WO2020052441A1 - 目标分类方法和相关设备 - Google Patents
目标分类方法和相关设备 Download PDFInfo
- Publication number
- WO2020052441A1 WO2020052441A1 PCT/CN2019/103227 CN2019103227W WO2020052441A1 WO 2020052441 A1 WO2020052441 A1 WO 2020052441A1 CN 2019103227 W CN2019103227 W CN 2019103227W WO 2020052441 A1 WO2020052441 A1 WO 2020052441A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- signal
- target
- energy
- micro
- motion
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
- G01S7/411—Identification of targets based on measurements of radar reflectivity
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
Definitions
- the present application relates to the technical field of target classification, and in particular, to a target classification method and related equipment.
- Radar can be applied in many fields of industry, such as automotive electronics, drones, etc. Radar can achieve different functions in different fields. Such as radar can achieve ranging, angle measurement, speed measurement, altimetry and other functions.
- radar can be divided into a variety of radars such as lidar, millimeter wave radar. The accuracy measured by each radar is different.
- the embodiments of the present application provide a target classification method and related equipment, which can implement accurate classification of long-distance targets by using radar.
- an embodiment of the present application provides a target classification method, including:
- the related features of the target include micro-motion features
- the classifying the target based on the related characteristics of the target includes:
- the target is classified as a target in a micro-motion state or a target in a non-micro-motion state.
- the micro-motion feature includes at least one of the following:
- the range entropy feature is used to represent the uncertainty of the distance between the target and the radar
- the noise energy ratio feature is used to represent a ratio of the energy of the noise signal to the energy of the Doppler signal in the Doppler signal;
- the fretting energy ratio feature is used to represent the energy of the fretting signal and the energy of the Doppler signal in the Doppler signal, or the ratio of the fretting energy to the energy of the noise signal .
- obtaining the relevant characteristics of the target based on the echo signal includes:
- the echo signal Processing the echo signal to obtain a two-dimensional signal; wherein the two-dimensional signal includes a distance signal and a Doppler signal; the Doppler signal includes a micro motion signal and a noise signal;
- obtaining the micro motion characteristics of the target based on the two-dimensional signal includes:
- formula (1) is:
- feature1 is used to represent the distance entropy feature
- M is used to represent the number of frames of the acquired echo signal
- k is used to represent the frame number of the acquired echo signal
- c (k) is used to represent the echo of the kth frame
- c (k) is determined according to formula (2);
- formula (2) is:
- N is used to indicate the number of frames included in the calculation window
- n is used to indicate any frame in the calculation window
- range (k) is used to indicate the distance value represented by the distance signal in the echo signal of the k-th frame
- Range (n) is used to represent a distance value represented by a distance signal in an echo signal of the n-th frame in the calculation window; wherein N is an integer greater than or equal to 1.
- obtaining the micro motion characteristics of the target based on the two-dimensional signal includes:
- obtaining the micro motion characteristics of the target based on the two-dimensional signal includes:
- Determining the target according to a ratio of the energy of the micro-motion signal to the energy of the Doppler signal, or the ratio of the energy of the micro-motion signal to the energy of the noise signal in the Doppler signal The characteristics of the fretting energy ratio.
- the classifying the target according to related characteristics of the target includes:
- the classification parameter formula is obtained by training any one of the following classifiers:
- Support vector machine SVM
- correlation vector machine RVM
- KNN K nearest neighbor classification algorithm
- the relevant characteristics of the target include RCS characteristics
- the RCS feature is used to indicate a reflection degree of a target to the radar signal.
- an embodiment of the present application provides a target classification device, including:
- a transceiver module for transmitting a radar signal to detect a target in the environment; obtaining an echo signal of the target based on the radar signal feedback;
- a processing module configured to obtain relevant characteristics of the target according to the echo signal; and classify the target according to the relevant characteristics of the target.
- the related characteristics of the target include micro-motion characteristics
- the processing module classifies the target according to the related characteristics of the target, including:
- the target is classified as a target in a micro-motion state or a target in a non-micro-motion state.
- the micro-motion feature includes at least one of the following:
- the range entropy feature is used to represent the uncertainty of the distance between the target and the radar
- the noise energy ratio feature is used to represent a ratio of the energy of the noise signal to the energy of the Doppler signal in the Doppler signal;
- the fretting energy ratio feature is used to represent the energy of the fretting signal and the energy of the Doppler signal in the Doppler signal, or the ratio of the fretting energy to the energy of the noise signal .
- the processing module obtains relevant characteristics of the target according to the echo signal, including:
- the echo signal Processing the echo signal to obtain a two-dimensional signal; wherein the two-dimensional signal includes a distance signal and a Doppler signal; the Doppler signal includes a micro motion signal and a noise signal;
- the processing module obtains the micro-motion characteristics of the target according to the two-dimensional signal, including:
- formula (1) is:
- feature1 is used to represent the distance entropy feature
- M is used to represent the number of frames of the acquired echo signal
- k is used to represent the frame number of the acquired echo signal
- c (k) is used to represent the echo of the kth frame
- c (k) is determined according to formula (2);
- formula (2) is:
- N is used to indicate the number of frames included in the calculation window
- n is used to indicate any frame in the calculation window
- range (k) is used to indicate the distance value represented by the distance signal in the echo signal of the k-th frame
- Range (n) is used to represent a distance value represented by a distance signal in an echo signal of the n-th frame in the calculation window; wherein N is an integer greater than or equal to 1.
- the processing module obtains the micro-motion characteristics of the target according to the two-dimensional signal, including:
- the processing module obtains the micro-motion characteristics of the target according to the two-dimensional signal, including:
- Determining the target according to a ratio of the energy of the micro-motion signal to the energy of the Doppler signal, or the ratio of the energy of the micro-motion signal to the energy of the noise signal in the Doppler signal The characteristics of the fretting energy ratio.
- the processing module classifies the target according to related characteristics of the target, including:
- the classification parameter formula is obtained by training any one of the following classifiers:
- Support vector machine SVM
- correlation vector machine RVM
- KNN K nearest neighbor classification algorithm
- the relevant characteristics of the target include RCS characteristics
- the RCS feature is used to indicate a reflection degree of a target to the radar signal.
- an embodiment of the present application provides a radar, including:
- a processor connected to the transmitter and the receiver;
- a memory connected to the processor
- the transmitter is used for transmitting a radar signal to detect a target in the environment
- the receiver is configured to acquire an echo signal of the target based on the radar signal feedback
- the processor is configured to execute a computer program stored in the memory to implement the following steps:
- the related features of the target include micro-motion features
- the processor is configured to implement the classification of the target based on the related features of the target, and is specifically used to implement:
- the target is classified as a target in a micro-motion state or a target in a non-micro-motion state.
- the micro-motion feature includes at least one of the following:
- the range entropy feature is used to represent the uncertainty of the distance between the target and the radar
- the noise energy ratio feature is used to represent a ratio of the energy of the noise signal to the energy of the Doppler signal in the Doppler signal;
- the fretting energy ratio feature is used to represent the energy of the fretting signal and the energy of the Doppler signal in the Doppler signal, or the ratio of the fretting energy to the energy of the noise signal .
- the processor when the processor is configured to obtain relevant characteristics of the target according to the echo signal, the processor is specifically configured to implement:
- the echo signal Processing the echo signal to obtain a two-dimensional signal; wherein the two-dimensional signal includes a distance signal and a Doppler signal; the Doppler signal includes a micro motion signal and a noise signal;
- the processor is configured to obtain the micro-motion feature of the target according to the two-dimensional signal, and is specifically configured to implement:
- formula (1) is:
- feature1 is used to indicate the distance entropy feature
- M is used to indicate the number of frames of the acquired echo signal
- k is used to indicate the frame number of the acquired echo signal
- c (k) is used to indicate the echo of the kth frame.
- c (k) is determined according to formula (2);
- formula (2) is:
- N is used to indicate the number of frames included in the calculation window
- n is used to indicate any frame in the calculation window
- range (k) is used to indicate the distance value represented by the distance signal in the echo signal of the k-th frame
- Range (n) is used to represent a distance value represented by a distance signal in an echo signal of the n-th frame in the calculation window; wherein N is an integer greater than or equal to 1.
- the processor is configured to obtain the micro-motion feature of the target according to the two-dimensional signal, and is specifically configured to implement:
- the processor is configured to obtain the micro-motion feature of the target according to the two-dimensional signal, and is specifically configured to implement:
- Determining the target according to a ratio of the energy of the micro-motion signal to the energy of the Doppler signal, or the ratio of the energy of the micro-motion signal to the energy of the noise signal in the Doppler signal The characteristics of the fretting energy ratio.
- the processor is configured to classify the target according to related characteristics of the target, and is specifically configured to implement:
- the classification parameter formula is obtained by training any one of the following classifiers:
- Support vector machine SVM
- correlation vector machine RVM
- KNN K nearest neighbor classification algorithm
- the relevant characteristics of the target include RCS characteristics
- the RCS feature is used to indicate a reflection degree of a target to the radar signal.
- an embodiment of the present application provides a readable storage medium, characterized in that a computer program is stored in the readable storage medium, and when the computer program is executed by a processor, the first aspect may be implemented Either method.
- an object in the environment can be detected and an echo signal of the target based on the radar signal feedback can be obtained.
- relevant characteristics of the target can be obtained, and The related features of the target classify the target.
- the above method can realize the classification of the target in the case of being far away from the target.
- FIG. 1 is a schematic diagram of an application scenario according to an embodiment of the present application.
- FIG. 2 is a schematic structural diagram of a radar according to an embodiment of the present application.
- FIG. 3 is a schematic flowchart of a target classification method according to an embodiment of the present application.
- FIG. 4 is a schematic flowchart of another target classification method according to an embodiment of the present application.
- FIG. 5 is a schematic diagram of a module composition of a target classification device according to an embodiment of the present application.
- the radar may be mounted on an aircraft, and the aircraft may be an Unmanned Aerial Vehicle (UAV) or other aircraft.
- UAV Unmanned Aerial Vehicle
- the radar 100 can be installed on the bottom of the aircraft 102, which can be used to detect the environmental situation of the landing site 104, and can use the landing site as a target to classify it. For example, classify landing sites as ground or water. Therefore, the aircraft can be notified of the classification result, so that the aircraft can adjust the landing place and avoid falling into the water. Can improve the intelligence of aircraft landing.
- the radar can also achieve other functions, such as the aircraft performing altimetry on the air.
- the radar can be installed on the vehicle to detect targets in the surrounding environment of the vehicle and classify the targets.
- the radar can recognize the targets in the surrounding environment of the vehicle as roadblocks, railings, people, etc.
- the following describes a radar provided by an embodiment of the present application to implement a target classification method.
- FIG. 2 is a schematic structural diagram of a radar system according to an embodiment of the present application.
- the radar system 200 may include a processor 201, a transmitter 203, a receiver 205, a bus 207, an interface 209, a memory 211, and the like.
- the processor 201 is connected to the transmitter 203, the receiver 205, the memory 211, the interface 109, and the power supply system 207, respectively.
- the power supply system 207 can also be connected to other modules outside the processor 201 according to design requirements.
- each module may be connected to other modules except the processor 201 according to design requirements, which is not limited herein.
- the transmitter 203 may be connected with a transmitting antenna, and the transmitting antenna may be an antenna array or other antenna forms applied to a radar, which is not limited herein.
- the transmitter is used to transmit radar signals.
- the transmitter can transmit lidar signals, millimeter-wave radar signals, and the like.
- the transmitter can be used to transmit millimeter-wave radar signals at 77 GHz, 24 GHz, or other frequency bands, which is not limited herein.
- the transmitter 203 may include a transmission control unit, which is used to control instruction interaction with the processor 201, and may also control the transmitter to transmit radar signals.
- a transmission control unit which is used to control instruction interaction with the processor 201, and may also control the transmitter to transmit radar signals.
- the receiver 205 may be connected with a receiving antenna.
- the receiving antenna may be an antenna array or other antenna forms applied to a radar, which is not limited herein.
- the receiver 205 is configured to receive an echo signal of the radar signal transmitted by the transmitter 203 after being reflected by the target.
- the information carried in the echo signal can be used to reflect the characteristics, attributes, and motion characteristics of the target.
- the receiver 205 may receive an echo signal through a single channel or multiple channels.
- the receiver 205 may include a receiving control unit, which is used to control the implementation of the instruction interaction with the processor 201, and may also control the receiver to receive an echo signal and the like.
- a receiving control unit which is used to control the implementation of the instruction interaction with the processor 201, and may also control the receiver to receive an echo signal and the like.
- the receiver 205 may further include a processor for further processing the received echo signal, for example, processing the echo signal into a two-dimensional signal; or the receiver may send the echo signal to the processor Further processing is performed in the signal processor 2011 in 201, and the embodiment of the present application is not limited herein.
- the transmitter 203 and the receiver 205 may be independent devices, or the transmitter 203 and the receiver 205 may be integrated into one device as a front end of the radar system 200.
- the processor 201 may include a signal processor 2011 and a data processor 2013.
- the signal processor 2011 is used to process the echo signal
- the data processor 2013 is used to further process the processed echo signal to achieve classification of the target.
- functions implemented by the signal processor 2011 and functions implemented by the data processor 2013 may be implemented by independent processors or jointly by the processors, which are not limited herein.
- the processor may include a digital signal processor (Digital Signal Processing, DSP), a micro processor (Micro Processing Unit, MCU), an advanced reduced instruction set machine (Advanced RISC Machine, ARM), and the like.
- DSP Digital Signal Processing
- MCU Micro Processing Unit
- ARM Advanced reduced instruction set machine
- the processor may refer to a processor core or a processor chip.
- the above processor or processors may be implemented by a hardware chip.
- the hardware chip may be an application-specific integrated circuit (English: application-specific integrated circuit (abbreviation: ASIC)), a programmable logic device (English: programmable logic device (abbreviation: PLD)), or a combination thereof.
- the PLD may be a complex programmable logic device (English: complex programmable device, abbreviation: CPLD), a field programmable logic gate array (English: field-programmable gate array, abbreviation: FPGA), general array logic (English: generic array) logic, abbreviation: GAL) or any combination thereof.
- the transmitter 203, the receiver 205, and the processor 201 may be integrated into one hardware chip, or each may be implemented by an independent hardware chip, which is not limited herein.
- the power system 207 may include a power source and a power management module.
- the power supply can supply power to each module in the radar 200, and the power management module can be used to manage and control the power supply of each module.
- the interface 209 is used to enable the radar system 200 to communicate with other equipment or devices.
- the radar system 200 can transmit the target classification result to other devices or devices through the interface 209, so that other devices or devices can implement other functions based on the target classification results.
- the radar system 200 when the radar system 200 is installed in an aircraft, the radar system 200 may be connected to a flight control system, a main control system, or other control systems in the aircraft through an interface 209.
- the radar system 200 is connected to the main control system through the interface 209 as an example.
- the radar system 200 can transmit the target classification result to the main control system through the interface 209.
- the main control system can further determine whether the detected target is suitable for landing, or whether it needs to avoid the detected target. And can further control the aircraft to achieve the above functions.
- the interface 209 may include a serial peripheral interface (SPI) 2091, a controller area network (CAN) 2093, a universal asynchronous transmission interface (Universal Receiver / Transmitter, UART) 2095 Wait.
- SPI serial peripheral interface
- CAN controller area network
- UART universal asynchronous transmission interface
- the interface 209 may also include other communication interfaces or input / output interfaces, which is not limited herein.
- the memory 211 may include volatile memory (English: volatile memory), such as random access memory (English: random-access memory, abbreviation: RAM), such as static random access memory (English: static-access memory, abbreviation: SRAM), double data rate synchronous dynamic random access memory (English: Double Data Rate Synchronous Dynamic Random Access Memory, abbreviation: DDR SDRAM), etc .; the memory can also include non-volatile memory (English: non-volatile memory), For example, flash memory (English: flash memory), hard disk (English: hard disk drive (abbreviation: HDD)) or solid-state hard disk (English: solid-state drive (abbreviation: SSD)); the memory 211 may also include a combination of the above types of memory .
- volatile memory such as random access memory (English: random-access memory, abbreviation: RAM), such as static random access memory (English: static-access memory, abbreviation: SRAM), double data rate synchronous dynamic random access memory (English: Double Data Rate Synchronous Dynamic Random Access Memory,
- the memory may be an independent memory, or may be a memory inside a chip (such as a processor chip) or a module having a storage function.
- the memory may store computer programs (such as application programs, functional modules), computer instructions, operating systems, data, databases, and the like.
- the memory can be partitioned.
- the radar system 200 may also include other components.
- the radar system 200 may further include a classification trainer, etc., to implement online training of a classification model.
- the other components included in the radar system 200 will not be repeated here.
- FIG. 3 is a schematic flowchart of a target classification method according to an embodiment of the present application. As shown in FIG. 3, the method includes the following steps.
- step S301 a radar signal is transmitted to detect a target in the environment.
- Step S302 Acquire an echo signal of the target based on the radar signal feedback.
- the radar system may transmit a radar signal to the environment through the above-mentioned transmitter and transmitting antenna, so as to detect a target in the environment, and may receive an echo signal of the target based on the radar signal reflected through the receiving antenna and the receiver, and This echo signal can be sent to a processor for further processing.
- radar can detect targets at a long distance without being affected by ambient light.
- Step S303 Obtain the relevant characteristics of the target according to the echo signal.
- the radar can calculate the relevant characteristics of the target according to the echo signal.
- the relevant characteristics of the target can be used to classify the target. That is, the relevant characteristics of the target can be used to reflect the motion state of the target, the electromagnetic characteristics of the target, and the target attributes. Furthermore, the target can be classified based on the relevant characteristics of the target.
- the relevant characteristics of the target may include a micro-motion characteristic, and through the micro-motion characteristic, it can be determined whether the target is in a micro-motion state, and then the target can be classified into a target in a micro-motion state and a target in a non-micro-motion state.
- the micro-motion can be understood as the movement of the target in addition to its own movement, or the movement of local components on the target, such as the rotation of the propeller of an aircraft, the swing of an arm back and forth when a person walks, and the like. If the target is in the micro-motion state, it indicates that the target has micro-motion.
- the micro-motion state may include a wave state of the water surface; the non-micro-motion state may include a stationary state or a motion state of the target itself.
- the radar when used in an aircraft to classify the landing point as a target, it can be understood that the water surface is non-rigid and easily affected by environmental factors such as wind. It is in a micro-motion state and rigid. The ground is at a static state, and the micro-movement characteristics of the obtained target can be used to classify the target to determine whether the currently detected landing point is the ground or the water surface, so that the aircraft can further determine whether the landing can be performed.
- two targets that are in the micro-motion state can be classified based on their micro-motion characteristics.
- radar when used in an aircraft to classify and recognize the ground in the desert, it can distinguish between quicksand and water surface based on the micromovement characteristics of quicksand and the micromotion characteristics of water surface.
- the related features of the target may include at least one of the micro-movement feature of the target, the Radar Cross Section (RCS) feature (also referred to as the reflective surface feature of the target), and the like.
- RCS Radar Cross Section
- the RCS characteristics of the target can be used to reflect the target's reflection of the radar signal. Because the target reflects the radar signal differently according to its own attributes, the radar system can classify different targets based on the RCS characteristics.
- the above two features can be combined to classify whether the target is in a micromotion state, and the classification accuracy can be improved by combining the features.
- Step S304 classify the target according to the relevant characteristics of the target.
- the target may be classified based on the related characteristics of the target based on the classifier or the trained classification model to obtain a classification result.
- the classification result can be the classification attribute of the target.
- the classification attributes of the target include ground attributes or water surface attributes.
- the ground property of the target is used to characterize that the ground is rigid, that is, it moves without being affected by the environment, such as wind.
- the target can be classified online or offline according to the relevant characteristics of the target.
- achieving online classification of targets refers to inputting the relevant features of the targets into a radar-configured classifier, and the classifier outputs the classification attributes of the targets based on the obtained relevant features of the targets. Further, the correctness of the output results can also be fed back to the classifier, so that the classifier can adjust the classification algorithm. In this case, as the number of times the target is detected increases, the classifier output is more accurate.
- to achieve offline classification of the target means to use multiple related features and classifiers to train a classification model and pre-store the classification model into the radar. After obtaining the relevant features of the target, the relevant features and classification model can be used to obtain Classification results, such as the classification properties of the target.
- an object in the environment can be detected and an echo signal of the target based on the radar signal feedback can be obtained.
- relevant characteristics of the target can be obtained, and The related features of the target classify the target.
- the above method can realize the use of radar signals to classify targets when they are far away from the target.
- FIG. 4 is a schematic flowchart of another target classification method according to an embodiment of the present application. As shown in FIG. 4, the method includes at least the following steps.
- Step S401 transmitting a radar signal to detect a target in the environment, and obtaining an echo signal of the target based on the radar signal feedback.
- step S402 the echo signal is processed to obtain a two-dimensional signal.
- a 1-dimensional Fast Fourier Transform (1DFFT) may be performed on the echo signal to obtain the distance data between the radar and the target (also may be a distance signal).
- a two-dimensional Fast Fourier Transform (2DFFT) may be performed on the echo signal to obtain a Doppler signal.
- a two-dimensional signal including distance data and a Doppler signal can be obtained.
- step S403 the relevant features of the target are obtained according to the two-dimensional signal.
- the micro-motion characteristics of the target can be obtained according to the Doppler signal and / or the distance data in the two-dimensional signal.
- the Doppler signal in the two-dimensional signal may include a micro-Doppler signal. That is, according to the Doppler signal, the micro-motion characteristics of the target related to the Doppler signal can be obtained.
- the range entropy feature can be used to represent the uncertainty of the distance between the target and the radar.
- the distance entropy characteristics are relatively large.
- the distance entropy characteristics are relatively small. Differentiate whether the target is in a jog state.
- the distance entropy feature (feature1) can be determined by formula (1).
- formula (1) is:
- feature1 is used to represent the distance entropy feature
- M is used to represent the number of frames of the acquired echo signal
- k is used to represent the frame number of the acquired echo signal
- c (k) is used to represent the echo of the kth frame
- the radar system can continuously transmit radar signals and can continuously receive multiple frames of echo signals.
- the radar system can determine the distance from the target based on each frame of echo signals and its corresponding radar signal. It can be expressed as a distance signal or distance data.
- the distance signal can be normalized, such as by formula (2), and the results of the M normalized processes can be calculated by entropy, such as by formula ( 1) Perform entropy calculation, and then multiple distance entropy features.
- c (k) is determined according to formula (2).
- formula (2) is:
- N is used to indicate the number of frames included in the calculation window
- n is used to indicate any frame in the calculation window
- range (k) is used to indicate the distance value in the echo signal of the k-th frame
- range (n) It is used to represent the distance value in the echo signal of the n-th frame in the calculation window
- N is an integer greater than or equal to 1.
- N can be understood as calculating the number of frames contained or contained in the window, or calculating the window length of the window, which is determined by the number of frames. Second, the characteristics of noise energy ratio.
- the noise-to-energy ratio feature is used to represent a ratio of the energy of the noise signal to the energy of the Doppler signal in the Doppler signal.
- the larger the noise energy ratio characteristic the smaller the proportion of the micromotion energy in the Doppler signal, and the less likely the target is in the micromotion state; accordingly, the smaller the noise energy ratio characteristic, the more the The greater the proportion of the fretting energy in the Doppler signal, the more likely the target is in the fretting state.
- the micro motion signal in the Doppler signal may be determined first, and then the micro motion signal in the Doppler signal is removed using the CLEAN algorithm to obtain the noise signal.
- the micro-motion signal can be represented by the harmonics of the Doppler signal in the frequency domain or the time domain.
- the Doppler signal can be searched for the harmonic with the largest amplitude value to determine the micro-motion signal, and the CLEAN algorithm can be used.
- the micro-motion signal represented by the harmonic is subtracted from the Doppler signal to obtain a residual signal.
- the noise energy ratio can be determined by the ratio of the energy of the residual signal to the energy of the Doppler signal. If the CLEAN algorithm is performed q times on the Doppler signal in the above manner, where q is an integer greater than or equal to q, then q ratios can be obtained.
- the noise energy ratio characteristic can be represented by a vector including these q ratios.
- the target may be classified using the vector, or the target may be classified using one or more ratios in the vector.
- q may be preset or determined based on the number of harmonics in the echo signal, which is not limited herein.
- y is the energy signal represented by the subharmonic, which can be a time domain signal or a frequency domain signal
- A represents the amplitude in the Doppler signal
- ⁇ represents the phase
- j represents the time
- f c represents the Doppler Frequency
- M is the cumulative number of pulses.
- the energy ratio characteristic value can be determined by the following formula (4):
- R i represents the eigenvalue of the energy ratio based on the CLEAN algorithm
- 1 ⁇ i ⁇ q L represents the spectrum length of the Doppler signal
- S r (n) represents the Doppler signal.
- S r (n) represents the noise signal in the Doppler signal
- S i (n) represents the i-th time according to the Doppler signal (also called the original signal) and The residual signal obtained by subtracting the harmonic signals.
- feature2 is the energy ratio feature.
- the fretting energy ratio feature is used to represent a ratio of the energy of the fretting signal to the energy of the Doppler signal in the Doppler signal, or the energy of the fretting signal to the energy of the noise signal.
- the larger the characteristic of the micro motion energy ratio the greater the proportion of the micro motion signal in the Doppler signal, and the more likely the target is in the micro motion state.
- feature3 represents the characteristics of the fretting energy ratio
- the interval [f1, F1] represents the frequency band of the fretting signal energy.
- the interval may be preset, and the determination of the interval may be related to the application scenario involved. For example, the preset interval is different in different application scenarios.
- the interval [f2, F2] indicates the frequency band of the noise signal or the frequency band of the Doppler signal. Similarly, the interval may be preset or determined based on the interval [f1, F1].
- the interval [f2, F2] represents the frequency band of the noise signal
- the interval [f2, F2] can represent all or part of the frequency band of the noise signal
- f is the interval [f1, F1] or the interval [f2, F2]
- P (f) is the amplitude corresponding to f frequency.
- the RCS feature can also be obtained based on the two-dimensional signal.
- the manner of obtaining is not limited in the examples of the present application.
- Step S404 Determine a classification parameter formula corresponding to the relevant feature of the target.
- the different combinations of the related features correspond to different classification parameter formulas.
- classification parameter formula in the embodiment of the present application can be understood as a classification model.
- the corresponding relationship between the combination mode of the relevant features of the target and the classification parameter formula may be pre-stored in the radar system, as pre-stored in the memory 211 of the radar system shown in FIG. 2.
- the classification parameter formula corresponding to the related feature can be obtained according to the pre-stored correspondence.
- Step S405 classify the target according to the classification parameter formula and related characteristics of the target.
- one or more related features of the target may be used as input values of the classification parameter formula, and then the parameter values of the classification parameters may be calculated. Further, the classification interval in which the parameter value of the classification parameter falls can be determined, and if it falls within a certain classification interval, a target classification corresponding to the classification interval can be determined. The correspondence between the classification interval and the target classification may be obtained through pre-training, which is pre-stored in the radar. Alternatively, one or more related features of the target can be used as the input value of the classification parameter formula, and the classification result of the target can be directly obtained, and then the target classification can be determined.
- the target classification corresponding to the combination may be determined according to a combination of the calculated related features of the target.
- the combination of related features may include at least two related features, and the combination of related features may further improve the accuracy of classifying the target.
- the above-mentioned method for determining the classification attribute of the target may be implemented by a classifier in the radar, or by other devices in the radar, such as by a processor in the radar executing a corresponding program.
- the corresponding relationship, or the classification parameter formula may be obtained by the classifier through a training algorithm.
- the classifier includes multiple training algorithms, and the above-mentioned correspondence relationship or classification parameter formula can be obtained based on one or more training algorithms in the classifier.
- the training correspondence or classification parameter formula can be pre-stored in the radar.
- the radar detects the target in real time and calculates the relevant characteristics of the target, it can be combined with the pre-stored correspondence or according to the classifier or other device in the radar.
- Classification parameter formula to further obtain classification results.
- the classifier in the radar can directly obtain the classification result according to the relevant features of the target and the training algorithm.
- the classifier may include at least one of the following:
- SVM Support Vector Machine
- RVM Relative Vector Machine
- KNN K-nearest neighbor classification algorithm
- neural network etc.
- the radar can also output the classification result of the target to other devices, so that other devices can perform further processing according to the classification attributes of the target.
- the embodiment of the present application only uses the classification attribute of the target as the ground attribute and the water surface attribute as an example for description.
- the recognition and classification of the target in other application scenarios can also be implemented. Be limited.
- the radar is installed on the bottom of the aircraft and can be used to assist the aircraft in autonomous landing.
- the radar when the aircraft is in a scene that needs to land, the radar is triggered to emit radar signals.
- the radar transmits radar signals toward the landing point or the ground.
- the landing point or the ground After receiving the radar signal, the landing point or the ground reflects and reflects the radar signal.
- the signal is an echo signal.
- the radar After receiving the echo signal, the radar can obtain the relevant characteristics of the target based on the echo signal.
- the echo signal can be processed into a two-dimensional signal based on the above algorithm, and the target can be calculated based on the two-dimensional signal. Micro-movement features, RCS features, etc.
- At least one of a range entropy feature, an energy ratio feature, a fretting energy ratio feature, an RCS feature, and the like can be obtained, and then the landing ground can be classified according to the related features of the calculated target. Landing points are classified as either ground or water.
- the radar can transmit the classification result to the aircraft's flight control system, and the flight control system can determine whether to make a vertical landing based on the classification result. For example, when the classification result is on the ground, the flight control system may control the power unit of the aircraft for vertical landing. When the classification result is on the water, the flight control system stops the landing plan, or plans a new landing path.
- FIG. 5 is a schematic diagram of a module composition of a target classification device according to an embodiment of the present application.
- the target classification apparatus 500 may include a transceiver module 501 and a processing module 503.
- the transceiver module 501 is configured to transmit a radar signal to detect a target in the environment, and obtain an echo signal of the target based on the radar signal feedback;
- a processing module 503 is configured to obtain relevant characteristics of the target according to the echo signal; and classify the target according to the relevant characteristics of the target.
- the related features of the target include micro-motion features
- the processing module 503 classifies the target based on the related characteristics of the target, including:
- the target is classified as a target in a micro-motion state or a target in a non-micro-motion state.
- the micro-motion feature includes at least one of the following:
- the range entropy feature is used to represent the uncertainty of the distance between the target and the radar
- the noise energy ratio feature is used to represent a ratio of the energy of the noise signal to the energy of the Doppler signal in the Doppler signal;
- the fretting energy ratio feature is used to represent the energy of the fretting signal and the energy of the Doppler signal in the Doppler signal, or the ratio of the fretting energy to the energy of the noise signal .
- the processing module 503 obtains the relevant characteristics of the target according to the echo signal, including:
- the echo signal Processing the echo signal to obtain a two-dimensional signal; wherein the two-dimensional signal includes a distance signal and a Doppler signal; the Doppler signal includes a micro motion signal and a noise signal;
- the processing module 503 obtains the micro motion characteristics of the target according to the two-dimensional signal, including:
- formula (1) is:
- feature1 is used to indicate the distance entropy feature
- M is used to indicate the number of frames of the acquired echo signal
- k is used to indicate the frame number of the acquired echo signal
- c (k) is used to indicate the echo of the kth frame.
- c (k) is determined according to formula (2);
- formula (2) is:
- N is used to indicate the number of frames included in the calculation window
- n is used to indicate any frame in the calculation window
- range (k) is used to indicate the distance value represented by the distance signal in the echo signal of the k-th frame
- Range (n) is used to represent a distance value represented by a distance signal in an echo signal of the n-th frame in the calculation window; wherein N is an integer greater than or equal to 1.
- the processing module 503 obtains the micro motion characteristics of the target according to the two-dimensional signal, including:
- the processing module 503 obtains the micro motion characteristics of the target according to the two-dimensional signal, including:
- Determining the target according to a ratio of the energy of the micro-motion signal to the energy of the Doppler signal, or the ratio of the energy of the micro-motion signal to the energy of the noise signal in the Doppler signal The characteristics of the fretting energy ratio.
- the processing module 503 classifies the target according to related characteristics of the target, including:
- the classification parameter formula is obtained by training any one of the following classifiers:
- Support vector machine SVM
- correlation vector machine RVM
- KNN K nearest neighbor classification algorithm
- the relevant characteristics of the target include RCS characteristics
- the RCS feature is used to indicate a reflection degree of a target to the radar signal.
- target classification device may also include other functional modules, which is not limited herein.
- the above functional modules may be implemented by software, hardware or a combination thereof.
- the above-mentioned functional module may be implemented by a computer program, or the transmitting-receiving module in the above-mentioned functional module may be implemented by a transmitter or a receiver shown in FIG. 2, and the processing module in the above-mentioned functional module may be processed by the processing shown in FIG.
- the processor or the computer program implemented by the processor is not limited herein.
- the processor in the radar may be included in the processor, or included in the transmitter or receiver, or included in other devices, where at least one processor may For performing any one of the methods in the above embodiments.
- an embodiment of the present application further provides a readable storage medium, characterized in that a computer program is stored in the readable storage medium, and when the computer program is executed by a processor, the first aspect may be implemented Either method.
- an object in the environment can be detected and an echo signal of the target based on the radar signal feedback can be obtained.
- relevant characteristics of the target can be obtained, and The related features of the target classify the target.
- the above method can realize the classification of the target in the case of being far away from the target.
Landscapes
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
一种目标分类方法和相关设备。其中,方法包括:向雷达所在环境发射雷达信号,以检测环境中的目标(S301),并获取目标基于雷达信号反馈的回波信号(S302);根据回波信号,得到目标的相关特征(S303);根据目标的相关特征,对目标进行分类(S304)。可以实现利用雷达对远距离目标进行精确地分类。
Description
相关申请交叉引用
本申请要求于2018年9月14日申请的、申请号为201811076412.8、申请名称为“目标分类方法和相关设备”的中国专利申请的优先权,其全部内容通过引用结合在本申请中。
本申请涉及目标分类技术领域,尤其涉及一种目标分类方法和相关设备。
雷达目前可以应用于工业的多个领域,如汽车电子领域、无人机领域等。雷达可以在不同的领域实现不同的功能等。如雷达可以实现测距、测角度、测速、测高等多种功能。
随着雷达技术的发展,雷达可以分为激光雷达、毫米波雷达等多种雷达。各雷达所测量的精度不同。
当前,有多种方式可以实现对目标进行分类,如利用视觉***拍摄图像并对图像数据进行分析,以实现对图像中的目标进行分类。然而,这种方式受环境光影响较大,即环境光弱的情况下,会导致无法识别图像中的目标,也就无法对其进行分类,并且该种方式受距离影响,即远距离时,无法对目标进行识别。而雷达的信号传播特性并不受环境光和距离的影响。因此,如何利用雷达实现对远距离目标进行精确地检测并分类,成为本领域技术人员积极研究的课题。
发明内容
本申请实施例提供了一种目标分类方法和相关设备,可以实现利用雷达对远距离目标进行精确地分类。
第一方面,本申请实施例提供了一种目标分类方法,包括:
发射雷达信号,以检测环境中的目标;
获取所述目标基于所述雷达信号反馈的回波信号;
根据所述回波信号,得到所述目标的相关特征;
根据所述目标的相关特征,对所述目标进行分类。
可选的,所述目标的相关特征包括微动特征,所述根据所述目标的相关特征,对所述目标进行分类,包括:
根据所述目标的微动特征,将所述目标分类为处于微动状态的目标或者处于非微动状态的目标。
可选的,所述微动特征包括以下至少一个:
距离熵特征、噪声能量比特征、微动能量比特征;
其中,所述距离熵特征用于表示目标与雷达之间距离的不确定性;
所述噪声能量比特征用于表示所述多普勒信号中所述噪声信号的能量与所述多普勒信号的能量的比值;
所述微动能量比特征用于表示所述多普勒信号中所述微动信号的能量与所述多普勒信号的能量,或者所述微动信号的能量与所述噪声信号的能量的比值。
可选的,所述根据所述回波信号,得到所述目标的相关特征,包括:
对所述回波信号进行处理,以得到二维信号;其中,所述二维信号包括距离信号和多普勒信号;所述多普勒信号包括微动信号和噪声信号;
根据所述二维信号,得到所述目标的微动特征。
可选的,所述根据所述二维信号,得到所述目标的微动特征,包括:
根据公式(1),得到所述目标的距离熵特征;
其中,公式(1)为:
其中,feature1用于表示距离熵特征,M用于表示所获取的回波信号的帧数,k用于表示所获取的回波信号的帧序号,c(k)用于表示第k帧的回波信号中距离信号归一化的结果;其中,M为大于或等于1的整数。
可选的,所述c(k)是根据公式(2)确定的;
其中,公式(2)为:
其中,N用于表示计算窗口包括的帧数,n用于表示所述计算窗口中的任意一帧,range(k)用于表示第k帧的回波信号中的距离信号所表示的距离值;range(n)用于表示所述计算窗口内第n帧的回波信号中的距离信号所表示的距离值;其中,N为大于或等于1的整数。
可选的,所述根据所述二维信号,得到所述目标的微动特征,包括:
通过q次CLEAN算法去除所述二维信号中的多普勒信号包括的微动信号,以得到噪声信号,其中,q为大于或等于1的整数;
根据所述噪声信号的能量与所述多普勒信号的能量的比值,得到所述目标的噪声能量比特征。
可选的,所述根据所述二维信号,得到所述目标的微动特征,包括:
根据预设频段范围,确定所述二维信号中的多普勒信号中的与所述预设频段范围对应的微动信号;
根据所述微动信号的能量与所述多普勒信号的能量的比值,或者,所述微动信号的能量与所述多普勒信号中所述噪声信号的能量的比值,确定所述目标的微动能量比特征。
可选的,所述根据所述目标的相关特征,对所述目标进行分类,包括:
确定所述目标的相关特征对应的分类参数公式;
根据所述分类参数公式和所述目标的相关特征,对所述目标进行分类。
可选的,所述分类参数公式是由以下任意一种分类器训练得到的:
支持向量机(SVM)、相关向量机(RVM)、K最近邻分类算法(KNN)、神经元网络。
可选的,所述目标的相关特征包括RCS特征;
其中,所述RCS特征用于表示目标对所述雷达信号的反射程度。
第二方面,本申请实施例提供了一种目标分类装置,包括:
收发模块,用于发射雷达信号,以检测环境中的目标;获取所述目标基于 所述雷达信号反馈的回波信号;
处理模块,用于根据所述回波信号,得到所述目标的相关特征;根据所述目标的相关特征,对所述目标进行分类。
可选的,所述目标的相关特征包括微动特征,所述处理模块根据所述目标的相关特征,对所述目标进行分类,包括:
根据所述目标的微动特征,将所述目标分类为处于微动状态的目标或者处于非微动状态的目标。
可选的,所述微动特征包括以下至少一个:
距离熵特征、噪声能量比特征、微动能量比特征;
其中,所述距离熵特征用于表示目标与雷达之间距离的不确定性;
所述噪声能量比特征用于表示所述多普勒信号中所述噪声信号的能量与所述多普勒信号的能量的比值;
所述微动能量比特征用于表示所述多普勒信号中所述微动信号的能量与所述多普勒信号的能量,或者所述微动信号的能量与所述噪声信号的能量的比值。
可选的,所述处理模块根据所述回波信号,得到所述目标的相关特征,包括:
对所述回波信号进行处理,以得到二维信号;其中,所述二维信号包括距离信号和多普勒信号;所述多普勒信号包括微动信号和噪声信号;
根据所述二维信号,得到所述目标的微动特征。
可选的,所述处理模块根据所述二维信号,得到所述目标的微动特征,包括:
根据公式(1),得到所述目标的距离熵特征;
其中,公式(1)为:
其中,feature1用于表示距离熵特征,M用于表示所获取的回波信号的帧数,k用于表示所获取的回波信号的帧序号,c(k)用于表示第k帧的回波信号中距离信号归一化的结果;其中,M为大于或等于1的整数。
可选的,所述c(k)是根据公式(2)确定的;
其中,公式(2)为:
其中,N用于表示计算窗口包括的帧数,n用于表示所述计算窗口中的任意一帧,range(k)用于表示第k帧的回波信号中的距离信号所表示的距离值;range(n)用于表示所述计算窗口内第n帧的回波信号中的距离信号所表示的距离值;其中,N为大于或等于1的整数。
可选的,所述处理模块根据所述二维信号,得到所述目标的微动特征,包括:
通过q次CLEAN算法去除所述二维信号中的多普勒信号包括的微动信号,以得到噪声信号,其中,q为大于或等于1的整数;
根据所述噪声信号的能量与所述多普勒信号的能量的比值,得到所述目标的噪声能量比特征。
可选的,所述处理模块根据所述二维信号,得到所述目标的微动特征,包括:
根据预设频段范围,确定所述二维信号中的多普勒信号中的与所述预设频段范围对应的微动信号;
根据所述微动信号的能量与所述多普勒信号的能量的比值,或者,所述微动信号的能量与所述多普勒信号中所述噪声信号的能量的比值,确定所述目标的微动能量比特征。
可选的,所述处理模块根据所述目标的相关特征,对所述目标进行分类,包括:
确定所述目标的相关特征对应的分类参数公式;
根据所述分类参数公式和所述目标的相关特征,对所述目标进行分类。
可选的,所述分类参数公式是由以下任意一种分类器训练得到的:
支持向量机(SVM)、相关向量机(RVM)、K最近邻分类算法(KNN)、神经元网络。
可选的,所述目标的相关特征包括RCS特征;
其中,所述RCS特征用于表示目标对所述雷达信号的反射程度。
第三方面,本申请实施例提供了一种雷达,包括:
发射器;
接收器;
与所述发射器和所述接收器连接的处理器;以及
与所述处理器连接的存储器;
其中,所述发射器用于发射雷达信号,以检测环境中的目标;
所述接收器用于获取所述目标基于所述雷达信号反馈的回波信号;
所述处理器用于执行所述存储器中存储的计算机程序,以实现以下步骤:
根据所述回波信号,得到所述目标的相关特征;
根据所述目标的相关特征,对所述目标进行分类。
可选的,所述目标的相关特征包括微动特征,所述处理器用于实现根据所述目标的相关特征,对所述目标进行分类时,具体用于实现:
根据所述目标的微动特征,将所述目标分类为处于微动状态的目标或者处于非微动状态的目标。
可选的,所述微动特征包括以下至少一个:
距离熵特征、噪声能量比特征、微动能量比特征;
其中,所述距离熵特征用于表示目标与雷达之间距离的不确定性;
所述噪声能量比特征用于表示所述多普勒信号中所述噪声信号的能量与所述多普勒信号的能量的比值;
所述微动能量比特征用于表示所述多普勒信号中所述微动信号的能量与所述多普勒信号的能量,或者所述微动信号的能量与所述噪声信号的能量的比值。
可选的,所述处理器用于实现根据所述回波信号,得到所述目标的相关特征时,具体用于实现:
对所述回波信号进行处理,以得到二维信号;其中,所述二维信号包括距离信号和多普勒信号;所述多普勒信号包括微动信号和噪声信号;
根据所述二维信号,得到所述目标的微动特征。
可选的,所述处理器用于实现根据所述二维信号,得到所述目标的微动特征,具体用于实现:
根据公式(1),得到所述目标的距离熵特征;
其中,公式(1)为:
其中,feature1用于表示距离熵特征,M用于表示所获取的回波信号的帧数,k用于表示所获取的回波信号的帧序号,c(k)用于表示第k帧的回波信号中距离信号归一化的结果;其中,M为大于或等于1的整数。
可选的,所述c(k)是根据公式(2)确定的;
其中,公式(2)为:
其中,N用于表示计算窗口包括的帧数,n用于表示所述计算窗口中的任意一帧,range(k)用于表示第k帧的回波信号中的距离信号所表示的距离值;range(n)用于表示所述计算窗口内第n帧的回波信号中的距离信号所表示的距离值;其中,N为大于或等于1的整数。
可选的,所述处理器用于实现根据所述二维信号,得到所述目标的微动特征,具体用于实现:
通过q次CLEAN算法去除所述二维信号中的多普勒信号包括的微动信号,以得到噪声信号,其中,q为大于或等于1的整数;
根据所述噪声信号的能量与所述多普勒信号的能量的比值,得到所述目标的噪声能量比特征。
可选的,所述处理器用于实现根据所述二维信号,得到所述目标的微动特征,具体用于实现:
根据预设频段范围,确定所述二维信号中的多普勒信号中的与所述预设频段范围对应的微动信号;
根据所述微动信号的能量与所述多普勒信号的能量的比值,或者,所述微动信号的能量与所述多普勒信号中所述噪声信号的能量的比值,确定所述目标的微动能量比特征。
可选的,所述处理器用于实现根据所述目标的相关特征,对所述目标进行分类,具体用于实现:
确定所述目标的相关特征对应的分类参数公式;
根据所述分类参数公式和所述目标的相关特征,对所述目标进行分类。
可选的,所述分类参数公式是由以下任意一种分类器训练得到的:
支持向量机(SVM)、相关向量机(RVM)、K最近邻分类算法(KNN)、神经元网络。
可选的,所述目标的相关特征包括RCS特征;
其中,所述RCS特征用于表示目标对所述雷达信号的反射程度。
第四方面,本申请实施例提供了一种可读性存储介质,其特征在于,所述可读性存储介质中存储有计算机程序,所述计算机程序被处理器执行时,可以实现第一方面中的任意一种方法。
本申请实施例中,通过向雷达所在环境发射雷达信号,可以检测环境中的目标,并获取目标基于雷达信号反馈的回波信号,可以根据该回波信号,得到目标的相关特征,并可以根据所述目标的相关特征对目标进行分类。上述方式可以实现在离目标较远的情况下对目标进行分类。
图1是本申请实施例涉及的一种应用场景的示意图;
图2是本申请实施例提供的一种雷达的结构示意图;
图3是本申请实施例提供的一种目标分类方法的流程示意图;
图4是本申请实施例提供的另一种目标分类方法的流程示意图;
图5是本申请实施例提供的一种目标分类装置的模块组成示意图。
首先介绍几种本申请实施例涉及的应用场景。
请参阅图1,图1是本申请实施例涉及的一种应用场景。如图1所示,雷达可以安装在飞行器上,该飞行器可以是无人机(Unmanned Aerial Vehicle,UAV)或其他飞行器。在此场景下,雷达100可以安装在飞行器102的底部,其可以用于检测降落地点104的环境情况,并可以将降落地点作为目标,将其进行分类。例如,将降落地点分类为地面或水面。从而,可以将分类结果告知飞行器,以使飞行器调整降落地点,避免落入水面中。可以提升飞行器降落的智能化。进一步地,该雷达还可以实现其他功能,如飞行器在空中对其进行测高等。
或者,雷达可以安装于车辆上,用于检测车辆周围环境中的目标,并对目 标进行分类识别,例如,可以识别出车辆周围环境中的目标是路障、栏杆、人等。
当然,本申请实施例还可以涉及其他应用场景,在此不予赘述。
下面介绍本申请实施例提供的一种用以实现目标分类方法的雷达。
请参阅图2,图2是本申请实施例提供的一种雷达***的结构示意图。
如图2所示,雷达***200可以包括处理器201、发射器203、接收器205、总线207、接口209、存储器211等。
其中,处理器201分别与发射器203、接收器205、存储器211、接口109、电源***207连接。
当然,电源***207还可以根据设计需求与处理器201外的其他模块连接。或者,各模块根据设计需求,还可以与除处理器201之外的其他模块连接,在此不予限定。
其中,发射器203可以连接有发射天线,该发射天线可以是天线阵列或应用于雷达上的其他天线形式,在此不予限定。发射器用于发射雷达信号。例如,发射器可以发射激光雷达信号、毫米波雷达信号等。其中,发射器可以用于发射77GHz、24GHz或其他频段的毫米波雷达信号,在此不予限定。
可选的,发射器203可以包括发射控制单元,该控制单元用于控制实现与处理器201之间的指令交互,也可以控制发射器发射雷达信号等。
接收器205可以连接有接收天线。接收天线可以是天线阵列或应用于雷达上的其他天线形式,在此不予限定。接收器205用于接收发射器203发射的雷达信号经过目标反射后的回波信号,回波信号中携带的信息可以用于反应目标的特性、属性、运动特征等。接收器205可以通过单通道或多通道接收回波信号。
可选的,接收器205可以包括接收控制单元,该控制单元用于控制实现与处理器201之间的指令交互,也可以控制接收器接收回波信号等。
可选的,接收器205中还可以包括处理器,用于对接收的回波信号进行进一步处理,例如将回波信号处理成二维信号;或者,接收器可以将回波信号发送至处理器201中的信号处理器2011中进行进一步处理,本申请实施例在此不予限定。
发射器203与接收器205可以为独立的装置,或者,发射器203与接收器205可以集成为一个装置,作为雷达***200的前端。
处理器201可以包括信号处理器2011和数据处理器2013。
其中,信号处理器2011用于对回波信号进行处理,数据处理器2013用于对处理后的回波信号进行进一步处理,以实现对目标进行分类。
在此,信号处理器2011所实现的功能与数据处理器2013所实现的功能可以由独立的处理器实现,或者由处理器共同实现,在此不予限定。
处理器可以包括数字信号处理器(Digital Signal Processing,DSP)、微型处理器(Micro Processing Unit,MCU)、高级精简指令集机器(Advanced RISC Machine,ARM)等。在此,处理器可以是指处理器内核,或者处理器芯片。
上述处理器或处理器可以由硬件芯片实现。其中,硬件芯片可以是专用集成电路(英文:application-specific integrated circuit,缩写:ASIC),可编程逻辑器件(英文:programmable logic device,缩写:PLD)或其组合。上述PLD可以是复杂可编程逻辑器件(英文:complex programmable logic device,缩写:CPLD),现场可编程逻辑门阵列(英文:field-programmable gate array,缩写:FPGA),通用阵列逻辑(英文:generic array logic,缩写:GAL)或其任意组合。
进一步地,发射器203、接收器205、处理器201可以集成于一个硬件芯片中,或者各自由独立的硬件芯片实现,在此不予限定。
电源***207可以包括电源和电源管理模块。其中,电源可以为雷达200中的各模块供电,电源管理模块可以用于管理和控制各模块的供电情况。
接口209用于实现雷达***200与其他设备或装置进行通信。例如,雷达***200可以通过接口209将目标分类结果,传输给其他设备或装置,以使其他设备或装置基于该目标分类结果实现其他功能。
举例说明,当雷达***200安装于飞行器时,雷达***200可以通过接口209与飞行器中的飞行控制***、主控***或者其他控制***连接。在此以雷达***200通过接口209与主控***连接为例。雷达***200可以将目标分类结果通过接口209传输至主控***中,主控***通过分析该目标分类结果,可以进一步判断所检测目标是否适合降落,或者是否需要躲避所检测的目标等。并可以进一步地控制飞行器来实现上述功能。
其中,接口209可以包括串行外设接口(Serial Peripheral Interface Bus,SPI)2091、控制器局域网接口(Controller Area Network,CAN)2093、通用非同步首发传输接口(Universal Asynchronous Receiver/Transmitter,UART)2095等。当然,接口209还可以包括其他通信接口或输入输出接口,在此不予限定。
存储器211可以包括易失性存储器(英文:volatile memory),例如随机存取存储器(英文:random-access memory,缩写:RAM),如静态随机存取存储器(英文:static random-access memory,缩写:SRAM),双倍数据率同步动态随机存取存储器(英文:Double Data Rate Synchronous Dynamic Random Access Memory,缩写:DDR SDRAM)等;存储器也可以包括非易失性存储器(英文:non-volatile memory),例如快闪存储器(英文:flash memory),硬盘(英文:hard disk drive,缩写:HDD)或固态硬盘(英文:solid-state drive,缩写:SSD);存储器211还可以包括上述种类的存储器的组合。
存储器可以是独立的存储器,也可以是芯片(如处理器芯片)内部的存储器或某一具有存储功能的模块。
存储器中可以存储有计算机程序(如应用程序、功能模块)、计算机指令、操作***、数据、数据库等。存储器可以对其进行分区存储。
当然,雷达***200还可以包括其他部件,例如,在一些实现方式中,雷达***200还可以包括分类训练器等,以实现在线训练分类模型。对于雷达***200包括的其他部件,在此不予赘述。
基于上述雷达的应用场景,以及雷达的结构,下面介绍本申请实施例中的方法实施例。
请参阅图3,图3是本申请实施例提供的一种目标分类方法的流程示意图。如图3所示,该方法包括以下步骤。
步骤S301,发射雷达信号,以检测环境中的目标。
步骤S302,获取所述目标基于所述雷达信号反馈的回波信号。
示例性地,雷达***可以通过上述发射器和发射天线向所在环境发射雷达信号,以检测所述环境中的目标,并可以通过接收天线和接收器接收目标基于雷达信号反射的回波信号,并可以将该回波信号发送至处理器中做进一步处理。
相较于可以实现目标分类的视觉***,雷达可以实现远距离检测目标,并且不受环境光的影响。
步骤S303,根据所述回波信号,得到所述目标的相关特征。
示例性地,雷达可以根据回波信号,计算得到目标的相关特性。其中,目标的相关特性可以用于对目标进行分类。即目标的相关特性可以用于反应目标 的运动状态、目标的电磁特性、目标属性等。进而可以基于目标的相关特征,实现对目标进行分类。
例如,目标的相关特征可以包括微动特征,通过该微动特征,可以判断目标是否处于微动状态,进而可以将目标分类为处于微动状态的目标和处于非微动状态的目标。
本申请实施例中,微动可以理解为目标除自身运动外,附加在目标上的运动,或目标上局部部件的运动,例如,飞机螺旋桨的转动,人行走时手臂来回的摆动等。目标处于微动状态则表明目标具有微动。例如,微动状态可以包括水面的波动状态;非微动状态可以包括静止状态或者目标自身的运动状态。
一种应用场景中,当雷达应用于飞行器中,对降落点作为目标进行分类识别时,则可以理解为水面为非刚性,容易受环境,如风等因素的影响,其处于微动状态,刚性地面处于静止状态,进而可以利用得到的目标的微动特性,对目标进行分类,以确定当前检测到的降落点为地面或水面,进而可以使飞行器进一步判断是否能够进行降落。
或者,两个处于微动状态的目标,基于其微动特性地不同,可以对其进行分类。例如,另一种应用场景下,雷达应用于飞行器中,对沙漠中的地面进行分类识别时,可以基于流沙的微动特性和水面的微动特性,区别出流沙和水面。
本申请实施例中,目标的相关特征可以包括目标的微动特征、散射截面(Radar Cross Section,RCS)特征(也可称为目标的反射面特征)等中的至少一个。
目标的RCS特征可以用于反应目标对雷达信号的反射程度。由于目标根据其自身属性的不同,对雷达信号的反射程度不同,进而雷达***可以根据该RCS特征分类出不同的目标。
进一步地,上述两种特征可以结合,以分类目标是否处于微动状态,并可以通过特征结合,提升分类准确性。
对于目标的相关特性的具体得到方式,可以参加以下实施例。
步骤S304,根据所述目标的相关特征,对所述目标进行分类。
示例性地,当得到目标的相关特性后,可以基于分类器或训练好的分类模型,基于目标的相关特性对目标进行分类,以得到分类结果。其中,分类结果 可以是目标的分类属性。例如,在上述应用场景下,目标的分类属性包括地面属性或水面属性。其中,目标的地面属性用以表征地面为刚性地,即不受环境,如风等,的影响而运动。
可选的,可以根据目标的相关特征,对目标实现在线或离线分类。其中,对目标实现在线分类是指将目标的相关特征输入至雷达配置的分类器中,分类器基于获得的目标的相关特征,输出目标的分类属性。进一步地,还可以向分类器反馈输出结果的正确性,以使分类器调整分类算法。在此种情况下,随着检测目标的次数增多,分类器输出结果越准确。或者,对目标实现离线分类,是指利用多个相关特征以及分类器,训练出分类模型,将分类模型预存至雷达中,当得到目标的相关特征后,可以利用该相关特征和分类模型,得到分类结果,如目标的分类属性。
本申请实施例中,通过向雷达所在环境发射雷达信号,可以检测环境中的目标,并获取目标基于雷达信号反馈的回波信号,可以根据该回波信号,得到目标的相关特征,并可以根据所述目标的相关特征对目标进行分类。上述方式可以实现利用雷达信号在离目标较远的情况下对目标进行分类。
请参阅图4,图4是本申请实施例提供的另一种目标分类方法的流程示意图。如图4所示,该方法至少包括以下步骤。
步骤S401,发射雷达信号,以检测所述环境中的目标,并获取所述目标基于所述雷达信号反馈的回波信号。
本申请实施例的实现方式可以参见上述实施例中的相关描述,在此不予赘述。
步骤S402,对所述回波信号进行处理,以得到二维信号。
示例性地,可以对回波信号进行1维快速傅里叶变换(1-Dimensional Fast Fourier Transform,1DFFT),以得到雷达与目标之间的距离数据(也可以成为距离信号)。进一步地,可以对回波信号进行二维快速傅里叶变换(2-Dimensional Fast Fourier Transform,2DFFT),以得到多普勒信号。进而可以得到包括距离数据和多普勒信号的二维信号。
步骤S403,根据所述二维信号,得到所述目标的相关特征。
示例性地,可以根据二维信号中的多普勒信号和/或距离数据,得到目标的 微动特征。在此,二维信号中的多普勒信号可以包括微多普勒信号。即可以根据多普勒信号,得到目标的与多普勒信号相关的微动特征。
下面介绍几种本申请实施例中的微动特征的获得方式。
一、距离熵特征。
其中,距离熵特征可以用于表示目标与雷达之间距离的不确定性。距离熵特征中的熵值越大,则表明目标与雷达之间距离的不确定性越大;同理,距离熵特征中的熵值越小,则表明目标与雷达之间距离的不确定性越小。对于处于微动状态的目标,如水面,其距离熵特征则较大,相对地,对于处于非微动状态的目标,如地面,则其距离熵特征则较小,进而可以根据距离熵特征来区别目标是否处于微动状态。
一种实现方式中,距离熵特征(feature1)可以通过公式(1)确定。
其中,公式(1)为:
其中,feature1用于表示距离熵特征,M用于表示所获取的回波信号的帧数,k用于表示所获取的回波信号的帧序号,c(k)用于表示第k帧的回波信号中距离数据归一化的结果;其中,M为大于或等于1的整数。
具体地,雷达***可以持续地发射雷达信号,并可以连续接收多帧回波信号,雷达***可以根据每一帧回波信号和其对应的雷达信号,来确定与目标之间的距离,该距离可以表示为距离信号或距离数据,进一步地,可以对该距离信号进行归一化处理,如通过公式(2)进行处理,并将M个归一化处理的结果进行熵计算,如通过公式(1)进行熵计算,进而可以多个距离熵特征。
具体的,c(k)是根据公式(2)确定的。
其中,公式(2)为:
其中,N用于表示计算窗口包括的帧数,n用于表示所述计算窗口中的任意一帧,range(k)用于表示第k帧的回波信号中的距离值;range(n)用于表示所述计算窗口内第n帧的回波信号中的距离值;N为大于或等于1的整数。
N可以理解为计算窗口所容纳或所包含的帧数,或者,计算窗口的窗口长度,该窗口长度是通过帧数确定的。二、噪声能量比特征。
噪声能量比特征用于表示所述多普勒信号中所述噪声信号的能量与所述多普勒信号的能量的比值。其中,噪声能量比特征越大,则表明多普勒信号中的微动能量占比越小,进而目标处于微动状态的可能性越小;相应地,噪声能量比特征越小,则表明多普勒信号中的微动能量占比越大,进而目标处于微动状态的可能性越大。
具体的,如果需要得到多普勒信号中的噪声信号,可以首先确定多普勒信号中的微动信号,进而利用CLEAN算法将多普勒信号中的微动信号去除,以得到噪声信号。
具体的,微动信号可以通过多普勒信号在频域或时域内的谐波表示,可以通过在多普勒信号中寻找幅度值最大的谐波,以确定微动信号,并可以通过CLEAN算法将该谐波代表的微动信号从多普勒信号中减去,以得到剩余信号。噪声能量比可以通过该剩余信号的能量与多普勒信号的能量的比值来确定。若通过上述方式对多普勒信号进行q次CLEAN算法,q为大于或等于的整数,则可以得到q个比值,进而,噪声能量比特征可以由包括这q个比值的向量表示。
进一步地,可以利用该向量对目标进行分类,或利用向量中的一个或多个比值对目标进行分类。
其中,q可以是预设的,或者是基于回波信号中的谐波数量确定的,在此不予限定。
示例性地,可以利用以下公式(3)来确定谐波:
其中,y为该次谐波所表示的能量信号,其可以是时域信号或频域信号,A表示多普勒信号中的幅值,Φ表示相位,j表示时间,f
c表示多普勒频率,M为脉冲累积数。
能量比特征值可以通过以下公式(4)确定:
其中,R
i表示第i次基于CLEAN算法得到的能量比特征值,1≤i≤q,,L表示多普勒信号的频谱长度,S
r(n)表示多普勒信号,当f1至f2频段用于表示噪声信号所在频段时,S
r(n)表示多普勒信号中的噪声信号,S
i(n)表示第i次根据上述多普勒信号(也可以被称为原始信号)与谐波信号相减得到的剩余信号。
进而,能量比特征可以通过以下公式(5)予以表达:
feature2=(R
1,R
2,…,R
i);
其中,feature2即为能量比特征。
三、微动能量比特征。
微动能量比特征用于表示所述多普勒信号中所述微动信号的能量与所述多普勒信号的能量,或者所述微动信号的能量与所述噪声信号的能量的比值。其中,微动能量比特征越大,则表明多普勒信号中的微动信号占比越大,进而目标处于微动状态的可能性越大。
微动能量比特征可以通过以下公式(6)得到:
其中,feature3表示微动能量比特征,区间[f1,F1]表示微动信号能量集中频段,该区间可以是预设的,其中,该区间的确定可以与所涉及的应用场景相关。如不同的应用场景下,该预设区间不同。区间[f2,F2]表示噪声信号的频段,或者表示多普勒信号的频段。同样地,该区间可以是预设地,或者是基于区间[f1,F1]确定的。当区间[f2,F2]表示噪声信号的频段时,区间[f2,F2]可以是表示噪声信号的全部或其中部分频段,f为在区间[f1,F1]或在区间[f2,F2]的任意一个频点,P(f)为f频点对应的幅值。
通过上述方式可以获取目标的微动特征中的一种或多种。
可选的,还可以根据二维信号,得到RCS特征。对于得到方式,本申请实施例不予限定。
综上,可以根据二维信号,得到以下相关特征中的至少一种:
距离熵特征、能量比特征、微动能量比特征、RCS特征。
步骤S404,确定所述目标的相关特征对应的分类参数公式。
示例性地,由于得到的目标的相关特征具有不同的组合方式,因此,相关特征的不同的组合方式对应不同的分类参数公式。
在此,本申请实施例中的分类参数公式可以理解为是分类模型。
一种实现方式中,目标的相关特征的组合方式与分类参数公式的对应关系可以预存在雷达***中,如预存在图2所示雷达***的存储器211中。可以根据步骤S403得到的相关特征的组合,如组合中包括距离熵特征、RCS特征,则可以根据预存的对应关系,得到相关特征对应的分类参数公式。
步骤S405,根据所述分类参数公式和所述目标的相关特征,对所述目标进行分类。
示例性地,可以将目标的一个或多个相关特征作为分类参数公式的输入值,进而可以计算得到分类参数的参数值。进一步地,可以判断该分类参数的参数值所落入的分类区间,如果其落入某一分类区间内,则可以确定与该分类区间对应的目标分类。其中,分类区间与目标分类的对应关系可以是预训练得到的,其预存储在雷达中。或者,可以将目标的一个或多个相关特征作为分类参数公式的输入值,并可以直接得到目标的分类结果,进而可以确定目标分类。
目标的相关特征与分类参数公式之间存在对应关系。例如,针对不同的应用场景,得到的目标的相关特征不同,进而获取的分类参数公式不同。
具体的,可以根据计算的到的目标的相关特征的组合,来确定与该组合对应的目标分类。
相关特征的组合可以包括至少两种相关特征,相关特征的组合可以进一步地提升对目标分类的准确性。
上述确定目标的分类属性的方式可以通过雷达中的分类器实现,或者通过雷达中的其他装置实现,如通过雷达中的处理器执行相应程序实现。
上述对应关系,或者,上述分类参数公式可以是分类器通过训练算法得到的。其中,分类器中包括多种训练算法,可以基于分类器中的一种或多种训练算法得到上述对应关系或分类参数公式。
训练得到的对应关系或分类参数公式可以预存储在雷达中,当雷达实时检测到目标,并计算得到目标的相关特征后,可以根据雷达中的分类器或其他装 置,结合预存储的对应关系或分类参数公式来进一步地得到分类结果。或者,雷达中的分类器根据目标的相关特征和训练算法,直接得到分类结果。
本申请实施例中,分类器可以包括以下至少一种:
支持向量机(Support Vector Machine,SVM)、相关向量机(Relative Vector Machine,RVM)、K最近邻分类算法(k-nearest neighbors algorithm,KNN)、神经元网络等。
进一步地,雷达还可以输出目标的分类结果至其他装置,以便于其他装置根据目标的分类属性来做进一步处理。
本申请实施例仅以目标的分类属性为地面属性和水面属性为例进行说明,当然,通过本申请实施例中描述的实现方式,还可以实现其他应用场景下对目标的识别分类,在此不予限定。
下面结合上述实施方法,对本申请实施例涉及的一种应用场景进行说明。
如图1所示,雷达安装于飞行器底部,可以用于辅助飞行器进行自主降落。
例如,飞行器处于需要降落的场景下,触发雷达发射雷达信号,例如,雷达朝向降落点或降落地面发射雷达信号,降落点或降落地面在接收到该雷达信号后,对该雷达信号进行反射,反射的信号为回波信号。雷达在接收到该回波信号后,可以根据回波信号,得到目标的相关特征,具体的,可以基于上述算法,将回波信号处理成为二维信号,并可以根据二维信号,计算得到目标的微动特征、RCS特征等。例如,可以得到距离熵特征、能量比特征、微动能量比特征、RCS特征等中的至少一种,进而可以根据计算得到的目标的相关特征,对降落地面进行分类。将降落点分类为地面或水面中的一种。雷达可以将分类结果传输至飞行器的飞行控制***,飞行控制***可以根据分类结果来确定是否进行垂直降落。例如,当分类结果为地面时,则飞行控制***可以控制飞行器的动力装置进行垂直降落,当分类结果为水面时,则飞行控制***停止降落计划,或者规划新的降落路径等。
当然,上述应用场景仅为示例性地,对于本申请实施例涉及的其他应用场景,在此不予赘述。
请参阅图5,图5是本申请实施例提供的一种目标分类装置的模块组成示意图。如图5所示,该目标分类装置500可以包括收发模块501和处理模块503。
其中,收发模块501,用于发射雷达信号,以检测环境中的目标;获取所述目标基于所述雷达信号反馈的回波信号;
处理模块503,用于根据所述回波信号,得到所述目标的相关特征;根据所述目标的相关特征,对所述目标进行分类。
可选的,所述目标的相关特征包括微动特征,所述处理模块503根据所述目标的相关特征,对所述目标进行分类,包括:
根据所述目标的微动特征,将所述目标分类为处于微动状态的目标或者处于非微动状态的目标。
可选的,所述微动特征包括以下至少一个:
距离熵特征、噪声能量比特征、微动能量比特征;
其中,所述距离熵特征用于表示目标与雷达之间距离的不确定性;
所述噪声能量比特征用于表示所述多普勒信号中所述噪声信号的能量与所述多普勒信号的能量的比值;
所述微动能量比特征用于表示所述多普勒信号中所述微动信号的能量与所述多普勒信号的能量,或者所述微动信号的能量与所述噪声信号的能量的比值。
可选的,所述处理模块503根据所述回波信号,得到所述目标的相关特征,包括:
对所述回波信号进行处理,以得到二维信号;其中,所述二维信号包括距离信号和多普勒信号;所述多普勒信号包括微动信号和噪声信号;
根据所述二维信号,得到所述目标的微动特征。
可选的,所述处理模块503根据所述二维信号,得到所述目标的微动特征,包括:
根据公式(1),得到所述目标的距离熵特征;
其中,公式(1)为:
其中,feature1用于表示距离熵特征,M用于表示所获取的回波信号的帧数,k用于表示所获取的回波信号的帧序号,c(k)用于表示第k帧的回波信号中距离信号归一化的结果;其中,M为大于或等于1的整数。
可选的,所述c(k)是根据公式(2)确定的;
其中,公式(2)为:
其中,N用于表示计算窗口包括的帧数,n用于表示所述计算窗口中的任意一帧,range(k)用于表示第k帧的回波信号中的距离信号所表示的距离值;range(n)用于表示所述计算窗口内第n帧的回波信号中的距离信号所表示的距离值;其中,N为大于或等于1的整数。
可选的,所述处理模块503根据所述二维信号,得到所述目标的微动特征,包括:
通过q次CLEAN算法去除所述二维信号中的多普勒信号包括的微动信号,以得到噪声信号,其中,q为大于或等于1的整数;
根据所述噪声信号的能量与所述多普勒信号的能量的比值,得到所述目标的噪声能量比特征。
可选的,所述处理模块503根据所述二维信号,得到所述目标的微动特征,包括:
根据预设频段范围,确定所述二维信号中的多普勒信号中的与所述预设频段范围对应的微动信号;
根据所述微动信号的能量与所述多普勒信号的能量的比值,或者,所述微动信号的能量与所述多普勒信号中所述噪声信号的能量的比值,确定所述目标的微动能量比特征。
可选的,所述处理模块503根据所述目标的相关特征,对所述目标进行分类,包括:
确定所述目标的相关特征对应的分类参数公式;
根据所述分类参数公式和所述目标的相关特征,对所述目标进行分类。
可选的,所述分类参数公式是由以下任意一种分类器训练得到的:
支持向量机(SVM)、相关向量机(RVM)、K最近邻分类算法(KNN)、神经元网络。
可选的,所述目标的相关特征包括RCS特征;
其中,所述RCS特征用于表示目标对所述雷达信号的反射程度。
当然,目标分类装置还可以包括其他功能模块,在此不予限定。
上述功能模块可以由软件、硬件或其结合实现。例如,上述功能模块可以由计算机程序实现,或者,上述功能模块中的收发模块可以由由图2所示的发射器或接收器实现,上述功能模块中的处理模块可以由图2所示的处理器或者由其处理器执行计算机程序实现等,在此不予限定。
结合图2所示的雷达结构和上述实施方式,雷达中的处理器可以包括在处理器中,或者包括在发射器或接收器中,或者包括在其他装置中,其中,至少1个处理器可以用于执行上述实施例中的任意一种方法。
进一步地,本申请实施例还提供了一种可读性存储介质,其特征在于,所述可读性存储介质中存储有计算机程序,所述计算机程序被处理器执行时,可以实现第一方面中的任意一种方法。
本申请实施例中,通过向雷达所在环境发射雷达信号,可以检测环境中的目标,并获取目标基于雷达信号反馈的回波信号,可以根据该回波信号,得到目标的相关特征,并可以根据所述目标的相关特征对目标进行分类。上述方式可以实现在离目标较远的情况下对目标进行分类。
以上所述,仅为本发明的具体实施方式,但本发明的保护范围并不局限于此,任何熟悉本技术领域的技术人员在本发明揭露的技术范围内,可轻易想到变化或替换,都应涵盖在本发明的保护范围之内。因此,本发明的保护范围应以所述权利要求的保护范围为准。
Claims (22)
- 一种目标分类方法,其特征在于,包括:发射雷达信号,以检测环境中的目标;获取所述目标基于所述雷达信号反馈的回波信号;根据所述回波信号,得到所述目标的相关特征;根据所述目标的相关特征,对所述目标进行分类。
- 根据权利要求1所述的方法,其特征在于,所述目标的相关特征包括微动特征,所述根据所述目标的相关特征,对所述目标进行分类,包括:根据所述目标的微动特征,将所述目标分类为处于微动状态的目标或者处于非微动状态的目标。
- 根据权利要求2所述的方法,其特征在于,所述微动特征包括以下至少一个:距离熵特征、噪声能量比特征、微动能量比特征;其中,所述距离熵特征用于表示目标与雷达之间距离的不确定性;所述噪声能量比特征用于表示所述多普勒信号中所述噪声信号的能量与所述多普勒信号的能量的比值;所述微动能量比特征用于表示所述多普勒信号中所述微动信号的能量与所述多普勒信号的能量,或者所述微动信号的能量与所述噪声信号的能量的比值。
- 根据权利要求3所述的方法,其特征在于,所述根据所述回波信号,得到所述目标的相关特征,包括:对所述回波信号进行处理,以得到二维信号;其中,所述二维信号包括距离信号和多普勒信号;所述多普勒信号包括微动信号和噪声信号;根据所述二维信号,得到所述目标的微动特征。
- 根据权利要求4所述的方法,其特征在于,所述根据所述二维信号,得到所述目标的微动特征,包括:通过q次CLEAN算法去除所述二维信号中的多普勒信号包括的微动信号,以得到噪声信号,其中,q为大于或等于1的整数;根据所述噪声信号的能量与所述多普勒信号的能量的比值,得到所述目标的噪声能量比特征。
- 根据权利要求4所述的方法,其特征在于,所述根据所述二维信号,得到所述目标的微动特征,包括:根据预设频段范围,确定所述二维信号中的多普勒信号中的与所述预设频段范围对应的微动信号;根据所述微动信号的能量与所述多普勒信号的能量的比值,或者,所述微动信号的能量与所述多普勒信号中所述噪声信号的能量的比值,确定所述目标的微动能量比特征。
- 根据权利要求1-8任一项所述的方法,其特征在于,所述根据所述目标的相关特征,对所述目标进行分类,包括:确定所述目标的相关特征对应的分类参数公式;根据所述分类参数公式和所述目标的相关特征,对所述目标进行分类。
- 根据权利要求9所述的方法,其特征在于,所述分类参数公式是由以下任意一种分类器训练得到的:支持向量机(SVM)、相关向量机(RVM)、K最近邻分类算法(KNN)、神经元网络。
- 根据权利要求1-10任一项所述的方法,其特征在于,所述目标的相关 特征包括RCS特征;其中,所述RCS特征用于表示目标对所述雷达信号的反射程度。
- 一种雷达***,其特征在于,包括:发射器;接收器;与所述发射器和所述接收器连接的处理器;以及与所述处理器连接的存储器;其中,所述发射器用于发射雷达信号,以检测环境中的目标;所述接收器用于获取所述目标基于所述雷达信号反馈的回波信号;所述处理器用于执行所述存储器中存储的计算机程序,以实现以下步骤:根据所述回波信号,得到所述目标的相关特征;根据所述目标的相关特征,对所述目标进行分类。
- 根据权利要求12所述的雷达***,其特征在于,所述目标的相关特征包括微动特征,所述处理器用于实现根据所述目标的相关特征,对所述目标进行分类时,具体用于实现:根据所述目标的微动特征,将所述目标分类为处于微动状态的目标或者处于非微动状态的目标。
- 根据权利要求13所述的雷达***,其特征在于,所述微动特征包括以下至少一个:距离熵特征、噪声能量比特征、微动能量比特征;其中,所述距离熵特征用于表示目标与雷达之间距离的不确定性;所述噪声能量比特征用于表示所述多普勒信号中所述噪声信号的能量与所述多普勒信号的能量的比值;所述微动能量比特征用于表示所述多普勒信号中所述微动信号的能量与所述多普勒信号的能量,或者所述微动信号的能量与所述噪声信号的能量的比值。
- 根据权利要求14所述的雷达***,其特征在于,所述处理器用于实现根据所述回波信号,得到所述目标的相关特征时,具体用于实现:对所述回波信号进行处理,以得到二维信号;其中,所述二维信号包括距离信号和多普勒信号;所述多普勒信号包括微动信号和噪声信号;根据所述二维信号,得到所述目标的微动特征。
- 根据权利要求15所述的雷达***,其特征在于,所述处理器用于实现根据所述二维信号,得到所述目标的微动特征,具体用于实现:通过q次CLEAN算法去除所述二维信号中的多普勒信号包括的微动信号,以得到噪声信号,其中,q为大于或等于1的整数;根据所述噪声信号的能量与所述多普勒信号的能量的比值,得到所述目标的噪声能量比特征。
- 根据权利要求15所述的雷达***,其特征在于,所述处理器用于实现根据所述二维信号,得到所述目标的微动特征,具体用于实现:根据预设频段范围,确定所述二维信号中的多普勒信号中的与所述预设频段范围对应的微动信号;根据所述微动信号的能量与所述多普勒信号的能量的比值,或者,所述微动信号的能量与所述多普勒信号中所述噪声信号的能量的比值,确定所述目标的微动能量比特征。
- 根据权利要求12-19任一项所述的雷达***,其特征在于,所述处理器用于实现根据所述目标的相关特征,对所述目标进行分类,具体用于实现:确定所述目标的相关特征对应的分类参数公式;根据所述分类参数公式和所述目标的相关特征,对所述目标进行分类。
- 根据权利要求21所述的雷达***,其特征在于,所述分类参数公式是由以下任意一种分类器训练得到的:支持向量机(SVM)、相关向量机(RVM)、K最近邻分类算法(KNN)、神经元网络。
- 根据权利要求12-21任一项所述的雷达***,其特征在于,所述目标的相关特征包括RCS特征;其中,所述RCS特征用于表示目标对所述雷达信号的反射程度。
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811076412.8A CN110907906B (zh) | 2018-09-14 | 2018-09-14 | 目标分类方法和相关设备 |
CN201811076412.8 | 2018-09-14 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2020052441A1 true WO2020052441A1 (zh) | 2020-03-19 |
Family
ID=69777340
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2019/103227 WO2020052441A1 (zh) | 2018-09-14 | 2019-08-29 | 目标分类方法和相关设备 |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN110907906B (zh) |
WO (1) | WO2020052441A1 (zh) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112926526A (zh) * | 2021-03-30 | 2021-06-08 | 矽典微电子(上海)有限公司 | 基于毫米波雷达的停车检测方法及*** |
CN117250594A (zh) * | 2023-11-17 | 2023-12-19 | 南京威翔科技有限公司 | 一种雷达目标分类识别方法 |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112034464B (zh) * | 2020-08-31 | 2024-06-25 | 上海英恒电子有限公司 | 一种目标分类方法 |
CN113050057B (zh) * | 2021-03-18 | 2022-12-30 | 森思泰克河北科技有限公司 | 一种人员检测方法、装置及终端设备 |
CN113341407B (zh) * | 2021-06-02 | 2024-02-06 | 中国水产科学研究院南海水产研究所 | 一种基于雷达探测的渔业捕捞追踪***及方法 |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103914703A (zh) * | 2014-05-12 | 2014-07-09 | 西安电子科技大学 | 一种行人与车辆微动目标的分类识别方法 |
EP3349038A1 (en) * | 2017-01-12 | 2018-07-18 | Delphi Technologies, Inc. | Method to classify objects in a vehicle environment based on radar detections |
Family Cites Families (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
FR2669116B1 (fr) * | 1990-11-09 | 1993-04-23 | Thomson Csf | Procede de reconnaissance d'une cible aerienne a partir de son echo radar. |
SE509733C2 (sv) * | 1996-07-05 | 1999-03-01 | Foersvarets Forskningsanstalt | Sätt att detektera och klassificera objekt med hjälp av radar |
CN105866755B (zh) * | 2016-05-30 | 2018-06-26 | 中国人民解放军国防科学技术大学 | 微波暗室内脉冲体制雷达目标回波信息重构方法 |
CN206209102U (zh) * | 2016-09-22 | 2017-05-31 | 北京聚速微波技术有限公司 | 基于微波雷达的周界防护*** |
CN107886121A (zh) * | 2017-11-03 | 2018-04-06 | 北京清瑞维航技术发展有限公司 | 基于多波段雷达的目标识别方法、装置及*** |
CN108008366B (zh) * | 2017-12-01 | 2020-08-04 | 北京润科通用技术有限公司 | 一种雷达目标回波模拟方法及*** |
CN108051813B (zh) * | 2017-12-04 | 2021-12-07 | 湖南华诺星空电子技术有限公司 | 用于低空多目标分类识别的雷达探测***及方法 |
CN108107413B (zh) * | 2018-01-09 | 2020-10-27 | 中国空空导弹研究院 | 一种雷达目标模拟器校准*** |
CN108459311A (zh) * | 2018-03-22 | 2018-08-28 | 三明学院 | 基于Hough变换的卷积神经网络下微多普勒目标分类方法 |
-
2018
- 2018-09-14 CN CN201811076412.8A patent/CN110907906B/zh active Active
-
2019
- 2019-08-29 WO PCT/CN2019/103227 patent/WO2020052441A1/zh active Application Filing
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103914703A (zh) * | 2014-05-12 | 2014-07-09 | 西安电子科技大学 | 一种行人与车辆微动目标的分类识别方法 |
EP3349038A1 (en) * | 2017-01-12 | 2018-07-18 | Delphi Technologies, Inc. | Method to classify objects in a vehicle environment based on radar detections |
Non-Patent Citations (4)
Title |
---|
HOU, QINGKAI: "Methods And Applications of Compressed Sensing Based Isar Imaging for Space Target", INFORMATION & TECHNOLOGY, CHINA DOCTORAL DISSERTATIONS FULL-TEXT DATABASE(ELECTRONIC JOURNAL), 28 February 2017 (2017-02-28), pages 103 ; 104, ISSN: 1674-022X * |
LI, YAN-BING ET AL.: "Ground Targets Classification Based on Micro-Doppler Effect", JOURNAL OF ELECTRONICS & INFORMATION TECHNOLOGY, vol. 32, no. 12, 31 December 2010 (2010-12-31), pages 2848 - 2853, XP055695493, ISSN: 1009-5896 * |
LIN, PING: "SAR/GMTI Vehicle Target Classification Based on Micro-doppler Feature", INFORMATION & TECHNOLOGY, CHINA MASTER'S THESES FULL-TEXT DATABASE(ELECTRONIC JOURNAL), 28 February 2018 (2018-02-28), pages I ; 6 - 7 ; 10, ISSN: 1674-0246 * |
YANG, LEI ET AL.: "Study on Moving Vehicles Classification Based on Micro-Doppler Signatures", FIRE CONTROL RADAR TECHNOLOGY, vol. 43, no. 3, 30 September 2014 (2014-09-30), pages 36 - 39 ; 58, ISSN: 1008-8652 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112926526A (zh) * | 2021-03-30 | 2021-06-08 | 矽典微电子(上海)有限公司 | 基于毫米波雷达的停车检测方法及*** |
CN112926526B (zh) * | 2021-03-30 | 2023-12-29 | 矽典微电子(上海)有限公司 | 基于毫米波雷达的停车检测方法及*** |
CN117250594A (zh) * | 2023-11-17 | 2023-12-19 | 南京威翔科技有限公司 | 一种雷达目标分类识别方法 |
CN117250594B (zh) * | 2023-11-17 | 2024-01-30 | 南京威翔科技有限公司 | 一种雷达目标分类识别方法 |
Also Published As
Publication number | Publication date |
---|---|
CN110907906A (zh) | 2020-03-24 |
CN110907906B (zh) | 2023-01-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2020052441A1 (zh) | 目标分类方法和相关设备 | |
Ezuma et al. | Radar cross section based statistical recognition of UAVs at microwave frequencies | |
WO2019119195A1 (zh) | 目标信号检测方法、设备、无人机及农业无人机 | |
US12044796B2 (en) | Method and apparatus for identifying behavior of target, and radar system | |
US20230333209A1 (en) | Gesture recognition method and apparatus | |
Bhatia et al. | Object classification technique for mmWave FMCW radars using range-FFT features | |
Tivive et al. | An improved SVD-based wall clutter mitigation method for through-the-wall radar imaging | |
CN115061113B (zh) | 用于雷达的目标检测模型训练方法、装置及存储介质 | |
Ezuma et al. | Comparative analysis of radar-cross-section-based UAV recognition techniques | |
US20190187251A1 (en) | Systems and methods for improving radar output | |
KR20220141748A (ko) | 레이더 신호로부터 표적 정보를 추출하는 방법 및 컴퓨터 판독가능 저장 매체 | |
Huang et al. | Yolo-ore: A deep learning-aided object recognition approach for radar systems | |
US20190187253A1 (en) | Systems and methods for improving lidar output | |
WO2022093565A1 (en) | Feature extraction for remote sensing detections | |
Belyaev et al. | Object detection in an urban environment using 77GHz radar | |
WO2021218347A1 (zh) | 一种聚类的方法和装置 | |
US11280899B2 (en) | Target recognition from SAR data using range profiles and a long short-term memory (LSTM) network | |
CN109738890A (zh) | 一种基于弹载双基sar距离多普勒图像生成地距图的方法 | |
CN115131756A (zh) | 一种目标检测方法及装置 | |
Lee et al. | Identification of a flying multi-rotor platform by high resolution ISAR through an experimental analysis | |
Ezuma | UAV detection and classification using radar, radio frequency and machine learning techniques | |
Gomez et al. | Classification of Radar Targets Based on Micro-Doppler Features Using High Frequency High Resolution Radar Signatures | |
Li et al. | Dynamic gesture recognition method based on millimeter-wave radar | |
Mardiev et al. | Convolutional Neural Networks for Processing Micro-Doppler Signatures and Range-Azimuth Radar Maps of Frequency Modulated Continuous Wave Radars | |
Raimondi et al. | mmDetect: YOLO-based Processing of mm-Wave Radar Data for Detecting Moving People |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 19861129 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 19861129 Country of ref document: EP Kind code of ref document: A1 |