Ultrasonic radar noise recognition method and system

Abstract

An ultrasonic radar noise recognition method and system are presented. The method includes: transmitting ultrasonic waves by an ultrasonic radar and receiving corresponding echoes; extracting echo data from the received echoes to form raw data; processing the raw data according to noise echo features to obtain feature data, wherein the noise echo features represent waveform features associated with echoes generated by noise; and inputting the feature data into a trained noise classifier to obtain a noise recognition result.

Description

Technical Field

[0001] This application generally relates to the field of ultrasonic radar technology, and more particularly to an ultrasonic radar noise identification method and system. Background Technology

[0002] Ultrasonic radar is widely used in various fields. For example, in the field of autonomous driving, ultrasonic radar is used in reversing radar systems or automatic parking systems. Compared with other vehicle sensors, ultrasonic radar has advantages such as strong penetration, mature and stable detection technology, and low cost.

[0003] However, ultrasonic ranging is susceptible to interference from environmental clutter or other frequencies approaching the ultrasonic radar, which can cause the ultrasonic radar to fail to accurately determine the distance to obstacles or misidentify noise as obstacles. Therefore, the detection and identification of ultrasonic radar noise is of great significance.

[0004] Current conventional ultrasonic radar noise identification methods primarily rely on improved ultrasonic wave emission and filtering hardware circuits combined with advanced denoising algorithms (such as wavelet denoising algorithms) to detect ultrasonic radar noise. However, such methods are often only effective for specific types of noise identification, while their performance is not significant for other types of noise. Furthermore, the denoising algorithm computation is complex and computationally intensive, making it unsuitable for practical engineering applications. Another common ultrasonic radar noise identification method determines the intensity of interference by detecting the number of interference pulses before the ultrasonic radar detects an obstacle, and distinguishes between obstacles and noise by adding a random emission delay during the ultrasonic sensor's emission detection process. This method can detect environmental interference before ultrasonic radar detection, but it cannot detect newly occurring interference during the ultrasonic radar detection process. Moreover, even with increased random emission delays, it is still insufficient to accurately distinguish between obstacles and noise when noise interference is strong.

[0005] Therefore, finding an ultrasonic radar noise identification method that is applicable to different noise types of ultrasonic radar and is easy to implement in engineering is a current research hotspot in the field of ultrasonic radar for automatic parking.

[0006] In view of this, it is desirable to provide an improved ultrasonic radar noise identification method and system. Summary of the Invention

[0007] The following provides a brief overview of one or more aspects to offer a basic understanding of them. This overview is not an exhaustive summary of all conceived aspects, nor is it intended to identify the key or decisive elements of all aspects, nor to define the scope of any or all aspects. Its sole purpose is to present some concepts of one or more aspects in a simplified form as an introduction to the more detailed description that follows.

[0008] This application provides an ultrasonic radar noise identification method, comprising: emitting ultrasonic waves through an ultrasonic radar and receiving corresponding echoes; extracting echo data from the received echoes to form raw data; processing the raw data according to noise echo characteristics to obtain feature data, wherein the noise echo characteristics represent waveform features associated with the echoes generated by noise; and inputting the feature data into a trained noise classifier to obtain noise identification results.

[0009] In some embodiments, the noise identification result indicates the presence of noise, and if noise is present, further indicates the noise type and / or noise intensity.

[0010] In some embodiments, the noise echo characteristics include at least one of the following: the detection distance distribution of the echo in multiple ultrasonic radar detections; the aspect ratio of the echo; the fluctuation information of the echo within a preset distance range; or secondary echo information.

[0011] In some embodiments, the noise classifier includes one of the following: support vector machine (SVM), decision tree, or logistic regression classifier.

[0012] In some embodiments, the noise classifier is trained by: emitting ultrasonic waves in a test scenario and receiving the corresponding echoes; extracting echo data from the received echoes to form a first training set; processing the first training set according to noise echo characteristics to obtain a second training set; labeling the second training set using known noise information in the test scenario to obtain a third training set; training the noise classifier using a machine learning algorithm based on the third training set; and completing the training of the noise classifier when the noise recognition accuracy reaches the expected target.

[0013] In some embodiments, achieving the desired noise recognition accuracy includes at least one of the following: a noise recognition rate higher than a first threshold; and a noise false recognition rate lower than a second threshold.

[0014] In some embodiments, the test scenario includes multiple driving scenarios and multiple noise scenarios. The multiple driving scenarios include at least one of the following: parking scenario, normal driving scenario, high-speed driving scenario, or congestion scenario. The multiple noise scenarios include at least one of the following: no noise scenario, noisy scenario with different noise types, or noisy scenario with different noise intensities.

[0015] This application also provides an ultrasonic radar noise identification system, comprising: an ultrasonic transceiver module configured to transmit ultrasonic waves and receive corresponding echoes; an echo data module configured to extract echo data from the received echoes to form raw data; a feature data module configured to process the raw data according to noise echo features to obtain feature data, wherein the noise echo features represent waveform features associated with the echoes generated by noise; and a noise identification module configured to input the feature data into a trained noise classifier to obtain a noise identification result.

[0016] In some embodiments, the noise echo characteristics include at least one of the following: the detection distance distribution of the echo in multiple ultrasonic radar detections; the aspect ratio of the echo; the fluctuation information of the echo within a preset distance range; or secondary echo information.

[0017] This application also provides an apparatus including a processor and a memory, the memory storing program instructions; the processor executes the program instructions to implement the aforementioned ultrasonic radar noise identification method.

[0018] The technical solution of this application effectively overcomes the problems of limited noise type and intensity identification and high computational load in existing technologies. The technical solution of this application can efficiently and accurately identify different noise types and intensities, and is suitable for real-time vehicle scenarios. Attached Figure Description

[0019] The features, essence, and advantages of this application will become more apparent when understood in conjunction with the accompanying drawings, which provide a detailed description. In the drawings, the same reference numerals are consistently used. It should be noted that the described drawings are schematic and non-limiting. Some components in the drawings may be enlarged and are not drawn to scale for illustrative purposes.

[0020] Figure 1 A schematic diagram of the system architecture for ultrasonic radar noise identification of this application is shown.

[0021] Figures 2-4 A schematic diagram illustrating the characteristics of various types of noise is shown.

[0022] Figure 5 An ultrasonic radar noise identification method according to various aspects of this application is shown.

[0023] Figure 6 A schematic diagram of the SVM algorithm is shown.

[0024] Figure 7 An exemplary process for training a noise classifier according to various aspects of this application is shown.

[0025] Figure 8 A block diagram of an ultrasonic radar noise identification system according to various aspects of this application is shown.

[0026] Figure 9 A block diagram of an apparatus including an ultrasonic radar noise identification system according to various aspects of this application is shown. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this application clearer, the application is further described in detail below with reference to specific embodiments and the accompanying drawings. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the described exemplary embodiments. However, it will be apparent to those skilled in the art that the described embodiments can be practiced without some or all of these specific details. In other exemplary embodiments, well-known structures have not been described in detail to avoid unnecessarily obscuring the concepts of this application. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application. Furthermore, the various aspects described in the embodiments can be combined arbitrarily without conflict.

[0028] As mentioned above, existing ultrasonic radar noise identification methods are not applicable to different noise types, and they are computationally intensive and difficult to implement. Furthermore, existing identification methods can only detect noise before ultrasonic detection, and cannot detect noise during the ultrasonic radar detection process in real time.

[0029] Therefore, this application proposes an improved ultrasonic radar noise identification method and system. The technical solution of this application can identify a wider range of noise types and can detect noise interference experienced by the ultrasonic radar in real time during ultrasonic radar detection. Furthermore, this application is applicable to engineering implementation and suitable for automotive scenarios. In real-vehicle testing, the technical solution of this application can accurately distinguish between real obstacles and noise echoes, demonstrating high accuracy.

[0030] Figure 1 A schematic diagram 100 of the system architecture for ultrasonic radar noise identification according to this application is shown.

[0031] As shown in the figure, the system architecture for ultrasonic radar noise identification in this application mainly includes two stages: the training stage and the application stage.

[0032] During the training phase, the classifier is trained using a training dataset. Specifically, training datasets for various scenarios are first acquired, features are extracted from them, and the extracted features are labeled using prior known information. Then, the labeled features are input into the classifier for training.

[0033] In a preferred embodiment, a machine learning algorithm can be used to train the classifier, as described below. Figure 7 Further description.

[0034] After training, the trained classifier is obtained.

[0035] Trained classifiers can be imported into various practical applications. For example, in the field of vehicle driving, trained classifiers can be imported into autonomous driving systems (e.g., integrated into or communicatively coupled to autonomous driving systems). During vehicle driving, real-vehicle ultrasonic radar data can be acquired and fed into the trained classifier to obtain noise recognition results.

[0036] Based on the identification results, noise and real obstacles can be distinguished. If noise is identified, its type and intensity can be further determined. If the identification results indicate that only real obstacles exist and there is no noise interference, information related to the obstacles, such as their distance, can be further determined.

[0037] It should be noted that although most of the content of this application is explained in conjunction with vehicle driving scenarios (especially parking scenarios), the technical solutions of this application are applicable to any other suitable scenarios, such as other vehicle driving scenarios or other fields of application of ultrasonic radar outside the field of vehicle driving.

[0038] Ultrasonic radar is susceptible to interference from various noise sources during operation. Based on the different noise sources and their intensity, the noise types of ultrasonic radar can be categorized as: random noise, light noise, and heavy noise.

[0039] Figures 2-4 A schematic diagram illustrating the characteristics of various types of noise is shown.

[0040] Random noise generally refers to the slight interference caused by surrounding environmental factors such as air guns, rain, and other sound sources to ultrasonic radar.

[0041] Light noise generally refers to interference from ultrasonic radar caused by ultrasonic radar on other vehicles in the vicinity or by other ultrasonic radar on the same vehicle.

[0042] Heavy noise generally refers to a situation where ultrasonic radar is significantly affected by interference from the surrounding environment, making it unable to determine the distance to actual obstacles. It should be noted that the above noise classification is merely exemplary and not restrictive. In practical implementations, other noise classification methods can also be used.

[0043] In order to identify different noise types and provide early warnings based on the intensity of noise interference, this application provides an effective ultrasonic radar noise identification mechanism based on machine learning and noise echo characteristics. By analyzing the ultrasonic radar detection data obtained from real vehicle testing, several characteristics of ultrasonic radar noise were obtained: (1) For random noise, the interference intensity of the noise source is relatively low. Due to the randomness of ultrasonic radar noise, the noise appears randomly at different detection distances; (2) For heavy noise, the interference intensity of the noise source is relatively strong. Within a certain distance range, the detection echo obtained by ultrasonic radar will have multiple peaks and troughs, that is, the detection echo fluctuates continuously, and the echo of the real obstacle is "submerged" in it; (3) For co-frequency interference generated by other ultrasonic radars, the noise echo generated by co-frequency interference has the characteristics of being "thin" and "tall". Therefore, the noise echo generated by co-frequency interference can be effectively distinguished from the echo of the real obstacle by characteristics such as distance, the ratio of echo height to echo width, and whether there is a secondary echo. At this time, the type of noise is light noise.

[0044] Figure 2 A schematic diagram 200 illustrating the characteristics of random noise is shown. For example... Figure 2 As shown, because random noise is relatively weak, the echo generated by the noise will jump at different detection distances. That is, the echo generated by the noise will randomly appear at different detection distances in different detections, such as... Figure 2 As shown in the rhombus. In contrast, the echoes generated by obstacles are concentrated at similar detection distances in different detections, such as... Figure 2 The circle in the diagram illustrates this. It can be seen that, based on the detection distance distribution of the echo during multiple detection processes, random noise and real obstacles can be distinguished.

[0045] Figure 3 A schematic diagram 300 illustrating the characteristics of heavy noise is shown. For example... Figure 3 As shown, due to the higher intensity of heavy noise, the echoes generated by the noise will exhibit significant frequency fluctuations within a certain distance range, while the echoes generated by real obstacles will be "submerged" in the continuous fluctuations of the noise echoes. Therefore, it is evident that heavy noise can be distinguished from real obstacles based on the fluctuation information of the echoes within a certain distance (e.g., fluctuation amplitude, fluctuation frequency, etc.).

[0046] Figure 4 A schematic diagram 400 illustrates the characteristics of co-channel interference noise (also referred to as light noise in this paper). When the co-channel interference is strong, the co-channel interference noise will produce a relatively "tall" and "thin" echo, such as... Figure 4As shown in the diamond diagram, the echoes generated by co-channel interference noise have a relatively large aspect ratio. In contrast, the echoes generated by obstacles have a smaller aspect ratio. Therefore, based on the aspect ratio of the echoes, co-channel interference noise can be distinguished from real obstacles.

[0047] Furthermore, noise does not produce a noticeable secondary echo, while real obstacles do. Therefore, noise and real obstacles can also be distinguished based on secondary echo information (such as the presence of a secondary echo).

[0048] Figure 5 An ultrasonic radar noise identification method 500 according to various aspects of this application is shown.

[0049] Method 500 begins at step 505. In step 505, ultrasonic waves are emitted by an ultrasonic radar and the corresponding echoes are received.

[0050] Taking automatic parking as an example, ultrasonic radar can be an on-board component or integrated into an on-board component.

[0051] In practical implementation, ultrasonic radar emits ultrasonic waves within its FOV (Field of View). When multiple ultrasonic radars are installed, their positions can be determined based on the FOV to cover a larger area. After emitting the ultrasonic waves, the ultrasonic radars can receive the reflected echoes within the FOV.

[0052] In step 510, echo data is extracted from the received echo to form raw data.

[0053] For example, echo data of interest (e.g., echo width, echo height, distance information, time information, etc.) can be extracted from the waveform signal of the received echo to form raw data.

[0054] In step 515, the original data is processed according to the noise echo characteristics to obtain feature data, wherein the noise echo characteristics represent waveform features associated with the echoes generated by the noise.

[0055] In one embodiment of this application, the noise echo characteristics may include at least one of the following: the detection distance distribution of the echo in multiple ultrasonic radar detections; the aspect ratio of the echo; the fluctuation information of the echo within a preset distance range; or secondary echo information.

[0056] As mentioned above, the echoes generated by noise and those generated by real obstacles exhibit differences in detection distance distribution, aspect ratio, fluctuation information within a preset distance, and secondary echo information during multiple detections. Therefore, based on these characteristics, noise can be distinguished from real obstacles, thereby identifying the presence of radar noise interference.

[0057] It should be noted that the noise echo characteristics described above are merely exemplary and not limiting. In specific implementations, those skilled in the art can employ other appropriate noise echo characteristics. For example, noise echo characteristics in the time domain other than those described above can be used, or noise echo characteristics in the frequency domain can be used.

[0058] In step 520, the feature data is input into the trained noise classifier to obtain the noise recognition result.

[0059] In one embodiment of this application, the noise classifier can be any suitable classifier in the field of machine learning, such as support vector machine (SVM), decision tree, logistic regression classifier, etc.

[0060] SVM is a binary classification model. Its basic idea is to find the separating hyperplane that correctly divides the training dataset and maximizes the geometric margin. The learning strategy of SVM is to maximize the margin, which can be formalized as solving a convex quadratic programming problem, and is also equivalent to minimizing a regularized hinge loss function. The SVM algorithm is an optimization algorithm for solving convex quadratic programming. SVM and its algorithm will be discussed in conjunction with... Figure 6 Further description is required.

[0061] A decision tree is a tree structure (which can be binary or non-binary) where each non-leaf node represents a test on a feature attribute, each branch represents the output of that feature attribute over a certain value range, and each leaf node represents a class. During learning, a decision tree model is built using training data based on the principle of minimizing the loss function; during prediction, the decision tree model is used to classify new data.

[0062] The idea behind logistic regression classifiers is to first fit a decision boundary (not limited to linear, but also polynomial), and then establish a relationship between this boundary and the probability of classification, thereby obtaining the probability in the case of binary classification.

[0063] In one embodiment of this application, a machine learning algorithm can be used to train the noise classifier. The training of the noise classifier will be described below. Figure 7 Further description is required.

[0064] In one embodiment of this application, the noise identification result indicates whether noise exists, and if noise exists, further indicates the noise type and / or noise intensity.

[0065] If there is no noise, it means that the received echo signals are all generated by the reflection of ultrasonic waves by real obstacles. Therefore, obstacle information (such as the obstacle's location, size, etc.) can be further determined.

[0066] If noise is present, it indicates that the received echo signal contains noise interference. Further information about the noise, such as its type and / or intensity, can then be determined.

[0067] In optional embodiments, a corresponding warning can be issued based on the noise identification results. For example, the type and / or intensity of the identified noise can be displayed to the user. In some implementations, an audible warning can also be issued if the noise intensity exceeds a preset threshold.

[0068] As can be seen from method 500, the technical solution of this application can identify different types and intensities of noise, such as random noise, light noise, and heavy noise. Simultaneously, the technical solution of this application can monitor noise in real time during ultrasonic detection, overcoming the limitation of current ultrasonic radar noise detection methods that can only be performed before detection. Furthermore, compared to denoising algorithms, the noise identification method of this application has reduced computational complexity and is easier to implement in engineering.

[0069] Based on real-vehicle testing, the technical solution of this application achieves a recognition rate of over 80% for both light and heavy noise, with a false recognition rate of less than 10%. Furthermore, the technical solution of this application also demonstrates good recognition rates for co-channel interference noise at different distances, especially for co-channel interference noise beyond 2 meters, achieving a recognition rate of over 85%. Therefore, it is evident that the technical solution of this application can accurately identify various types of ultrasonic radar noise.

[0070] Figure 6 A schematic diagram of the SVM algorithm is shown in Figure 600.

[0071] The SVM algorithm is a supervised learning model and related machine learning algorithm used in classification and regression analysis. It is a novel few-shot learning method with a solid theoretical foundation, exhibiting good robustness and the ability to solve classification and regression problems with high-dimensional features.

[0072] The basic idea of ​​the SVM algorithm is to find the separating hyperplane that correctly partitions the training dataset and maximizes the geometric margin. For example... Figure 6 As shown, w·x+b=0 is the separating hyperplane. For linearly separable datasets, there are infinitely many such hyperplanes (i.e., perceptrons), but the separating hyperplane with the largest geometric margin is unique.

[0073] For linear classification problems, the SVM algorithm steps are as follows:

[0074] Input: Training dataset T = {(x1, y1), (x2, y2), ..., (x...} N y N Where, x i ∈R n y i ∈{+1,-1},i=1,2,...N;x i Let y be the i-th eigenvector. i This is a class identifier.

[0075] Output: Separating hyperplane and classification decision function

[0076] (1) Select the penalty parameter C > 0, construct and solve the convex quadratic programming problem.

[0077]

[0078]

[0079] Where α i Using Lagrange multipliers, the optimal solution to the above programming problem is obtained.

[0080] (2) Calculation

[0081]

[0082] Choose α * A component Meet the conditions calculate:

[0083]

[0084] (3) Find the separating hyperplane

[0085] ω * ·x+b * =0

[0086] And the classification decision function: f(x) = sign(ω) * ·x+b * )

[0087] For nonlinear classification problems, a nonlinear transformation is used to convert them into linear classification problems in a feature space of a certain dimension. A linear support vector machine (SVM) is then learned in this high-dimensional feature space. In the dual problem of the linear SVM learning process, the inner product is replaced by a kernel function K(x, z), and the solution yields the nonlinear SVM. Specifically, the steps of the nonlinear SVM algorithm are as follows:

[0088] Input: Training dataset T = {(x1, y1), (x2, y2), ..., (x...} N y N Where, x i ∈R n y i ∈{+1,-1},i=1,2,...N;

[0089] Output: Separating hyperplane and classification decision function

[0090] (1) Select a suitable kernel function K(x, z) and a penalty parameter C > 0, construct and solve the convex quadratic programming problem.

[0091]

[0092] Obtain the optimal solution

[0093] (2) Calculation

[0094] Choose α * A component Meet the conditions calculate

[0095]

[0096] (3) Classification decision function:

[0097] It should be noted that the specific implementation of the SVM algorithm is well-known in the field of machine learning, and will not be elaborated here.

[0098] Figure 7 An exemplary process 700 for training a noise classifier according to various aspects of this application is shown.

[0099] Process 700 begins at step 705. In step 705, ultrasonic waves are emitted in the test scenario and the corresponding echoes are received.

[0100] In one embodiment of this disclosure, the test scenario may include multiple driving scenarios and multiple noise scenarios. For example, the multiple driving scenarios may include at least one of the following: parking scenario, normal driving scenario, high-speed driving scenario, or congested scenario. The multiple noise scenarios may include at least one of the following: no noise scenario, noisy scenario with different noise types, or noisy scenario with different noise intensities.

[0101] It should be noted that the above division of driving and noise scenarios is merely exemplary and not restrictive. In specific implementations, those skilled in the art can make any other suitable scenario divisions based on the actual situation.

[0102] During training, each driving scenario can be combined with multiple noise scenarios. For example, for parking scenarios, you can set up parking scenarios with no noise, parking scenarios with random noise, parking scenarios with light noise, parking scenarios with heavy noise, and so on. By making the test scenarios cover a variety of possible driving and noise scenarios, you can better simulate real-world conditions, thereby improving the training effect.

[0103] In step 710, echo data is extracted from the received echoes to form the first training set.

[0104] For example, the first training set can have the following form:

[0105] T1={(x Eco_Num_1 y Eco_Heigt_1 ), (x Eco_Num_2 y Eco_Heigt_2 ), ..., (x Eco_Num_N y Eco_Heigt_N )}

[0106] Where T1 represents the first training set, x Eco_Num_i y represents the detection distance of the noise-generated echo at different detection times. Eco_Height_i This indicates the corresponding echo height. It should be noted that this is merely an example of the extracted echo data and not a limitation. In specific implementations, those skilled in the art can extract different forms of echo data from the received echoes as needed.

[0107] In step 715, the first training set is processed according to the noise echo characteristics to obtain the second training set.

[0108] In one embodiment of this disclosure, the noise echo characteristics may include at least one of the following: the detection distance distribution of the echo in multiple ultrasonic radar detections; the aspect ratio of the echo; the fluctuation information of the echo within a preset distance range; or secondary echo information.

[0109] For example, the aforementioned noise echo features can be extracted from the first training set to form the second training set. In a specific implementation, one or more noise echo features can be extracted from the aforementioned noise echo features.

[0110] In one example implementation, the second training set can have the following form:

[0111] T2={(x1, y1), (x2, y2),..., (x N y N )}

[0112] Where T2 represents the second training set, (x i y i() represents the echo features extracted at different detection times. As an example, x i It can represent the echo distance value detected in different detections, y i It can represent the aspect ratio of the corresponding echo.

[0113] It should be noted that the above only shows one example of the second training set. In a specific implementation, the second training set can take different forms depending on the selected noise echo features. For example, if the selected noise echo features are the fluctuation information of the echo within a preset distance range, then x in the above formula will be... i It can represent the echo distance value detected in different detections, while y i It can represent the fluctuation amplitude of the corresponding echo.

[0114] As mentioned above, based on the noise echo characteristics described above, noise can be distinguished from real obstacles, thereby identifying whether radar noise interference exists.

[0115] In step 720, the second training set is labeled using known noise information from the test scenario to obtain the third training set.

[0116] Since the noise information in the test scenario is known, this prior knowledge can be used to label the second training set. For example, the data in the second training set can be labeled as noisy or noiseless. For noisy cases, further noise information can be labeled, such as noise source information, noise type, and / or noise intensity. The labeled second training set constitutes the third training set.

[0117] In one example implementation, the third training set can have the following form:

[0118] T3={(x1, y1, z1), (x2, y2, z2),..., (x N y N , z N )}

[0119] Where T3 represents the third training set, (x i y i ) represents the echo features extracted at different detection times (e.g., x i y represents the echo distance value detected in different detections. i (This indicates the aspect ratio of the corresponding echo), while z i This can represent noise information (e.g., the presence or absence of noise, noise type, and / or noise intensity). In step 725, a noise classifier is trained using a machine learning algorithm based on the third training set.

[0120] For example, when the noise classifier is a linear SVM, the SVM algorithm can be used to train the noise classifier. After training, the feature separating hyperplane ω is obtained. * .x+b * =0 and the classification decision function f(x) = sign(ω) * ·x+b * If the noise classifier is of another type, the corresponding machine learning algorithm can be used for training.

[0121] In step 730, when the noise recognition accuracy reaches the expected target, the training of the noise classifier is completed.

[0122] In one embodiment of this application, achieving the expected target in noise recognition accuracy may include at least one of the following: the noise recognition rate is higher than a first threshold; and the noise false recognition rate is lower than a second threshold.

[0123] In practical implementation, the first and second thresholds can be set empirically, experimentally, or theoretically. Furthermore, it should be noted that the noise recognition rate and noise false recognition rate are provided as examples of noise recognition accuracy. In actual implementation, those skilled in the art can use other forms of measurement to characterize noise recognition accuracy as needed.

[0124] For example, noise recognition rate and noise false recognition rate can be incorporated into the loss function during training. When the loss function meets a threshold condition, the noise recognition accuracy can be considered to have reached the expected goal, and training can then be terminated.

[0125] Figure 8 A block diagram of an ultrasonic radar noise identification system 800 according to various aspects of this application is shown.

[0126] See Figure 8 The system 800 may include an ultrasonic transceiver module 802, an echo data module 804, a feature data module 806, and a noise identification module 808. Each of these modules may be directly or indirectly connected to or communicate with each other on one or more buses 810.

[0127] The ultrasonic transceiver module 802 can transmit ultrasonic waves and receive the corresponding echoes.

[0128] In one embodiment of this application, the ultrasonic transceiver module 802 may be an ultrasonic radar installed in a vehicle. The ultrasonic radar can emit ultrasonic waves within the field of view (FOV) and receive the reflected echoes.

[0129] The echo data module 804 can extract echo data from the received echo to form raw data.

[0130] The feature data module 806 can process the original data according to the noise echo characteristics to obtain feature data, wherein the noise echo characteristics represent the waveform characteristics associated with the echo generated by the noise.

[0131] In one embodiment of this application, the noise echo characteristics include at least one of the following: the detection distance distribution of the echo in multiple ultrasonic radar detections; the aspect ratio of the echo; the fluctuation information of the echo within a preset distance range; or secondary echo information.

[0132] In one embodiment of this application, the noise classifier includes one of the following: SVM, decision tree, or logistic regression classifier.

[0133] In one embodiment of this application, the noise classifier is trained by the following operations: emitting ultrasonic waves in a test scenario and receiving corresponding echoes; extracting echo data from the received echoes to form a first training set; processing the first training set according to noise echo characteristics to obtain a second training set; labeling the second training set using known noise information in the test scenario to obtain a third training set; training the noise classifier using a machine learning algorithm based on the third training set; and completing the training of the noise classifier when the noise recognition accuracy reaches the expected target.

[0134] The noise recognition module 808 can input feature data into a trained noise classifier to obtain noise recognition results.

[0135] In one embodiment of this application, the noise identification result indicates whether noise exists, and if noise exists, further indicates the noise type and / or noise intensity.

[0136] Although Figure 8 The diagram illustrates specific modules of system 800, but it should be understood that these modules are exemplary and not limiting. In different implementations, one or more of these modules may be combined, split, removed, or additional modules may be added. For example, in some implementations, the ultrasonic transceiver module 802 may be split into a transmitting module and a receiving module (not shown in the diagram). In some implementations, system 800 may also include additional modules. For example, system 800 may include a warning module (not shown in the diagram) that can issue corresponding warnings based on noise recognition results. In vehicle driving scenarios, this can further improve safety and enhance the user experience.

[0137] Figure 9 A block diagram of an apparatus 900 including an ultrasonic radar noise identification system according to various aspects of this application is shown.

[0138] This device illustrates a general hardware environment in which this application can be applied according to exemplary embodiments thereof.

[0139] Now refer to Figure 9 Device 900 is described, which is an exemplary embodiment of a hardware device that can be applied to various aspects of this application. Device 900 can be any machine configured to perform processing and / or computation, and can be, but is not limited to, a workstation, server, desktop computer, laptop computer, tablet computer, personal digital assistant (PDA), smartphone, or any combination thereof.

[0140] Device 900 may include components that can be connected to or communicate with bus 912 via one or more interfaces. For example, device 900 may include bus 912, processor 902, memory 904, input device 908, and output device 910, etc.

[0141] Processor 902 can be any type of processor and may include, but is not limited to, general-purpose processors and / or special-purpose processors (e.g., special processing chips), intelligent hardware devices (e.g., general-purpose processors, DSPs, CPUs, microcontrollers, ASICs, FPGAs, programmable logic devices, discrete gate or transistor logic components, discrete hardware components, or any combination thereof). In some cases, processor 902 may be configured to use a memory controller to operate a memory array. In other cases, a memory controller (not shown) may be integrated into processor 902. Processor 902 may be responsible for managing the bus and general processing, including executing software stored in memory. Processor 902 may also be configured to perform various functions related to ultrasonic radar noise identification as described herein. For example, processor 902 may be configured to: emit ultrasonic waves via ultrasonic radar and receive corresponding echoes; extract echo data from the received echoes to form raw data; process the raw data according to noise echo characteristics to obtain feature data, wherein the noise echo characteristics represent waveform features associated with the echoes generated by noise; and input the feature data into a trained noise classifier to obtain noise identification results.

[0142] Memory 904 can be any storage device capable of storing data. Memory 804 may include, but is not limited to, disk drives, optical storage devices, solid-state storage, floppy disks, hard disks, magnetic tapes or any other magnetic media, optical discs or any other optical media, ROM (read-only memory), RAM (random access memory), cache memory and / or any other memory chip or cartridge, and / or any other medium from which a computer can read data, instructions and / or code. Memory 904 may store computer-executable software 906 including computer-readable instructions that, when executed, cause a processor to perform the various functions described herein related to ultrasonic radar noise identification.

[0143] Input device 908 can be any type of device that can be used to input information.

[0144] The output device 910 can be any type of device used for outputting information. In one case, the output device 910 can be any type of output device capable of displaying information.

[0145] The detailed description above, in conjunction with the accompanying drawings, describes examples but does not represent all examples that can be implemented or fall within the scope of the claims. The terms "example" and "exemplary" are used in this specification to mean "serving as an example, instance, or illustration" and do not imply "superiority or superiority over other examples."

[0146] Throughout this specification, the terms "an embodiment" or "an embodiment" mean that a particular feature, structure, or characteristic described in connection with that embodiment is included in at least one embodiment of this application. Therefore, the use of these phrases may refer to more than one embodiment. Furthermore, the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0147] The preceding description is provided to enable any person skilled in the art to practice the various aspects described herein. Various modifications to these aspects will readily be understood by those skilled in the art, and the universal principles defined herein can be applied to other aspects. Therefore, the claims are not intended to be limited to the aspects shown herein, but should be granted the full scope consistent with the language of the claims, wherein references to the singular form of an element, unless specifically stated otherwise, are not intended to mean “one and only one,” but rather “one or more.” Unless specifically stated otherwise, the term “some” refers to one or more. All structural and functional equivalents of the elements of the various aspects described throughout this application that are now or hereafter known to a person skilled in the art are expressly incorporated herein by reference and are intended to be covered by the claims.

[0148] It should also be noted that these embodiments may be described as processes depicted as flowcharts, flow diagrams, structure diagrams, or block diagrams. Although a flowchart may describe the operations as a sequential process, many of these operations can be executed in parallel or concurrently. Furthermore, the order of these operations can be rearranged.

[0149] While various embodiments have been described and illustrated, it should be understood that the embodiments are not limited to the precise configurations and components described above. Various modifications, substitutions, and improvements that will be apparent to those skilled in the art can be made to the arrangement, operation, and details of the apparatus disclosed herein without departing from the scope of the claims.

Claims

1. An ultrasonic radar noise identification method, comprising: transmitting ultrasonic waves and receiving corresponding echoes by an ultrasonic radar; extracting echo data from the received echoes to form raw data; processing the raw data according to noise echo features to obtain feature data, wherein the noise echo features represent waveform features associated with echoes generated by noise; and inputting the feature data into a trained noise classifier to obtain a noise identification result, wherein the noise echo features comprise at least one of: a distribution of detection distances of echoes in multiple ultrasonic radar detections; a height-width ratio of echoes; fluctuation information of echoes within a preset distance range; or second echo information.

2. The method of claim 1, wherein, The noise identification result indicates whether there is noise, and further indicates a noise type and / or a noise intensity in the case where there is noise.

3. The method of claim 1, wherein, The noise classifier comprises one of: a support vector machine (SVM), a decision tree, or a logistic regression classifier.

4. The method of claim 1, wherein, The noise classifier is trained by: transmitting ultrasonic waves and receiving corresponding echoes in a test scenario; extracting echo data from the received echoes to form a first training set; processing the first training set according to noise echo features to obtain a second training set; labeling the second training set using known noise information in the test scenario to obtain a third training set; training the noise classifier based on the third training set using a machine learning algorithm; and when a noise identification accuracy reaches an expected target, completing the training of the noise classifier. The noise identification accuracy reaching the expected target comprises at least one of:

5. The method of claim 4, wherein, a noise identification rate being higher than a first threshold; and a noise misidentification rate being lower than a second threshold. The test scenario comprises multiple driving scenarios and multiple noise scenarios, the multiple driving scenarios comprising at least one of: a parking scenario, a normal driving scenario, a high-speed driving scenario, or a congestion scenario, and the multiple noise scenarios comprising at least one of: a noise-free scenario, a noisy scenario with different noise types, or a noisy scenario with different noise intensities.

6. The method of claim 4, wherein, 7. An ultrasonic radar noise identification system, comprising: an ultrasonic transceiver module configured to transmit ultrasonic waves and receive corresponding echoes; an echo data module configured to extract echo data from the received echoes to form raw data; a feature data module configured to process the raw data according to noise echo features to obtain feature data, wherein the noise echo features represent waveform features associated with echoes generated by noise; and a noise identification module configured to input the feature data into a trained noise classifier to obtain a noise identification result, wherein the noise echo features comprise at least one of: a distribution of detection distances of echoes in multiple ultrasonic radar detections; a height-width ratio of echoes; fluctuation information of echoes within a preset distance range; or second echo information. ​ ​ 8. An ultrasonic radar noise identification device comprising a processor and a memory, the memory storing program instructions; the processor executing the program instructions to implement the method of any one of claims 1 to 6.

Images