Ultrasonic radar signal identification method and system

Abstract

The invention relates to an ultrasonic radar signal identification method and system, and belongs to the technical field of vehicle perception identification, and the method comprises the steps: obtaining a first original signal detected by an ultrasonic radar; identifying the first original signal through a pre-constructed noise classification model to obtain a first ultrasonic radar signal; calculating the power spectral density of the first ultrasonic radar signal, and dynamically updating a noise baseline according to the power spectral density to obtain a second ultrasonic radar signal; and identifying an effective ultrasonic radar signal of the second ultrasonic radar signal through a pre-constructed environment filtering mapping model to obtain an ultrasonic radar signal. According to the invention, interference noise in the environment can be effectively removed, and the accuracy and reliability of ultrasonic radar signal identification are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of signal recognition, and in particular to an ultrasonic radar signal recognition method and system. BACKGROUND

[0002] In the field of modern automobile intelligent driving and industrial automation, environmental perception capability has become a core factor determining the safety and reliability of the perception vehicle and automation system. Ultrasonic radar has become one of the core perception devices in key scenarios such as distance detection and obstacle recognition, due to its unique advantages of controllable cost, high precision in short distance detection, and being unaffected by light and adverse weather (such as fog and rain).

[0003] Ultrasonic radar measures distance or identifies obstacles by emitting and receiving ultrasonic signals. During the propagation of ultrasonic waves in the environment, due to the sensitivity of ultrasonic radar to environmental noise, it is easily disturbed in a noisy environment, leading to misidentification and affecting the accuracy and operational safety of the system. To address the problem of environmental interference, traditional methods mainly use hardware filtering and frequency adjustment for filtering processing, which can reduce the interference of environmental noise, but cannot eliminate the misidentification phenomenon, and it is difficult to meet the requirements of automobile intelligent driving and industrial automation for the stability and accuracy of the perception system. SUMMARY

[0004] To solve the above problems existing in the prior art, the present application provides an ultrasonic radar signal recognition method and system, which uses a pre-constructed noise model to identify and classify different types of sound waves, and distinguishes ultrasonic radar signals from environmental noise; dynamically updates the noise baseline according to the power spectral density, improves the accuracy of the noise baseline, effectively removes interfering sound waves, and retains effective ultrasonic radar signals; based on a pre-constructed environmental filtering mapping model, effective features are extracted to improve the accuracy and reliability of ultrasonic radar signal recognition.

[0005] In a first aspect, an ultrasonic radar signal recognition method is provided, applied to a vehicle. The method comprises: obtaining a first original signal detected by an ultrasonic radar; identifying the first original signal to obtain a first ultrasonic radar signal through a pre-constructed noise classification model; calculating the power spectral density of the first ultrasonic radar signal, and dynamically updating the noise baseline according to the power spectral density to obtain a second ultrasonic radar signal; identifying the effective ultrasonic radar signal of the second ultrasonic radar signal through a pre-constructed environmental filtering mapping model to obtain an ultrasonic radar signal.

[0006] In the embodiments of the present application, the first ultrasonic radar signal is obtained by classifying and identifying the first original signal detected by the ultrasonic radar through the pre-constructed noise classification model, the ultrasonic radar signal is distinguished from the environmental noise (for example, mechanical vibration noise, wind noise, electromagnetic interference noise, etc.), the accurate identification and exclusion of the interference sound wave are realized. The second ultrasonic radar signal is obtained by updating the noise baseline according to the calculated power spectral density, the stability of the noise baseline is improved, and the detectability of the ultrasonic radar signal is improved. The ultrasonic radar signal is obtained by identifying the effective ultrasonic radar signal of the second ultrasonic radar signal through the pre-constructed environmental filtering mapping model, the interference sound wave is effectively removed, the effective ultrasonic radar signal is retained, and the identification accuracy of the ultrasonic radar signal is improved.

[0007] In combination with the first aspect, in some implementations of the first aspect, the method for constructing the noise classification model is: obtaining environmental noise data under different environmental conditions; preprocessing and feature extraction are performed on the environmental noise data to obtain different environmental noise features; different environmental noise features are formed into a data set, and are divided into a training set, a validation set and a test set; an initial noise classification model is obtained by training a machine learning model through the training set; an evaluation result is obtained by evaluating the initial noise classification model through the validation set and the test set; the initial noise classification model parameters are adjusted according to the evaluation result to obtain the noise classification model.

[0008] In the embodiments of the present application, the noise classification model is constructed in the manner of training and verifying the machine learning model (for example, a deep learning model, a random forest model, etc.) with the environmental noise data, the noise classification model is optimized according to the evaluation result, and the classification performance of the noise classification model can be improved. The environmental noise can be accurately identified and excluded through the constructed noise classification model, and the ultrasonic radar signal is distinguished from the environmental noise.

[0009] In combination with the first aspect, in some implementations of the first aspect, the method for calculating the power spectral density of the first ultrasonic radar signal is: the first signal is obtained by performing detrending and mean removal processing on the first ultrasonic radar signal; the second signal is obtained by transforming the first signal from the time domain to the frequency domain through Fourier transform; the second signal is divided into multiple signals, the power spectral density of each signal is calculated, and the average value of the power spectral densities of the multiple signals is taken as the power spectral density of the first ultrasonic radar signal.

[0010] In the embodiments of the present application, when calculating the power spectral density, first, the first signal is obtained through de-trending and de-meaning preprocessing to reduce the influence of low-frequency noise. Then, the first signal is transformed from the time domain to the frequency domain through Fourier transform to obtain the second signal. The second signal obtained through Fourier transform can intuitively disassemble the signal frequency composition to retain the target frequency signal and remove the interference frequency signal. The second signal is divided into multiple segments of signals, the power spectral density of each segment of signal is calculated respectively, and the mean value is taken as the power spectral density of the first ultrasonic radar signal, thereby improving the accuracy of the power spectral density calculation.

[0011] In combination with the first aspect, in some implementations of the first aspect, the method for updating the noise baseline according to the power spectral density to obtain the second ultrasonic radar signal is: calculating the power spectral density of the signal in a preset sliding window; judging whether the power spectral density is greater than or equal to the current noise baseline; when the power spectral density is greater than or equal to the current noise baseline, taking the current noise baseline as the new noise baseline; and when the power spectral density is less than the current noise baseline, taking the power spectral density of the signal as the new noise baseline.

[0012] In the embodiments of the present application, the sliding window technology is used to calculate the power spectral density of the signal, and the noise baseline is determined by comparing the size of the power spectral density and the current noise baseline. The noise baseline can be dynamically updated, and the stability of the noise baseline is improved.

[0013] In combination with the first aspect, in some implementations of the first aspect, the environmental parameters of the environmental filtering mapping model include the signal-to-noise ratio, the noise bandwidth, the burst noise duration, and the effective ultrasonic radar signal, and the filtering parameters of the environmental filtering mapping model include the convergence factor, the filter length, and the filter coefficient. The method for adjusting the filtering parameters of the environmental filtering mapping model is: judging whether the signal-to-noise ratio is less than a set signal-to-noise ratio, and increasing the convergence factor when the signal-to-noise ratio is less than the set signal-to-noise ratio; judging whether the noise bandwidth is greater than a set bandwidth, and shortening the filter length when the noise bandwidth is greater than the set bandwidth; freezing the filter coefficient update when the environmental filtering mapping model detects the effective ultrasonic radar signal.

[0014] In the embodiments of the present application, the size of the signal-to-noise ratio is judged to be greater than or less than the preset signal-to-noise ratio, the convergence factor is adjusted, and when the signal-to-noise ratio is less than the preset signal-to-noise ratio, the convergence factor is increased to accelerate the convergence of the filter. The size of the noise bandwidth is judged to be greater than or less than the preset bandwidth, the filter length is adjusted, and when the noise bandwidth is greater than the preset bandwidth, the filter length is shortened to reduce the processing delay. When the effective ultrasonic radar signal is detected by the environmental filtering mapping model, the filter coefficient update is frozen to prevent the effective ultrasonic radar signal from being mistakenly suppressed, thereby improving the identification accuracy of the effective ultrasonic radar signal.

[0015] In combination with the first aspect, in some implementations of the first aspect, when the burst noise duration is greater than the preset time, the environmental filtering mapping model uses a radiological projection algorithm for filtering.

[0016] In the embodiments of the present application, when the burst noise duration is greater than the preset time, the environmental filtering mapping model uses a radiological projection algorithm for filtering, thereby improving the stability of the environmental filtering mapping model filtering and ensuring the accuracy of identifying effective signals.

[0017] In combination with the first aspect, in some implementations of the first aspect, the identification method further comprises: performing effectiveness evaluation on the ultrasonic radar signal; and the effectiveness evaluation method comprises: acquiring a second original signal collected by a camera and a third original signal detected by a laser radar; spatially and temporally aligning the second original signal, the third original signal, and the ultrasonic radar signal to obtain a camera initial signal and a laser radar initial signal; performing motion blur compensation on the camera initial signal to obtain a camera signal, and performing rain and fog compensation on the laser radar initial signal to obtain a laser radar signal; evaluating the comprehensive confidence of the ultrasonic radar signal, the camera signal, and the laser radar signal by using a pre-constructed confidence evaluation model; judging whether the ultrasonic radar signal is an effective signal according to the comprehensive confidence.

[0018] In the embodiments of the present application, the effectiveness of the ultrasonic radar signal is judged by the comprehensive confidence evaluated by the pre-constructed confidence evaluation model in combination with the signals collected by the camera and the laser radar, which can verify the authenticity of the ultrasonic radar signal and reduce the false recognition rate.

[0019] In combination with the first aspect, in some implementations of the first aspect, the method for spatial and temporal alignment is: temporally aligning the second original signal, the third original signal, and the ultrasonic radar signal based on a PTP protocol; transforming the spatial coordinates of the camera, the laser radar, and the ultrasonic radar to the same coordinate system.

[0020] In the embodiments of the present application, the time reference of the data collected by multiple sensors is unified through the Precision Time Protocol (PTP), so that the "time stamp" of the data collected by different sensors is based on the same clock, and the time deviation is eliminated. After time alignment, the "spatial coordinates" of different sensors (i.e., ultrasonic radar, camera and laser radar) are mapped to the same coordinate system (usually radar coordinate system or vehicle body coordinate system), and the spatial deviation caused by the installation position and viewing angle difference of the sensors is eliminated.

[0021] In combination with the first aspect, in some implementations of the first aspect, the method for performing motion blur compensation on the camera initial signal to obtain the camera signal is: obtaining a vehicle speed; determining whether the vehicle speed is greater than a set vehicle speed; when the vehicle speed is greater than the set vehicle speed, performing deconvolution restoration on the camera initial signal based on the constructed point spread function matrix; performing feature enhancement on the deconvolution-restored camera initial signal to obtain the camera signal.

[0022] In the embodiments of the present application, whether the vehicle speed is greater than the set vehicle speed is used as a trigger condition for motion blur compensation. When the vehicle speed is greater than the set vehicle speed, the deconvolution restoration is performed on the camera initial signal based on the constructed point spread function matrix, and the feature enhancement is performed on the deconvolution-restored camera initial signal to improve the target contour clarity, so as to realize the motion blur compensation on the camera initial signal, improve the accuracy of the camera signal, and solve the recognition failure problem caused by the camera image blur in the high-speed scene.

[0023] In combination with the first aspect, in some implementations of the first aspect, the method for performing rain and fog compensation on the laser radar initial signal to obtain the laser radar signal is: obtaining a meteorological environment parameter related to rain and fog; determining whether the meteorological environment parameter is within a set meteorological environment parameter range; when the meteorological environment parameter is outside the set meteorological environment parameter range, performing reverse intensity compensation on the laser radar initial signal based on a constructed laser radar attenuation model to obtain the laser radar initial signal in a non-attenuation state; filtering the laser radar initial signal in the non-attenuation state to remove signals with an intensity less than a set intensity threshold to obtain the laser radar signal.

[0024] In the embodiments of the present application, whether the meteorological environment parameter related to rain and fog is within the set meteorological environment parameter range is taken as a trigger condition for rain and fog compensation. When the meteorological environment parameter exceeds the set meteorological environment parameter range, the initial signal of the laser radar is compensated in reverse intensity based on the constructed laser radar attenuation model to restore the signal intensity in the non-attenuation state. After the initial signal of the laser radar is compensated in reverse intensity, some false signals (such as signals reflected only by raindrops) caused by rain and fog scattering may still remain, so the noise signals are removed by removing the signals with an intensity less than a set intensity threshold, and the real target laser radar signals are retained.

[0025] In a second aspect, an ultrasonic radar signal identification system is provided, which is applied to a vehicle. The system comprises: An acquisition module is configured to acquire a first original signal detected by an ultrasonic radar; A noise classification module is provided with a noise classification model, which is configured to identify the first original signal to obtain a first ultrasonic radar signal; A calculation module is configured to calculate the power spectral density of the first ultrasonic radar signal; An update module is configured to dynamically update a noise baseline according to the power spectral density to obtain a second ultrasonic radar signal; An identification module is provided with an environment filtering mapping model, which is configured to identify the effective ultrasonic radar signal of the second ultrasonic radar signal to obtain an ultrasonic radar signal.

[0026] In combination with the second aspect, in some implementations of the second aspect, the data acquisition module is further configured to acquire environment noise data under different environmental conditions.

[0027] In combination with the second aspect, in some implementations of the second aspect, the identification system further comprises a noise classification model construction module, which is configured to construct a noise classification model.

[0028] In combination with the second aspect, in some implementations of the second aspect, the noise classification model construction module comprises: A preprocessing submodule is configured to preprocess the environment noise data; A feature extraction submodule is configured to extract features from the preprocessed environment noise data to obtain different environment noise features; A data set generation submodule is configured to form a data set from the different environment noise features, and divide the data set into a training set, a validation set, and a test set; A training and validation submodule is configured to train a machine learning model through the training set to obtain an initial noise classification model, and evaluate the initial noise classification model through the validation set and the test set to obtain an evaluation result; The parameter adjusting sub-module adjusts initial noise classification model parameters according to the evaluation result to obtain the noise classification model.

[0029] With reference to the second aspect, in some implementations of the second aspect, the recognition system further includes an environment filtering mapping model construction module configured to construct an environment filtering mapping model.

[0030] With reference to the second aspect, in some implementations of the second aspect, the data acquisition module is further configured to acquire a second original signal collected by a camera and a third original signal detected by a laser radar, and to acquire a vehicle speed and a meteorological environment parameter related to rain and fog.

[0031] With reference to the second aspect, in some implementations of the second aspect, the recognition system further includes a space-time alignment module configured to perform space-time alignment on the second original signal, the third original signal, and an ultrasonic radar signal to obtain a camera initial signal and a laser radar initial signal.

[0032] With reference to the second aspect, in some implementations of the second aspect, the recognition system further includes a judgment module configured to judge whether the vehicle speed is greater than a set vehicle speed and whether the meteorological environment parameter is within a set range of meteorological environment parameters.

[0033] With reference to the second aspect, in some implementations of the second aspect, the recognition system further includes a compensation module configured to, when the vehicle speed is greater than the set vehicle speed, perform deconvolution restoration on the camera initial signal based on a constructed point spread function matrix using a linear filtering algorithm or an iterative nonlinear algorithm; and perform feature enhancement on the deconvolution-restored camera initial signal to obtain a camera signal.

[0034] With reference to the second aspect, in some implementations of the second aspect, the compensation module is further configured to, when the meteorological environment parameter is outside the set range of meteorological environment parameters, perform reverse intensity compensation on the laser radar initial signal based on a constructed laser radar attenuation model to obtain a laser radar initial signal in a non-attenuation state; and filter the laser radar initial signal in the non-attenuation state to remove signals with an intensity less than a set intensity threshold to obtain a laser radar signal.

[0035] With reference to the second aspect, in some implementations of the second aspect, the recognition system further includes a confidence evaluation model construction module configured to construct a confidence evaluation model.

[0036] With reference to the second aspect, in some implementations of the second aspect, the recognition system further comprises an evaluation module configured to evaluate the comprehensive confidence of the ultrasonic radar signal, the camera signal and the laser radar signal by using a pre-constructed confidence evaluation model.

[0037] With reference to the second aspect, in some implementations of the second aspect, the judging module is further configured to judge whether the ultrasonic radar signal is a valid signal according to the comprehensive confidence.

[0038] In a third aspect, a computer program product is provided, which comprises computer program code, and when the computer program code is run on a computer, the computer program code causes the computer to execute the ultrasonic radar signal recognition method of the first aspect.

[0039] In a fourth aspect, a computer readable storage medium is provided, which stores computer program code, and when the computer program code is executed by one or more processors, the computer program code causes an apparatus comprising the one or more processors to execute the ultrasonic radar signal recognition method of the first aspect.

[0040] In a fifth aspect, an embodiment of the present application provides a chip system, which comprises a processor, and the processor is configured to invoke computer program or computer instruction stored in a memory, so as to cause the processor to execute the ultrasonic radar signal recognition method of the first aspect.

[0041] In a sixth aspect, an embodiment of the present application provides an electronic device, which comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the electronic device implements the ultrasonic radar signal recognition method of the first aspect.

[0042] In a seventh aspect, a vehicle is provided. The vehicle comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the processor implements the ultrasonic radar signal recognition method of the first aspect; or the vehicle comprises the ultrasonic radar signal recognition system of the second aspect, or the computer readable storage medium of the fourth aspect, or the chip system of the fifth aspect, or the electronic device of the sixth aspect.

[0043] The technical scheme provided by the embodiments of the present application has at least the following beneficial effects: The ultrasonic wave radar signal identification method and system provided by the embodiment of the application are applied to a vehicle. The first original signal detected by the ultrasonic wave radar is classified and identified by a pre-constructed noise classification model to obtain a first ultrasonic wave radar signal, the ultrasonic wave radar signal is distinguished from environmental noise, accurate identification and exclusion of interference sound waves are realized. The second ultrasonic wave radar signal is obtained by updating the noise baseline according to the calculated power spectral density, the stability of the noise baseline is improved, and the detectability of the ultrasonic wave radar signal is improved. The effective ultrasonic wave radar signal of the second ultrasonic wave radar signal is identified by a pre-constructed environmental filtering mapping model to obtain an ultrasonic wave radar signal, interference sound waves are effectively removed, effective ultrasonic wave radar signals are retained, and the identification accuracy and reliability of the ultrasonic wave radar signal are improved.

[0044] The above description is only a summary of the technical solutions of the application. In order to more clearly understand the technical means of the application, the content of the specification can be implemented, and in order to make the above and other purposes, characteristics and advantages of the application more obvious, the specific embodiments of the application are described below. BRIEF DESCRIPTION OF DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the application, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0046] Figure 1 The ultrasonic wave radar signal identification method flowchart of the embodiment of the application.

[0047] Figure 2 The construction method flowchart of the noise classification model of the embodiment of the application.

[0048] Figure 3 The method flowchart for calculating the power spectral density of the first ultrasonic wave radar signal of the embodiment of the application.

[0049] Figure 4 The method flowchart for updating the noise baseline according to the power spectral density to obtain the second ultrasonic wave radar signal of the embodiment of the application.

[0050] Figure 5 The method flowchart for adjusting the filtering parameters of the environmental filtering mapping model of the embodiment of the application.

[0051] Figure 6 The method flowchart for the effectiveness evaluation of the embodiment of the application.

[0052] Figure 7A method flowchart for time-space alignment of an embodiment of the present application.

[0053] Figure 8 A method flowchart for motion blur compensation of an initial camera signal to obtain a camera signal of an embodiment of the present application.

[0054] Figure 9 A method flowchart for rain and fog compensation of an initial laser radar signal to obtain a laser radar signal of an embodiment of the present application.

[0055] Figure 10 A method flowchart for judging whether an ultrasonic radar signal is a valid signal according to a comprehensive confidence level of an embodiment of the present application.

[0056] Figure 11 An architecture diagram of an ultrasonic radar signal recognition system of an embodiment of the present application.

[0057] Figure 12 A structure diagram of a noise classification model construction module of an embodiment of the present application.

[0058] Figure 13 An architecture diagram of a vehicle of an embodiment of the present application.

[0059] In the figure, 100, an ultrasonic radar signal recognition system, 101, an acquisition module, 102, a noise classification module, 1021, a noise classification model, 103, a calculation module, 104, an update module, 105, a recognition module, 1051, an environment filtering mapping model, 106, a noise classification model construction module, 1061, a preprocessing submodule, 1062, a feature extraction submodule, 1063, a dataset generation submodule, 1064, a training and verification submodule, 1065, a parameter adjustment submodule, 107, an environment filtering mapping model construction module, 108, a time-space alignment module, 109, a judgment module, 110, a compensation module, 111, a confidence level evaluation model construction module, 112, an evaluation module, 200, a vehicle, 201, a memory, 202, a processor, 203, a computer program. DETAILED DESCRIPTION

[0060] In order to make the technical problems to be solved by the present application, technical solutions and beneficial effects more clearly understood, the present application will be further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0061] The prefix words such as "first", "second" in the embodiments of the present application are only used to distinguish different description objects, and have no limiting effect on the position, order, priority, quantity or content of the described objects. The use of ordinal words such as prefixes in the embodiments of the present application does not constitute a limitation on the described objects, and the description of the described objects should be referred to the description of the context in the claims or embodiments, and should not constitute redundant limitation because of the use of such prefix words. In addition, in the description of the embodiments, unless otherwise stated, the meaning of "multiple" is two or more.

[0062] The technical solutions in the embodiments of the present application will be described below in combination with the drawings in the embodiments of the present application. In the description of the embodiments of the present application, unless otherwise stated, " / " represents the meaning of or, for example, A / B can represent A or B; "and / or" in this paper only describes the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B, which means that there are three cases of A alone, A and B together, and B alone.

[0063] In several embodiments provided in the embodiments of the present application, it should be understood that the disclosed system and method can be implemented in other ways. For example, the above-described device embodiments are only schematic, for example, the division of the units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the shown or discussed objects can be indirect coupling or communication connection through some interface, device or unit, which can be electrical, mechanical or other forms.

[0064] In the background of rapid iteration of modern automobile intelligent driving and industrial automation technology, environmental perception capability has become the core requirement of safety and reliability of vehicles and automated equipment. Ultrasonic radar, with its unique advantages of controllable cost, high precision in short distance detection, and immunity to light and adverse weather (such as fog and rain), has become one of the core perception devices in key scenarios such as distance detection and obstacle identification. In the automotive field, ultrasonic radar is widely integrated in automatic parking and low-speed collision warning systems, which provides key data support for driving decision-making by detecting the distance between the vehicle and surrounding pedestrians, objects (such as walls, other vehicles, etc.) in real time. In the field of industrial automation, ultrasonic radar is applied to production line material positioning, mechanical arm obstacle avoidance, warehouse robot path planning and other links to ensure the precise cooperation and safe operation of equipment in complex working space.

[0065] However, with the continuous increase in the complexity of application scenarios, the technical limitations of ultrasonic radar are increasingly prominent. The problem of misrecognition caused by environmental sound interference has become a core bottleneck restricting the further breakthrough of ultrasonic radar performance. From the technical principle, ultrasonic radar periodically transmits ultrasonic signals of a specific frequency (usually 20 kHz-200 kHz), receives the echo signals reflected by target objects, calculates the distance by combining the sound wave propagation speed and the time difference, and judges the characteristics and attributes of the objects by combining the echo characteristics. However, this working mode is highly sensitive to environmental sound. In a busy traffic section, the wideband noise generated by vehicle honking, engine roaring, tire friction with the ground, and other vehicle ultrasonic radar signals of the same frequency or near frequency are easily superimposed with the target echo signal. In an industrial workshop, the vibration noise of large machinery, the airflow noise of compressed air equipment, and the friction noise of pipeline transmission devices will also invade the ultrasonic radar receiving system, causing the ultrasonic radar to misjudge the environmental noise as obstacle echo, and thus causing problems such as false braking of intelligent driving systems and false shutdown of industrial equipment, seriously threatening the accuracy and safety of vehicles and automated equipment.

[0066] To address the above interference problems, traditional methods mainly use hardware filtering and frequency adjustment techniques for filtering. The hardware filtering scheme integrates high-pass, low-pass or band-pass filters in the receiving circuit of the ultrasonic radar to filter out environmental noise signals outside the frequency range. For example, for the commonly used 40 kHz operating frequency of automotive ultrasonic radar, a band-pass filter with a center frequency of 40 kHz is designed to suppress non-target frequency interference waves. The frequency adjustment scheme changes the transmission frequency of the ultrasonic radar to avoid known high-frequency interference source frequencies. For example, in an industrial workshop, if it is detected that the noise of a certain machine is concentrated at 50 kHz, the frequency of the ultrasonic radar can be adjusted to 60 kHz to reduce signal overlap. However, practice shows that such traditional methods have significant limitations: on the one hand, the interference sources in complex environments often exhibit wideband and time-varying characteristics. For example, the frequencies of ultrasonic radar of different vehicles in a traffic scene may fluctuate between 38 kHz and 42 kHz, making it difficult for a single filter to cover all interference frequencies, and sudden noises such as sudden vehicle collision sounds can easily break through the filtering threshold. On the other hand, the adjustable range of frequency adjustment is limited by the hardware performance of the ultrasonic radar, and excessive adjustment may lead to a decrease in the detection distance and resolution of the ultrasonic radar. At the same time, when the frequencies of multiple interference sources are widely distributed, the frequency adjustment scheme cannot achieve "full avoidance". Therefore, existing traditional techniques can only alleviate the interference problem to a certain extent and cannot fundamentally eliminate the misrecognition phenomenon, making it difficult to meet the high requirements of intelligent driving and industrial automation for the stability and accuracy of the perception system.

[0067] Based on the above application scenarios, the present application proposes an ultrasonic radar signal recognition method.

[0068] Figure 1 is a schematic flowchart of an ultrasonic radar signal recognition method provided by the embodiment. The method is applicable to a vehicle. The method comprises the following steps.

[0069] S1, obtaining a first original signal detected by an ultrasonic radar.

[0070] It should be noted that when the ultrasonic radar measures a distance or identifies an object, the transmission process of the ultrasonic signal emitted by the ultrasonic radar before reaching the receiving system will be disturbed by other sound waves existing in the environment, so that the received ultrasonic radar signal includes environmental noise (for example, mechanical vibration noise, wind noise, electromagnetic interference noise, etc.). Therefore, the first original signal detected by the ultrasonic radar contains environmental noise.

[0071] S2, identifying the first original signal to obtain a first ultrasonic radar signal through a pre-constructed noise classification model.

[0072] In the embodiment of the application, the first original signal detected by the ultrasonic radar is classified and identified through the pre-constructed noise classification model to obtain the first ultrasonic radar signal, so as to distinguish the ultrasonic radar signal from the environmental noise (for example, mechanical vibration noise, wind noise, electromagnetic interference noise, etc.), and accurately identify and exclude the interfering sound waves.

[0073] In an embodiment of the application, referring to Figure 2 , the construction method of the noise classification model is as follows: S21, obtaining environmental noise data under different environmental conditions.

[0074] S22, pre-processing and feature extraction are performed on the environmental noise data to obtain different environmental noise features.

[0075] S23, forming a data set from the different environmental noise features, and dividing the data set into a training set, a validation set and a test set.

[0076] Specifically, the data set can be divided according to the ratio of training set:validation set:test set=7:1.5:1.5.

[0077] S24, training a machine learning model through the training set to obtain an initial noise classification model.

[0078] Specifically, the machine learning model can be a deep learning model, or a support vector machine model or a random forest model.

[0079] S25, evaluating the initial noise classification model through the validation set and the test set to obtain an evaluation result.

[0080] S26, adjusting the initial noise classification model parameter according to the evaluation result to obtain the noise classification model.

[0081] In the embodiments of the present application, a machine learning model is introduced, the machine learning model is trained by the acquired environmental noise data under different environmental conditions, and the performance of the machine learning model is verified. The performance evaluation result is used to adjust the parameters of the machine learning model, so as to improve the performance of the noise classification model. The constructed noise classification model can accurately identify and exclude environmental noise, and distinguish the ultrasonic radar signal from the environmental noise.

[0082] S3, calculating the power spectral density of the first ultrasonic radar signal, and dynamically updating the noise baseline according to the power spectral density to obtain the second ultrasonic radar signal.

[0083] The power spectral density is used to describe the distribution rule of the average power of the power signal in the frequency domain, and is commonly used to analyze the frequency characteristics of the power signal (such as an environmental noise signal, a power grid voltage signal, etc.). In the embodiments of the present application, the power spectral density is introduced, the noise baseline is updated according to the calculated power spectral density to obtain the second ultrasonic radar signal, the stability of the noise baseline is improved, and the detectability of the ultrasonic radar signal is improved.

[0084] In an embodiment of the present application, referring to Figure 3 , the method for calculating the power spectral density of the first ultrasonic radar signal is as follows: S311, performing detrending and mean removal processing on the first ultrasonic radar signal to obtain a first signal.

[0085] In the embodiments of the present application, the first ultrasonic signal is subjected to detrending and mean removal preprocessing, which can reduce the influence of low-frequency noise.

[0086] S312, transforming the first signal from the time domain to the frequency domain by Fourier transform to obtain a second signal.

[0087] In the embodiments of the present application, the first signal is transformed from the time domain to the frequency domain by Fourier transform to obtain a second signal. The second signal obtained by Fourier transform can directly and intuitively decompose the signal frequency composition, so as to retain the target frequency signal and remove the interference frequency signal.

[0088] S313, dividing the second signal into multiple segment signals, calculating the power spectral density of each segment signal, and taking the average value of the power spectral densities of the multiple segment signals as the power spectral density of the first ultrasonic radar signal.

[0089] In the embodiments of the present application, the second signal is divided into multiple segment signals, the power spectral density of each segment signal is calculated, and the average value is taken as the power spectral density of the first ultrasonic radar signal, which can improve the accuracy of the power spectral density calculation.

[0090] In an embodiment of the present application, referring to Figure 4 , the method for obtaining the second ultrasonic wave radar signal according to the noise baseline updated by the power spectral density is: S321, calculating the power spectral density of the signal in the preset sliding window.

[0091] Specifically, a window is defined, which slides in the signal, and the power spectral density in the window is calculated.

[0092] S322, judging whether the power spectral density is greater than or equal to the current noise baseline.

[0093] Specifically, the power spectral density calculated in the non-signal region (i.e. the part containing only noise) of the signal is taken as the initial noise baseline. The power spectral density of multiple continuous signals can be calculated through the sliding window. When the power spectral density of the first signal is compared with the current noise baseline, the current noise baseline is the initial noise baseline, and when the power spectral density of the nth signal is compared with the current power spectral density, the current noise baseline is the new noise baseline obtained by comparing the power spectral density of the (n-1)th signal with the current noise baseline corresponding to the (n-1)th signal. Wherein, n≥2.

[0094] S323, when the power spectral density is greater than or equal to the current noise baseline, the current noise baseline is taken as the new noise baseline; when the power spectral density is less than the current noise baseline, the power spectral density of the signal is taken as the new noise baseline.

[0095] The noise baseline is a core concept in the fields of signal processing, measurement science, electronic engineering, etc., and its essence is the inherent noise level reference generated by the system itself or the environment when there is no target signal input. It is a key reference for distinguishing "effective signal" and "useless noise". Only when the strength of the target signal is significantly higher than the noise baseline, the signal has detectability and analysis value. Any measurement or signal transmission system (such as electronic circuits, sensors, communication equipment, laboratory instruments, etc.) cannot achieve "absolute noiselessness". Even without actively inputting target signals (such as microphones not collecting sound, radars not detecting targets, sensors not detecting physical quantities), the internal electronic components of the system (such as the thermal noise of resistors) and external environmental interference (such as electromagnetic radiation) will still produce random fluctuations, and the average level or statistical characteristics of these fluctuations are the "noise baseline".

[0096] In the embodiment of the present application, the sliding window technology is used to calculate the power spectral density of the signal, and the noise baseline is determined by comparing the size of the power spectral density and the current noise baseline, which can dynamically update the noise baseline and improve the stability of the noise baseline.

[0097] S4, identifying the effective ultrasonic wave radar signal of the second ultrasonic wave radar signal through the pre-constructed environmental filtering mapping model to obtain the ultrasonic wave radar signal.

[0098] In the embodiment of the present application, the effective ultrasonic radar signal of the second ultrasonic radar signal is obtained by identifying the effective ultrasonic radar signal of the second ultrasonic radar signal through the pre-constructed environment filtering mapping model, the interference sound wave is effectively removed, and the effective ultrasonic radar signal is retained.

[0099] In an embodiment of the present application, the environment parameters of the environment filtering mapping model include signal-to-noise ratio, noise bandwidth, burst noise duration, and effective ultrasonic radar signal, and the filtering parameters of the environment filtering mapping model include convergence factor, filter length, and filter coefficient.

[0100] Referring to Figure 5 , the method for adjusting the filtering parameters of the environment filtering mapping model is: S41, judging whether the signal-to-noise ratio is less than a set signal-to-noise ratio, and increasing the convergence factor when the signal-to-noise ratio is less than the set signal-to-noise ratio.

[0101] In the embodiment of the present application, the convergence factor is adjusted by judging the size of the signal-to-noise ratio and the set signal-to-noise ratio, and the convergence factor is increased when the signal-to-noise ratio is less than the set signal-to-noise ratio, so as to accelerate the convergence of the filter.

[0102] For example, the set signal-to-noise ratio is 5 dB, the convergence factor is increased to 0.3 when the signal-to-noise ratio is less than 5 dB, and the convergence of the filter is accelerated.

[0103] S42, judging whether the noise bandwidth is greater than a set bandwidth, and shortening the filter length when the noise bandwidth is greater than the set bandwidth.

[0104] In the embodiment of the present application, the filter length is adjusted by judging the size of the noise bandwidth and the set bandwidth, and the filter length is shortened when the noise bandwidth is greater than the set bandwidth, so as to reduce the processing delay.

[0105] For example, the set bandwidth is 10 kHz, and the filter length is shortened to 32 orders when the noise bandwidth is greater than 10 kHz, so as to reduce the processing delay.

[0106] S43, freezing the filter coefficient update when the environment filtering mapping model detects the effective ultrasonic radar signal.

[0107] Specifically, when the filter coefficient is frozen, the freezing duration needs to cover the entire duration (microsecond level) of the effective ultrasonic radar signal, and is delayed for 1-2 sampling periods after the effective ultrasonic radar signal is received, so as to ensure that the effective ultrasonic radar signal passes through the entire filter completely.

[0108] In the embodiment of the present application, the filter coefficient update is frozen when the environment filtering mapping model detects the effective ultrasonic radar signal, so as to prevent the effective ultrasonic radar signal from being mistakenly suppressed.

[0109] It should be noted that the effective ultrasonic radar signal is a signal containing a 40Hz carrier component and having a pulse width of microseconds.

[0110] In an embodiment of the present application, when the duration of the burst noise is greater than the set time, the environmental filtering mapping model uses the radiological projection algorithm for filtering.

[0111] The radiological projection algorithm is essentially through the analysis of the characteristics of the signal in different "projection dimensions", positioning and separating the noise components with significant differences from the effective signal characteristics. In an embodiment of the present application, when the duration of the burst noise (such as the whistle noise) is greater than the set time, the environmental filtering mapping model filters through the radiological projection algorithm, improves the stability of the environmental filtering model filtering, improves the fidelity of the effective signal, and ensures the accuracy of identifying the effective signal.

[0112] For example, the set time is 50ms, when the duration of the vehicle whistle is greater than 50ms, the environmental filtering mapping model uses the radiological projection algorithm for filtering, which improves the stability.

[0113] In an embodiment of the present application, the identification method further comprises: performing effectiveness evaluation on the ultrasonic radar signal.

[0114] Specifically, referring to Figure 6 , the effectiveness evaluation method comprises: S51, acquiring a second original signal collected by a camera and a third original signal detected by a laser radar.

[0115] S52, time and space alignment of the second original signal, the third original signal and the ultrasonic radar signal to obtain a camera initial signal and a laser radar initial signal.

[0116] In an embodiment of the present application, referring to Figure 7 , the method for time and space alignment is: S521, time alignment of the second original signal, the third original signal and the ultrasonic radar signal based on the PTP protocol.

[0117] Specifically, the method for time alignment is: (1) Connect all sensors to a GPS receiver to obtain UTC (Universal Time Coordinated) time stamp.

[0118] (2) The master clock (usually the central controller) sends a synchronization message (such as a sync message) based on the PTP protocol through Ethernet or CAN bus.

[0119] (3) The slave clock (i.e. ultrasonic radar, camera, laser radar) records the message receiving time and calculates the offset from the master clock.

[0120] (4) Measure network delay by Delay_Req / Delay_Resp message to achieve ±100ns level time synchronization.

[0121] (5) Attach the original data of ultrasonic radar, camera, and laser radar with synchronous time stamp.

[0122] Unify the time reference of all sensors through the Precision Time Protocol (PTP), so that the "time stamp" of data collected by different sensors is based on the same clock, eliminating time deviation.

[0123] S522, transform the spatial coordinates of the camera, laser radar, and ultrasonic radar to the same coordinate system (usually radar coordinate or vehicle body coordinate system).

[0124] Specifically, the method of spatial alignment is: (1) In a flat, open, static environment (no moving target), place a high-precision calibration board (such as a checkerboard calibration board, with known three-dimensional coordinates of each corner point, unit: m), ensure that the calibration board is in the field of view of all sensors (i.e. the camera can clearly capture the checkerboard, and the ultrasonic radar and laser radar can scan the point cloud of the calibration board).

[0125] (2) Ultrasonic radar, laser radar, and camera simultaneously collect calibration board data and extract feature points.

[0126] (3) Solve the coordinate transformation matrix by minimizing the re-projection error.

[0127] (4) Map the spatial coordinates of ultrasonic radar, laser radar, and camera to the same coordinate system based on the coordinate transformation matrix, and compensate for the spatial deviation caused by target motion.

[0128] After time alignment, map the "spatial coordinates" of different sensors (i.e. ultrasonic radar, camera, and laser radar) to the same coordinate system (usually radar coordinate system or vehicle body coordinate system), eliminate spatial misalignment caused by physical location differences and acquisition delays of sensors, and ensure that multi-source data is fused under the same space-time reference.

[0129] For example, convert the camera coordinate (Xc, Yc, Zc) to the radar coordinate system (Xu, Yu, Zu) by the following formula.

[0130] (Xu, Yu, Zu) = Tc·(Xc, Yc, Zc) + Δt·v In the formula, Tc is the coordinate transformation matrix, Δt is the acquisition time difference between the camera and the ultrasonic radar, compensated by the PTP protocol synchronization; v is the target motion speed, estimated by multi-frame data through the optical flow method.

[0131] S53, performing motion blur compensation on the camera initial signal to obtain a camera signal, and performing rain and fog compensation on the lidar initial signal to obtain a lidar signal.

[0132] Specifically, in an embodiment of the present application, referring to Figure 8 The method for performing motion blur compensation on the camera initial signal to obtain a camera signal is as follows: S5311, obtaining a vehicle speed.

[0133] S5312, determining whether the vehicle speed is greater than a set vehicle speed.

[0134] S5313, when the vehicle speed is greater than the set vehicle speed, performing deconvolution restoration on the camera initial signal based on a constructed point spread function matrix.

[0135] Specifically, the method for constructing the point spread function matrix is as follows: According to a given formula PSF(x, y) = δ(y-tan(θ)x) wherein (x, y) are pixel point coordinates; δ(·) is a Dirac function (1 only when y=tan(θ)x, and 0 otherwise); is a normalization coefficient, which ensures that the energy sum of the PSF function (i.e., the point spread function) is 1; L is a blur length, and θ is a motion direction.

[0136] L=v·t exp wherein v is the vehicle speed, t exp is an exposure time.

[0137] Specifically, a linear filtering algorithm or an iterative nonlinear algorithm can be used to perform deconvolution restoration on the camera initial signal.

[0138] S5314, performing feature enhancement (e.g., using a Laplacian operator) on the deconvolution-restored camera initial signal to obtain a camera signal.

[0139] When a vehicle is running at a high speed, the image captured by the camera is prone to blurring, which leads to camera image recognition failure. In an embodiment of the present application, whether the vehicle speed is greater than a set vehicle speed is taken as a trigger condition for motion blur compensation. When the vehicle speed is greater than the set vehicle speed (e.g., 30 km / h), deconvolution restoration is performed on the camera initial signal based on a constructed point spread function matrix, feature enhancement is performed on the deconvolution-restored camera initial signal to improve target contour clarity, motion blur compensation is performed on the camera initial signal, and the accuracy of the camera signal is improved, thereby solving the problem of recognition failure caused by camera image blurring in a high-speed scenario.

[0140] In an embodiment of the present application, referring toFigure 9 The method for rain and fog compensation of the initial signal of the laser radar to obtain the laser radar signal is: S5321, obtaining a meteorological environment parameter related to rain and fog.

[0141] Specifically, the meteorological environment parameter includes but is not limited to relative humidity, visibility, raindrop intensity, etc.

[0142] S5322, determining whether the meteorological environment parameter is within a set meteorological environment parameter range.

[0143] S5323, when the meteorological environment parameter is out of the set meteorological environment parameter range, performing reverse intensity compensation on the initial signal of the laser radar based on the constructed laser radar attenuation model to obtain the initial signal of the laser radar in the non-attenuation state. When the meteorological environment parameter is within the set meteorological environment parameter range, no rain and fog compensation is performed.

[0144] Specifically, the laser radar attenuation model is expressed as: I=I0e -β雨雾d In the formula, I is the echo intensity, I0 is the laser radar laser emission power, βrain fog is the rain and fog attenuation coefficient, and d is the distance.

[0145] Exemplarily, when any one of the following conditions is met, the reverse intensity compensation is performed on the initial signal of the laser radar.

[0146] 1. The visibility is less than 200 meters, and the duration is greater than or equal to 2s; 2. The relative humidity is greater than 90%, and the raindrop intensity is greater than or equal to the light rain level.

[0147] S5324, filtering the initial signal of the laser radar in the non-attenuation state, and removing the signal with an intensity less than a set intensity threshold to obtain the laser radar signal.

[0148] In the rain, fog, haze and other adverse weather conditions, the laser signal emitted by the laser radar will be severely attenuated due to the scattering and absorption of raindrops and fog droplets, resulting in a decrease in the intensity of the return signal and an increase in the noise of the point cloud data, and even the loss of point cloud information of distant targets (such as vehicles and pedestrians), affecting the reliability of automatic driving, environmental perception and other systems. In the embodiments of the present application, whether the weather environment parameters related to rain and fog are within the set weather environment parameter range is used as the trigger condition for rain and fog compensation. When the weather environment parameters exceed the set weather environment parameter range, the initial signal of the laser radar is compensated in reverse intensity based on the constructed laser radar attenuation model to restore the signal intensity in the non-attenuation state. After the initial signal of the laser radar is compensated in reverse intensity, some false signals (such as signals reflected only by raindrops) caused by rain and fog scattering will still remain, so such noise signals are removed by the limiting condition of removing signals with an intensity less than a set intensity threshold, and the real target laser radar signal is retained.

[0149] S54, evaluate the comprehensive confidence of the ultrasonic radar signal, the camera signal and the laser radar signal through a pre-constructed confidence evaluation model.

[0150] The confidence evaluation model is represented as: C 综合 =αC 超声 +βC 视觉 +γC 激光 In the formula, C 综合 is the comprehensive confidence, C 超声 is the confidence of the ultrasonic radar signal, and α is the weight coefficient of the ultrasonic radar signal. C 视觉 is the confidence of the camera signal, and β is the weight coefficient of the camera signal. C 激光 is the confidence of the laser radar signal, and γ is the weight coefficient of the ultrasonic radar signal.

[0151] The weight coefficients satisfy:

[0152] =SSIM×light compensation factor

[0153] In the formula, f noise is the environmental noise frequency, SSIM is the structural similarity coefficient, k is the rain and fog influence coefficient, and is the rain and fog attenuation coefficient.

[0154] S55, determine whether the ultrasonic radar signal is a valid signal according to the comprehensive confidence.

[0155] Specifically, refer to Figure 10, the method for determining whether the ultrasonic radar signal is a valid signal according to the comprehensive confidence is: S551, determining whether the comprehensive confidence is greater than or equal to a set confidence; S552, determining that the ultrasonic radar signal is a valid signal when the comprehensive confidence is greater than or equal to the set confidence, and determining that the ultrasonic radar signal is an invalid signal when the comprehensive confidence is less than the set confidence.

[0156] For example, the set confidence is 0.7, the comprehensive confidence is 0.8, which is greater than the set confidence, and the obtained ultrasonic radar signal is a valid signal.

[0157] For example, the set confidence is 0.7, the comprehensive confidence is 0.65, which is less than the set confidence, and the obtained ultrasonic radar signal is an invalid signal and needs to be re-detected.

[0158] In the embodiments of the present application, the validity of the ultrasonic radar signal is determined by the comprehensive confidence evaluated by the pre-constructed confidence evaluation model combined with the signals collected by the camera and the laser radar, which can verify the authenticity of the ultrasonic radar signal and reduce the false recognition rate.

[0159] For example, an ultrasonic radar signal recognition method includes the following steps: S1, obtaining a first original signal detected by an ultrasonic radar, a second original signal collected by a camera, and a third original signal detected by a laser radar.

[0160] S2, identifying the first original signal to obtain a first ultrasonic radar signal through a pre-constructed noise classification model.

[0161] S3, calculating the power spectral density of the first ultrasonic radar signal, and dynamically updating the noise baseline according to the power spectral density to obtain a second ultrasonic radar signal.

[0162] S4, identifying the effective ultrasonic radar signal of the second ultrasonic radar signal to obtain an ultrasonic radar signal through a pre-constructed environment filtering mapping model.

[0163] S5, performing time and space alignment on the second original signal, the third original signal, and the ultrasonic radar signal to obtain a camera initial signal and a laser radar initial signal.

[0164] S6, performing motion blur compensation on the camera initial signal to obtain a camera signal, and performing rain and fog compensation on the laser radar initial signal to obtain a laser radar signal.

[0165] S7, evaluating the comprehensive confidence of the ultrasonic radar signal, the camera signal, and the laser radar signal through a pre-constructed confidence evaluation model.

[0166] S8, determine whether the ultrasonic radar signal is a valid signal according to the comprehensive confidence. When the comprehensive confidence is greater than or equal to a set confidence, the ultrasonic signal is determined to be a valid signal. When the comprehensive confidence is less than the set confidence, the ultrasonic signal is determined to be an invalid signal, and re-detection is performed.

[0167] The embodiment of the present application provides an ultrasonic radar signal identification system, which is suitable for a vehicle. Figure 11 Fig. 1 shows a structural schematic diagram of the ultrasonic radar signal identification system.

[0168] The ultrasonic radar signal identification system 100 comprises: An acquisition module 101 is configured to acquire a first original signal detected by an ultrasonic radar; A noise classification module 102 is provided with a noise classification model 1021, which is configured to identify the first original signal to obtain a first ultrasonic radar signal; A calculation module 103 is configured to calculate a power spectral density of the first ultrasonic radar signal; An updating module 104 is configured to dynamically update a noise baseline according to the power spectral density to obtain a second ultrasonic radar signal; An identification module 105 is provided with an environment filtering mapping model 1051, which is configured to identify an effective ultrasonic radar signal of the second ultrasonic radar signal to obtain an ultrasonic radar signal.

[0169] In an embodiment of the present application, the data acquisition module 101 is further configured to acquire environment noise data under different environment conditions.

[0170] Continuing to refer to Figure 11 In an embodiment of the present application, the identification system further comprises a noise classification model construction module 106, which is configured to construct a noise classification model.

[0171] In an embodiment of the present application, referring to Figure 12 The noise classification model construction module 106 comprises: A preprocessing sub-module 1061 is configured to pre-process the environment noise data; A feature extraction sub-module 1062 is configured to extract features from the pre-processed environment noise data to obtain different environment noise features; A data set generation sub-module 1063 is configured to form a data set from the different environment noise features, and divide the data set into a training set, a validation set and a test set; A training and validation sub-module 1064 is configured to train a machine learning model by using the training set to obtain an initial noise classification model, and evaluate the initial noise classification model by using the validation set and the test set to obtain an evaluation result; The parameter adjusting sub-module 1065 adjusts the initial noise classification model parameter according to the evaluation result to obtain the noise classification model.

[0172] In an embodiment of the present application, the recognition system further comprises an environment filtering mapping model construction module 107, which is configured to construct an environment filtering mapping model.

[0173] In an embodiment of the present application, the data acquisition module 101 is further configured to acquire a second original signal collected by a camera and a third original signal detected by a laser radar, and to acquire a vehicle speed and a meteorological environment parameter related to rain and fog.

[0174] Continuing to refer to Figure 11 In an embodiment of the present application, the recognition system further comprises a space-time alignment module 108, which is configured to perform space-time alignment on the second original signal, the third original signal and an ultrasonic radar signal to obtain a camera initial signal and a laser radar initial signal.

[0175] Continuing to refer to Figure 11 In an embodiment of the present application, the recognition system further comprises a judgment module 109, which is configured to judge whether the vehicle speed is greater than a set vehicle speed and whether the meteorological environment parameter is within a set meteorological environment parameter range.

[0176] Continuing to refer to Figure 11 In an embodiment of the present application, the recognition system further comprises a compensation module 110, which is configured to, when the vehicle speed is greater than the set vehicle speed, perform deconvolution restoration on the camera initial signal based on a constructed point spread function matrix by using a linear filtering algorithm or an iterative nonlinear algorithm; and to perform feature enhancement on the deconvolution-restored camera initial signal to obtain a camera signal.

[0177] In an embodiment of the present application, the compensation module 110 is further configured to, when the meteorological environment parameter is out of the set meteorological environment parameter range, perform reverse intensity compensation on the laser radar initial signal based on a constructed laser radar attenuation model to obtain a laser radar initial signal in a non-attenuation state; and to filter the laser radar initial signal in the non-attenuation state to remove signals with an intensity less than a set intensity threshold to obtain a laser radar signal.

[0178] Continuing to refer to Figure 11 In an embodiment of the present application, the recognition system further comprises a confidence evaluation model construction module 111, which is configured to construct a confidence evaluation model.

[0179] Continuing to refer to Figure 11In an embodiment of the present application, the recognition system further comprises an evaluation module 112, which evaluates the comprehensive confidence of the ultrasonic radar signal, the camera signal and the laser radar signal by using a pre-constructed confidence evaluation model.

[0180] In an embodiment of the present application, the judgment module 109 is further configured to judge whether the ultrasonic radar signal is a valid signal according to the comprehensive confidence.

[0181] The embodiment of the present application further provides a computer program product, which comprises computer program codes. When the computer program codes are run on a computer, the computer is caused to execute the ultrasonic radar signal recognition method according to the above-mentioned embodiments. The computer program can be loaded on a vehicle system.

[0182] The embodiment of the present application further provides a computer readable storage medium, which stores program codes. When the program codes are run on a processor, the device comprising the processor is caused to execute the ultrasonic radar signal recognition method according to the above-mentioned embodiments. The processor running the computer readable storage medium can be loaded on a vehicle system.

[0183] It should be understood that when the modules or units described herein are implemented in software, they can be entirely or partially implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present application are entirely or partially generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer readable storage medium or transferred from one computer readable storage medium to another computer readable storage medium. The computer readable storage medium can be any available medium accessible by a computer or a data storage device such as a server, data center, etc. integrated with one or more available media sets. The available media can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid state disk (SSD)), etc.

[0184] The embodiment of the present application provides a chip system, which comprises a processor, or a chip system comprising a memory and a processor, for calling computer programs or computer instructions stored in the memory, so that the processor executes the ultrasonic radar signal recognition method according to the above-mentioned embodiments. The chip system can be a single chip or a chip module composed of multiple chips. The chip system can be loaded on a vehicle system.

[0185] The embodiment of the present application provides an electronic device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the electronic device realizes the ultrasonic radar signal identification method related to the above embodiment. The electronic device can be carried on a vehicle system.

[0186] The embodiment of the present application provides a vehicle.

[0187] For example, referring to Figure 13 The vehicle 200 comprises a memory 201, a processor 202 and a computer program 203 stored in the memory 201 and executable on the processor 202, and when the processor 202 executes the computer program 203, the processor 202 realizes the ultrasonic radar signal identification method related to the above embodiment.

[0188] For example, the vehicle 200 can comprise an acquisition module, a noise classification module, a calculation module, an updating module and an identification module, and the acquisition module, the noise classification module, the calculation module, the updating module and the identification module are integrated in the processor.

[0189] The acquisition module is used for acquiring a first original signal detected by an ultrasonic radar; The noise classification module is provided with a noise classification model, and the noise classification model is used for identifying the first original signal to obtain a first ultrasonic radar signal; The calculation module is used for calculating a power spectral density of the first ultrasonic radar signal; The updating module is used for dynamically updating a noise baseline according to the power spectral density to obtain a second ultrasonic radar signal; The identification module is provided with an environment filtering mapping model, and the environment filtering mapping model is used for identifying an effective ultrasonic radar signal of the second ultrasonic radar signal to obtain an ultrasonic radar signal.

[0190] Those skilled in the art can realize that the modules, units and method steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software mode depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0191] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, and any skilled person in the art can easily think of changes or replacements within the technical range disclosed in the present application, which should be covered. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for identifying ultrasonic radar signals, characterized in that, Applied to vehicles, the method includes: Acquire the first raw signal detected by ultrasonic radar; The first ultrasonic radar signal is obtained by identifying the first original signal through a pre-built noise classification model; Calculate the power spectral density of the first ultrasonic radar signal, and dynamically update the noise baseline based on the power spectral density to obtain the second ultrasonic radar signal; The effective ultrasonic radar signal of the second ultrasonic radar signal is obtained by identifying the effective ultrasonic radar signal through a pre-built environmental filtering mapping model.

2. The ultrasonic radar signal identification method according to claim 1, characterized in that, The method for constructing a noise classification model is as follows: Acquire environmental noise data under different environmental conditions; The environmental noise data is preprocessed and feature extracted to obtain different environmental noise features; Different environmental noise characteristics are used to form a dataset, which is then divided into a training set, a validation set, and a test set. The initial noise classification model is obtained by training the machine learning model using the training set. The evaluation results were obtained by evaluating the initial noise classification model using the validation set and the test set. The noise classification model is obtained by adjusting the parameters of the initial noise classification model based on the evaluation results.

3. The ultrasonic radar signal identification method according to claim 1, characterized in that, The method for calculating the power spectral density of the first ultrasonic radar signal is as follows: The first ultrasonic radar signal is detrended and mean-removed to obtain the first signal. The first signal is transformed from the time domain to the frequency domain using Fourier transform to obtain the second signal; The second signal is divided into multiple signal segments, and the power spectral density of each segment is calculated. The average power spectral density of the multiple signal segments is taken as the power spectral density of the first ultrasonic radar signal.

4. The ultrasonic radar signal identification method according to claim 1 or 3, characterized in that, The method for obtaining the second ultrasonic radar signal by updating the noise baseline based on the power spectral density is as follows: Calculate the power spectral density of the signal within the preset sliding window; Determine whether the power spectral density is greater than or equal to the current noise baseline; When the power spectral density is greater than or equal to the current noise baseline, the current noise baseline is used as the new noise baseline; When the power spectral density is less than the current noise baseline, the power spectral density of the signal is used as the new noise baseline.

5. The ultrasonic radar signal identification method according to claim 1, characterized in that, The environmental parameters of the environmental filtering mapping model include signal-to-noise ratio, noise bandwidth, burst noise duration, and effective ultrasonic radar signal. The filtering parameters of the environmental filtering mapping model include convergence factor, filter length, and filter coefficients. The method for adjusting the filtering parameters of the environmental filtering mapping model is as follows: Determine if the signal-to-noise ratio (SNR) is less than the set SNR; if the SNR is less than the set SNR, increase the convergence factor. Determine if the noise bandwidth is greater than the set bandwidth; if the noise bandwidth is greater than the set bandwidth, shorten the filter length. When the environmental filtering mapping model detects a valid ultrasonic radar signal, the filter coefficients are frozen and updated.

6. The ultrasonic radar signal identification method according to claim 1, characterized in that, The identification method further includes: When the duration of sudden noise exceeds a set time, the environmental filtering mapping model uses a radial projection algorithm for filtering.

7. The ultrasonic radar signal identification method according to claim 1, characterized in that, The identification method further includes: evaluating the effectiveness of the ultrasonic radar signal; the method for evaluating the effectiveness includes: Acquire the second raw signal captured by the camera and the third raw signal detected by the lidar; The second and third original signals are spatiotemporally aligned with the ultrasonic radar signal to obtain the camera initial signal and the lidar initial signal; Motion blur compensation is applied to the initial signal from the camera to obtain the camera signal, and rain and fog compensation is applied to the initial signal from the lidar to obtain the lidar signal. The overall confidence level of ultrasonic radar signals, camera signals, and lidar signals is evaluated using a pre-built confidence assessment model. The validity of an ultrasonic radar signal is determined based on the overall confidence level.

8. The ultrasonic radar signal identification method according to claim 7, characterized in that, The method for obtaining the camera signal by performing motion blur compensation on the initial camera signal is as follows: Get vehicle speed; Determine if the vehicle speed is greater than the set speed; When the vehicle speed exceeds the set speed, the initial signal from the camera is deconvolved and restored based on the constructed point spread function matrix. The camera signal is obtained by performing feature enhancement on the initial camera signal after deconvolution restoration.

9. The ultrasonic radar signal identification method according to claim 7, characterized in that, The method for obtaining the lidar signal by performing rain and fog compensation on the initial lidar signal is as follows: Obtain meteorological and environmental parameters related to rain and fog; Determine whether the meteorological environmental parameters are within the set range; When the meteorological environmental parameters exceed the set range, the initial signal of the lidar is obtained by reverse intensity compensation based on the constructed lidar attenuation model, resulting in the lidar initial signal in the no-attenuation state. The initial LiDAR signal under no attenuation is filtered out, and signals with intensity less than a set intensity threshold are removed to obtain the LiDAR signal.

10. An ultrasonic radar signal identification system, characterized in that, Applied to vehicles, including: The acquisition module is used to acquire the first raw signal detected by the ultrasonic radar; The noise classification module includes a noise classification model, which is used to identify the first original signal to obtain the first ultrasonic radar signal. A calculation module is used to calculate the power spectral density of the first ultrasonic radar signal; The update module is used to dynamically update the noise baseline based on the power spectral density to obtain the second ultrasonic radar signal; The identification module includes an environmental filtering mapping model, which is used to identify the effective ultrasonic radar signal of the second ultrasonic radar signal to obtain the ultrasonic radar signal.

Images