Apparatus and method for determining the position of an object sensed by radar using spatial voxelization

Abstract

According to the present disclosure, a radar sensing file in which a reflected signal obtained by receiving a radio wave signal transmitted by a radar and reflected by an object is stored is loaded, a voxel corresponding to the position of a point stored in the radar sensing file is searched, and the reflected signal intensity of the point is accumulated and stored in the voxel ID of the searched voxel to generate a dictionary; a voxelization step, the radar sensing file is loaded, a voxel corresponding to the position of a point stored in the radar sensing file is searched, and when the ID of the searched voxel is the same as the voxel in which the accumulated value of the reflected signal intensity is the largest in the dictionary, the position coordinates of the point are separately stored; a point extraction step, and based on the position coordinates of the separately stored points, a step of determining the position of a reference point serving as a calibration reference are included, and a method for determining the position of an object sensed by a radar using spatial voxelization is provided, and the reference point of the radar for calibration can be accurately determined.

Description

Technical Field

[0001] The present disclosure relates to an apparatus and method for determining the position of an object sensed by radar using spatial voxelization.

Background Art

[0002] Radar outputs a radio wave signal and receives the signal reflected by an object. Depending on the characteristics of the object's surface, the radio wave signal is reflected in various directions and intensities. When various objects exist in space, the reflected signal may be received in various directions and intensities. Therefore, complex calculations must be performed to analyze the reflected signal received by the radar and select the position of the object.

[0003] On the other hand, with the development of autonomous driving-related technologies, technologies for calibrating sensors such as radar, lidar, and cameras are under development. Calibration is to adjust so that radar, lidar, and cameras recognize the position of a specific object as the same position. In the calibration process, the radar has a problem that it is difficult to obtain the position of a specific object based on the characteristics that the reflected signal is received in various directions and intensities. Such characteristics require that no other objects exist in the space where calibration is performed. Therefore, space and another tool are required for calibration.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] The present disclosure provides an apparatus and method for determining the position of an object sensed by radar using spatial voxelization. [Means for solving the problem]

[0006] A method for determining the position of an object sensed by radar using spatial voxelization according to this disclosure may include: a voxelization step of loading a radar sensing file stored by receiving and saving reflected signals that are returned when radio wave signals transmitted by radar are reflected by an object, searching for voxels corresponding to the positions of points stored in the radar sensing file, and generating a dictionary by accumulating and saving the reflected signal intensity of the points to the voxel ID of the searched voxel; a point extraction step of loading a radar sensing file, searching for voxels corresponding to the positions of points stored in the radar sensing file, and separately saving the position coordinates of the points if the ID of the searched voxel is the same as the voxel with the largest accumulated value of the reflected signal intensity in the dictionary; and a step of determining the position of a reference point to be used as a calibration reference based on the position coordinates of the separately saved points.

[0007] According to one embodiment, the radar sensing file may be generated when the radar receives a reflected signal and stores the position coordinates of the points and the reflected signal intensity in the form of point cloud data.

[0008] According to one embodiment, the voxelization step may include a step of setting parameters including voxel size, lower and upper threshold values ​​for reflected signal intensity, and spatial range for generating a voxel grid; a filtering step of excluding points whose reflected signal intensity does not fall within the range from the lower threshold to the upper threshold; a step of generating a voxel grid based on the voxel size and spatial range; and a step of searching the voxel grid for voxels corresponding to the locations of points stored in the radar sensing file, accumulating and storing the reflected signal intensity of the points in the voxel ID of the searched voxels to generate a dictionary.

[0009] According to one embodiment, the step of generating the dictionary may include: searching for a voxel from the voxel grid that corresponds to the location of a point stored in the radar sensing file; determining whether the voxel ID of the searched voxel exists in the dictionary; if the voxel ID of the searched voxel does not exist in the dictionary, adding the voxel ID of the searched voxel to the dictionary, storing the reflected signal intensity in the voxel ID, storing the count of points recorded in the voxel ID, and storing the average reflected signal intensity by dividing the reflected signal intensity by the count; and if the voxel ID of the searched voxel exists in the dictionary, accumulating and storing the reflected signal intensity in the voxel ID of the searched voxel in the dictionary, accumulating and storing the count of points recorded in the voxel ID, and storing the average reflected signal intensity by dividing the reflected signal intensity by the count.

[0010] According to one embodiment, the point extraction step may include: setting parameters including voxel size, lower and upper threshold values ​​for reflected signal intensity, and spatial range for generating a voxel grid; filtering points that do not fall within the range from the lower threshold to the upper threshold; generating a voxel grid based on the voxel size and spatial range; and loading a radar sensing file, searching for voxels corresponding to the locations of points stored in the radar sensing file, separately saving the position coordinates of the point if the voxel ID of the searched voxel is the same as the voxel ID of the top voxel with the largest cumulative value of reflected signal intensity in the dictionary, and extracting the point of the top voxel.

[0011] According to one embodiment, the step of extracting the point of the top voxel may include the steps of: searching the voxel grid for a voxel corresponding to the location of the point stored in the radar sensing file; determining whether the voxel ID of the searched voxel matches the voxel ID of the top voxel in the dictionary; and, if the voxel ID of the top voxel matches the voxel ID of the searched voxel, storing the location of the point in the voxel ID of the top voxel.

[0012] According to one example, the step of extracting the point of the top voxel further includes a step of determining whether the current file containing the point matches the target file when the voxel ID of the retrieved voxel matches the voxel ID of the top voxel in the dictionary, and if the current file matches the target file, the step of saving the position of the point is performed, and if the current file does not match the target file, the step of saving the position of the point is not required.

[0013] The apparatus for determining the position of an object sensed by radar using spatial voxelization according to this disclosure may include a radar that transmits radio signals and receives reflected signals that are reflected back by an object, and a determination module that analyzes a radar sensing file storing the reflected signals received by the radar using spatial voxelization to determine the position of a reflector that serves as a calibration reference.

[0014] According to one embodiment, the decision module includes a processor that executes program code, a storage unit connected to the processor so as to be able to send and receive data and store the program code, a radar sensing file storing reflected signals received by the radar, and a dictionary, and an input / output unit connected to the processor so as to be able to send and receive data and receive parameters necessary for voxelization, wherein the program code may be written to perform one of the methods for determining the position of an object sensed by the radar using the spatial voxelization.

[0015] The features and advantages of the present disclosure will become more apparent from the following detailed description based on the accompanying drawings.

[0016] In this description, terms and words used in this specification and the claims should not be construed in a normal and dictionary sense, but should be construed in a meaning and concept that conforms to the technical idea of the present disclosure based on the principle that the inventor can appropriately define the concept of the terms in order to explain his invention in the best way.

Effects of the Invention

[0017] According to the present disclosure, the position of a reference object for calibration can be determined in an environment where various objects exist using one radar.

Brief Description of the Drawings

[0018] [Figure 1] FIG. showing an environment in which a device for determining the position of an object sensed by radar using spatial voxelization according to an embodiment is used.

[0019] [Figure 2] FIG. showing a device for determining the position of an object sensed by radar using spatial voxelization according to an embodiment.

[0020] [Figure 3] FIG. showing a flowchart of a method for determining the position of an object sensed by radar using spatial voxelization according to an embodiment.

[0021] [Figure 4] FIG. explaining data for storing reflected signals received by a radar according to an embodiment.

[0022] <​​​​​​ [Figure 6] This is a diagram illustrating a voxel dictionary related to a specific example.

[0024] [Figure 7] This diagram illustrates the extraction of key points and the determination of reference points related to a specific example.

[0025] [Figure 8] This diagram illustrates the calibration process for a specific example.

[0026] [Figure 9] This diagram illustrates the method for extracting reference points during calibration. [Modes for carrying out the invention]

[0027] The purposes, advantages, and features of this disclosure will become more apparent from the following detailed description and preferred examples with reference to the attached drawings, but this disclosure is not limited thereto. Furthermore, in describing this disclosure, if it is determined that specific descriptions of relevant prior art would obscure the essence of this disclosure, such detailed descriptions will be omitted.

[0028] When assigning reference numerals to elements in a drawing, it should be noted that identical elements should be given the same reference numeral whenever possible, even if they appear in other drawings, and similar elements should be given similar reference numerals.

[0029] The terminology used to describe one specific example of this disclosure is not intended to limit the disclosure. Unless otherwise specified in the context, singular expressions can be understood to include plural expressions.

[0030] The drawings may be shown in a schematic or exaggerated manner for the purpose of illustrating the actual examples.

[0031] In this document, expressions such as "possess," "may possess," "include," or "may include" refer to the existence of the relevant feature (for example, numerical values, functions, operations, or components such as parts), and do not exclude the existence of further features.

[0032] Terms such as "one," "other," "another," "first," and "second" are used to distinguish one component from other components, and the components are not limited by these terms.

[0033] Terms indicating direction, such as up, down, left, right, X-axis, Y-axis, and Z-axis, are merely for explanatory convenience. It should be understood that different expressions are possible depending on the observer's position, the object's position, and other factors.

[0034] The examples and accompanying drawings described herein are not intended to limit this disclosure to any particular embodiment. This disclosure should be understood to include various modifications, equivalents, and / or alternatives to the examples.

[0035] An embodiment of this disclosure will be described in detail below with reference to the attached drawings.

[0036] Figure 1 shows an environment En in which a device 1, which determines the position of an object sensed by radar using spatial voxelization, is used, according to one embodiment. Figure 2 shows a device 1, which determines the position of an object sensed by radar using spatial voxelization, according to one embodiment.

[0037] An embodiment of the present disclosure, a device 1 that determines the position of an object sensed by radar using spatial voxelization, may be used in an environment En in which other objects Ob are present in the same space as the radar 20 and reflector 10. For example, the environment En may be a space in which various objects Ob such as desks, chairs, partitions, clocks, picture frames, etc., are placed. The environment En of an embodiment of the present disclosure may include a state in which the radar 20 and reflector 10 are placed together in a space in which various objects Ob are placed.

[0038] Generally, the environment En in which calibration is performed using radar 20, lidar, and camera is a space in which only a reflector 10 exists to be used as a reference. If various objects exist in the space, the radio wave signals of radar 20 are scattered, making it difficult to accurately distinguish the reference object. Therefore, conventionally, a method is used in which two or more radars 20 are used to determine the position of the reflector 10.

[0039] In contrast, one embodiment of this disclosure can accurately distinguish the position of the reflector 10 even when various objects Ob are present in the environment En. Furthermore, the position of the reflector 10 can be accurately determined using only one radar 20. Therefore, when performing calibration, only a single radar 20 needs to be prepared, minimizing equipment costs, and since it can be performed in a normal office space, the cost of securing space can be minimized.

[0040] An embodiment of the apparatus 1 for determining the position of an object sensed by radar using spatial voxelization may include a radar 20 and a determination module 30. An embodiment of the apparatus 1 for determining the position of an object sensed by radar using spatial voxelization may further include a reflector 10.

[0041] The radar 20 may utilize a single radar unit. The radar 20 can transmit radio signals and receive reflected signals that are reflected back from an object. The radar 20 can transmit radio signals with a frequency of 79 GHz. The radio signals transmitted by the radar 20 may have other frequencies. The reflected signals may be received by the radar 20 along various paths and at various intensities depending on the shape and material of the object, the distance from the radar 20, and various other factors. The radar 20 can output the position coordinates of the point from which the received reflected signal originated and the reflected signal intensity. The point refers to the part of the object from which the reflected signal originated. The position coordinates may include values ​​for the X, Y, and Z axes. The reflected signal intensity refers to the intensity of the reflected signal that is reflected from an object and received by the radar 20. The reflected signal intensity may be the radar cross-section (RCS) of the radar 20.

[0042] The reflector 10 is an object made of a material that reflects radio wave signals well. The reflector 10 may include a corner reflector. The reflector 10 may be formed in a manner that reflects the radio wave signals output by the radar 20 toward the radar 20 with minimal loss. The reflector 10 is positioned to be used as a reference point for calibration relative to the radar 20.

[0043] The decision module 30 can generate a radar sensing file F by saving the position coordinates and reflected signal intensity of points received from the radar 20. The decision module 30 can also receive the radar sensing file F generated using the radar 20. The decision module 30 can analyze the radar sensing file F, which stores the reflected signals received by the radar 20, using spatial voxelization to determine the position of the reflector 10, which will serve as the calibration reference.

[0044] The decision module 30 may include a computer device. For example, the decision module 30 may include a PC, notebook PC, server computer, tablet PC, or other device capable of performing information processing functions. The decision module 30 is connected to the radar 20 and can receive the position coordinates of points and the reflected signal intensity output from the radar 20.

[0045] The decision module 30 may include a processor 310 that executes program code, a storage unit 320 connected to the processor 310 for data transmission and reception and storing the program code, a radar sensing file F storing reflected signals received by the radar 20, and a dictionary, and an input / output unit 340 connected to the processor 310 for data transmission and reception and receiving parameters necessary for voxelization. The decision module 30 may further include a communication unit 330 connected to the processor 310 for data transmission and reception and sending and receiving data with an external device. Here, the program code may be written to perform a method of determining the position of an object sensed by the radar using spatial voxelization.

[0046] The processor 310 may include elements capable of information processing. The processor 310 may be a CPU, GPU, AP, or various other computing elements. The decision module 30 may include multiple processors 310. The multiple processors 310 may be linked to send and receive data from each other. The processor 310 can perform a method for determining the position of an object sensed by radar using spatial voxelization according to an example created in program code.

[0047] The memory unit 320 can store data necessary for performing a method of determining the position of an object sensed by radar using spatial voxelization. The memory unit 320 may include RAM, ROM, memory, hard disk, cloud storage, etc. The memory unit 320 can store program code created to perform each step of the method of determining the position of an object sensed by radar using spatial voxelization. The program code may be executed by the processor 310. The memory unit 320 can store radar sensing files F, voxel dictionaries, separately stored point position coordinates, parameters, and other data.

[0048] The input / output unit 340 may include a keyboard, mouse, touchpad, touchscreen, pen, etc., for receiving input from the user. The input / output unit 340 may also include a display, speaker, etc., for displaying information to the user. The input / output unit 340 can provide the user with a screen for inputting parameters and can visually present analysis results.

[0049] The communication unit 330 can connect to a communication network and send and receive data. The processor 310 can download radar sensing files F via the communication unit 330 and transmit analysis results. The communication unit 330 can utilize various communication methods such as 5G, 6G, satellite communication, Wi-Fi, Bluetooth, LAN, WAN, Ethernet, IP4, and IP6.

[0050] Figure 3 is a flowchart illustrating a method for determining the position of an object sensed by radar using spatial voxelization, as an example.

[0051] The processor 310 performs a method for determining the position of an object sensed by the radar 20 using spatial voxelization, according to one embodiment. The method for determining the position of an object sensed by the radar 20 using spatial voxelization, according to one embodiment, may include a voxelization step (S200) in which a radar sensing file F is stored by receiving reflected signals that are returned when radio wave signals transmitted by the radar 20 are reflected by an object, a voxelization step (S300) in which a voxel corresponding to the position of a point stored in the radar sensing file F is searched for, and a dictionary is generated by accumulating and storing the reflected signal intensity of the point in the voxel ID of the searched voxel; a point extraction step (S300) in which the radar sensing file F is loaded, a voxel corresponding to the position of a point stored in the radar sensing file F is searched for, and the position coordinates of the point are separately stored if the ID of the searched voxel is the same as the voxel with the largest accumulated value of the reflected signal intensity in the dictionary; and a step (S400) in which the position of a reference point to be used as a reference for calibration is determined based on the position coordinates of the separately stored point. The voxelization stage (S200), the point extraction stage (S300), and the stage of determining the position of the reference point (S400) may be performed by the processor 310.

[0052] The voxelization stage (S200) is the process of determining the corresponding voxel for a point stored in the radar sensing file F, matching the point information to the voxel, and storing it in a dictionary. The dictionary can store information about the points contained in the voxel. The dictionary can store voxel ID, reflected signal intensity, count, and average reflected signal intensity. The dictionary does not store the location of the points. If the dictionary stored the location of the points, the dictionary size would increase, and the data processing speed may slow down. When voxels are arranged in descending order of reflected signal intensity, the voxel with the greatest reflected signal intensity can be called the top voxel. The top voxel has the greatest accumulated reflected signal intensity, so it contains many points, and can be analyzed as having a high reflected signal intensity. Therefore, it can be analyzed as a space where the reflector 10 is likely to be located.

[0053] The point extraction stage (S300) is a process in which points stored in the radar sensing file F are read, a determination is made as to whether the points are contained in the top voxel stored in the dictionary, and only the points contained in the top voxel are extracted. By separating the voxelization stage (S200) for generating the dictionary and the point extraction stage (S300) for points contained in the top voxel, it is possible to improve the data processing speed and minimize the dictionary size. The position coordinates of the points extracted in the point extraction stage (S300) may be stored separately.

[0054] The step of determining the position of the reference point (S400) is the process of determining the position of the reflector 10 using the position coordinates of the points determined to be included in the top voxel. The position of the top voxel generally refers to the coordinates of the center point of the voxel. The position coordinates of the points extracted after being determined to be included in the top voxel are different from the coordinates of the center point of the top voxel. Therefore, in order to accurately determine the position of the reflector 10, the position coordinates of the extracted points are used. The position of the reflector 10 can serve as a reference point for calibration that matches sensing data from the radar 20, lidar, and camera.

[0055] Figure 4 illustrates the data stored by the radar 20 that receives the reflected signal, as per one concrete example. Refer to both Figures 3 and 4.

[0056] A method for determining the position of an object sensed by radar using spatial voxelization, according to one embodiment, may further include a sensing step (S100) in which the radar 20 transmits a radio signal and receives a reflected signal reflected by an object. The sensing step (S100) may further include the operation in which the determination module 30 organizes the data received from the radar 20 and generates a radar sensing file F. The radar sensing file F may be generated when the radar 20 receives the reflected signal and stores the position coordinates of the point and the reflected signal intensity in the form of point cloud data.

[0057] The raw data that the radar 20 receives from the reflected signal and provides to the decision module 30 may be in various forms. The data that the radar 20 provides to the decision module 30 may be X-axis, Y-axis, and Z-axis coordinates, and may be in various forms depending on the type of radar 20. The decision module 30 can organize the data received from the radar 20 as X-axis, Y-axis, Z-axis, and reflected signal intensity to generate a radar sensing file F. The radar sensing file F may be generated at a predetermined size or time.

[0058] The radar 20 may have a data acquisition cycle of 10 to 20 Hz. The radar 20 can receive reflected signals at each data acquisition cycle to generate data and provide it to the decision module 30, or generate a radar sensing file F. For example, the radar sensing file F may be generated by receiving reflected signals to radio signals at each data acquisition cycle of 10 to 20 Hz, and may be generated as a file with a length of 50 to 60 ms. The data acquisition cycle of the radar 20 may vary, and the length or size of the radar sensing file F may vary. Multiple radar sensing files F may be generated during the sensing stage (S100) using the radar 20.

[0059] Figure 4 shows the first radar sensing file F1 to the Nth radar sensing file FN, where N is a positive number. Refer to Table A in Figure 4, which is an enlarged view of the data for the fourth radar sensing file F4. The fourth radar sensing file F4 may include values ​​corresponding to the X, Y, and Z axes, and may also include values ​​corresponding to the reflected signal strength (RCS). Radar sensing file F may contain data in this form. Radar sensing file F generated by the decision module 30 may be stored in the storage unit 320.

[0060] The decision module 30 can load the radar sensing file F stored in the memory unit 320 and perform the voxelization step (S200).

[0061] The voxelization step (S200) may include a step of setting parameters including voxel size, lower and upper threshold values ​​for reflected signal intensity, and spatial range for generating a voxel grid (S210); a filtering step (S220) that excludes points whose reflected signal intensity does not fall within the range from the lower threshold to the upper threshold; a step of generating a voxel grid based on the voxel size and spatial range (S230); and a step of searching the voxel grid for voxels corresponding to the location of points stored in the radar sensing file F, accumulating and storing the reflected signal intensity of the points in the voxel ID of the searched voxels to generate a dictionary (S240).

[0062] The parameter setting stage (S210) allows for further configuration of the dictionary generation range DB.

[0063] The dictionary generation range database (DB) is the range of files that will be voxed in order to generate the dictionary. The dictionary generation range DB can be set by defining a target file (File_target) and the number of files (n) to select before and after the target file. For example, if the 8th radar sensing file F8 is selected as the target file and the number of files to select before and after it (n) is set to 6, then the dictionary generation range DB can be set from the 2nd radar sensing file F2 (File_target-n) to the 14th radar sensing file F14 (File_target+n).

[0064] Alternatively, the dictionary generation range DB can be configured by selecting a start file and an end file. For example, by selecting the second radar sensing file F2 as the start file and the fourteenth radar sensing file F14 as the end file, the second to fourteenth radar sensing files F2 to F14 can be configured as the dictionary generation range DB.

[0065] Figure 5 illustrates spatial voxelization in one specific example. Refer to both Figures 3 and 4.

[0066] The voxel size can be set at the parameter setting stage (S210). The voxel size may be expressed in R. The value of R can be entered by the user. The voxel may be in the form of a cube. A cubic voxel may contain eight vertices V1 to V8. The spatial coordinates of the first to eighth vertices V1 to V8 may be expressed as [-R,-R,-R] to [R,R,R]. In such a case, the distance between adjacent vertices may be 2R. The larger the voxel size set, the more points it can contain internally.

[0067] At the parameter setting stage (S210), the spatial range for generating the voxel grid can be set. The spatial range may be determined by the size of the space sensed by the radar 20. Preferably, the spatial range is set to be larger than the distance from the radar 20 to the reflector 10. The spatial range may be set using the values ​​of the X, Y, and Z axes.

[0068] Once the voxel size and spatial range are set, a voxel grid can be generated. A voxel grid divides space into multiple voxels. Figure 5 shows a portion of the voxel grid relative to the origin. Each of the multiple voxels contained in the voxel grid is given a voxel ID that distinguishes them. The voxel ID may be given in a predetermined format.

[0069] At the parameter setting stage (S210), the lower and upper thresholds for reflected signal strength can be set. The radio waves from the radar 20 are reflected by objects in various directions, are reflected several times, and are attenuated. Therefore, it is preferable to ignore reflected signals with a reflected signal strength below a predetermined lower threshold among the reflected signals received by the radar 20. Since the radio waves from the radar 20 are attenuated as they are reflected by objects, if the reflected signal strength is above the upper threshold, it may not be a reflected signal from the radio waves transmitted from the radar 20, and it is preferable to ignore it. The lower and upper thresholds for reflected signal strength may be input by the user. Alternatively, the lower and upper thresholds may be determined by the characteristics of the radar 20.

[0070] After setting the voxel size, spatial range, lower threshold, upper threshold, and dictionary generation range DB in the parameter setting stage (S210), the filtering stage (S220) can be performed.

[0071] The filtering stage (S220) is the process of excluding points stored in the radar sensing file F whose reflected signal intensity does not fall within the range from the lower threshold to the upper threshold. The processor 310 can compare the reflected signal intensity of a point with the lower threshold and the upper threshold and exclude points that do not fall within the range between the lower threshold and the upper threshold. The processor 310 can repeat the filtering stage (S220) for each loaded radar sensing file F.

[0072] After the filtering stage (S220), a voxel grid can be generated. The voxel grid may be generated based on the voxel size and spatial range. The processor 310 can repeatedly perform the step of generating a voxel grid (S230) for each loaded radar sensing file F. Once the voxel grid is generated, multiple voxels are generated within the spatial range, and the processor 310 can determine which voxel corresponds to the position coordinates of a point.

[0073] Once the voxel grid is generated, the dictionary generation step (S240) can be performed. The dictionary generation step (S240) is the process of determining which voxel corresponds to a point and recording the point's information in that voxel.

[0074] The dictionary generation step (S240) may include the steps of: searching for a voxel in the voxel grid that corresponds to the location of a point stored in the radar sensing file F (S241); determining whether the voxel ID of the searched voxel exists in the dictionary (S242); adding the voxel ID of the searched voxel to the dictionary if it does not exist in the dictionary, storing the reflected signal intensity in the voxel ID, storing the count of points recorded in the voxel ID, and saving the average reflected signal intensity by dividing the reflected signal intensity by the count (S243); and accumulating the reflected signal intensity in the dictionary if it exists in the dictionary, storing the reflected signal intensity accumulated in the voxel ID of the searched voxel, storing the count of points recorded in the voxel ID, and saving the average reflected signal intensity by dividing the reflected signal intensity by the count (S244).

[0075] The voxel search step (S241) is the process of determining which voxel corresponds to the location of a point stored in the radar sensing file F. The voxel grid has the positions of multiple voxels predetermined, and by comparing the position coordinates of a point with the position of a voxel, the voxel containing the point can be searched. For example, in Figure 5, by comparing the position coordinates of the first point P1 with the position and size of the voxel, it is possible to search which voxel it is contained in. The first point P1 may be contained in the first voxel Vx1, the second point P2 and the third point P3 may be contained in the second voxel Vx2, and the fourth point P4 and the fifth point P5 may be contained in the third voxel (Vx3). Since the radar sensor file contains multiple points, the voxel search step (S241) may be repeated for each point.

[0076] The voxel search step (S241) allows the voxel corresponding to the point to be identified. Each voxel is assigned a voxel ID. The step of determining whether the voxel ID exists in the dictionary (S242) is to determine whether the voxel ID of the voxel corresponding to the point is recorded in the dictionary. The dictionary may store multiple voxel IDs and information about points that match the voxel IDs. The processor 310 determines whether the voxel ID of the searched voxel matches a voxel ID stored in the dictionary.

[0077] Figure 6 illustrates a voxel dictionary for one specific example. Refer to Figures 3, 4, and 5.

[0078] If the voxel ID of the searched voxel does not exist in the dictionary, an appending step (S243) can be performed, and if the voxel ID of the searched voxel does exist in the dictionary, an accumulation step (S244) can be performed.

[0079] The addition step (S243) is the process of adding a new voxel ID that does not exist in the dictionary. In the addition step (S243), the searched voxel ID is newly added to the dictionary, and the point's reflected signal intensity, count, and average reflected signal intensity may be matched to the voxel ID and stored. For example, in Figure 6, if the voxel ID of the searched voxel is 2322, it does not match the voxel ID stored in the existing dictionary D1, so the addition step (S243) can be performed to add the point information to the dictionary. The dictionary (D2) after the addition step (S243) includes the voxel ID of the searched voxel, stores the value of the reflected signal intensity (RCS) (14.4), and since a new point has been added, the count is stored as 1, and the average reflected signal intensity (AVG) may be stored as the same value as the reflected signal intensity (14.4) because the count is 1.

[0080] The cumulative stage (S244) is the process of accumulating and storing point information in the voxel IDs that exist in the dictionary. If the same voxel ID as the searched voxel ID exists in the existing dictionary D1, it means that the other points are determined to correspond to the same voxel, and the point information has been stored in the dictionary. For example, in Figure 6, if the voxel ID of the searched voxel is 1401, there is a matching voxel ID among the voxel IDs stored in the existing dictionary D1. Therefore, the cumulative stage (S244) can be performed. In the dictionary (D2) after the cumulative stage (S244), the reflected signal intensity that has been matched to the voxel ID of the searched voxel has been accumulated and stored from 57.6 to 64.0, the count has been accumulated and stored from 3 to 4, and the average reflected signal intensity has been accumulated and stored from 19.2 to 16.0.

[0081] In the voxelization stage (S200), the parameter setting stage (S210) is performed once. Then, for each of the multiple radar sensing files F set in the dictionary generation range DB, the filtering stage (S220), the voxel grid generation stage (S230), and the dictionary generation stage (S240) are repeated. In the dictionary generation stage (S240), the voxel search stage (S241) is performed for each of the multiple points contained in one radar sensing file F, the voxel ID of the searched voxel is determined to exist in the dictionary (S242), the addition stage (S243), and the accumulation stage (S244) are repeated. By performing the voxelization stage (S200) for each of the multiple radar sensing files F set in the dictionary generation range DB, a dictionary can be obtained in which the reflected signal intensity of the points contained in the voxel is accumulated and stored.

[0082] The dictionary stores multiple voxel IDs and reflected signal strengths. By sorting them in descending order of reflected signal strength, the voxel with the highest reflected signal strength can be called the top voxel.

[0083] Refer to Figure 3. Once the dictionary is generated, the point extraction stage (S300) can be performed. Figure 7 is a diagram illustrating point extraction and reference point determination in one example. Refer to both Figure 3 and Figure 7.

[0084] The point extraction step (S300) may include a step of setting parameters including voxel size, lower and upper threshold values ​​for reflected signal intensity, and spatial range for generating a voxel grid (S310); a filtering step (S320) for excluding points whose reflected signal intensity does not fall within the range from the lower threshold to the upper threshold; a step of generating a voxel grid based on the voxel size and spatial range (S330); and a step of loading a radar sensing file F, searching for voxels corresponding to the locations of points stored in the radar sensing file F, and, if the voxel ID of the searched voxel is the same as the voxel ID of the top voxel with the largest cumulative value of reflected signal intensity in the dictionary, separately storing the position coordinates of the point and extracting the top voxel point (S340).

[0085] The step of setting parameters (S310) in the point extraction stage (S300) may be the same as the step of setting parameters (S210) in the voxelization stage (S200). Therefore, the step of setting parameters (S310) in the point extraction stage (S300) may be omitted. In the step of setting parameters (S310) in the point extraction stage (S300), the voxel size, the lower and upper threshold values ​​for reflected signal intensity, and the spatial range can be set, and the range of the radar sensing file F from which points are extracted (point extraction range EB) can be set. The point extraction range EB may be set to be the same as the dictionary generation range DB for generating the dictionary in the voxelization stage (S200), or it may be set to be different.

[0086] In the point extraction stage (S300), the filtering stage (S320) may be performed in the same manner as the filtering stage (S220) in the voxelization stage (S200). Points stored in the radar sensing file F, which is set as the point extraction range EB, can be excluded if their reflected signal intensity does not fall within the range of the lower and upper thresholds.

[0087] The step of generating a voxel grid (S330) in the point extraction step (S300) may be performed in the same manner as the step of generating a voxel grid (S230) in the voxelization step (S200). A grid can be generated using voxels of a set voxel size within a set spatial range.

[0088] The step of extracting the top voxel point (S340) is a process of searching for a voxel corresponding to the point stored in the radar sensing file F, which is set as the point extraction range EB, and extracting the point if the searched voxel is the same as the top voxel. The step of extracting the top voxel point (S340) may include the steps of searching for a voxel corresponding to the position of the point stored in the radar sensing file F from the voxel grid (S341), determining whether the voxel ID of the searched voxel matches the voxel ID of the top voxel in the dictionary (S342), and saving the position of the point to the voxel ID of the top voxel if the voxel ID of the top voxel and the voxel ID of the searched voxel match (S343).

[0089] The step of searching for voxels (S341) in the step of extracting points of the top voxel (S340) is similar to the step of searching for voxels (S241) in the step of generating a dictionary (S240). The step of searching for voxels (S341) in the step of extracting points of the top voxel (S340) is the process of confirming the voxel corresponding to the location of the point in the radar sensing file F set in the point extraction range EB. Once the voxel corresponding to the point is found, the voxel ID of the voxel can be confirmed.

[0090] Next, the system determines whether the voxel ID of the searched voxel matches the voxel ID of the top voxel in the dictionary (S342). If the voxel ID of the searched voxel is different from the voxel ID of the top voxel, the system further searches for a voxel for the next point in the radar sensing file F (S241).

[0091] If the voxel ID of the searched voxel is the same as the voxel ID of the top voxel, the step of saving the point's position (S343) is performed. The step of saving the point's position (S343) is the process of separately saving the position coordinates of the points that are determined to be included in the top voxel. At this time, the position coordinates of the points can be saved by matching them with the voxel ID of the top voxel. If there are multiple points included in the top voxel, the separately saved data may contain the position coordinates of multiple points in a single top voxel. For example, from the visualization result in Figure 7, it can be confirmed that the voxel ID of the top voxel is 3027. Referring to the content A of the top voxel recorded in the dictionary, it can be confirmed that the voxel ID of the top voxel is 3027, the accumulated reflected signal intensity is 140.2, the count is 11, and the average reflected signal intensity is 12.7. If the voxel ID of the voxel corresponding to a point in radar sensing file F is 3027, then it is the same as the voxel ID of the top voxel. Therefore, the position coordinates of that point and the reflected signal intensity of that point can be saved separately. In Figure 7, referring to Content B, where the position coordinates and reflected signal intensity of that point are saved separately, it can be confirmed that the position coordinates and reflected signal intensity of multiple points are saved under voxel ID 3027.

[0092] In the point extraction stage (S300), the step of setting parameters (S210) is performed once. For each radar sensing file F set as the point extraction range EB, the filtering stage (S220), the voxel grid generation stage (S230), and the top voxel point extraction stage (S340) are repeated. In the top voxel point extraction stage (S340), for each of the multiple points contained in one radar sensing file F, the steps of searching for voxels (S241), determining whether the voxel ID of the searched voxel matches the voxel ID of the top voxel (S342), and saving the point position are repeated. When the point extraction stage (S300) is performed for each radar sensing file F set as the radar sensing file F from which points are extracted, the position coordinates of the points determined to be included in the top voxel can be saved separately.

[0093] Once the position coordinates of the point corresponding to the top voxel are extracted, the step of determining the position of the reference point (S400) can be performed. The reference point refers to the position of the reflector 10 which serves as the reference in calibration. In the calibration process, it is necessary to determine the position of the reflector 10 as a reference point in order to adjust the radar 20, lidar, and camera so that they measure a single reflector 10 as being at the same position. The method of determining the position of an object sensed by radar using spatial voxelization, as described in one embodiment, may be used to determine the position of the reference point measured by the radar 20 in the calibration process.

[0094] Referring to the determination of the reflector 10's position in Figure 7, the position of the reflector 10 (i.e., the position of the reference point) is calculated using the position coordinates of the points contained in the top voxel. The X-coordinate of the reflector 10 may be calculated as the average of the X-coordinates of the points contained in the top voxel, the Y-coordinate of the reflector 10 may be calculated as the average of the Y-coordinates of the points contained in the top voxel, and the Z-coordinate of the reflector 10 may be calculated as the average of the Z-coordinates of the points contained in the top voxel. At the position of the reflector 10, the distance (Dist) refers to the distance from the radar 20 to the reflector 10. The distance may be calculated using the position coordinates (X, Y, Z) of the reflector 10 by the following equation 1. The position coordinates of the reflector 10 are the coordinates of the reference point.

[0095]

number

[0096] Dist: Straight-line distance from radar 20 to reflector 10, X: X-coordinate of reflector 10, Y: Y-coordinate of reflector 10, Z: Z-coordinate of reflector 10

[0097] A method for determining the position of an object sensed by radar using spatial voxelization, according to one embodiment, may further include a step of visualizing the reflected signal intensity of voxels stored in a dictionary in order of magnitude. The visualization step may be performed by user selection after the dictionary generation step (S240). The visualization result VR can be provided in a manner in which voxels with high reflected signal intensity are displayed in a dark color and those with low reflected signal intensity are displayed in a light color. The visualization result VR may display the accumulated reflected signal intensity value and voxel ID. The user can refer to the visualization result VR to sequentially check the accumulated value of the reflected signal intensity of the points contained in the voxels.

[0098] Figure 8 is a diagram illustrating calibration in one specific example. Figure 9 is a diagram illustrating the method for extracting reference points at the time of calibration.

[0099] A method for determining the position of an object sensed by radar using spatial voxelization, as described in one embodiment, may be used for calibration. For example, calibration may be performed in a manner in which the radar 20, lidar, and camera sense a single reference point (reflector 10). The reference time for adjusting the position during the calibration process may be applied identically to all sensors (radar 20, lidar, and camera). Due to the characteristics of optical equipment, lidar and cameras have the advantage of being able to acquire clear data at a specific point in time. However, due to the characteristics of radio wave equipment, radar 20 needs to determine a reference point by aggregating data collected over a certain period rather than acquiring data at a specific point in time. Therefore, it may not be possible to determine an accurate reference point using only the target file (File_target) corresponding to the calibration reference time. For example, the length of the target file may not be sufficiently long. A method for determining the position of an object sensed by radar using spatial voxelization, as described in one embodiment, can use a method in which a dictionary is generated using multiple radar sensing files F, and the points included in the target file corresponding to the calibration reference time are extracted to determine the reference point. Therefore, the position of the reference point measured by radar 20 at the calibration reference time can be determined.

[0100] In a method for determining the position of an object sensed by radar using spatial voxelization according to one embodiment, the step of extracting the top voxel point (S340) may further include a step of determining whether the current file containing the point matches the target file (S344) if the voxel ID of the retrieved voxel matches the voxel ID of the top voxel in the dictionary. If the current file matches the target file (Y), the step of saving the point's position (S343) is performed, and if the current file does not match the target file (Y), the step of saving the point's position (S343) is not performed.

[0101] Here, the dictionary generation range DB in the dictionary generation stage (S240) and the point extraction range EB in the point extraction stage (S300) may be different. For example, as shown in Figure 8, the first dictionary generation range DB1 may be radar sensing files F43 to F49, numbered 43 to 49, which correspond to the calibration reference time, and the point extraction range EB may be radar sensing file F53, number 53, which is a single target file (File_target). By generating a dictionary using the radar sensing files F in the first dictionary generation range DB1, the position of the reference point can be determined by extracting the points corresponding to the top voxel from the points included in the target file.

[0102] Alternatively, at the dictionary generation stage (S240), the dictionary generation range DB may include the target file corresponding to the calibration reference time, and at the point extraction stage (S300), the point extraction range EB may further include files different from the target file. For example, as shown in Figure 8, the second dictionary generation range DB2 includes the target file (File_target), and the point extraction range EB also includes the target file (File_target). A dictionary can be generated using the radar sensing file F in the second dictionary generation range DB2, and points that correspond to the top voxel and the target file (File_target) can be extracted from the points included in the point extraction range EB to determine the position of the reference point. Here, points that correspond to the top voxel but are included in the radar sensing file F that is not the target file (File_target) are not extracted.

[0103] Using the described method, the reference point for the radar 20 to be used for calibration can be determined. The method and apparatus according to one embodiment have the advantage that the reference point can be determined in an everyday environment where various objects exist, thus eliminating the need to prepare a separate space for calibration.

[0104] The above has provided a detailed explanation of this disclosure with specific examples. These examples are provided to illustrate the disclosure specifically, and the disclosure is not limited thereto. It will be clear that the technical ideas of this disclosure can be modified and improved upon by those with ordinary skill in the art.

[0105] Any simple modification or alteration of this disclosure falls within the scope of this disclosure, and the specific scope of protection of this disclosure will be clarified by the attached claims. [Explanation of Symbols]

[0106] 1: A device that uses spatial voxelization to determine the position of an object sensed by radar.

[0107] En: Environment

[0108] Ob: object

[0109] 10:Reflector

[0110] 20: Radar

[0111] 30: Decision Module

[0112] 310: Processor

[0113] 320: Storage section

[0114] 330: Communications Department

[0115] 340: Input / output section

Claims

1. A voxelization step involves loading a radar sensing file that has been stored by receiving reflected signals from an object after the radar has transmitted radio signals and returned, searching for voxels corresponding to the positions of points stored in the radar sensing file, and generating a dictionary by accumulating the reflected signal intensity of the points in an aggregate manner and saving it as a reference point setting dictionary for calibration to the voxel ID of the searched voxel, A point extraction step includes loading the radar sensing file, searching for voxels corresponding to the location of points stored in the radar sensing file, and separately saving the position coordinates of the point as a point to be used for calculating the calibration reference point if the ID of the searched voxel is the same as the voxel with the largest cumulative value of the reflected signal intensity in the dictionary. A method for determining the position of an object sensed by radar using spatial voxelization, comprising: a reference point position determination step of determining the position of a reference point for radar coordinate calibration based on the position coordinates of a separately stored point.

2. The aforementioned radar sensing file is A method for determining the position of an object sensed by a radar using spatial voxelization according to claim 1, which is generated by the radar receiving a reflected signal and storing the position coordinates of the point and the reflected signal intensity in the form of point cloud data.

3. The voxelization step is, A step of setting parameters including voxel size, lower and upper thresholds for reflected signal intensity, and the spatial range for generating the voxel grid, A filtering step of excluding points whose reflected signal intensity does not fall within the range from the lower threshold to the upper threshold, A step of generating a voxel grid based on the voxel size and spatial range, A method for determining the position of an object sensed by radar using spatial voxelization according to claim 2, comprising the steps of: searching the voxel grid for a voxel corresponding to the position of a point stored in the radar sensing file; accumulating and storing the reflected signal intensity of the point in the voxel ID of the searched voxel to generate a dictionary.

4. The step of generating the aforementioned dictionary is: The steps include: searching the voxel grid for a voxel corresponding to the location of a point stored in the radar sensing file; The step of determining whether the voxel ID of the searched voxel exists in the dictionary, If the voxel ID of the searched voxel does not exist in the dictionary, an additional step is to add the voxel ID of the searched voxel to the dictionary, store the reflected signal intensity in the voxel ID, store the count of points recorded in the voxel ID, and store the average of the reflected signal intensity by dividing the reflected signal intensity by the count. A method for determining the position of an object sensed by radar using spatial voxelization according to claim 3, comprising: an accumulation step in which, if the voxel ID of the searched voxel exists in the dictionary, the reflected signal intensity is accumulated and stored in the voxel ID of the searched voxel in the dictionary, the count of points recorded in the voxel ID is accumulated and stored, and the average of the reflected signal intensity is stored by dividing the reflected signal intensity by the count.

5. The aforementioned point extraction step is, A step of setting parameters including voxel size, lower and upper thresholds for reflected signal intensity, and the spatial range for generating the voxel grid, A filtering step of excluding points whose reflected signal intensity does not fall within the range from the lower threshold to the upper threshold, The steps include generating a voxel grid based on the voxel size and spatial range, and A method for determining the position of an object sensed by radar using spatial voxelization according to claim 2, comprising the steps of: loading a radar sensing file; searching for a voxel corresponding to the position of a point stored in the radar sensing file; separately storing the position coordinates of the point if the voxel ID of the searched voxel is the same as the voxel ID of the top voxel with the largest cumulative value of reflected signal intensity in the dictionary; and extracting the point of the top voxel.

6. The step of extracting the points of the top voxel is as follows: The steps include: searching the voxel grid for a voxel corresponding to the location of a point stored in the radar sensing file; The step of determining whether the voxel ID of the searched voxel matches the voxel ID of the top voxel in the dictionary, A method for determining the position of a radar-sensed object using spatial voxelization according to claim 5, comprising the step of storing the position of the point in the voxel ID of the top voxel when the voxel ID of the top voxel matches the voxel ID of the searched voxel.

7. The step of extracting the points of the top voxel is as follows: The process further includes the step of determining whether the current radar sensing file containing the point matches the target radar sensing file corresponding to the calibration reference time, if the voxel ID of the searched voxel matches the voxel ID of the top voxel in the dictionary. A method for determining the position of an object sensed by radar using spatial voxelization according to claim 6, wherein the step of saving the position of the point is performed when the current radar sensing file matches the target radar sensing file, and the step of saving the position of the point is not performed when the current radar sensing file does not match the target radar sensing file.

8. A radar transmits radio signals and receives reflected signals that are reflected back by objects, The system includes: a determination module that analyzes the reflected signals for setting a coordinate reference for at least one radar sensing file storing the reflected signals received by the radar, using spatial voxelization, and determines the position of a reflector that will serve as a calibration reference to be used for the coordinate reference, based on the results of the analysis by spatial voxelization; The aforementioned decision module is A processor that executes program code, A storage unit is connected to the processor in a manner that enables data transmission and reception, and stores the program code, a radar sensing file containing reflected signals received by the radar, and a dictionary. The processor is connected to an input / output unit capable of transmitting and receiving data, and includes an input / output unit that receives parameters necessary for voxelization. The aforementioned program code is: A device for determining the position of a radar-sensed object using spatial voxelization, which is constructed to perform any one of the methods described in claims 1 to 7.

Images