Reflective object recognition method and device

Abstract

The present invention discloses a reflective object recognition method and device. The method determines the seed area according to the intensity value in the point cloud data, and can accurately obtain the area with high reflectivity and accurately locate the reflective point of the reflective object. The center of the cluster can also be accurately determined, eliminating the problem of unclear clustering results caused by inaccurate determination of the center point in general clustering algorithms. The various reflective points are diffused by the diffusion clustering algorithm to obtain a clear candidate area range. The diffusion clustering algorithm does not need to classify or cluster all the points in the regional point cloud data, but diffuses outward with the seed area, which can eliminate the problem of unclear candidate areas caused by clustering all points in other clustering algorithms. The shape of each candidate area is determined by the curvature of each candidate area, so that the shape of the candidate area can be quickly identified, reducing the burden on the hardware equipment while improving accuracy.

Description

Technical Field

[0001] The present invention relates to the field of reflective recognition, and in particular to a reflective object recognition method and device. Background Art

[0002] Automated Guided Vehicles (AGVs) are essential transportation tools in automated warehousing and logistics production. AGV navigation and positioning is a crucial technology in automated warehousing and logistics. Common industrial AGV navigation and positioning methods include magnetic navigation, laser navigation, and QR code navigation. Laser navigation, based on reflective columns and reflective panels, is widely used in AGVs due to its high positioning accuracy, low maintenance, easy installation, and relatively low cost.

[0003] There are certain differences between reflectors and reflective columns in terms of recognition distance and recognizable angle range. After comprehensively considering factors such as ease of use and positioning recognition stability, reflectors and reflective columns will be used in combination to improve AGV navigation and positioning accuracy.

[0004] However, in the existing technology, it is impossible to identify the shape of the reflective object, and it is also impossible to distinguish between the reflective plate and the reflective column, resulting in poor accuracy in AGV navigation and positioning. Summary of the Invention

[0005] The present invention provides a reflective object recognition method and device to solve the problem that the shape of reflective objects cannot be recognized, resulting in low accuracy in AGV navigation and positioning.

[0006] In a first aspect, the present application provides a method for identifying reflective objects, comprising:

[0007] Obtaining regional point cloud data, and determining a number of regions in the regional point cloud data whose intensity values ​​are greater than a preset value as a number of seed regions;

[0008] According to the diffusion clustering algorithm, the regional point cloud data is clustered with several seed regions as the center to obtain several candidate regions;

[0009] The curvature of each candidate region is calculated, the shape of each candidate region is determined according to the curvature of each candidate region, and the shape of the reflective object corresponding to each candidate region is determined according to the shape of each candidate region.

[0010] In this way, by determining the seed area based on the intensity value in the point cloud data, the area with high reflectivity can be accurately obtained. The reflective points of the reflective object can be accurately located according to the reflective characteristics of the reflective plate and the reflective column. And when the reflective points are accurately obtained as described above, the center of the cluster can be accurately determined, eliminating the problem of unclear clustering results caused by inaccurate determination of the center point in the general clustering algorithm. By diffusing each reflective point through the diffusion clustering algorithm, a clear candidate area range can be obtained. The diffusion clustering algorithm is different from the general clustering algorithm. The diffusion clustering algorithm does not need to classify or cluster all the points in the regional point cloud data, but diffuses outward with the seed area. This can eliminate the problem of unclear candidate areas caused by clustering all points in other clustering algorithms. Furthermore, the shape of each candidate area is determined by the curvature of each candidate area, which makes it simple and convenient to identify the shape of the candidate area, reducing the burden on the hardware equipment while improving accuracy.

[0011] Furthermore, after determining the shape of the reflective object corresponding to each candidate area according to the shape of each candidate area, the method further includes:

[0012] When there is a candidate area with a shape of a reflective column, the candidate area is fitted according to the least squares circle fitting method to obtain the center position of the reflective column;

[0013] When there is a candidate area whose shape is a reflector, the candidate area is fitted according to the least squares straight line fitting method to obtain the midpoint position and normal angle of the reflector.

[0014] In this way, the corresponding minimum multiplication method is used based on the shape of the candidate area to determine the center of the circle and the center of the reflector. This can simplify the process of identifying and locating reflectors and reflectors in the point cloud, and enhance the adaptability and real-time performance of the reflector and reflector identification method.

[0015] Furthermore, the obtaining of regional point cloud data is specifically as follows:

[0016] Obtain radar point cloud data according to the radar detection device, and eliminate point cloud data exceeding the recognition distance in the radar point cloud data according to the recognition distance of the radar detection device;

[0017] According to the real-time motion data of AGV, the radar point cloud data is calibrated to obtain regional point cloud data.

[0018] Eliminating point cloud data that exceeds the recognition distance from the radar point cloud data can eliminate point cloud data with low accuracy, reduce the interference of noise on subsequent recognition, and reduce the amount of point cloud data that needs to be recognized and processed subsequently, thereby increasing the efficiency of data processing and the real-time performance of subsequent image recognition.

[0019] In this way, the radar point cloud data is corrected according to the real-time motion data of the AGV such as the wheel speed meter, which can eliminate the errors in the collected data caused by the movement of the AGV.

[0020] Furthermore, the several regions in the regional point cloud data whose intensity values ​​are greater than a preset value are determined as several seed regions, specifically:

[0021] The first point cloud data having a reflectivity greater than a calibration value of the radar detection device in the regional point cloud data is determined, and the connected first point cloud data is determined as a seed area.

[0022] Determining the seed area in this way can ensure that the reflectivity of the seed area is greater than that of the non-seed area, thereby accurately locating the reflective object and preventing possible candidate areas from being missed during subsequent diffusion clustering based on the seed area.

[0023] Furthermore, the diffusion clustering algorithm is used to cluster the regional point cloud data with several seed regions as the center to obtain several candidate regions, specifically:

[0024] Determine each seed area as a number of central areas in the cluster, and continuously determine whether the difference between the intensity value of the point cloud data connected to the periphery of each central area boundary point and the intensity value of the central area boundary point is less than a preset intensity difference, until the difference between the intensity value of the point cloud data of the non-central area without the central area boundary point connection and the intensity value of the central area boundary point is less than the preset intensity difference, then determine each central area as the corresponding candidate area;

[0025] Among them, in each judgment, if the difference between the intensity value of the point cloud data of the non-central area connected to the boundary point of the central area and the intensity value of the boundary point of the central area is less than the preset intensity difference, the point cloud data of the non-central area is added to the central area.

[0026] In this way, the seed area is diffused by the diffusion clustering algorithm, and the outline and size of the candidate area can be accurately obtained. The central area is expanded each time by judging the difference between the boundary points and the internal points, which can prevent points with low reflectivity from being added to the candidate area.

[0027] Furthermore, the curvature of each candidate region is calculated, and the shape of each candidate region is determined according to the curvature of each candidate region, specifically:

[0028] According to the long side of each candidate area, all point cloud data in each candidate area are split into upper and lower parts;

[0029] According to the position information of the point cloud data in each candidate area, the slopes of the upper part and the lower part of each candidate area are calculated respectively, and the slope difference between the upper part and the lower part of each candidate area is calculated;

[0030] Determine whether the slope difference between the upper and lower parts of each candidate area is greater than a preset difference;

[0031] If the difference between the slopes of the upper and lower parts of the candidate area is less than or equal to the preset difference, determining that the shape of the candidate area is a reflector;

[0032] If the difference in slope between the upper part and the lower part of the candidate area is greater than a preset difference, the shape of the candidate area is determined to be a reflective column.

[0033] In this way, the shape of the candidate area can be identified simply and quickly by calculating the slope of each candidate area, reducing the computing power consumption of the device in identifying reflective objects, increasing the recognition speed of reflective objects while ensuring accuracy, and achieving the recognition of reflective boards and reflective columns, thereby combining the respective advantages of reflective columns and reflective boards to ensure the reliability of avg navigation positioning.

[0034] Furthermore, according to the diffusion clustering algorithm, the regional point cloud data is clustered with several seed regions as the center to obtain several candidate regions, which then includes:

[0035] Determine whether the size of each candidate area meets the preset size range;

[0036] If the size of a candidate region does not meet the preset size range, the candidate region is determined to be an invalid region;

[0037] If there is a candidate region whose size meets the preset size range, the candidate region is determined to be the final candidate region.

[0038] By setting a preset size range, we can eliminate the possibility of mistakenly identifying strong reflective objects such as glass and stainless steel as candidate areas. This can eliminate the interference of debris and more accurately locate the reflective objects that need to be located.

[0039] In a second aspect, the present application provides a reflective object recognition device, comprising: a seed region module, a candidate region module, and a shape determination module;

[0040] The seed region module is used to obtain regional point cloud data and determine a number of regions in the regional point cloud data whose intensity values ​​are greater than a preset value as a number of seed regions;

[0041] The candidate region module is used to cluster the regional point cloud data with several seed regions as the center according to the diffusion clustering algorithm to obtain several candidate regions;

[0042] The shape determination module is used to calculate the curvature of each candidate area, determine the shape of each candidate area according to the curvature of each candidate area, and determine the shape of the reflective object corresponding to each candidate area according to the shape of each candidate area.

[0043] Furthermore, the reflective object recognition device further includes: a circle center determination module and a midpoint determination module;

[0044] The circle center determination module is used to fit the candidate area according to the least squares circle fitting method when there is a candidate area with a shape of a reflective column, so as to obtain the center position of the reflective column;

[0045] The midpoint determination module is used to fit the candidate area according to the least squares straight line fitting method when the shape of the candidate area is a reflector, so as to obtain the midpoint position and normal angle of the reflector.

[0046] Furthermore, the seed region module includes a radar point cloud data unit and a regional point cloud data unit;

[0047] The radar point cloud data unit is used to obtain radar point cloud data according to the radar detection device, and eliminate point cloud data that exceeds the recognition distance in the radar point cloud data according to the recognition distance of the radar detection device;

[0048] The regional point cloud data unit is used to calibrate the radar point cloud data according to the real-time motion data of the AGV to obtain regional point cloud data.

[0049] Furthermore, the seed region module includes a seed region determination unit;

[0050] The seed region determination unit is used to determine the first point cloud data in the regional point cloud data whose reflectivity is greater than the calibration value of the radar detection device, and determine the connected first point cloud data as a seed region.

[0051] Furthermore, the candidate region module includes: a continuous judgment unit;

[0052] The continuous judgment unit is used to determine each seed area as a number of central areas in the cluster, and continuously judge whether the intensity difference between the point cloud data of the peripheral connection of the boundary points of each central area and the intensity of the boundary points of the central area is less than a preset intensity difference, until the intensity difference between the point cloud data of the non-central area without the boundary points of the central area and the intensity of the boundary points of the central area is less than the preset intensity difference, and then determine each central area as the corresponding candidate area;

[0053] In each judgment, if the intensity difference between the point cloud data of the non-central area connected to the boundary points of the central area and the intensity of the boundary points of the central area is less than the preset intensity difference, the point cloud data of the non-central area is added to the central area.

[0054] Furthermore, the shape determination module includes: a splitting unit, a slope unit, a first judgment unit and a shape recognition unit;

[0055] The splitting unit is used to split all point cloud data in each candidate area into an upper part and a lower part according to the long side of each candidate area;

[0056] The slope unit is used to calculate the slope of the upper part and the lower part of each candidate area according to the position information of the point cloud data in each candidate area, and calculate the slope difference between the upper part and the lower part of each candidate area;

[0057] The first judging unit is used to judge whether the difference between the slopes of the upper part and the lower part of each candidate area is greater than a preset difference;

[0058] The shape recognition unit is configured to determine that the shape of the candidate area is a reflector when the first judgment unit determines that the difference in slope between the upper part and the lower part of the candidate area is less than or equal to a preset difference;

[0059] The shape recognition unit is further configured to determine that the shape of the candidate area is a reflective column when the first judgment unit determines that the slope difference between the upper part and the lower part of the candidate area is greater than a preset difference.

[0060] Furthermore, the reflective object recognition device further includes: a second judgment module and a candidate area determination module;

[0061] The second judgment module is used to judge whether the size of each candidate area meets a preset size range;

[0062] The candidate region determination module is configured to determine that a candidate region is an invalid region when the second judgment module determines that a size of the candidate region does not meet a preset size range;

[0063] The candidate region determination module is further configured to determine that, when the second judgment module determines that there is a candidate region whose size satisfies a preset size range, the candidate region is the final candidate region.

[0064] In this way, by determining the seed area based on the intensity value in the point cloud data, the area with high reflectivity can be accurately obtained. The reflective points of the reflective object can be accurately located according to the reflective characteristics of the reflective plate and the reflective column. And when the reflective points are accurately obtained as described above, the center of the cluster can be accurately determined, eliminating the problem of unclear clustering results caused by inaccurate determination of the center point in the general clustering algorithm. By diffusing each reflective point through the diffusion clustering algorithm, a clear candidate area range can be obtained. The diffusion clustering algorithm is different from the general clustering algorithm. The diffusion clustering algorithm does not need to classify or cluster all the points in the regional point cloud data, but diffuses outward with the seed area. This can eliminate the problem of unclear candidate areas caused by clustering all points in other clustering algorithms. Furthermore, the shape of each candidate area is determined by the curvature of each candidate area, which makes it simple and convenient to identify the shape of the candidate area, reducing the burden on the hardware equipment while improving accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 : A schematic flow chart of an embodiment of a reflective object recognition method provided by the present invention;

[0066] Figure 2 : A module structure diagram of an embodiment of a reflective object recognition device provided by the present invention. DETAILED DESCRIPTION

[0067] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0068] In the description of this specification, the reference terms "one embodiment," "some embodiments," "example," "specific example," or "some examples" mean that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. Moreover, the specific features, structures, materials, or characteristics described may be combined in any appropriate manner in any one or more embodiments or examples. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification, as well as features of different embodiments or examples, unless they are mutually inconsistent.

[0069] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. Throughout the description of this application, "plurality" means two or more, unless otherwise specifically defined.

[0070] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present application belong.

[0071] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0072] Example 1

[0073] Please refer to Figure 1 , a reflective object recognition method provided by an embodiment of the present invention includes steps S1 to S3, each of which is specifically as follows:

[0074] Step S1: obtaining regional point cloud data, and determining a number of regions in the regional point cloud data whose intensity values ​​are greater than a preset value as a number of seed regions;

[0075] In this embodiment, the obtaining of regional point cloud data is specifically as follows:

[0076] Obtain radar point cloud data according to the radar detection device, and eliminate point cloud data exceeding the recognition distance in the radar point cloud data according to the recognition distance of the radar detection device;

[0077] In this embodiment, the recognition distance of the radar detection device can be determined by the detection performance of the radar detection device, or the recognition distance of the radar detection device can be determined according to the distance that the AGV needs to detect and recognize during movement.

[0078] Furthermore, in this embodiment, the radar detection device may be mounted on the AGV or may not be mounted on the AGV and only controls the movement of the AGV.

[0079] According to the real-time motion data of AGV, the radar point cloud data is calibrated to obtain regional point cloud data.

[0080] Eliminating point cloud data that exceeds the recognition distance from the radar point cloud data can eliminate point cloud data with low accuracy, reduce the interference of noise on subsequent recognition, and reduce the amount of point cloud data that needs to be recognized and processed subsequently, thereby increasing the efficiency of data processing and the real-time performance of subsequent image recognition.

[0081] In this way, the radar point cloud data is corrected according to the real-time motion data of the AGV such as the wheel speed meter, which can eliminate the errors in the collected data caused by the movement of the AGV.

[0082] In this embodiment, the several regions in the regional point cloud data whose intensity values ​​are greater than a preset value are determined as several seed regions, specifically:

[0083] The first point cloud data having a reflectivity greater than a calibration value of the radar detection device in the regional point cloud data is determined, and the connected first point cloud data is determined as a seed area.

[0084] Determining the seed area in this way can ensure that the reflectivity of the seed area is greater than that of the non-seed area, thereby accurately locating the reflective object and preventing possible candidate areas from being missed during subsequent diffusion clustering based on the seed area.

[0085] Step S2: clustering the regional point cloud data with several seed regions as the center according to the diffusion clustering algorithm to obtain several candidate regions;

[0086] In this embodiment, the diffusion clustering algorithm is used to cluster the regional point cloud data with several seed regions as the center to obtain several candidate regions, specifically:

[0087] Determine each seed area as a number of central areas in the cluster, and continuously determine whether the difference between the intensity value of the point cloud data connected to the periphery of each central area boundary point and the intensity value of the central area boundary point is less than a preset intensity difference, until the difference between the intensity value of the point cloud data of the non-central area without the central area boundary point connection and the intensity value of the central area boundary point is less than the preset intensity difference, then determine each central area as the corresponding candidate area;

[0088] Among them, in each judgment, if the difference between the intensity value of the point cloud data of the non-central area connected to the boundary point of the central area and the intensity value of the boundary point of the central area is less than the preset intensity difference, the point cloud data of the non-central area is added to the central area.

[0089] In another embodiment, other clustering algorithms or classification algorithms can be used to identify the central area. However, this application uses a diffusion clustering algorithm to diffuse the seed area, which can accurately obtain the outline and size of the candidate area. Each time, the central area is expanded by determining the difference between the boundary points and the internal points, which can prevent points with low reflectivity from being added to the candidate area.

[0090] In this embodiment, the regional point cloud data is clustered with several seed regions as the center according to the diffusion clustering algorithm to obtain several candidate regions, and then the following steps are included:

[0091] Determine whether the size of each candidate area meets the preset size range;

[0092] If the size of a candidate region does not meet the preset size range, the candidate region is determined to be an invalid region;

[0093] If there is a candidate region whose size meets the preset size range, the candidate region is determined to be the final candidate region.

[0094] By setting a preset size range, we can eliminate the possibility of mistakenly identifying strong reflective objects such as glass and stainless steel as candidate areas. This can eliminate the interference of debris and more accurately locate the reflective objects that need to be located.

[0095] In one embodiment, the preset size range can be determined based on the debris that needs to be removed from the surrounding environment, or can be determined based on the size of the reflective column and the reflective plate.

[0096] Step S3: Calculate the curvature of each candidate region, determine the shape of each candidate region according to the curvature of each candidate region, and determine the shape of the reflective object corresponding to each candidate region according to the shape of each candidate region.

[0097] In this embodiment, the curvature of each candidate region is calculated, and the shape of each candidate region is determined according to the curvature of each candidate region, specifically:

[0098] According to the long side of each candidate area, all point cloud data in each candidate area are split into upper and lower parts;

[0099] According to the position information of the point cloud data in each candidate area, the slopes of the upper part and the lower part of each candidate area are calculated respectively, and the slope difference between the upper part and the lower part of each candidate area is calculated;

[0100] Determine whether the slope difference between the upper and lower parts of each candidate area is greater than a preset difference;

[0101] If the difference between the slopes of the upper and lower parts of the candidate area is less than or equal to the preset difference, determining that the shape of the candidate area is a reflector;

[0102] If the difference in slope between the upper part and the lower part of the candidate area is greater than a preset difference, the shape of the candidate area is determined to be a reflective column.

[0103] In this way, the shape of the candidate area can be identified simply and quickly by calculating the slope of each candidate area, reducing the computing power consumption of the device in identifying reflective objects, increasing the recognition speed of reflective objects while ensuring accuracy, and achieving the recognition of reflective boards and reflective columns, thereby combining the respective advantages of reflective columns and reflective boards to ensure the reliability of avg navigation positioning.

[0104] In a specific embodiment, the preset difference may be 0.

[0105] Furthermore, after determining the shape of the reflective object corresponding to each candidate area according to the shape of each candidate area, the process includes steps S4 to S5:

[0106] Step S4: When there is a candidate area with a shape of a reflective column, fitting the candidate area according to the least squares circle fitting method to obtain the center position of the reflective column;

[0107] Step S5: When there is a candidate area whose shape is a reflector, fitting is performed on the candidate area according to the least squares straight line fitting method to obtain the midpoint position and normal angle of the reflector.

[0108] In this way, the corresponding minimum multiplication method is used based on the shape of the candidate area to determine the center of the circle and the center of the reflector. This can simplify the process of identifying and locating reflectors and reflectors in the point cloud, and enhance the adaptability and real-time performance of the reflector and reflector identification method.

[0109] Example 2

[0110] Please refer to Figure 2 , which is a module structure diagram of an embodiment of a reflective object recognition device provided by the present invention.

[0111] A reflective object recognition device includes: a seed region module 210, a candidate region module 220, and a shape determination module 230;

[0112] The seed region module 210 is used to obtain regional point cloud data and determine a number of regions in the regional point cloud data whose intensity values ​​are greater than a preset value as a number of seed regions;

[0113] The candidate region module 220 is used to cluster the regional point cloud data with several seed regions as the center according to the diffusion clustering algorithm to obtain several candidate regions;

[0114] The shape determination module 230 is used to calculate the curvature of each candidate region, determine the shape of each candidate region according to the curvature of each candidate region, and determine the shape of the reflective object corresponding to each candidate region according to the shape of each candidate region.

[0115] Furthermore, the reflective object recognition device further includes: a circle center determination module 240 and a midpoint determination module 250;

[0116] The circle center determination module 240 is used to fit the candidate area according to the least squares circle fitting method when there is a candidate area with a shape of a reflective column, so as to obtain the center position of the reflective column;

[0117] The midpoint determination module 250 is used to fit the candidate area according to the least squares straight line fitting method when the shape of the candidate area is a reflector, so as to obtain the midpoint position and normal angle of the reflector.

[0118] Furthermore, the seed region module 210 includes a radar point cloud data unit 211 and a region point cloud data unit 212;

[0119] The radar point cloud data unit 211 is used to obtain radar point cloud data according to the radar detection device, and eliminate point cloud data that exceeds the recognition distance in the radar point cloud data according to the recognition distance of the radar detection device;

[0120] The regional point cloud data unit 212 is used to calibrate the radar point cloud data according to the real-time motion data of the AGV to obtain regional point cloud data.

[0121] Furthermore, the seed region module 210 includes a seed region determination unit 213;

[0122] The seed region determining unit 213 is configured to determine the first point cloud data in the regional point cloud data whose reflectivity is greater than a calibration value of the radar detection device, and determine the connected first point cloud data as a seed region.

[0123] Furthermore, the candidate region module 220 includes: a continuous determination unit 221;

[0124] The continuous judgment unit 221 is used to determine each seed area as a number of central areas in the cluster, and continuously judge whether the intensity difference between the point cloud data of the peripheral connection of the boundary points of each central area and the intensity of the boundary points of the central area is less than a preset intensity difference, until the intensity difference between the point cloud data of the non-central area without the boundary points of the central area and the intensity of the boundary points of the central area is less than the preset intensity difference, and then determine each central area as the corresponding candidate area;

[0125] In each judgment, if the intensity difference between the point cloud data of the non-central area connected to the boundary points of the central area and the intensity of the boundary points of the central area is less than the preset intensity difference, the point cloud data of the non-central area is added to the central area.

[0126] Furthermore, the shape determination module 230 includes: a splitting unit 231, a slope unit 232, a first judgment unit 233 and a shape recognition unit 234;

[0127] The splitting unit 231 is used to split all point cloud data in each candidate area into an upper part and a lower part according to the long side of each candidate area;

[0128] The slope unit 232 is used to calculate the slope of the upper part and the lower part of each candidate area according to the position information of the point cloud data in each candidate area, and calculate the slope difference between the upper part and the lower part of each candidate area;

[0129] The first judging unit 233 is used to judge whether the difference between the slopes of the upper part and the lower part of each candidate area is greater than a preset difference;

[0130] The shape recognition unit 234 is configured to determine that the shape of the candidate area is a reflector when the first judgment unit 233 determines that the slope difference between the upper part and the lower part of the candidate area is less than or equal to a preset difference;

[0131] The shape recognition unit 234 is further configured to determine that the shape of the candidate area is a reflective column when the first judgment unit 233 determines that the slope difference between the upper and lower parts of the candidate area is greater than a preset difference.

[0132] Furthermore, the reflective object recognition device further includes: a second judgment module 260 and a candidate area determination module 270;

[0133] The second judging module 260 is used to judge whether the size of each candidate area meets a preset size range;

[0134] The candidate region determination module 270 is configured to determine that a candidate region is an invalid region when the second judgment module 260 determines that the size of the candidate region does not meet a preset size range;

[0135] The candidate region determining module 270 is further configured to determine that, when the second judging module 260 determines that a candidate region exists whose size satisfies a preset size range, the candidate region is a final candidate region.

[0136] In this way, by determining the seed area based on the intensity value in the point cloud data, the area with high reflectivity can be accurately obtained. The reflective points of the reflective object can be accurately located according to the reflective characteristics of the reflective plate and the reflective column. And when the reflective points are accurately obtained as described above, the center of the cluster can be accurately determined, eliminating the problem of unclear clustering results caused by inaccurate determination of the center point in the general clustering algorithm. By diffusing each reflective point through the diffusion clustering algorithm, a clear candidate area range can be obtained. The diffusion clustering algorithm is different from the general clustering algorithm. The diffusion clustering algorithm does not need to classify or cluster all the points in the regional point cloud data, but diffuses outward with the seed area. This can eliminate the problem of unclear candidate areas caused by clustering all points in other clustering algorithms. Furthermore, the shape of each candidate area is determined by the curvature of each candidate area, which makes it simple and convenient to identify the shape of the candidate area, reducing the burden on the hardware equipment while improving accuracy.

[0137] The specific embodiments described above further illustrate the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for identifying reflective objects, characterized in that: include: Obtaining regional point cloud data, and determining a number of regions in the regional point cloud data whose intensity values ​​are greater than a preset value as a number of seed regions; According to the diffusion clustering algorithm, the regional point cloud data is clustered with several seed regions as the center to obtain several candidate regions; Calculating the curvature of each candidate region, determining the shape of each candidate region according to the curvature of each candidate region, and determining the shape of the reflective object corresponding to each candidate region according to the shape of each candidate region; According to the diffusion clustering algorithm, the regional point cloud data is clustered with several seed regions as the center to obtain several candidate regions, specifically: Determine each seed area as a number of central areas in the cluster, and continuously determine whether the difference between the intensity value of the point cloud data connected to the periphery of each central area boundary point and the intensity value of the central area boundary point is less than a preset intensity difference, until the difference between the intensity value of the point cloud data of the non-central area without the central area boundary point connection and the intensity value of the central area boundary point is less than the preset intensity difference, then determine each central area as the corresponding candidate area; Among them, in each judgment, if the difference between the intensity value of the point cloud data of the non-central area connected to the boundary point of the central area and the intensity value of the boundary point of the central area is less than the preset intensity difference, the point cloud data of the non-central area is added to the central area; The curvature of each candidate region is calculated, and the shape of each candidate region is determined according to the curvature of each candidate region, specifically: According to the long side of each candidate area, all point cloud data in each candidate area are split into upper and lower parts; According to the position information of the point cloud data in each candidate area, the slopes of the upper part and the lower part of each candidate area are calculated respectively, and the slope difference between the upper part and the lower part of each candidate area is calculated; Determine whether the slope difference between the upper and lower parts of each candidate area is greater than a preset difference; If the difference between the slopes of the upper and lower parts of the candidate area is less than or equal to the preset difference, determining that the shape of the candidate area is a reflector; If the difference in slope between the upper part and the lower part of the candidate area is greater than a preset difference, the shape of the candidate area is determined to be a reflective column.

2. The reflective object recognition method according to claim 1, characterized in that: After determining the shape of the reflective object corresponding to each candidate area according to the shape of each candidate area, the following steps are included: When there is a candidate area with a shape of a reflective column, the candidate area is fitted according to the least squares circle fitting method to obtain the center position of the reflective column; When there is a candidate area whose shape is a reflector, the candidate area is fitted according to the least squares straight line fitting method to obtain the midpoint position and normal angle of the reflector.

3. The reflective object recognition method according to claim 1, characterized in that: The obtaining of regional point cloud data is specifically as follows: Obtain radar point cloud data according to the radar detection device, and eliminate point cloud data exceeding the recognition distance in the radar point cloud data according to the recognition distance of the radar detection device; According to the real-time motion data of AGV, the radar point cloud data is calibrated to obtain regional point cloud data.

4. The reflective object recognition method according to claim 1, characterized in that: The method of determining a plurality of regions in the regional point cloud data whose intensity values ​​are greater than a preset value as a plurality of seed regions is specifically as follows: The first point cloud data having a reflectivity greater than a calibration value of the radar detection device in the regional point cloud data is determined, and the connected first point cloud data is determined as a seed area.

5. The reflective object recognition method according to claim 1, characterized in that: According to the diffusion clustering algorithm, the regional point cloud data is clustered with several seed regions as the center to obtain several candidate regions, and then the following steps are included: Determine whether the size of each candidate area meets the preset size range; If the size of a candidate region does not meet the preset size range, the candidate region is determined to be an invalid region; If there is a candidate region whose size meets the preset size range, the candidate region is determined to be the final candidate region.

6. A reflective object recognition device, characterized in that: include: Seed region module, candidate region module and shape determination module; The seed region module is used to obtain regional point cloud data and determine a number of regions in the regional point cloud data whose intensity values ​​are greater than a preset value as a number of seed regions; The candidate region module is used to cluster the regional point cloud data with several seed regions as the center according to the diffusion clustering algorithm to obtain several candidate regions; The shape determination module is used to calculate the curvature of each candidate area, determine the shape of each candidate area according to the curvature of each candidate area, and determine the shape of the reflective object corresponding to each candidate area according to the shape of each candidate area; The candidate area module includes: a continuous judgment unit; The continuous judgment unit is used to determine each seed area as a number of central areas in the cluster, and continuously judge whether the difference between the intensity value of the point cloud data of the peripheral connection of the boundary points of each central area and the intensity value of the boundary points of the central area is less than a preset intensity difference, until the difference between the intensity value of the point cloud data of the non-central area without the boundary points of the central area and the intensity value of the boundary points of the central area is less than the preset intensity difference, and then determine each central area as the corresponding candidate area; Among them, in each judgment, if the difference between the intensity value of the point cloud data of the non-central area connected to the boundary point of the central area and the intensity value of the boundary point of the central area is less than the preset intensity difference, the point cloud data of the non-central area is added to the central area; The shape determination module includes: a splitting unit, a slope unit, a first judgment unit and a shape recognition unit; The splitting unit is used to split all point cloud data in each candidate area into an upper part and a lower part according to the long side of each candidate area; The slope unit is used to calculate the slope of the upper part and the lower part of each candidate area according to the position information of the point cloud data in each candidate area, and calculate the slope difference between the upper part and the lower part of each candidate area; The first judging unit is used to judge whether the difference between the slopes of the upper part and the lower part of each candidate area is greater than a preset difference; The shape recognition unit is configured to determine that the shape of the candidate area is a reflector when the first judgment unit determines that the difference in slope between the upper part and the lower part of the candidate area is less than or equal to a preset difference; The shape recognition unit is further configured to determine that the shape of the candidate area is a reflective column when the first judgment unit determines that the slope difference between the upper part and the lower part of the candidate area is greater than a preset difference.

7. The reflective object recognition device according to claim 6, characterized in that: Also includes: Circle center determination module and midpoint determination module; The circle center determination module is used to fit the candidate area according to the least squares circle fitting method when there is a candidate area with a shape of a reflective column, so as to obtain the center position of the reflective column; The midpoint determination module is used to fit the candidate area according to the least squares straight line fitting method when the shape of the candidate area is a reflector, so as to obtain the midpoint position and normal angle of the reflector.

8. The reflective object recognition device according to claim 6, characterized in that: The seed area module includes a radar point cloud data unit and a regional point cloud data unit; The radar point cloud data unit is used to obtain radar point cloud data according to the radar detection device, and eliminate point cloud data that exceeds the recognition distance in the radar point cloud data according to the recognition distance of the radar detection device; The regional point cloud data unit is used to calibrate the radar point cloud data according to the real-time motion data of the AGV to obtain regional point cloud data.

9. The reflective object recognition device according to claim 6, characterized in that: The seed region module includes a seed region determination unit; The seed region determination unit is used to determine the first point cloud data in the regional point cloud data whose reflectivity is greater than the calibration value of the radar detection device, and determine the connected first point cloud data as a seed region.

10. The reflective object recognition device according to claim 6, characterized in that: Also includes: a second judgment module and a candidate region determination module; The second judgment module is used to judge whether the size of each candidate area meets a preset size range; The candidate region determination module is configured to determine that a candidate region is an invalid region when the second judgment module determines that a size of the candidate region does not meet a preset size range; The candidate region determination module is further configured to determine that, when the second judgment module determines that there is a candidate region whose size satisfies a preset size range, the candidate region is the final candidate region.

Images