lidar sensors for autonomous mapping and localization have significant limitations. The main reasons are that the narrow environments restrict the lidar's field of view, and the environmental features are often monotonous and repetitive, affecting point cloud matching; such scenarios are generally referred to as feature-degraded scenarios. Furthermore, these scenarios are typically located indoors or even underground, making GNSS signals difficult for assisted localization, thus preventing robots from performing long-term, autonomous operations in these environments.

[0003] For similar scenarios, existing solutions mainly fall into the following categories: 1. Combining UWB base stations for positioning and navigation, such as CN118258392A which uses UWB base stations to obtain location information in long corridors. However, due to the poor penetration capability of UWB signals, such solutions are difficult to implement in environments with obstructions such as curves and electric doors. 2. Combining visual sensors for positioning and navigation, such as CN115014338A which uses industrial cameras combined with QR code images and LiDAR to provide mapping and positioning accuracy. However, visual solutions are affected by the ambient light intensity, resulting in insufficient environmental adaptability and reliability. 3. Using reflective markers to assist in positioning, but the markers cannot be uniquely identified. For example, CN117782056A uses reflective stickers for degraded scene-assisted positioning. This type of method is based on the relative positional relationship between reflective markers and does not save the global map position of the markers. If the recursive relationship is accidentally lost during operation, there is still a risk of positioning loss. 4. In addition to the above single methods, there are also solutions that integrate multiple additional sensors for fusion positioning. Compared with other methods, this has higher reliability, but it increases the cost and structural complexity of robot equipment, making it less feasible.

[0004] In summary, existing technical solutions have significant drawbacks in degraded scenarios such as tunnels, underground utility tunnels, and long corridors, including large on-site implementation workload, poor reliability, and complex equipment. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method, device and storage medium for localization of degraded scenes based on reflective patterns. It uses LiDAR to identify reflective patterns in the robot's surrounding environment and performs auxiliary localization based on the unique code of the reflective pattern. It does not rely on other additional hardware and has the characteristics of high reliability and wide applicability.

[0006] To achieve the above objectives, the present invention is implemented using the following technical solution:

[0007] In a first aspect, the present invention provides a method for locating degraded scenes based on reflective patterns, comprising:

[0008] Multiple reflective patterns are set within the feature degradation scene according to a preset interval, and each reflective pattern is uniquely encoded; each reflective pattern contains multiple reflective areas arranged differently, and the reflective pattern is uniquely encoded according to the relative position distribution of the reflective areas it contains.

[0009] Each reflective pattern with a unique code is pre-calibrated to obtain the global map position of each reflective pattern;

[0010] Obtain the global map location of effective reflective patterns in the environment surrounding the lost robot;

[0011] The global map position of the lost robot is obtained based on the global map position of the effective reflective pattern and the relative position of the lost robot to the effective reflective pattern.

[0012] Optionally, the reflective pattern is a reflective rectangular array, in which multiple spaced square reflective strips are present;

[0013] The square reflective strips are distributed at the three vertices of the reflective rectangular array and in other areas. The square reflective strips at the vertices are used to determine the size and orientation of the reflective rectangular array.

[0014] The square reflective strips distributed on a set of diagonals in the reflective rectangular array are all in the horizontal direction.

[0015] Optionally, the areas other than the top corner of the reflective rectangular array are characterized by the presence or absence of square reflective strips corresponding to binary 1 and 0, and the encoding of the reflective rectangular array is composed by placing each binary 1 and 0 in the corresponding position according to a preset sorting rule.

[0016] Optionally, obtaining the global map location of the effective reflective pattern in the environment surrounding the lost robot includes:

[0017] The three-dimensional point cloud information of the surrounding environment is obtained by installing a lidar on the robot that has lost its location, and the three-dimensional point cloud information is reduced in dimension to obtain a two-dimensional planar point cloud.

[0018] The two-dimensional planar point cloud is identified, and the presence of a valid reflective pattern is determined by combining the reflective pattern characteristics.

[0019] If a valid reflective pattern exists, its associated pre-calibrated global map location is obtained based on its corresponding code.

[0020] Optionally, the step of acquiring three-dimensional point cloud information of the surrounding environment and performing dimensionality reduction processing on the three-dimensional point cloud information to obtain a two-dimensional planar point cloud includes:

[0021] Based on the characteristics of reflective patterns in the 3D point cloud information, 3D point cloud preprocessing is performed to obtain effective point cloud information;

[0022] Obtain the point cloud center and normal vector from the valid point cloud information, and determine the direction of the reflective pattern within it;

[0023] Based on the direction of the reflective pattern, the 3D point cloud is translated to the point cloud center located at the origin of the 3D coordinate system, and the 3D point cloud is rotated so that the point cloud normal vector coincides with the z-axis. The point cloud is described by the x and y coordinates to obtain a 2D planar point cloud.

[0024] Optionally, the step of performing three-dimensional point cloud preprocessing based on the features of the reflective pattern in the three-dimensional point cloud information to obtain effective point cloud information includes:

[0025] The 3D point cloud is filtered based on point cloud timestamp, point confidence, point distance, and point intensity. The filtering criteria include: the single-frame point cloud timestamp must be within a reasonable time range; the point confidence must be within a reasonable range; the point distance must be within a reasonable range; and the point intensity must be greater than the minimum intensity threshold.

[0026] Distortion correction is performed on the filtered 3D point cloud based on local location information; where local location information is the leg odometry coordinates of the lost robot, and the point cloud with the smallest difference in timestamps is matched for coordinate correction.

[0027] The voxel mesh method was used to downsample the distortion-free 3D point cloud; the voxel size was set according to the size of the square reflective strip.

[0028] The Eulerian clustering method was used to cluster the downsampled 3D point cloud; the clustering tolerance was selected based on the length of the square reflective strips and the spacing between the square reflective strips, so that the reflective rectangular array as a whole belongs to the same class;

[0029] The RANSAC algorithm is used to perform point cloud plane segmentation on the clustered 3D point cloud to obtain an effective 3D point cloud; the plane thickness is determined based on the flatness of the wall surface in the environment around the robot that lost its localization.

[0030] Optionally, the step of identifying the two-dimensional planar point cloud and determining whether it contains a valid reflective pattern based on reflective pattern features includes:

[0031] Obtain the smallest rectangle that can enclose the point cloud from the two-dimensional planar point cloud, and determine whether the size of the reflective rectangle array matches the size of the smallest rectangle;

[0032] The minimum rectangle is divided equally according to the maximum number of square reflective strips contained in the reflective rectangle array. The center position of the point cloud in each sub-rectangle is calculated, and it is determined whether the spacing between the square reflective strips matches the distance between the center positions of adjacent sub-rectangles.

[0033] If the reflected rectangular array is found to not meet any of the matching rules, it is determined that the lidar field of view does not contain a complete reflected pattern.

[0034] Optionally, obtaining the associated pre-calibrated global map location based on the code corresponding to the effective reflective pattern includes:

[0035] The two-dimensional point cloud of the reflective rectangular array is divided according to the size and spacing of the square reflective strips to obtain multiple sub-regions;

[0036] If the number of points in the point cloud of a sub-region meets the preset threshold, it is determined that there is a reflective stripe in the sub-region and it corresponds to binary 1; otherwise, it is determined that there is no reflective stripe in the sub-region and it corresponds to binary 0.

[0037] The two-dimensional point cloud of the reflective rectangular array is binary encoded by combining the encoding rules of the reflective rectangular array;

[0038] The global map position of the center of the two-dimensional point cloud of the reflective rectangular array is obtained based on the pre-calibrated position information associated with the binary code.

[0039] Secondly, the present invention provides a degradation scene positioning device based on reflective patterns, comprising:

[0040] Reflective pattern encoding module: used to set multiple reflective patterns in the feature degradation scene according to a preset interval and to uniquely encode each reflective pattern; wherein, each reflective pattern contains multiple reflective areas arranged differently, and the reflective pattern is uniquely encoded according to the relative position distribution of the reflective areas it contains;

[0041] Location pre-calibration module: used to pre-calibrate each reflective pattern with a unique code and obtain the global map location of each reflective pattern;

[0042] Reference object localization module: used to obtain the global map location of effective reflective patterns in the environment surrounding the robot that has lost its localization;

[0043] Positioning recovery module: used to obtain the global map position of the lost robot based on the global map position of the effective reflective pattern and the relative position of the lost robot to the effective reflective pattern.

[0044] Thirdly, the present invention provides a computer storage medium having a computer program stored thereon, which, when executed by a processor, implements the degradation scene localization method based on reflective patterns as described in any of the first aspects.

[0045] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: by setting reflective patterns in feature degradation scenarios and encoding the reflective patterns by distinguishing the arrangement of reflective areas within the reflective patterns, each reflective pattern has a unique identity and is associated with a corresponding global location. Using these uniquely encoded reflective patterns as a reference for the positioning lost robot in feature degradation scenarios, a positioning recovery method with wide adaptability, simple implementation and high reliability can be achieved without relying on additional hardware. Attached Figure Description

[0046] Figure 1 The diagram shown is a flowchart of a degradation scene localization method based on reflective patterns in one embodiment of the present invention.

[0047] Figure 2 The diagram shown is an application scenario illustration of the reflective pattern positioning method in one embodiment of the present invention.

[0048] Figure 3 The diagram shown is an example of a square reflective strip being pasted in a reflective rectangular array according to one embodiment of the present invention;

[0049] Figure 4 The diagram shown is a flowchart of a method for obtaining the global position of a reflective pattern in one embodiment of the present invention;

[0050] Figure 5 The image shown is a schematic diagram of a two-dimensional planar point cloud of a reflective pattern in one embodiment of the present invention. Detailed Implementation

[0051] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0052] Example 1

[0053] like Figure 1 As shown, this embodiment provides a method for locating degraded scenes based on reflective patterns, specifically including the following steps:

[0054] S1. Before the robot starts its automatic operation phase, reflective patterns are affixed at regular intervals in the feature degradation scene, and the reflective patterns are pre-calibrated to obtain their accurate global map positions.

[0055] The reflective patterns used are applied in the following scenarios: Figure 2In the aforementioned degradation scenario, the reflective pattern should be affixed to the side wall in the direction the robot is moving. The reflective pattern is a reflective rectangular array composed of square reflective blocks, with square reflective strips affixed to the three square reflective blocks at the top corners, and no square reflective strips affixed to the remaining square reflective block at the top corner, used to determine the overall size and orientation of the reflective rectangular array.

[0056] The reflective rectangular array, excluding the top corners, consists of square reflective blocks with or without reflective strips. The presence or absence of reflective strips represents binary 1 and 0 respectively, and the square reflective strips at different positions correspond to different bits of the complete binary code in a specified order. Furthermore, regardless of whether reflective strips are affixed to the square reflective blocks, adjacent square reflective blocks should be spaced a certain distance apart.

[0057] The reflective rectangular array should be rotated 45° at its center so that a set of diagonals where square reflective strips must be affixed are in a horizontal direction, ensuring that the entire array is presented in the lidar's field of view to the maximum extent when the robot identifies the three vertices.

[0058] Taking a 3×3 reflective rectangular array as an example, the reflective rectangular array after being posted should be as follows: Figure 3 As shown, reflective strips must be affixed to gray areas, but not to white areas. The areas marked with an "X" pattern are used for binary encoding.

[0059] The method for pre-calibrating the reflective pattern is as follows: when the robot is positioned normally, control its movement until the reflective rectangular array is completely contained in the lidar point cloud. At this time, combine the global map position and the relative position of the reflective strip point cloud to calculate the global map position of the reflective strip, record it and associate it with the binary code corresponding to the reflective pattern.

[0060] S2. When the robot loses its localization during operation, it acquires the 3D point cloud information of the robot's surrounding environment. If the information contains a valid reflective pattern, it obtains the pre-calibrated global map position of the reflective pattern based on the pattern. Figure 4 As shown.

[0061] S2.1. When the robot loses its positioning during operation, the reflective pattern positioning algorithm starts to work, acquiring and processing three-dimensional point cloud data through a high-precision lidar installed on the robot.

[0062] S2.1.1 Perform three-dimensional point cloud preprocessing based on reflective properties and pattern features, including point cloud screening, point cloud distortion correction, point cloud downsampling, point cloud clustering, and point cloud plane segmentation.

[0063] In this embodiment, the point cloud screening method is based on the 3D point cloud used by the LiDAR. The screening criteria include that the single-frame point cloud timestamp must be within a reasonable time range, the point confidence level must be within a reasonable range, the point distance must be within a reasonable range, and the point intensity must be greater than the minimum intensity threshold. The point cloud distortion correction method corrects the distortion of the LiDAR point cloud caused by robot movement by using local position information. The position information used is the robot's leg odometry coordinates. The LiDAR point cloud with the smallest timestamp difference is matched for coordinate transformation. The point cloud downsampling method adopts the voxel mesh method. The parameter selection needs to consider the size of the square reflective strips. The point cloud clustering method adopts the Eulerian clustering method. The clustering tolerance parameter is selected according to the length of the square reflective strips and the spacing of the reflective strips in the direction, ensuring that the entire reflective rectangular array belongs to the same class. The point cloud plane segmentation method uses the RANSAC algorithm for plane segmentation. The plane thickness is determined according to the flatness of the wall surface on site and is not greater than 0.05m.

[0064] S2.1.2 Obtain the point cloud center and normal vector from the point cloud after planar segmentation, determine the application direction of the reflective pattern, and reduce the 3D planar point cloud to a 2D point cloud through rotation and translation. The 2D point cloud effect is as follows. Figure 5 As shown.

[0065] The point cloud dimensionality reduction method determines whether the reflective pattern is affixed to the left or right wall of the robot by the direction of the normal, and then determines the rotation direction of the point cloud to ensure that the pattern does not become a mirror image.

[0066] Point cloud dimensionality reduction methods use rotation and translation to center the point cloud at the origin, aligning the point cloud plane normal vector with the z-axis. In this case, the point cloud can be described using only x and y coordinates, facilitating subsequent pattern feature identification. The transformation process can be represented by the following formula:

[0067] ,

[0068] in, This is the two-dimensional point cloud obtained after coordinate transformation. This is the 3D point cloud before coordinate transformation. It is a 3×3 identity matrix. Let the coordinates be the centroid coordinates of the point cloud plane. The rotation matrix is ​​calculated as follows:

[0069] ,

[0070] ,

[0071] ,

[0072] ,

[0073] in, It is a plane normal vector. The z-axis unit vector. For the required rotation angle, The axis of rotation required to rotate the plane normal vector to coincide with the z-axis unit vector.

[0074] S2.2. Identify the point cloud information and determine whether it contains a valid reflective pattern based on the reflective pattern features.

[0075] Based on the two-dimensional planar point cloud, obtain the smallest rectangle that can enclose the point cloud, and determine whether the length and width of the rectangle conform to the size of the reflective rectangle array. Divide the smallest rectangle that encloses the planar point cloud into equal parts according to the number of the largest square reflective strips contained in the reflective rectangle array, calculate the center position of the point cloud in each sub-rectangle, and determine whether the distance between the center positions of adjacent sub-rectangles conforms to the size of the square reflective strips and their spacing.

[0076] If local features of the reflective pattern are identified but not all features are not found, such as insufficient length and width of the enclosing rectangle, insufficient distance between the centers of adjacent sub-rectangles, or insufficient number of corner points, it is considered that the lidar field of view does not completely contain the reflective rectangle array, and the robot's pose is adjusted so that the lidar field of view is aligned with the center of the point cloud.

[0077] S2.3. If a valid reflective pattern exists, decode it to obtain its associated pre-calibrated global position.

[0078] S2.3.1 Divide the two-dimensional point cloud of the reflective rectangular array according to the relevant size and spacing of the square reflective strips. If the number of point clouds in a sub-region meets the threshold, it is considered that the sub-region has a reflective strip. Calculate the binary code by combining the correspondence between the reflective pattern position and the number of binary bits in step S1.

[0079] S2.3.2 For binary encoding queries, the corresponding location information is the global map location of the center of the reflective rectangular array point cloud.

[0080] S3. Based on the global map position of the reflective pattern and the relative position of the robot and the reflective pattern, and by matching the surrounding point cloud, obtain the best global map position of the robot and restore the robot's positioning status.

[0081] The global map position of the reflective pattern is the position associated with the identified binary code during pre-calibration, and the local position of the reflective pattern relative to the robot is the position of the centroid of the reflective pattern's point cloud relative to the robot. The robot SLAM localization program combines the global position of the reflective pattern and the relative position of the robot and the reflective pattern to perform ICP point cloud registration. If a threshold is reached, localization is considered successful, and localization returns to normal.

[0082] Example 2

[0083] This embodiment provides a degradation scene positioning device based on reflective patterns, including:

[0084] Reflective pattern encoding module: used to set multiple reflective patterns in the feature degradation scene according to a preset interval and to uniquely encode each reflective pattern; wherein, each reflective pattern contains multiple reflective areas arranged differently, and the reflective pattern is uniquely encoded according to the relative position distribution of the reflective areas it contains;

[0085] Location pre-calibration module: used to pre-calibrate each reflective pattern with a unique code and obtain the global map location of each reflective pattern;

[0086] Reference object localization module: used to obtain the global map location of effective reflective patterns in the environment surrounding the robot that has lost its localization;

[0087] Positioning recovery module: used to obtain the global map position of the lost robot based on the global map position of the effective reflective pattern and the relative position of the lost robot to the effective reflective pattern.

[0088] The device provided in this embodiment can execute the degradation scene localization method based on reflective patterns provided in any step of Embodiment 1, and has the corresponding functional modules and beneficial effects of the execution method.

[0089] Example 3

[0090] This embodiment provides a computer storage medium storing a computer program. When the computer program is executed by a processor, it implements the degradation scene localization method based on reflective patterns as provided in any step of Embodiment 1.

[0091] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0092] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0093] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0094] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0095] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims. All of these forms are within the protection scope of the present invention.

Claims

1. A method for locating degraded scenes based on reflective patterns, characterized in that, include: Multiple reflective patterns are set within the feature degradation scene according to a preset interval, and each reflective pattern is uniquely encoded; each reflective pattern contains multiple reflective areas arranged differently, and the reflective pattern is uniquely encoded according to the relative position distribution of the reflective areas it contains. Each reflective pattern with a unique code is pre-calibrated to obtain the global map position of each reflective pattern; Obtain the global map location of effective reflective patterns in the environment surrounding the lost robot; The global map position of the lost robot is obtained based on the global map position of the effective reflective pattern and the relative position of the lost robot to the effective reflective pattern.

2. The degradation scene localization method based on reflective patterns according to claim 1, characterized in that, The reflective pattern is a reflective rectangular array, in which multiple square reflective strips are spaced apart; The square reflective strips are distributed at the three vertices of the reflective rectangular array and in other areas. The square reflective strips at the vertices are used to determine the size and orientation of the reflective rectangular array. The square reflective strips distributed on a set of diagonals in the reflective rectangular array are all in the horizontal direction.

3. The degradation scene localization method based on reflective patterns according to claim 2, characterized in that, The reflective rectangular array, except for the top corner, is characterized by the presence or absence of square reflective strips corresponding to binary 1 and 0. The unique code of the reflective rectangular array is composed of each binary 1 and 0 placed in the corresponding position according to a preset sorting rule.

4. The degradation scene localization method based on reflective patterns according to claim 3, characterized in that, The global map location of the effective reflective patterns in the environment surrounding the lost robot includes: The three-dimensional point cloud information of the surrounding environment is obtained by installing a lidar on the robot that has lost its location, and the three-dimensional point cloud information is reduced in dimension to obtain a two-dimensional planar point cloud. The two-dimensional planar point cloud is identified, and the presence of a valid reflective pattern is determined by combining the reflective pattern characteristics. If a valid reflective pattern exists, its associated pre-calibrated global map location is obtained based on its corresponding code.

5. The degradation scene localization method based on reflective patterns according to claim 4, characterized in that, The process of acquiring three-dimensional point cloud information of the surrounding environment and performing dimensionality reduction processing on the three-dimensional point cloud information to obtain a two-dimensional planar point cloud includes: Based on the characteristics of reflective patterns in the 3D point cloud information, 3D point cloud preprocessing is performed to obtain effective point cloud information; Obtain the point cloud center and normal vector from the valid point cloud information, and determine the pasting direction of the reflective pattern; Based on the direction of the reflective pattern, the 3D point cloud is translated to the point cloud center located at the origin of the 3D coordinate system, and the 3D point cloud is rotated so that the point cloud normal vector coincides with the z-axis. The point cloud is described by the x and y coordinates to obtain a 2D planar point cloud.

6. The degradation scene localization method based on reflective patterns according to claim 5, characterized in that, The step of performing 3D point cloud preprocessing based on the characteristics of reflective patterns in 3D point cloud information to obtain effective point cloud information includes: The 3D point cloud is filtered based on point cloud timestamp, point confidence, point distance, and point intensity. The filtering criteria include: the single-frame point cloud timestamp must be within a reasonable time range; the point confidence must be within a reasonable range; the point distance must be within a reasonable range; and the point intensity must be greater than the minimum intensity threshold. Distortion correction is performed on the filtered 3D point cloud based on local location information; where local location information is the leg odometry coordinates of the lost robot, and the point cloud with the smallest difference in timestamps is matched for coordinate correction. The voxel mesh method was used to downsample the distortion-free 3D point cloud; the voxel size was set according to the size of the square reflective strip. The Eulerian clustering method was used to cluster the downsampled 3D point cloud; the clustering tolerance was selected based on the length of the square reflective strips and the spacing between the square reflective strips, so that the reflective rectangular array as a whole belongs to the same class; The RANSAC algorithm is used to perform point cloud plane segmentation on the clustered 3D point cloud to obtain an effective 3D point cloud; the plane thickness is determined based on the flatness of the wall surface in the environment around the robot that lost its localization.

7. The degradation scene localization method based on reflective patterns according to claim 4, characterized in that, The process of identifying two-dimensional planar point clouds and determining whether they contain valid reflective patterns based on reflective pattern features includes: Obtain the smallest rectangle that can enclose the point cloud from the two-dimensional planar point cloud, and determine whether the size of the reflective rectangle array matches the size of the smallest rectangle; The minimum rectangle is divided equally according to the maximum number of square reflective strips contained in the reflective rectangle array. The center position of the point cloud in each sub-rectangle is calculated, and it is determined whether the spacing between the square reflective strips matches the distance between the center positions of adjacent sub-rectangles. If the reflected rectangular array is found to not meet any of the matching rules, it is determined that the lidar field of view does not contain a complete reflected pattern.

8. The degradation scene localization method based on reflective patterns according to claim 4, characterized in that, The step of obtaining the associated pre-calibrated global map location based on the code corresponding to the effective reflective pattern includes: The two-dimensional point cloud of the reflective rectangular array is divided according to the size and spacing of the square reflective strips to obtain multiple sub-regions; If the number of points in the point cloud of a sub-region meets the preset threshold, it is determined that there is a reflective stripe in the sub-region and it corresponds to binary 1; otherwise, it is determined that there is no reflective stripe in the sub-region and it corresponds to binary 0. The two-dimensional point cloud of the reflective rectangular array is binary encoded by combining the encoding rules of the reflective rectangular array; The global map position of the center of the two-dimensional point cloud of the reflective rectangular array is obtained based on the pre-calibrated position information associated with the binary code.

9. A degraded scene positioning device based on reflective patterns, characterized in that, include: Reflective pattern encoding module: used to set multiple reflective patterns in the feature degradation scene according to a preset interval and to uniquely encode each reflective pattern; wherein, each reflective pattern contains multiple reflective areas arranged differently, and the reflective pattern is uniquely encoded according to the relative position distribution of the reflective areas it contains; Location pre-calibration module: used to pre-calibrate each reflective pattern with a unique code and obtain the global map location of each reflective pattern; Reference object localization module: used to obtain the global map location of effective reflective patterns in the environment surrounding the robot that has lost its localization; Positioning recovery module: used to obtain the global map position of the lost robot based on the global map position of the effective reflective pattern and the relative position of the lost robot to the effective reflective pattern.

10. A computer storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the degradation scene localization method based on reflective patterns as described in any one of claims 1-8.

Images