Object classification method and apparatus, electronic device, and storage medium

Abstract

An object classification method and apparatus, an electronic device, and a storage medium, relating to the technical field of radar sensors. The method comprises: in a preset simulation environment, identifying a plurality of objects by means of a hybrid radar sensor (S100); in the preset simulation environment, determining the nearest object closest to a target object and, in a polar coordinate system using the hybrid radar sensor as the pole, separately determining first polar coordinates corresponding to the target object and second polar coordinates corresponding to the nearest object (S200); according to the first polar coordinates and the second polar coordinates, calculating a distance difference value and an angle difference value between the target object and the nearest object (S300); by means of a preset Gaussian distribution variance, obtaining a Gaussian distribution function of the target object in the polar coordinate system (S400); according to the distance difference value, the angle difference value, a radar resolution and the Gaussian distribution function, calculating an outcome of probability that the target object and the nearest object are identified as the same object (S500); and, according to the outcome of probability, determining a classification result of the target object (S600). By means of the Gaussian variance, the method outputs an object classification probability in the hybrid radar sensor, thereby extending radar application scenarios while ensuring calculation efficiency.

Description

Object classification method, device, electronic device and storage medium

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This disclosure claims priority to Chinese patent application number 202310913111.0, filed with the State Intellectual Property Office of China on July 25, 2023, entitled “A method, device, electronic device and storage medium for object classification,” the entire contents of which are incorporated herein by reference. Technical Field

[0003] The present disclosure relates to the technical field of radar sensors, and in particular to an object classification method, an object classification device, an electronic device, and a storage medium. Background Art

[0004] In autonomous driving technology, on-board millimeter-wave sensors are used for target obstacle detection. When designing autonomous driving simulation software, in order to simulate the algorithm of autonomous driving vehicles for perception data, it is necessary to simulate the detection data of millimeter-wave radar. At the same time, the simulated millimeter-wave radar results need to meet the characteristics of real millimeter-wave radar, one of which is resolution. Millimeter waves cannot classify two objects that are too close in distance or angle relative to the sensor as one object. Due to the accuracy and error of millimeter-wave radar, this phenomenon will occur probabilistically. Among the relevant technical means, hybrid sensors or physical sensors are usually used to complete the detection simulation of target obstacles.

[0005] Since the output data of physical sensors are calculated through physical processes for each frame, if we want to output the recognition probability, it is usually obtained by measuring and statistically analyzing multiple frames of simulation data in an experimental setting. It is impossible to output the probability in real time. Although the hybrid model has relatively high computational efficiency, it does not support the calculation of the probability or possibility of the target obstacle being correctly classified.

[0006] Summary of the Invention

[0007] In view of this, the present disclosure provides at least one object classification method, object classification device, electronic device, and storage medium. Through Gaussian variance, the output of object classification probability is realized in a hybrid radar sensor, while ensuring computational efficiency and extending radar application scenarios.

[0008] The present disclosure mainly includes at least the following aspects:

[0009] Some embodiments of the present disclosure provide an object classification method, which may include:

[0010] In a preset simulation environment, multiple objects are identified by the hybrid radar sensor, and the nearest neighboring object closest to the target object in the preset simulation environment is determined. The first polar coordinate corresponding to the target object and the second polar coordinate corresponding to the nearest neighboring object are respectively determined in a polar coordinate system with the hybrid radar sensor as the pole. The distance difference and the angle difference between the target object and the nearest neighboring object are calculated based on the first polar coordinate and the second polar coordinate. The Gaussian distribution function of the target object in the polar coordinate system is obtained through the preset Gaussian distribution variance. The possibility that the target object and the nearest neighboring object are identified as the same object is calculated based on the distance difference, the angle difference, the radar resolution and the Gaussian distribution function. The classification result of the target object is determined based on the possibility result.

[0011] In one possible implementation, the first polar coordinate may include a first distance value and a first angle value, and the second polar coordinate may include a second distance value and a second angle value, wherein the distance difference and the angle difference between the target object and the nearest object may be determined in the following manner: the difference between the first distance value and the second distance value is determined as the distance difference between the target object and the nearest object; and the difference between the first angle value and the second angle value is determined as the angle difference between the target object and the nearest object.

[0012] In one possible embodiment, the preset Gaussian distribution variance may include a distance Gaussian distribution variance and an angle Gaussian distribution variance, and the Gaussian distribution function may include a distance Gaussian distribution function and an angle Gaussian distribution function, wherein the Gaussian distribution function of the target object in the polar coordinate system can be obtained in the following manner: adding the distance Gaussian distribution variance to the first distance value to obtain the distance Gaussian distribution function, the horizontal axis of the distance Gaussian distribution function represents the predicted distance value of the target object under the influence of the distance Gaussian distribution variance, and the vertical axis of the distance Gaussian distribution function represents the probability of occurrence of the predicted distance value; adding the angle Gaussian distribution variance to the first angle value to obtain the angle Gaussian distribution function, the horizontal axis of the angle Gaussian distribution function represents the multiple predicted angle values ​​of the target object under the influence of the angle Gaussian distribution variance, and the vertical axis of the angle Gaussian distribution function represents the probability of occurrence of the predicted angle.

[0013] In one possible embodiment, the radar resolution may include distance resolution and angle resolution, wherein the possibility result can be determined in the following manner: determining the distance range in which the target object and the nearest neighboring object cannot be correctly classified based on the distance difference and the distance resolution; calculating a first probability that the predicted distance of the target object is within the distance range based on the distance Gaussian distribution variance, the distance range and the distance resolution, the first probability indicating the possibility that the target object and the nearest neighboring object cannot be correctly classified in the distance dimension; determining the angle range in which the target object and the nearest neighboring object cannot be correctly classified based on the angle difference and the angle resolution; calculating a second probability that the predicted angle of the target object is within the angle range based on the angle Gaussian distribution variance, the angle range and the angle resolution, the second probability indicating the possibility that the target object and the nearest neighboring object cannot be correctly classified in the angle dimension; and calculating the possibility result that the target object and the nearest neighboring object are identified as the same object based on the first probability and the second probability.

[0014] In a possible implementation, the first probability may be determined by the following formula:

[0015] In this formula, P R represents the first probability, R represents the distance difference between the target object and the nearest object, σ1 2 is the variance of the distance Gaussian distribution, R+α represents the upper limit of the distance range, R-α represents the lower limit of the distance range, and α represents the distance resolution.

[0016] The second probability can be determined by the following formula:

[0017] In this formula, P θ represents the second probability, θ represents the angular difference between the target object and the nearest object, σ2 2 is the variance of the angular Gaussian distribution, θ+β represents the upper limit of the angular range, θ-β represents the lower limit of the angular range, and β represents the angular resolution.

[0018] In one possible embodiment, the step of calculating the likelihood that the target object and the nearest neighboring object are identified as the same object based on the first probability and the second probability may include: calculating the product between the first probability and the second probability; and determining the difference between 1 and the product as the likelihood that the target object and the nearest neighboring object are identified as the same object.

[0019] In one possible embodiment, the step of determining the classification result of the target object based on the possibility result may include: comparing the possibility result with a first preset threshold; if the possibility result is greater than or equal to the first preset threshold, determining that the radar sensor correctly classifies the target object; if the possibility result is less than the first preset threshold, determining whether the possibility result is less than a second preset threshold; if the possibility result is less than the second preset threshold, determining that the radar sensor incorrectly classifies the target object; if the possibility result is greater than or equal to the second preset threshold, determining that the radar sensor inaccurately classifies the target object.

[0020] Some other embodiments of the present disclosure further provide an object classification device, which may include:

[0021] An identification module, wherein the identification module is configured to: obtain multiple objects by identifying a hybrid radar sensor in a preset simulation environment; a first determination module, wherein the first determination module is configured to: determine the nearest neighboring object closest to the target object in the preset simulation environment, and respectively determine the first polar coordinates corresponding to the target object and the second polar coordinates corresponding to the nearest neighboring object in a polar coordinate system with the radar sensor as the pole; a first calculation module, wherein the first calculation module is configured to: calculate the distance difference and angle difference between the target object and the nearest neighboring object based on the first polar coordinates and the second polar coordinates; a second determination module, wherein the second determination module is configured to: obtain the Gaussian distribution function of the target object in the polar coordinate system by presetting the Gaussian distribution variance; a second calculation module, wherein the second calculation module is configured to: calculate the possibility result that the target object and the nearest neighboring object are identified as the same object based on the distance difference, the angle difference, the radar resolution and the Gaussian distribution function; a third determination module, wherein the third determination module is configured to: determine the classification result of the target object based on the possibility result.

[0022] In one possible implementation, the first polar coordinate may include a first distance value and a first angle value, and the second polar coordinate may include a second distance value and a second angle value, wherein the first calculation module may further be configured to: determine the distance difference and angle difference between the target object and the nearest object in the following manner: determine the difference between the first distance value and the second distance value as the distance difference between the target object and the nearest object; and determine the difference between the first angle value and the second angle value as the angle difference between the target object and the nearest object.

[0023] In one possible embodiment, the preset Gaussian distribution variance may include a distance Gaussian distribution variance and an angle Gaussian distribution variance, and the Gaussian distribution function may include a distance Gaussian distribution function and an angle Gaussian distribution function, wherein the second determination module may also be configured to: obtain the Gaussian distribution function of the target object in the polar coordinate system in the following manner: add the distance Gaussian distribution variance to the first distance value to obtain the distance Gaussian distribution function, the horizontal axis of the distance Gaussian distribution function represents the predicted distance value of the target object under the influence of the distance Gaussian distribution variance, and the vertical axis of the distance Gaussian distribution function represents the probability of occurrence of the predicted distance value; add the angle Gaussian distribution variance to the first angle value to obtain the angle Gaussian distribution function, the horizontal axis of the angle Gaussian distribution function represents the multiple predicted angle values ​​of the target object under the influence of the angle Gaussian distribution variance, and the vertical axis of the angle Gaussian distribution function represents the probability of occurrence of the predicted angle.

[0024] In one possible embodiment, the radar resolution may include distance resolution and angle resolution, wherein the second calculation module may further be configured to: determine the possibility result in the following manner: determine the distance range in which the target object and the nearest neighboring object cannot be correctly classified based on the distance difference and the distance resolution; calculate a first probability that the predicted distance of the target object is within the distance range based on the distance Gaussian distribution variance, the distance range and the distance resolution, the first probability indicating the possibility that the target object and the nearest neighboring object cannot be correctly classified in the distance dimension; determine the angle range in which the target object and the nearest neighboring object cannot be correctly classified based on the angle difference and the angle resolution; calculate a second probability that the predicted angle of the target object is within the angle range based on the angle Gaussian distribution variance, the angle range and the angle resolution, the second probability indicating the possibility that the target object and the nearest neighboring object cannot be correctly classified in the angle dimension; calculate the possibility result that the target object and the nearest neighboring object are identified as the same object based on the first probability and the second probability.

[0025] In a possible implementation, the second calculation module may be further configured to determine the first probability using the following formula:

[0026] In this formula, P R represents the first probability, R represents the distance difference between the target object and the nearest object, σ1 2 is the variance of the distance Gaussian distribution, R+α represents the upper limit of the distance range, R-α represents the lower limit of the distance range, and α represents the distance resolution.

[0027] The second probability can be determined by the following formula:

[0028] In this formula, P θrepresents the second probability, θ represents the angular difference between the target object and the nearest object, σ2 2 is the variance of the angular Gaussian distribution, θ+β represents the upper limit of the angular range, θ-β represents the lower limit of the angular range, and β represents the angular resolution.

[0029] In a possible embodiment, the second calculation module can also be configured as follows: based on the first probability and the second probability, the step of calculating the possibility that the target object and the nearest object are identified as the same object can include: calculating the product between the first probability and the second probability; and determining the difference between 1 and the product as the possibility that the target object and the nearest object are identified as the same object.

[0030] In a possible embodiment, the third determination module can also be configured as follows: based on the possibility result, the step of determining the classification result of the target object may include: comparing the possibility result with a first preset threshold; if the possibility result is greater than or equal to the first preset threshold, determining that the radar sensor classifies the target object correctly; if the possibility result is less than the first preset threshold, determining whether the possibility result is less than a second preset threshold; if the possibility result is less than the second preset threshold, determining that the radar sensor classifies the target object incorrectly; if the possibility result is greater than or equal to the second preset threshold, determining that the radar sensor classifies the target object inaccurately.

[0031] Some further embodiments of the present disclosure further provide an electronic device, which may include: a processor, a memory, and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor and the memory communicate through the bus, and the machine-readable instructions are executed by the processor to execute the steps of the object classification method in some of the above embodiments or any possible implementation of some of the above embodiments.

[0032] Some further embodiments of the present disclosure further provide a computer-readable storage medium, which may store a computer program. When the computer program is executed by a processor, the steps of the object classification method in some of the above embodiments or any possible implementation of the above embodiments are executed.

[0033] The embodiments of the present disclosure provide an object classification method, an object classification device, an electronic device, and a storage medium, including: identifying multiple objects through a hybrid radar sensor, determining the first polar coordinates corresponding to the target object and the second polar coordinates corresponding to the nearest object, calculating the distance difference and angle difference between the target object and the nearest object based on the first polar coordinates and the second polar coordinates, obtaining the Gaussian distribution function of the target object in the polar coordinate system by presetting the Gaussian distribution variance, calculating the probability that the target object and the nearest object are identified as the same object based on the distance difference, angle difference, radar resolution, and Gaussian distribution function, and determining the classification result of the target object based on the probability result. The present disclosure implements the output of object classification probability in the hybrid radar sensor through Gaussian variance, extending the radar application scenarios while ensuring computational efficiency.

[0034] The present disclosure is beneficial in at least the following ways:

[0035] 1. The output of the probability of an object being correctly classified is realized, bypassing some tedious physical measurements and processes, and directly applying it to high-efficiency hybrid sensors, avoiding the impact of non-real-time simulation of physical sensors on simulation technology verification.

[0036] 2. The target classification process is simplified, making data output faster, and the radar's recognition and classification capabilities are quantified using concepts such as the probability of correct classification, which can be applied to more autonomous driving algorithm simulation verification.

[0037] In order to make the above-mentioned objectives, features and advantages of the present disclosure more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present disclosure and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.

[0039] FIG1 shows a flow chart of an object classification method provided by an embodiment of the present disclosure;

[0040] FIG2 shows a schematic diagram of a distance Gaussian distribution function provided by an embodiment of the present disclosure;

[0041] FIG3 shows a schematic structural diagram of an object classification device provided by an embodiment of the present disclosure;

[0042] FIG4 shows a schematic structural diagram of an electronic device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solutions and advantages of the embodiments of the present disclosure clearer, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. It should be understood that the drawings in the present disclosure are only for the purpose of illustration and description and are not used to limit the scope of protection of the present disclosure. In addition, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in the present disclosure illustrate operations implemented according to some embodiments of the present disclosure. It should be understood that the operations of the flowchart can be implemented out of sequence, and steps that do not have a logical context relationship can be reversed in order or implemented simultaneously. In addition, those skilled in the art, guided by the contents of the present disclosure, can add one or more other operations to the flowchart, or remove one or more operations from the flowchart.

[0044] In addition, the described embodiments are only a part of the embodiments of the present disclosure, rather than all of the embodiments. The components of the embodiments of the present disclosure generally described and shown in the drawings herein can be arranged and designed in a variety of different configurations. Therefore, the following detailed description of the embodiments of the present disclosure provided in the drawings is not intended to limit the scope of the disclosure claimed for protection, but merely represents selected embodiments of the present disclosure. Based on the embodiments of the present disclosure, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present disclosure.

[0045] Related technologies usually use physical sensors and hybrid sensors to measure obstacles in simulation environments. Specifically:

[0046] Hybrid radar sensors are low-level sensors that focus on embodying the physical characteristics of sensors. Compared to more precise physical sensors that rely more on millimeter wave principles, hybrid sensors approximate many millimeter wave characteristics and directly add them to the data. Their advantage is fast computing speed, but hybrid sensors do not have the ability to output probability parameters.

[0047] Physical sensors need to use methods such as ray tracing to perform object recognition or obstacle measurement. This requires considering the sensor's signal transmission and reception, as well as more other sensor parameters. Although this restores the principle of sensor operation, it requires a larger amount of computing power and modeling of environmental materials, resulting in low classification calculation efficiency of physical sensors.

[0048] Moreover, since the output data of physical sensors are calculated through physical processes for each frame, they cannot output probability parameters in real time. For example, if the probability of two target objects being correctly classified is 90%, then its performance in the physical sensor is: out of 100 frames of simulation data, 90 frames of simulation data are identified as two target objects, and 10 frames of simulation data are identified as one target object. In other words, for physical sensors, it is impossible to output the probability of the target being correctly classified within the real-time frame.

[0049] Based on this, the disclosed embodiments provide an object classification method that enables a hybrid radar sensor to output parameters indicating the likelihood of a target being correctly classified in real time. This method extends radar application scenarios while ensuring computational efficiency. The details are as follows:

[0050] Please refer to Figure 1, which shows a flow chart of an object classification method provided by an embodiment of the present disclosure. As shown in Figure 1, the object classification method provided by an embodiment of the present disclosure may include the following steps:

[0051] S100: In a preset simulation environment, multiple objects are identified by a hybrid radar sensor.

[0052] In a specific embodiment, the hybrid radar sensor may be a hybrid millimeter-wave radar sensor, the preset simulation environment may be a pre-built autonomous driving simulation environment, the hybrid radar sensor is located on a main vehicle in the autonomous driving simulation environment, and the multiple objects may be vehicles or other obstacles. By controlling the hybrid radar sensor to scan the autonomous driving simulation environment, the multiple objects within the detection range of the hybrid radar sensor may be determined based on the detection data of the hybrid radar sensor.

[0053] S200: Determine a nearest neighboring object in a preset simulation environment, and respectively determine a first polar coordinate corresponding to the target object and a second polar coordinate corresponding to the nearest neighboring object in a polar coordinate system with the hybrid radar sensor as a pole.

[0054] In the present disclosure, any one of multiple objects can be selected as the target object. In the present disclosure, based on the detection data output by the radar sensor, the position coordinates of each object in the rectangular coordinate system with the radar sensor as the origin can be determined. Based on the position coordinates corresponding to each object, the distance between the target object and other objects can be calculated. Then, the object corresponding to the minimum distance value is determined as the nearest object corresponding to the target object.

[0055] In the present disclosure, polar coordinate conversion is performed on the position coordinates corresponding to the target object and the nearest object respectively, so as to obtain the first polar coordinates corresponding to the target object and the second polar coordinates corresponding to the nearest object, wherein the first polar coordinates may include a first distance value and a first angle value, and the second polar coordinates may include a second distance value and a second angle value.

[0056] Specifically, the first polar coordinate can be expressed as Q1(R1, θ1), where R1 represents a first distance value and θ1 represents a first angle value. The second polar coordinate can be expressed as Q2(R2, θ2), where R2 represents a second distance value and θ2 represents a second angle value.

[0057] S300 , calculating the distance difference and the angle difference between the target object and the nearest neighboring object according to the first polar coordinates and the second polar coordinates.

[0058] In a preferred embodiment, the distance difference and angle difference between the target object and the nearest neighboring object can be determined by:

[0059] The difference between the first distance value and the second distance value is determined as the distance difference between the target object and the nearest object, and the difference between the first angle value and the second angle value is determined as the angle difference between the target object and the nearest object.

[0060] Specifically, the distance difference R=|R1-R2|, and the angle difference θ=|θ1-θ2|.

[0061] S400 , obtaining a Gaussian distribution function of the target object in a polar coordinate system by presetting a Gaussian distribution variance.

[0062] In the present disclosure, the measurement errors that may occur in the target object are taken into account, but the measurement errors corresponding to objects other than the target object are not considered (that is, the measurement errors corresponding to other objects are regarded as zero). Therefore, a preset Gaussian distribution variance is only introduced in the target object. The preset Gaussian distribution variance can include the distance Gaussian distribution variance and the angle Gaussian distribution variance from the two error dimensions of distance and angle. The Gaussian distribution function can include the distance Gaussian distribution function and the angle Gaussian distribution function.

[0063] In a preferred embodiment, the Gaussian distribution function of the target object in the polar coordinate system can be obtained by:

[0064] The distance Gaussian distribution variance is added to the first distance value to obtain a distance Gaussian distribution function, the horizontal axis of the distance Gaussian distribution function represents the predicted distance value of the target object under the influence of the distance Gaussian distribution variance, and the vertical axis of the distance Gaussian distribution function represents the probability of occurrence of the predicted distance value. The angle Gaussian distribution variance is added to the first angle value to obtain an angle Gaussian distribution function, the horizontal axis of the angle Gaussian distribution function represents multiple predicted angle values ​​of the target object under the influence of the angle Gaussian distribution variance, and the vertical axis of the angle Gaussian distribution function represents the probability of occurrence of the predicted angle.

[0065] Please refer to Figure 2, which shows a schematic diagram of a distance Gaussian distribution function provided by an embodiment of the present disclosure. As shown in Figure 2, Figure 2 shows that the variance of the distance Gaussian distribution is σ1 2 , the normal distribution curve L corresponding to the distance Gaussian distribution function with mathematical expectation μ, the horizontal axis R represents the first distance value in the distance Gaussian distribution with variance σ1 2 The predicted distance value obtained under the influence of , the vertical axis P represents the probability of the predicted distance value. At this time, with the target object Object1 as the center point, that is, the target object Object1 is at the position shown in Figure 2, and the nearest object Object2 is at the position relative to the target object Object1 as shown in Figure 2. Due to the symmetry of the Gaussian distribution, the relative distance is directly taken as positive.

[0066] S500 , calculating the likelihood that the target object and the nearest neighboring object are identified as the same object based on the distance difference, the angle difference, the radar resolution, and the Gaussian distribution function.

[0067] Among them, radar resolution can include distance resolution and angle resolution.

[0068] In a preferred embodiment, the likelihood result can be determined by:

[0069] Based on the distance difference and distance resolution, the distance range in which the target object and the nearest object cannot be correctly classified is determined. Based on the variance of the distance Gaussian distribution, the distance range and the distance resolution, the first probability that the predicted distance of the target object is within the distance range is calculated. The first probability represents the possibility that the target object and the nearest object cannot be correctly classified in the distance dimension.

[0070] As shown in Figure 2, the width between the dotted lines LL1 and LL2 is twice the distance resolution α, and the determined distance range is (R-α, R+α). Therefore, based on the randomness of the predicted distance of the target object Object1, it is only necessary to calculate the first probability corresponding to the target object Object1 falling within the distance range to know the probability that the target object Object1 cannot be correctly classified, that is, the first probability.

[0071] If the predicted distance value of the target object falls within the distance range formed by the dotted lines LL1 and LL2, it is determined that the target object cannot be correctly classified. Failure to correctly classify means that the hybrid millimeter-wave radar identifies the target object Object1 and the nearest object Object2 as the same object.

[0072] Based on the angle difference and angular resolution, the angular range in which the target object and the nearest neighboring object cannot be correctly classified is determined. According to the variance of the angular Gaussian distribution, the angular range and the angular resolution, the second probability that the predicted angle of the target object is within the angular range is calculated. The second probability represents the possibility that the target object and the nearest neighboring object cannot be correctly classified in the angular dimension.

[0073] The method for determining the second probability is the same as the method for determining the first probability, and will not be described in detail here.

[0074] The probability result that the target object and the nearest neighbor object are identified as the same object is calculated based on the first probability and the second probability.

[0075] In another preferred embodiment, the first probability can be determined by the following formula:

[0076] In this formula, P R represents the first probability, R represents the distance difference between the target object and the nearest object, σ1 2 is the variance of the distance Gaussian distribution, R+α represents the upper limit of the distance range, R-α represents the lower limit of the distance range, and α represents the distance resolution.

[0077] The second probability can be determined by the following formula:

[0078] In this formula, P θ represents the second probability, θ represents the angular difference between the target object and the nearest object, σ2 2 is the variance of the angular Gaussian distribution, θ+β represents the upper limit of the angular range, θ-β represents the lower limit of the angular range, and β represents the angular resolution.

[0079] In a specific embodiment, the step of calculating the probability that the target object and the nearest neighbor object are identified as the same object based on the first probability and the second probability may include:

[0080] The product of the first probability and the second probability is calculated, and the difference between 1 and the product is determined as the probability result that the target object and the nearest neighbor object are identified as the same object.

[0081] Specifically, it can be calculated by the following formula: P = 1-P R P θ

[0082] Among them, P represents the probability that the target object and the nearest neighbor object are identified as the same object.

[0083] In the present disclosure, the above process is repeated for all objects within the detection range of the hybrid millimeter wave sensor, and the output of the possibility of the target being correctly classified (i.e., the possibility result) can be added to each object, thereby achieving the function of the hybrid millimeter wave sensor outputting probability parameters.

[0084] S600: Determine the classification result of the target object according to the possibility result.

[0085] In a preferred embodiment, step S600 may include:

[0086] The likelihood result is compared with a first preset threshold. If the likelihood result is greater than or equal to the first preset threshold, it is determined that the radar sensor correctly classifies the target object. If the likelihood result is less than the first preset threshold, it is determined whether the likelihood result is less than a second preset threshold. If the likelihood result is less than the second preset threshold, it is determined that the radar sensor incorrectly classifies the target object. If the likelihood result is greater than or equal to the second preset threshold, it is determined that the radar sensor inaccurately classifies the target object.

[0087] Based on the same application concept, the embodiment of the present disclosure also provides an object classification device corresponding to the object classification method provided in the above embodiment. Since the principle of solving the problem by the device in the embodiment of the present disclosure is similar to the object classification method in the above embodiment of the present disclosure, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be repeated.

[0088] Please refer to Figure 3, which shows a schematic diagram of the structure of an object classification device provided by an embodiment of the present disclosure. As shown in Figure 3, the object classification device may include:

[0089] The recognition module 700 is configured to: in a preset simulation environment, identify multiple objects using a hybrid radar sensor.

[0090] The first determination module 710 is configured to: determine the nearest neighboring object closest to the target object in a preset simulation environment, and respectively determine the first polar coordinates corresponding to the target object and the second polar coordinates corresponding to the nearest neighboring object in a polar coordinate system with the hybrid radar sensor as the pole.

[0091] The first calculation module 720 is configured to calculate the distance difference and the angle difference between the target object and the nearest object according to the first polar coordinates and the second polar coordinates.

[0092] The second determining module 730 is configured to obtain a Gaussian distribution function of the target object in a polar coordinate system by presetting a Gaussian distribution variance.

[0093] The second calculation module 740 is configured to calculate the probability that the target object and the nearest neighbor object are identified as the same object based on the distance difference, the angle difference, the radar resolution, and the Gaussian distribution function.

[0094] The third determination module 750 is configured to determine the classification result of the target object according to the possibility result.

[0095] Based on the same application concept, please refer to Figure 4, which shows a schematic diagram of the structure of an electronic device provided in accordance with an embodiment of the present disclosure. As shown in Figure 4, electronic device 800 may include: a processor 810, a memory 820, and a bus 830. Memory 820 stores machine-readable instructions executable by processor 810. When electronic device 800 is in operation, processor 810 and memory 820 communicate via bus 830. The machine-readable instructions are executed by processor 810 to execute the steps of the object classification method provided in any of the above embodiments.

[0096] Based on the same application concept, the embodiment of the present disclosure further provides a computer-readable storage medium, which may store a computer program. When the computer program is executed by a processor, the steps of the object classification method provided in the above embodiment are executed.

[0097] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system and device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. In the several embodiments provided in the present disclosure, it should be understood that the disclosed system, device and method can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0098] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0099] In addition, each functional unit in each embodiment of the present disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0100] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present disclosure, or the part that contributes to the relevant technology or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present disclosure. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0101] The above are only specific embodiments of the present disclosure, but the scope of protection of the present disclosure is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this disclosure should be included in the scope of protection of the present disclosure. Therefore, the scope of protection of the present disclosure should be based on the scope of protection of the claims. Industrial Applicability

[0102] The present disclosure provides an object classification method, device, electronic device, and storage medium, relating to the field of radar sensor technology, including: identifying multiple objects through a hybrid radar sensor, determining a first polar coordinate corresponding to the target object and a second polar coordinate corresponding to the nearest object, calculating the distance difference and angle difference between the target object and the nearest object based on the first polar coordinate and the second polar coordinate, obtaining a Gaussian distribution function of the target object in a polar coordinate system by presetting a Gaussian distribution variance, calculating the probability that the target object and the nearest object are identified as the same object based on the distance difference, angle difference, radar resolution, and Gaussian distribution function, and determining the classification result of the target object based on the probability result. The present disclosure implements the output of object classification probability in a hybrid radar sensor through Gaussian variance, extending the radar application scenarios while ensuring computational efficiency.

[0103] Furthermore, it is understood that the object classification apparatus of the present disclosure is reproducible and can be used in a variety of industrial applications. For example, the object classification apparatus of the present disclosure can be used in any application that requires an object classification apparatus to classify a target object.

Claims

1. A method for classifying objects, wherein: The object classification method comprises: In a preset simulation environment, multiple objects are identified by hybrid radar sensors; Determine a nearest neighbor object in a preset simulation environment that is closest to the target object, and respectively determine a first polar coordinate corresponding to the target object and a second polar coordinate corresponding to the nearest neighbor object in a polar coordinate system with the hybrid radar sensor as a pole; Calculate the distance difference and the angle difference between the target object and the nearest object according to the first polar coordinates and the second polar coordinates; By presetting the Gaussian distribution variance, a Gaussian distribution function of the target object in the polar coordinate system is obtained; Calculating, based on the distance difference, the angle difference, the radar resolution, and the Gaussian distribution function, a probability result that the target object and the nearest object are identified as the same object; According to the possibility result, a classification result of the target object is determined.

2. The object classification method according to claim 1, wherein: The first polar coordinates include a first distance value and a first angle value, and the second polar coordinates include a second distance value and a second angle value. The distance difference and angle difference between the target object and the nearest object are determined in the following manner: Determine a difference between the first distance value and the second distance value as a distance difference between the target object and the nearest object; A difference between the first angle value and the second angle value is determined as an angle difference between the target object and the nearest object.

3. The object classification method according to claim 2, wherein: The preset Gaussian distribution variance includes a distance Gaussian distribution variance and an angle Gaussian distribution variance, and the Gaussian distribution function includes a distance Gaussian distribution function and an angle Gaussian distribution function. The Gaussian distribution function of the target object in the polar coordinate system is obtained in the following manner: Adding the distance Gaussian distribution variance to the first distance value to obtain a distance Gaussian distribution function, wherein the horizontal axis of the distance Gaussian distribution function represents a predicted distance value of the target object under the influence of the distance Gaussian distribution variance, and the vertical axis of the distance Gaussian distribution function represents an occurrence probability of the predicted distance value; The angle Gaussian distribution variance is added to the first angle value to obtain an angle Gaussian distribution function, wherein the horizontal axis of the angle Gaussian distribution function represents multiple predicted angle values ​​of the target object under the influence of the angle Gaussian distribution variance, and the vertical axis of the angle Gaussian distribution function represents the predicted angle occurrence probability.

4. The object classification method according to claim 3, wherein: The radar resolution includes distance resolution and angle resolution. The possible results are determined in the following manner: Determining, according to the distance difference and the distance resolution, a distance range within which the target object and the nearest object cannot be correctly classified; Calculating a first probability that the predicted distance of the target object is within the distance range according to the distance Gaussian distribution variance, the distance range, and the distance resolution, wherein the first probability indicates a possibility that the target object and the nearest object cannot be correctly classified under the distance dimension; Based on the angle difference and the angle resolution, determining an angle range in which the target object and the nearest object cannot be correctly classified; Calculating, according to the variance of the angular Gaussian distribution, the angular range, and the angular resolution, a second probability that the predicted angle of the target object is within the angular range, the second probability indicating the possibility that the target object and the nearest object cannot be correctly classified in the angular dimension; A probability result that the target object and the nearest object are identified as the same object is calculated based on the first probability and the second probability.

5. The object classification method according to claim 4, wherein: The first probability is determined by the following formula: In this formula, P R represents the first probability, R represents the distance difference between the target object and the nearest object, σ1 2 is the distance Gaussian distribution variance, R+α represents the upper limit of the distance range, R-α represents the lower limit of the distance range, and α represents the distance resolution; The second probability is determined by the following formula: In this formula, P θ represents the second probability, θ represents the angle difference between the target object and the nearest object, σ2 2 is the angular Gaussian distribution variance, θ+β represents the upper limit of the angular range, θ-β represents the lower limit of the angular range, and β represents the angular resolution.

6. The object classification method according to claim 5, wherein: The step of calculating the probability result that the target object and the nearest object are identified as the same object according to the first probability and the second probability comprises: calculating the product between the first probability and the second probability; The difference between 1 and the product is determined as a possibility result that the target object and the nearest neighbor object are identified as the same object.

7. The object classification method according to any one of claims 1 to 6, wherein: The step of determining the classification result of the target object according to the possibility result includes: The probability result is compared with a first preset threshold value, and if the probability result is greater than or equal to the first If a preset threshold is reached, it is determined that the radar sensor correctly classifies the target object; If the likelihood result is less than the first preset threshold, determining whether the likelihood result is less than a second preset threshold; If the likelihood result is less than the second preset threshold, determining that the radar sensor misclassifies the target object; If the likelihood result is greater than or equal to the second preset threshold, it is determined that the radar sensor classifies the target object inaccurately.

8. An object classification device, wherein: The object classification device comprises: An identification module, wherein the identification module is configured to: identify multiple objects in a preset simulation environment through a hybrid radar sensor; A first determination module, wherein the first determination module is configured to: determine a nearest neighbor object in a preset simulation environment that is closest to the target object, and respectively determine a first polar coordinate corresponding to the target object and a second polar coordinate corresponding to the nearest neighbor object in a polar coordinate system with the hybrid radar sensor as a pole; A first calculation module, wherein the first calculation module is configured to: calculate a distance difference and an angle difference between the target object and the nearest object according to the first polar coordinates and the second polar coordinates; A second determination module, wherein the second determination module is configured to: obtain a Gaussian distribution function of the target object in the polar coordinate system by presetting a Gaussian distribution variance; A second calculation module, the second calculation module is configured to: calculate a probability result that the target object and the nearest object are identified as the same object according to the distance difference, the angle difference, the radar resolution and the Gaussian distribution function; A third determination module is configured to: determine a classification result of the target object according to the possibility result.

9. The object classification device according to claim 8, wherein: The first polar coordinates include a first distance value and a first angle value, and the second polar coordinates include a second distance value and a second angle value. The first calculation module is further configured to determine the distance difference and the angle difference between the target object and the nearest object in the following manner: Determine a difference between the first distance value and the second distance value as a distance difference between the target object and the nearest object; A difference between the first angle value and the second angle value is determined as an angle difference between the target object and the nearest object.

10. The object classification device according to claim 9, wherein: The preset Gaussian distribution variance includes a distance Gaussian distribution variance and an angle Gaussian distribution variance, and the Gaussian distribution function includes a distance Gaussian distribution function and an angle Gaussian distribution function. Wherein, the second determination module is further configured to: obtain the Gaussian distribution function of the target object in the polar coordinate system by: Adding the distance Gaussian distribution variance to the first distance value to obtain a distance Gaussian distribution function, wherein the horizontal axis of the distance Gaussian distribution function represents a predicted distance value of the target object under the influence of the distance Gaussian distribution variance, and the vertical axis of the distance Gaussian distribution function represents an occurrence probability of the predicted distance value; The angle Gaussian distribution variance is added to the first angle value to obtain an angle Gaussian distribution function, wherein the horizontal axis of the angle Gaussian distribution function represents multiple predicted angle values ​​of the target object under the influence of the angle Gaussian distribution variance, and the vertical axis of the angle Gaussian distribution function represents the predicted angle occurrence probability.

11. The object classification device according to claim 10, wherein: The radar resolution includes distance resolution and angle resolution. The second calculation module is further configured to determine the possibility result in the following manner: Determining, according to the distance difference and the distance resolution, a distance range within which the target object and the nearest object cannot be correctly classified; Calculating a first probability that the predicted distance of the target object is within the distance range according to the distance Gaussian distribution variance, the distance range, and the distance resolution, wherein the first probability indicates a possibility that the target object and the nearest object cannot be correctly classified under the distance dimension; Based on the angle difference and the angle resolution, determining an angle range in which the target object and the nearest object cannot be correctly classified; Calculating, according to the variance of the angular Gaussian distribution, the angular range, and the angular resolution, a second probability that the predicted angle of the target object is within the angular range, the second probability indicating the possibility that the target object and the nearest object cannot be correctly classified in the angular dimension; A probability result that the target object and the nearest object are identified as the same object is calculated based on the first probability and the second probability.

12. The object classification device according to claim 11, wherein: The second calculation module is further configured to determine the first probability by the following formula: In this formula, P R represents the first probability, R represents the distance difference between the target object and the nearest object, σ1 2 is the distance Gaussian distribution variance, R+α represents the upper limit of the distance range, R-α represents the lower limit of the distance range, and α represents the distance resolution; The second probability is determined by the following formula: In this formula, P θ represents the second probability, θ represents the angle difference between the target object and the nearest object, σ2 2 is the angular Gaussian distribution variance, θ+β represents the upper limit of the angular range, θ-β represents the lower limit of the angular range, and β represents the angular resolution.

13. The object classification device according to claim 12, wherein: The second calculation module is further configured to: calculate the probability result that the target object and the nearest object are identified as the same object according to the first probability and the second probability, comprising: calculating the product between the first probability and the second probability; The difference between 1 and the product is determined as a possibility result that the target object and the nearest neighbor object are identified as the same object.

14. The object classification device according to any one of claims 8 to 13, wherein: The third determination module is further configured to: determine the classification result of the target object according to the possibility result, including: Comparing the possibility result with a first preset threshold, and if the possibility result is greater than or equal to the first preset threshold, determining that the radar sensor correctly classifies the target object; If the likelihood result is less than the first preset threshold, determining whether the likelihood result is less than a second preset threshold; If the likelihood result is less than the second preset threshold, determining that the radar sensor misclassifies the target object; If the likelihood result is greater than or equal to the second preset threshold, it is determined that the radar sensor classifies the target object inaccurately.

15. An electronic device, wherein: The electronic device includes: a processor, a memory and a bus, the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate through the bus, and the machine-readable instructions are executed by the processor to execute the steps of the object classification method as described in any one of claims 1 to 7.

16. A computer-readable storage medium, wherein: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the object classification method according to any one of claims 1 to 7 are executed.

Images