Obstacle detection method and system based on frequency modulated continuous wave laser radar

Abstract

The invention relates to the technical field of laser radars, in particular to an obstacle detection method and system based on a frequency-modulated continuous wave laser radar, and the method comprises the steps: transmitting a chaotic frequency-modulated continuous wave signal to a detection region, and obtaining a reflection echo signal of a target object; performing decoupling calculation on the difference frequency electric signal to obtain a target time frequency distribution matrix; according to the target time-frequency distribution matrix, predicting the Doppler frequency shift amount of the target object at the next moment; based on the chaos frequency modulation adjustment signal, obtaining a distance measurement deviation compensation amount; the chaos frequency modulation adjustment signal is obtained by dynamically adjusting the chaos frequency modulation slope of the chaos frequency modulation continuous wave signal through the Doppler frequency shift amount; and compensating the difference frequency electric signal frequency value of the target object according to the distance measurement deviation compensation amount and the environmental noise data so as to solve obstacle distance speed detection information. According to the invention, the Doppler frequency shift amount is predicted in real time and the chaotic frequency modulation continuous wave signal parameters are dynamically adjusted, so that the accurate compensation of the distance measurement deviation is realized, and the accuracy of obstacle distance speed detection is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of lidar, and particularly to an obstacle detection method and system based on a frequency-modulated continuous-wave lidar. Background Art

[0002] In many fields such as autonomous driving, robot navigation, and security monitoring, obstacle detection technology plays a crucial role. As a high-precision detection device, the frequency-modulated continuous-wave lidar has been widely used in obstacle detection due to its advantages such as high resolution and strong anti-interference ability.

[0003] However, there are still some problems to be solved urgently in the existing obstacle detection methods based on frequency-modulated continuous-wave lidar. During the detection process, when the target object is in a moving state, due to the influence of the Doppler effect, the frequency of the echo signal received by the lidar will shift. This frequency shift will cause a deviation in the ranging result, thereby affecting the accuracy of obstacle distance measurement. Especially in the case where the target moving speed is relatively fast or the moving direction is complex, the ranging deviation caused by the frequency shift will be more obvious, reducing the reliability of obstacle detection. In addition, when dealing with the Doppler frequency shift, the existing methods often lack real-time performance and dynamic adjustment ability. Traditional frequency shift compensation methods are usually based on fixed parameters or preset models, and it is difficult to flexibly adjust according to the real-time motion state of the target. This leads to poor compensation effects when the target motion state changes in practical applications, and the ranging deviation caused by the frequency shift cannot be effectively eliminated, thus affecting the accuracy and efficiency of obstacle detection.

[0004] In summary, there are many deficiencies in the existing obstacle detection methods. There is an urgent need for an obstacle detection method based on a frequency-modulated continuous-wave lidar that can predict the Doppler frequency shift amount in real time and dynamically adjust the signal parameters to compensate for the ranging deviation, improving the detection accuracy and reliability in complex scenarios. Summary of the Invention

[0005] To solve the above technical problems, the present invention provides an obstacle detection method and system based on a frequency-modulated continuous-wave lidar.

[0006] In a first aspect, the present invention provides an obstacle detection method based on a frequency-modulated continuous-wave lidar, and the method includes the following steps: Transmit a chaotic frequency-modulated continuous-wave signal with non-periodic characteristics to a detection area, and obtain the reflected echo signal of the target object; Perform decoupling calculation on the difference-frequency electrical signal to obtain the target time-frequency distribution matrix; wherein, the difference-frequency electrical signal is extracted by performing coherent optical mixing on the reflected echo signal and the chaotic frequency-modulated continuous-wave signal; Predicting the Doppler frequency shift of the target object at the next moment according to the target time-frequency distribution matrix; Based on the chaotic frequency modulation adjustment signal, a ranging deviation compensation amount is calculated; wherein the chaotic frequency modulation adjustment signal is obtained by dynamically adjusting the chaotic frequency modulation slope of the chaotic frequency modulation continuous wave signal by the Doppler frequency shift amount; The target object difference frequency electric signal frequency value is compensated according to the ranging deviation compensation amount and the environmental noise data to calculate the obstacle distance and speed detection information.

[0007] In a further embodiment, the step of emitting a chaotic frequency-modulated continuous wave signal having a non-periodic characteristic to the detection area to obtain a reflected echo signal of the target object includes: Generate an initial chaotic sequence with non-periodic characteristics, and dynamically adjust the instantaneous frequency offset of the initial chaotic sequence through a phase modulator to generate a chaotic frequency-modulated continuous wave signal; The chaotic frequency-modulated continuous wave signal is modulated by adopting a hybrid modulation method combining orthogonal frequency division multiplexing and high-order orthogonal amplitude modulation to obtain a modulated chaotic frequency-modulated continuous wave signal; The modulated chaotic frequency-modulated continuous wave signal is loaded onto a multi-beam transducer array for space-time coding, and a chaotic encrypted positioning pilot sequence is embedded into the chaotic frequency-modulated continuous wave signal loaded onto the multi-beam transducer array to obtain a chaotic frequency-modulated continuous wave transmission signal; The chaotic frequency-modulated continuous wave transmission signal is directionally transmitted to the detection area through the multi-beam transducer array, and the reflected echo signal of the target object reflected from the detection area is received.

[0008] In a further embodiment, the process of acquiring the difference frequency electrical signal is specifically as follows: performing time domain alignment on the reflected echo signal according to timestamp information of the reflected echo signal and the chaotic frequency modulation continuous wave signal to obtain a target reflected echo time domain aligned signal; Using a mixer to perform coherent optical mixing on the target reflected echo time-domain aligned signal and the chaotic frequency-modulated continuous wave signal to obtain an original mixed signal; The high-frequency component in the original mixed signal is filtered out by a low-pass filter to extract the difference frequency electrical signal.

[0009] In a further embodiment, the target time-frequency distribution matrix is extracted by a deep neural network, and the deep neural network includes a first channel branch and a second channel branch adopting a dual-channel parallel branch architecture, the first channel branch is a branch structure composed of a cascade of a long short-term memory network and a three-dimensional convolutional neural network, and the second channel branch is a branch structure based on fractional Fourier transform.

[0010] In a further embodiment, the step of decoupling and calculating the difference-frequency electrical signal to obtain the target time-frequency distribution matrix includes: According to the amplitude range of the difference-frequency electrical signal, the difference-frequency electrical signal is normalized by the maximum-minimum normalization algorithm to obtain a normalized difference-frequency electrical signal; The long short-term memory network is used to capture the temporal dependence relationship in the normalized difference-frequency electrical signal, and the temporal features of the difference-frequency electrical signal are extracted; The three-dimensional convolutional neural network is used to perform space-frequency domain convolution operation on the temporal features of the difference-frequency electrical signal to obtain a spatio-temporal frequency joint feature vector; The spatio-temporal frequency joint feature vector is input into the fully connected layer for non-linear transformation to obtain a channel state parameter vector; Based on the channel state parameter vector, the optimal fractional Fourier transform order is calculated by using the gradient descent method, and the normalized difference-frequency electrical signal is subjected to fractional Fourier transform according to the optimal fractional Fourier transform order to obtain a preliminary time-frequency distribution matrix; The morphological filtering algorithm is used to perform spatial domain filtering on the preliminary time-frequency distribution matrix to obtain the target time-frequency distribution matrix.

[0011] In a further embodiment, the step of predicting the Doppler frequency shift amount of the target object at the next moment according to the target time-frequency distribution matrix: The peak search algorithm is used to search for energy concentration points in the target time-frequency distribution matrix, and the characteristic parameters of the energy concentration points are extracted; the characteristic parameters of the energy concentration points include the time center position, the frequency center value, the energy intensity, and the time-frequency distribution variance; The state vector of the Kalman filter is defined, and the state equation is established according to the state vector, and the frequency center position and the time center position in the characteristic parameters of the energy concentration point are used as the observation variables to establish the observation equation; the state vector includes the target distance, the radial velocity, and the radial acceleration to be estimated; The state equation and the observation equation are integrated to form a target motion state space model; According to the characteristic parameters of the energy concentration point, the target motion state is initially estimated by using the target motion state space model to obtain the initial state estimation value at the previous moment; According to the target state estimation value at the previous moment and the state equation, the target state estimation value at the current moment is predicted by using the Kalman filter; The radial velocity information is calculated according to the target state estimation value at the current moment, and the radial velocity information is converted into a frequency shift amount by using the Doppler effect principle to obtain the Doppler frequency shift amount at the next moment.

[0012] In a further embodiment, the process of obtaining the chaotic frequency modulation adjustment signal is specifically as follows: Process the chaotic frequency modulation continuous wave signal using a spectrum analysis method to extract the initial chaotic frequency modulation slope; Calculate the chaotic frequency modulation slope adjustment amount according to the Doppler frequency shift amount, and obtain the corrected chaotic frequency modulation slope value based on the chaotic frequency modulation slope adjustment amount and the initial chaotic frequency modulation slope; Generate a chaotic frequency modulation adjustment signal using the corrected chaotic frequency modulation slope value and a preset signal center frequency based on the logistic chaotic map.

[0013] In a further embodiment, the steps of calculating the ranging deviation compensation amount based on the chaotic frequency modulation adjustment signal include: Use Hilbert transform to extract the first instantaneous frequency sequence of the chaotic frequency modulation adjustment signal and the second instantaneous frequency sequence of a preset standard chaotic frequency modulation signal; Align the first instantaneous frequency sequence and the second instantaneous frequency sequence in the time domain to obtain the corresponding first time-domain aligned instantaneous frequency sequence and second time-domain aligned instantaneous frequency sequence; Perform a point-by-point difference operation on the first time-domain aligned instantaneous frequency sequence and the second time-domain aligned instantaneous frequency sequence to obtain a frequency difference sequence; Calculate the ranging deviation at each moment according to the frequency difference sequence and the laser wavelength to obtain the ranging deviation compensation amount.

[0014] In a further embodiment, the steps of compensating the frequency value of the target object's difference frequency electrical signal according to the ranging deviation compensation amount and environmental noise data to calculate the obstacle distance and speed detection information include: Collect environmental noise data in real time, and perform preliminary compensation on the original frequency value of the difference frequency electrical signal according to the ranging deviation compensation amount to obtain the compensated frequency value of the difference frequency electrical signal; Perform power spectral density estimation on the environmental noise data to obtain the environmental noise power spectral density, and construct an adaptive filter using the least mean square error algorithm according to the environmental noise power spectral density; Use the adaptive filter to correct the noise of the compensated frequency value of the difference frequency electrical signal to obtain the corrected frequency value of the difference frequency electrical signal; Perform a fractional Fourier transform on the corrected frequency value of the difference frequency electrical signal to obtain the frequency offset characteristics of the difference frequency electrical signal in the fractional Fourier transform domain; Calculate the initial distance estimate value according to the frequency offset characteristics of the difference frequency electrical signal, the frequency modulation bandwidth, and the frequency modulation period, and calculate the target speed according to the deviation between the corrected frequency value of the difference frequency electrical signal and the original frequency value of the difference frequency electrical signal; Obtain the obstacle distance-velocity detection information based on the initial distance estimation value, the target velocity, and the geometric position of the frequency-modulated continuous-wave lidar.

[0015] In a second aspect, the present invention provides an obstacle detection system based on a frequency-modulated continuous-wave lidar. The system includes: A signal acquisition module, configured to transmit a chaotic frequency-modulated continuous-wave signal with non-periodic characteristics to a detection area and acquire the reflected echo signal of a target object. A difference frequency analysis module, configured to perform decoupling calculation on the difference frequency electrical signal to obtain a target time-frequency distribution matrix; wherein, the difference frequency electrical signal is extracted by performing coherent optical mixing on the reflected echo signal and the chaotic frequency-modulated continuous-wave signal. A frequency shift estimation module, configured to predict the Doppler frequency shift amount of the target object at the next moment according to the target time-frequency distribution matrix. A deviation compensation module, configured to calculate a ranging deviation compensation amount based on a chaotic frequency modulation adjustment signal; wherein, the chaotic frequency modulation adjustment signal is obtained by dynamically adjusting the chaotic frequency modulation slope of the chaotic frequency-modulated continuous-wave signal by the Doppler frequency shift amount. A target detection module, configured to compensate the frequency value of the difference frequency electrical signal of the target object according to the ranging deviation compensation amount and environmental noise data to calculate the obstacle distance-velocity detection information.

[0016] The present invention provides an obstacle detection method and system based on a frequency-modulated continuous-wave lidar. The method includes transmitting a chaotic frequency-modulated continuous-wave signal with non-periodic characteristics to a detection area to acquire the reflected echo signal of a target object; performing decoupling calculation on the difference frequency electrical signal to obtain a target time-frequency distribution matrix; wherein, the difference frequency electrical signal is extracted by performing coherent optical mixing on the reflected echo signal and the chaotic frequency-modulated continuous-wave signal; predicting the Doppler frequency shift amount of the target object at the next moment according to the target time-frequency distribution matrix; calculating a ranging deviation compensation amount based on a chaotic frequency modulation adjustment signal; wherein, the chaotic frequency modulation adjustment signal is obtained by dynamically adjusting the chaotic frequency modulation slope of the chaotic frequency-modulated continuous-wave signal by the Doppler frequency shift amount; compensating the frequency value of the difference frequency electrical signal of the target object according to the ranging deviation compensation amount and environmental noise data to calculate the obstacle distance-velocity detection information. Compared with the prior art, by predicting the Doppler frequency shift amount in real time and dynamically adjusting the parameters of the chaotic frequency-modulated continuous-wave signal, the method realizes precise compensation for the ranging deviation, significantly improves the accuracy of obstacle distance-velocity detection in complex scenarios, and can meet the requirements of high-dynamic scenarios such as autonomous driving. Description of the Drawings

[0017] Figure 1 is a schematic flowchart of an obstacle detection method based on a frequency-modulated continuous-wave lidar provided by an embodiment of the present invention; Figure 2 It is a block diagram of an obstacle detection system based on a frequency-modulated continuous-wave lidar provided by an embodiment of the present invention.

[0018] Explanation of reference numerals: 101, signal acquisition module; 102, difference frequency analysis module; 103, frequency shift estimation module; 104, deviation compensation module; 105, target detection module. Specific embodiments

[0019] The following specifically clarifies the implementation manner of the present invention in conjunction with the accompanying drawings. The presentation of the embodiments is only for illustrative purposes and should not be construed as a limitation of the present invention. The accompanying drawings are for reference and illustration only and do not constitute a limitation on the scope of patent protection of the present invention, because many changes can be made to the present invention without departing from its spirit and scope.

[0020] Figure 1 It is a schematic flowchart of an obstacle detection method based on a frequency-modulated continuous-wave lidar provided by an embodiment of the present invention. An embodiment of the present invention provides an obstacle detection method based on a frequency-modulated continuous-wave lidar, as Figure 1 shown, the method includes the following steps: S1. Transmit a chaotic frequency-modulated continuous-wave signal with non-periodic characteristics to the detection area to obtain the reflected echo signal of the target object.

[0021] In some embodiments, the step of transmitting a chaotic frequency-modulated continuous-wave signal with non-periodic characteristics to the detection area to obtain the reflected echo signal of the target object includes: Using a chaotic sequence generator to generate an initial chaotic sequence with non-periodic characteristics, and dynamically adjusting the instantaneous frequency offset of the initial chaotic sequence through a phase modulator to generate a chaotic frequency-modulated continuous-wave signal; Modulating the chaotic frequency-modulated continuous-wave signal by a hybrid modulation method combining orthogonal frequency division multiplexing and high-order quadrature amplitude modulation to obtain a modulated chaotic frequency-modulated continuous-wave signal; Loading the modulated chaotic frequency-modulated continuous-wave signal onto a multi-beam transducer array for space-time coding, and embedding a chaotic-encrypted positioning pilot sequence in the chaotic frequency-modulated continuous-wave signal loaded onto the multi-beam transducer array to obtain a chaotic frequency-modulated continuous-wave transmission signal; Directly transmitting the chaotic frequency-modulated continuous-wave transmission signal to the detection area through the multi-beam transducer array, and receiving the reflected echo signal of the target object reflected from the detection area.

[0022] Specifically, in this embodiment, the logistic mapping algorithm is used as a chaotic sequence generator to generate an initial chaotic sequence with aperiodic characteristics by means of logistic mapping. After receiving the initial chaotic sequence, the phase modulator maps the numerical values of the initial chaotic sequence into instantaneous frequency offsets, and modulates the phase of the continuous-wave laser of the initial chaotic sequence according to the instantaneous frequency offsets, so as to generate a chaotic frequency modulation continuous-wave signal according to the instantaneous frequency of the modulated laser signal. Since orthogonal frequency division multiplexing is a multi-carrier modulation technology, it divides a high-speed data stream into multiple low-speed sub-data streams, and then modulates these sub-data streams onto multiple mutually orthogonal sub-carriers for transmission respectively, which can improve the anti-interference ability and spectrum utilization rate of the signal. Therefore, in this embodiment, the generated chaotic frequency modulation continuous-wave signal is segmented at a preset time interval to obtain multiple signal segments, and a fast Fourier transform is performed on each signal segment to convert each signal segment into a frequency-domain signal. The spectrum is divided into multiple sub-carriers in the frequency domain, and each sub-carrier corresponds to a sub-data stream. In this embodiment, according to the frequency characteristics of each sub-carrier, the spectrum energy of the chaotic frequency modulation continuous-wave signal is allocated to each sub-carrier, so as to obtain a signal after orthogonal frequency division multiplexing processing. The signal after orthogonal frequency division multiplexing processing is distributed on multiple orthogonal sub-carriers in the frequency domain. For the signal on each sub-carrier after orthogonal frequency division multiplexing processing, in this embodiment, modulation is performed according to the constellation diagram rules of high-order quadrature amplitude modulation. For example, for 16-QAM modulation, there are sixteen constellation points on the constellation diagram, and each constellation point corresponds to four bits of information. In this embodiment, the amplitude and phase of the signal on the sub-carrier are adjusted to the corresponding constellation point positions, so that the signal contains specific bit information. Specifically, in this embodiment, according to the bit sequence to be transmitted, the corresponding constellation point is found in the constellation diagram, and then the amplitude and phase of the signal on the sub-carrier are adjusted to match the amplitude and phase of the constellation point, so as to obtain a signal after high-order quadrature amplitude modulation processing, and the inverse fast Fourier transform is performed on the high-order quadrature amplitude modulation processed signals of each sub-carrier to convert the frequency-domain signal back into a time-domain signal. After completing the orthogonal frequency division multiplexing and high-order quadrature amplitude modulation of each sub-carrier, the signals of all sub-carriers are combined to finally obtain a chaotic frequency modulation continuous-wave signal after hybrid modulation. In this embodiment, orthogonal frequency division multiplexing and high-order quadrature amplitude modulation are combined, making full use of the anti-interference and spectrum utilization rate advantages of orthogonal frequency division multiplexing and the high data transmission rate advantage of high-order quadrature amplitude modulation, and improving the transmission performance of the chaotic frequency modulation continuous-wave signal.

[0023] Space-time coding is a coding technique that combines the spatial domain and the time domain. By simultaneously transmitting encoded signals on multiple antennas (or transducers), spatial diversity and time diversity are utilized to improve the reliability and transmission efficiency of the signals. A multi-beam transducer array consists of multiple transducer units, which are arranged according to a preset geometric layout, such as a linear array or a circular array. Each transducer can be regarded as an antenna. By encoding the signals transmitted on different transducers, space-time coding can be achieved. In this embodiment, the modulated chaotic frequency-modulated continuous wave signal is distributed to each transducer of the multi-beam transducer array according to a preset coding rule. For example, in this embodiment, the Alamouti space-time coding scheme can be adopted to divide the modulated chaotic frequency-modulated continuous wave signal into two time slots. In the first time slot, the modulated chaotic frequency-modulated continuous wave signal is respectively transmitted to two transducers; in the second time slot, after operations such as conjugation and negation are performed on the modulated chaotic frequency-modulated continuous wave signal, it is then respectively transmitted to these two transducers, so that at the receiving end, the original signal can be restored by performing corresponding decoding processing on the received signals, and the anti-interference ability of the signals can be improved by using spatial diversity and time diversity, thereby obtaining the chaotic frequency-modulated continuous wave signal after space-time coding processing, and distributing the chaotic frequency-modulated continuous wave signal after space-time coding processing to each transducer of the multi-beam transducer array for transmission preparation.

[0024] Meanwhile, in this embodiment, a chaotic-encrypted positioning pilot sequence is generated. This positioning pilot sequence can adopt different chaotic equations (such as the Lorenz system) or perform exclusive OR operations on the chaotic sequence generated by the Logistic map, etc., to obtain the encrypted positioning pilot sequence. Then, the encrypted positioning pilot sequence is embedded into the chaotic frequency-modulated continuous wave signal after space-time coding processing. The embedding method can be to insert the encrypted positioning pilot sequence at a specific position of the signal. For example, in this embodiment, a section of the encrypted positioning pilot sequence is inserted at the beginning or end of the signal, thereby obtaining the chaotic frequency-modulated continuous wave transmission signal. The positioning pilot sequence proposed in this embodiment is used for signal synchronization and positioning at the receiving end, while chaotic encryption can improve the security of the positioning pilot sequence and prevent it from being maliciously interfered with or cracked. Chaotic encryption is to use the characteristics of the chaotic system to encrypt the pilot sequence, so that the encrypted pilot sequence has randomness and unpredictability.

[0025] Finally, in this embodiment, the transmission phases and amplitudes of the transducers in the multi-beam transducer array are calculated according to the direction and distance of the detection area. For example, according to the direction of the detection area, this embodiment uses a beamforming algorithm to determine the phase differences of the transducers, so that the signals transmitted by the transducers are coherently superimposed in the direction of the detection area to form a stronger beam. At the same time, the transmission amplitudes of the transducers are adjusted according to the distance of the detection area to ensure that the signals have sufficient energy in the detection area. The calculated transmission phases and amplitudes are respectively set on the transducers of the multi-beam transducer array, and then each transducer is driven to transmit a chaotic frequency-modulated continuous wave transmission signal. The transducer converts the chaotic frequency-modulated continuous wave transmission signal into an optical signal or an electromagnetic wave signal and transmits it to the detection area in the set beam direction. The transmitted signal propagates in space and covers each position in the detection area, waiting for the target object to reflect the echo signal. After transmitting the signal, the multi-beam transducer array continuously receives the signal reflected from the detection area. After the transducer array receives the reflected echo signal, it converts the optical signal or electromagnetic wave signal into an electrical signal. Since the direction and intensity of the reflected signal are related to the position and characteristics of the obstacle, the received signal will contain this information. This embodiment performs preprocessing operations such as amplification and filtering on the signal received by the multi-beam transducer array to improve the signal quality and signal-to-noise ratio, and uses the preprocessed signal as the reflected echo signal of the target object for subsequent signal processing and obstacle detection.

[0026] S2. Perform decoupling calculation on the difference-frequency electrical signal to obtain the target time-frequency distribution matrix; wherein, the difference-frequency electrical signal is extracted by performing coherent optical mixing on the reflected echo signal and the chaotic frequency-modulated continuous wave signal.

[0027] In some embodiments, the process of obtaining the difference-frequency electrical signal is specifically as follows: According to the timestamp information of the reflected echo signal and the chaotic frequency-modulated continuous wave signal, perform time-domain alignment on the reflected echo signal to obtain the time-domain aligned signal of the target reflected echo; Use a mixer to perform coherent optical mixing on the time-domain aligned signal of the target reflected echo and the chaotic frequency-modulated continuous wave signal to obtain the original mixed signal; Filter out the high-frequency components in the original mixed signal through a low-pass filter to extract the difference-frequency electrical signal.

[0028] Since there will be a certain time delay in the propagation of the reflected echo signal of the target object, compared with the originally transmitted chaotic frequency-modulated continuous wave signal, there will be an offset in time. Therefore, in this embodiment, the reflected echo signal of the target object is aligned in the time domain through timestamp information. Specifically, the chaotic frequency-modulated continuous wave signal in this embodiment represents the signal transmission time by marking a timestamp when it is transmitted. Similarly, the reflected echo signal of the target object represents the signal reception time by marking a timestamp when it is received. The difference between these two timestamps is the propagation time delay of the signal. In this embodiment, according to the timestamp information of transmission and reception, the time delay amount of the reflected echo signal of the target object relative to the chaotic frequency-modulated continuous wave signal is calculated, and according to the calculated time delay amount, a time-domain translation operation is performed on the reflected echo signal of the target object. If the time delay amount is positive, it means that the reflected echo signal of the target object is delayed relative to the chaotic frequency-modulated continuous wave signal, and the reflected echo signal of the target object needs to be translated to the left by the time delay amount on the time axis; if the time delay amount is negative, it means that the reflected echo signal of the target object arrives in advance, and this embodiment can be processed by buffering or filling with zero values, so as to obtain the time-domain aligned signal of the target reflected echo. The time-domain aligned signal of the target reflected echo is aligned with the chaotic frequency-modulated continuous wave signal in the time dimension.

[0029] Coherent optical mixing uses a mixer to perform a multiplication operation on two signals, thereby obtaining a mixed signal containing sum-frequency components and difference-frequency components. In a lidar system, the difference-frequency signal reflecting the target distance and speed information can be extracted through coherent mixing. Therefore, in this embodiment, the time-domain aligned reflected echo signal of the target object and the original chaotic frequency-modulated continuous wave signal are respectively input into the two input ports of the mixer, and the mixer internally performs a multiplication operation on the two input signals. Assuming that the reflected echo signal of the target object is , and the chaotic frequency-modulated continuous wave signal is , then the output signal after mixing is , and this multiplication operation will generate sum-frequency components and difference-frequency components Two main frequency components. In this embodiment, the output of the mixer is the original mixed-frequency signal, which contains two frequency components, the sum-frequency component and the difference-frequency component. Among them, the difference-frequency component includes the key information of the target distance and speed. Then, in this embodiment, the cut-off frequency of the low-pass filter is set according to the preset frequency range of the difference-frequency electrical signal. The cut-off frequency should be set above the highest frequency of the difference-frequency electrical signal to ensure that the difference-frequency signal can pass through, while filtering out all high-frequency components higher than this frequency. In this embodiment, the original mixed-frequency signal is input into the low-pass filter, and the low-pass filter will perform frequency analysis on the input original mixed-frequency signal, allowing signals with frequencies lower than the cut-off frequency to pass through, while suppressing signals with frequencies higher than the cut-off frequency, so as to retain the difference-frequency component (low frequency) in the original mixed-frequency signal and filter out the sum-frequency component (high frequency). After being processed by the low-pass filter, a difference-frequency electrical signal is output, and the difference-frequency electrical signal can reflect the distance and speed information between the target and the lidar.

[0030] In some embodiments, the target time-frequency distribution matrix is extracted by a deep neural network. The deep neural network includes a first channel branch and a second channel branch adopting a dual-channel parallel branch structure. The first channel branch is a branch structure formed by cascading a long short-term memory network and a three-dimensional convolutional neural network. The second channel branch is a branch structure based on the fractional Fourier transform. In this embodiment, the steps of decoupling and calculating the difference-frequency electrical signal to obtain the target time-frequency distribution matrix include: According to the amplitude range of the difference-frequency electrical signal, the difference-frequency electrical signal is normalized by the maximum-minimum normalization algorithm to obtain a normalized difference-frequency electrical signal; The long short-term memory network is used to capture the temporal dependence relationship in the normalized difference-frequency electrical signal and extract the temporal domain features of the difference-frequency electrical signal; The three-dimensional convolutional neural network is used to perform spatial-frequency domain convolution operation on the temporal domain features of the difference-frequency electrical signal to obtain a spatio-temporal frequency joint feature vector; The spatio-temporal frequency joint feature vector is input into a fully connected layer for non-linear transformation to obtain a channel state parameter vector; Based on the channel state parameter vector, the optimal fractional Fourier transform order is calculated by the gradient descent method, and the normalized difference-frequency electrical signal is subjected to fractional Fourier transform according to the optimal fractional Fourier transform order to obtain a preliminary time-frequency distribution matrix; The morphological filtering algorithm is used to perform spatial domain filtering on the preliminary time-frequency distribution matrix to obtain the target time-frequency distribution matrix.

[0031] Specifically, in this embodiment, by traversing all data points of the difference frequency electrical signal, the maximum amplitude value and the minimum amplitude value of the difference frequency electrical signal within a preset observation window are found. In this embodiment, the length of the observation window can be set to twice the current frequency modulation period, and the amplitude of each data point in the difference frequency electrical signal is normalized according to the maximum amplitude value and the minimum amplitude value to obtain a normalized difference frequency electrical signal. At the same time, in this embodiment, a deep neural network is constructed. The deep neural network adopts a dual-channel parallel branch structure. The first channel branch adopts a cascaded branch structure of a long short-term memory network and a three-dimensional convolutional neural network. The second channel branch adopts a branch structure based on the fractional Fourier transform. Among them, the cascaded branch structure of the long short-term memory network and the three-dimensional convolutional neural network includes a cascaded long short-term memory network and a three-dimensional convolutional neural network. The long short-term memory network is constructed according to the temporal characteristics of the difference frequency electrical signal. The long short-term memory network consists of multiple long short-term memory units. Each long short-term memory unit includes structures such as an input gate, a forget gate, and an output gate. The input gate controls the degree of new information entering the unit. The forget gate determines the degree of forgetting of old information. The output gate controls the contribution of the unit state to the output. During the training process, in this embodiment, the weights of these gates can be adjusted through a large amount of temporal data, so that the long short-term memory network can capture the long-term dependence relationship in the temporal data. In this embodiment, a three-dimensional convolutional neural network is cascaded behind the long short-term memory network. The three-dimensional convolutional neural network includes multiple three-dimensional convolutional layers and pooling layers. The three-dimensional convolutional layer performs convolutional operations on the data in the time dimension, frequency dimension, and spatial dimension using a three-dimensional convolutional kernel to extract the spatio-temporal features of the data. Among them, the width of the time convolutional kernel corresponds to three frequency modulation periods, and the frequency convolutional kernel covers seven adjacent frequency points. The pooling layer is used to reduce the dimension and computational amount of the data while retaining important features. The three-dimensional convolutional neural network can further extract the spatio-frequency domain features of the features output by the long short-term memory network. The second channel branch mainly processes the signal based on the fractional Fourier transform. The fractional Fourier transform is a generalized Fourier transform that can better handle the time-frequency localization problem. By performing fractional Fourier transform operations on the input difference frequency electrical signal, the features of the signal in the fractional Fourier domain are extracted.

[0032] In this embodiment, the standardized difference-frequency electrical signal is input into the long short-term memory network in time steps. Through the forget gate, input gate, and output gate mechanisms inside the long short-term memory network, the input standardized difference-frequency electrical signal is processed step by step in time. The output at each time step depends not only on the current input but also on the hidden state at the previous time step, thereby capturing the temporal dependence relationship. After being processed by the long short-term memory network, the time-domain features of the difference-frequency electrical signal containing multi-period dependence relationships are obtained. These time-domain features of the difference-frequency electrical signal reflect the variation law and dependence relationship of the signal in the time dimension. Then, this embodiment uses the time-domain features of the difference-frequency electrical signal output by the long short-term memory network as the input and performs convolution operations in the space-frequency domain using the three-dimensional convolutional layer of the three-dimensional convolutional neural network. The three-dimensional convolutional neural network is provided with multiple three-dimensional convolutional kernels, and each three-dimensional convolutional kernel slides in the three dimensions of time, space, and frequency, performing operations such as weighted summation on the features of each local area. After the convolution operation, this embodiment reduces the feature dimension through the pooling operation while retaining the main features. After being processed by multiple three-dimensional convolutional layers and pooling layers, a spatio-temporal frequency joint feature vector is obtained. This spatio-temporal frequency joint feature vector synthesizes the feature information of the signal in the three dimensions of time, space, and frequency, and can comprehensively describe the characteristics of the difference-frequency electrical signal in the spatio-temporal frequency joint domain, providing richer information for subsequent tasks such as target state estimation. This embodiment inputs the spatio-temporal frequency joint feature vector into the fully connected layer. In the fully connected layer, this embodiment performs a non-linear mapping on the input spatio-temporal frequency joint feature vector through the ReLU activation function. The ReLU activation function can set the negative values in the input spatio-temporal frequency joint feature vector to zero, retain the positive values, and perform a linear transformation on the positive values. This non-linear transformation enables the network to learn more complex feature representations and improve the expression ability of the model, thereby obtaining a channel state parameter vector characterizing the current channel state. This channel state parameter vector includes features such as Doppler shift, channel attenuation, and noise power. These parameters can reflect the state changes of the signal during transmission and provide a basis for subsequent calculations of the fractional Fourier transform order, etc.

[0033] In this embodiment, the initial fractional Fourier transform order is set according to the numerical distribution of the channel state parameter vector, and a loss function is constructed to measure the difference between the time-frequency distribution obtained at the current fractional Fourier transform order and the desired time-frequency distribution. The initial fractional Fourier transform order is continuously adjusted by the gradient descent method to maximize the energy concentration degree, and the optimal fractional Fourier transform order that makes the signal energy most concentrated is obtained. Specifically, in this embodiment, the gradient of the loss function with respect to the fractional Fourier transform order is calculated, and then the order value is adjusted along the direction of the gradient descent. For example, the amplitude of the order value adjustment for each iteration can be 0.01. After multiple iterations, the fractional Fourier transform order that minimizes the loss function is gradually found, and this order is the optimal fractional Fourier transform order. According to the calculated optimal fractional Fourier transform order, the standardized difference-frequency electrical signal is subjected to fractional Fourier transform. The fractional Fourier transform can decompose the standardized difference-frequency electrical signal into different fractional frequency components to obtain a two-dimensional preliminary time-frequency distribution matrix. The horizontal axis of this preliminary time-frequency distribution matrix can represent time, and the vertical axis represents the fractional frequency. Each element in the preliminary time-frequency distribution matrix represents the signal energy magnitude at the corresponding time and fractional frequency. This preliminary time-frequency distribution matrix reflects the time-frequency characteristics of the signal in the fractional domain.

[0034] Finally, in this embodiment, morphological filtering operations such as opening operation and closing operation are sequentially performed on the preliminary time-frequency distribution matrix. The opening operation is to erode first and then dilate, which can remove some small noise points and disconnected connections; the closing operation is to dilate first and then erode, which can fill some small holes and connect adjacent regions. For example, in the opening operation, in this embodiment, a cross-shaped structuring element is used to perform a dilation operation on the entire preliminary time-frequency distribution matrix. For each position, if all the pixels within the region covered by the structuring element are greater than a certain threshold, the pixel at that position is retained, otherwise it is set to zero; the dilation operation is to perform an erosion operation using a square structuring element, and the maximum pixel value within the region covered by the square structuring element is assigned to the pixel at the central position. Through the morphological filtering operation, the time-frequency distribution matrix is processed in the spatial domain to remove noise and unreasonable components. After the morphological filtering, a target time-frequency distribution matrix with discrete noise points removed is obtained. In the target time-frequency distribution matrix, the row vectors of the target time-frequency distribution matrix independently represent the target distance distribution, and the column vectors independently represent the speed distribution. Therefore, the distance and speed parameters can be independently analyzed, providing a more reliable basis for subsequent target recognition, positioning, and speed estimation, etc.

[0035] It should be noted that various information is coupled in the difference-frequency electrical signal. The various information not only includes motion state information such as the distance, speed, and radial acceleration of the target, but also includes interference factors such as environmental noise and channel characteristics. Through its non-linear mapping structure, the deep neural network can separate the useful target motion state-related features from these mixed signal features, removing the influence of noise and interference components. For example, the long short-term memory network captures temporal dependencies, extracts time-domain features, and decouples the relationship between the continuity of the signal in the time dimension and the coherence of target motion; the three-dimensional convolutional neural network performs convolutional operations in the spatio-frequency domain, further decomposing the time-domain features into spatial and frequency features, and separating the association between the frequency components and the target motion state. This process gradually separates the originally entangled various factors from the signal.

[0036] S3. Predict the Doppler frequency shift amount of the target object at the next moment according to the target time-frequency distribution matrix.

[0037] In some embodiments, the step of predicting the Doppler frequency shift amount of the target object at the next moment according to the target time-frequency distribution matrix: Adopt a peak search algorithm to search for energy accumulation points in the target time-frequency distribution matrix, and extract the characteristic parameters of the energy accumulation points; the characteristic parameters of the energy accumulation points include the time center position, the frequency center value, the energy intensity, and the time-frequency distribution variance; Define the state vector of the Kalman filter, establish a state equation according to the state vector, and use the frequency center position and the time center position in the characteristic parameters of the energy accumulation points as observation variables to establish an observation equation; the state vector includes the target distance, radial velocity, and radial acceleration to be estimated; Integrate the state equation and the observation equation to form a target motion state space model; According to the characteristic parameters of the energy accumulation points, use the target motion state space model to perform an initial estimate of the target motion state to obtain the initial state estimate value at the previous moment; According to the target state estimate value at the previous moment and the state equation, use the Kalman filter to predict the target state estimate value at the current moment; Calculate the radial velocity information according to the target state estimate value at the current moment, and use the Doppler effect principle to convert the radial velocity information into a frequency shift amount to obtain the Doppler frequency shift amount at the next moment.

[0038] Specifically, in this embodiment, each element of the target time-frequency distribution matrix is traversed, and a sliding window peak search is performed on the target time-frequency distribution matrix to calculate the energy values within the local region. These energy concentration points usually correspond to the characteristics of the target reflection signal. At the same time, this embodiment sets an energy threshold. For example, this embodiment can set the energy threshold to 30% of the maximum value of the target time-frequency distribution matrix. The local regions with energy values higher than this energy threshold are marked as potential energy concentration points, and the time center position, frequency center value, energy intensity, and time-frequency distribution variance of the energy concentration points are extracted to obtain an energy concentration point feature parameter set characterizing the target motion characteristics. Among them, the time center position is obtained by calculating the weighted average position of the energy concentration point on the time axis, and the weight of the time center position is the energy value at the corresponding position; the frequency center value is obtained by calculating the weighted average value of the energy concentration point on the frequency axis, and the weight of the frequency center value is the energy value at the corresponding frequency; the total energy intensity is obtained by summing all the energy values within the energy concentration point; the time-frequency distribution variance is obtained by calculating the variance of the energy concentration point in the time and frequency dimensions, and the time-frequency distribution variance reflects the degree of energy dispersion around the energy concentration point.

[0039] Then, this embodiment defines the target distance, radial velocity, and radial acceleration as the state vector. The three parameters of the target distance, radial velocity, and radial acceleration jointly describe the motion state of the target. According to the state vector, a state equation describing the continuous motion of the target is established using the target kinematic model. The state equation describes the relationship between the target state and time. It is assumed that the target motion is a uniformly accelerated motion in a short period. The target distance at the next moment is equal to the sum of the current target distance, the product of the current radial velocity and the time interval, and the product of half of the radial acceleration and the square of the time interval; the radial velocity at the next moment is equal to the sum of the current radial velocity and the product of the current radial acceleration and the time interval; the radial acceleration at the current moment is equal to the radial acceleration at the previous moment. The time interval refers to the time difference between two state estimations. At the same time, this embodiment uses the frequency center position and time center position in the energy concentration point feature parameters as the observation variables to establish an observation equation. In the observation equation, the observed value is equal to the sum of the observation noise and the product of the state vector and the observation matrix. Thus, this embodiment establishes the state equation and observation equation of the Kalman filter, providing a mathematical model for subsequent state estimation and prediction. This embodiment combines the state equation and the observation equation into a target motion state space model. Among them, the state equation describes the variation law of the target state with time, and the observation equation describes the relationship between the observed value and the target state. This embodiment sets the target motion state space model parameters, which include the process noise covariance matrix and the observation noise covariance matrix. These parameters reflect the uncertainty of the target motion state space model.

[0040] Next, in this embodiment, the extracted energy concentration point characteristic parameters (time center position and frequency center value) are used as the initial observation data. Using the target motion state space model, the target state is preliminarily estimated by the least squares method to obtain the initial state estimate value at the previous moment, which provides a starting point for the subsequent Kalman filter recursive estimation. The initial state estimate value at the previous moment includes the initial distance, initial radial velocity, and initial radial acceleration of the target. In this embodiment, the state equation is recursively advanced forward based on the state estimate value at the previous moment. Specifically, the target state estimate value obtained at the previous moment is used as the input, and the target state prediction value at the current moment is calculated according to the state equation. The target state prediction value at the current moment is equal to the sum of the initial state estimate value at the previous moment, the product of the state change rate at the previous moment, and the time interval. In this embodiment, according to the process noise covariance matrix, the covariance matrix of the predicted state is calculated. The covariance matrix of the predicted state reflects the uncertainty of the prediction, thereby obtaining the target state prediction value at the current moment and its covariance matrix. The target state prediction value includes the predicted value of the radial velocity at the current moment, which provides a basis for the subsequent update step. In this embodiment, the radial velocity component is extracted from the target state estimate value at the current moment. According to the Doppler effect principle, it is known that the Doppler frequency shift is proportional to the radial velocity. The Doppler frequency shift at the next moment is equal to the ratio of twice the product of the predicted value of the radial velocity and the carrier frequency to the speed of light. This Doppler frequency shift can be used for further tracking and prediction of the target motion state.

[0041] S4. Calculate the ranging deviation compensation amount based on the chaotic frequency modulation adjustment signal; wherein, the chaotic frequency modulation adjustment signal is obtained by dynamically adjusting the chaotic frequency modulation slope of the chaotic frequency modulation continuous wave signal by the Doppler frequency shift amount.

[0042] In some embodiments, the process of obtaining the chaotic frequency modulation adjustment signal is specifically as follows: Process the chaotic frequency modulation continuous wave signal by using a spectrum analysis method to extract the initial chaotic frequency modulation slope; Calculate the chaotic frequency modulation slope adjustment amount according to the Doppler frequency shift amount, and obtain the corrected value of the chaotic frequency modulation slope according to the chaotic frequency modulation slope adjustment amount and the initial chaotic frequency modulation slope; Generate the chaotic frequency modulation adjustment signal by using the logistic chaos mapping according to the corrected value of the chaotic frequency modulation slope and the preset signal center frequency.

[0043] Specifically, in this embodiment, the chaotic frequency-modulated continuous wave signal to be processed is used as the input, and the Fourier transform is performed on the chaotic frequency-modulated continuous wave signal to obtain the spectral distribution characteristics of the chaotic frequency-modulated continuous wave signal. The main frequency component and its change trend of the chaotic frequency-modulated continuous wave signal are identified in the spectral distribution characteristics. The spectrum of the chaotic frequency-modulated continuous wave signal is broadband and continuous. The initial chaotic frequency modulation slope can be obtained by measuring the ratio of the frequency change amount of the chaotic frequency-modulated continuous wave signal to the corresponding time change amount. The initial chaotic frequency modulation slope reflects the speed of frequency change of the chaotic frequency-modulated continuous wave signal before dynamic adjustment, providing a reference value for subsequent slope adjustment.

[0044] The Doppler frequency shift amount reflects the motion state of the target relative to the radar. In this embodiment, by adjusting the chaotic frequency modulation slope, the radar signal can better adapt to the motion characteristics of the target, improving the detection accuracy. The adjustment amount of the chaotic frequency modulation slope needs to be dynamically calculated according to the Doppler frequency shift amount. Specifically, in this embodiment, the amount that the chaotic frequency modulation slope needs to be adjusted is calculated based on the Doppler frequency shift amount to obtain the chaotic frequency modulation slope adjustment amount. The chaotic frequency modulation slope adjustment amount is proportional to the Doppler frequency shift amount, that is, the chaotic frequency modulation slope adjustment amount is equal to the product of the preset proportional coefficient and the Doppler frequency shift amount. The preset proportional coefficient is determined by the operating frequency of the radar, the speed of light, and signal processing parameters, etc. In this embodiment, the calculated chaotic frequency modulation slope adjustment amount is added to the initial chaotic frequency modulation slope to obtain the chaotic frequency modulation slope correction value. This chaotic frequency modulation slope correction value comprehensively considers the influence of target motion on the signal frequency, enabling the frequency modulation slope of the chaotic frequency-modulated continuous wave signal to better match the motion state of the target, improving the detection accuracy and adaptability.

[0045] The Logistic chaotic map is a commonly used chaotic mapping method. By iterative calculation, a sequence with chaotic characteristics can be generated. In this embodiment, an initial value is randomly selected between 0 and 1, and this initial value is used as the starting point for chaotic map iteration. Starting from the initial value, a series of chaotic sequence values are generated. These values are all distributed between 0 and 1 and have chaotic characteristics, that is, they are very sensitive to the initial value and control parameters and have no periodicity. In this embodiment, the generated chaotic sequence values are mapped to a frequency interval centered on the preset signal center frequency, and the upper and lower ranges are determined by the chaotic frequency modulation slope correction value. For example, the preset signal center frequency is , the chaotic frequency modulation slope correction value determines the range of frequency change. For each chaotic sequence value , the corresponding actual frequency , where, is the time variable. Thus, with the chaotic change of the chaotic sequence value , the actual frequency It will also perform dynamic modulation according to the chaotic frequency modulation slope correction value to generate a chaotic frequency modulation adjustment signal adapted to the target motion state. Through the above steps, the process of dynamically adjusting the chaotic frequency modulation slope of the chaotic frequency modulation continuous wave signal according to the Doppler frequency shift amount and obtaining the chaotic frequency modulation adjustment signal is realized. This process enables the lidar to better adapt to the change of the target motion state and improves the accuracy and robustness of obstacle detection.

[0046] In some embodiments, the step of calculating the ranging deviation compensation amount based on the chaotic frequency modulation adjustment signal includes: Using Hilbert transform to extract the first instantaneous frequency sequence of the chaotic frequency modulation adjustment signal and the second instantaneous frequency sequence of the preset standard chaotic frequency modulation signal; Aligning the first instantaneous frequency sequence and the second instantaneous frequency sequence in the time domain to obtain the corresponding first time-domain aligned instantaneous frequency sequence and second time-domain aligned instantaneous frequency sequence; Performing point-by-point difference operation on the first time-domain aligned instantaneous frequency sequence and the second time-domain aligned instantaneous frequency sequence to obtain a frequency difference sequence; Calculating the ranging deviation at each moment according to the frequency difference sequence and the laser wavelength to obtain the ranging deviation compensation amount.

[0047] Specifically, the Hilbert transform is a method for converting a time-domain signal into an analytic signal, which can extract the instantaneous frequency information of the signal. In this embodiment, the chaotic frequency modulation adjustment signal and the preset standard chaotic frequency modulation signal are respectively subjected to the Hilbert transform to obtain their respective analytic signals. The analytic signal contains the amplitude and phase information of the original signal. Then, in this embodiment, the phase demodulation is performed on the analytic signal of the chaotic frequency modulation adjustment signal to obtain the first instantaneous frequency sequence, and the phase demodulation is performed on the analytic signal of the preset standard chaotic frequency modulation signal to obtain the second instantaneous frequency sequence, so as to obtain the first instantaneous frequency sequence of the chaotic frequency modulation adjustment signal and the second instantaneous frequency sequence of the preset standard chaotic frequency modulation signal. Then, the first instantaneous frequency sequence and the second instantaneous frequency sequence are aligned in the time domain. The time-domain alignment is to ensure that the two instantaneous frequency sequences correspond exactly on the time axis, so as to eliminate the error caused by time asynchronization. Specifically, in this embodiment, the time axes of the first instantaneous frequency sequence and the second instantaneous frequency sequence are adjusted by interpolation or resampling techniques, so that the first instantaneous frequency sequence and the second instantaneous frequency sequence are aligned at time points, ensuring that the two sequences have corresponding frequency values at the same time point, and obtaining the first time-domain aligned instantaneous frequency sequence and the second time-domain aligned instantaneous frequency sequence. For example, if the time ranges of the first instantaneous frequency sequence and the second instantaneous frequency sequence are different, it is necessary to pad zeros to the shorter instantaneous frequency sequence or truncate the longer instantaneous frequency sequence to make their time lengths consistent; if the sampling rates of the first instantaneous frequency sequence and the second instantaneous frequency sequence are different, it is necessary to resample one of the instantaneous frequency sequences to make its sampling rate the same as that of the other instantaneous frequency sequence.

[0048] In this embodiment, the first time-domain aligned instantaneous frequency sequence and the second time-domain aligned instantaneous frequency sequence are used as inputs. The frequency values at each corresponding time point of the two sequences of the first time-domain aligned instantaneous frequency sequence and the second time-domain aligned instantaneous frequency sequence are differentially calculated to obtain frequency differences. All the frequency differences are combined into a frequency difference sequence. This frequency difference sequence reflects the frequency deviation between the chaotic frequency modulation adjustment signal and the preset standard signal. Finally, in this embodiment, ranging deviation calculation is performed according to the frequency difference sequence and the known laser wavelength. In a frequency-modulated continuous-wave lidar, the ranging deviation is proportional to the frequency difference and inversely proportional to the laser wavelength. For each frequency difference, in this embodiment, by calculating the ratio of the product of the frequency difference and the speed of light to the product of the laser wavelength and twice the frequency modulation slope, the ranging deviation is obtained. According to the product of the ranging deviation and the preset distance calibration coefficient, the ranging deviation compensation amount is obtained. All the ranging deviation compensation amounts are arranged in chronological order to form a ranging deviation compensation amount sequence. This ranging deviation compensation amount sequence reflects the distance measurement errors caused by the frequency difference between the chaotic frequency modulation adjustment signal and the preset standard chaotic frequency modulation signal at different time points. This compensation amount can be used to correct the ranging error caused by the frequency deviation and improve the accuracy of obstacle detection.

[0049] S5. Compensate the frequency value of the target object's difference-frequency electrical signal according to the ranging deviation compensation amount and environmental noise data to calculate the obstacle distance and speed detection information.

[0050] In some embodiments, the step of compensating the frequency value of the target object's difference-frequency electrical signal according to the ranging deviation compensation amount and environmental noise data to calculate the obstacle distance and speed detection information includes: Collect environmental noise data in real time, and perform preliminary compensation on the original frequency value of the difference-frequency electrical signal according to the ranging deviation compensation amount to obtain the compensated frequency value of the difference-frequency electrical signal; Perform power spectral density estimation on the environmental noise data to obtain the environmental noise power spectral density, and construct an adaptive filter according to the environmental noise power spectral density by using the least mean square error algorithm; Use the adaptive filter to correct the noise of the compensated frequency value of the difference-frequency electrical signal to obtain the corrected frequency value of the difference-frequency electrical signal; Perform fractional Fourier transform on the corrected frequency value of the difference-frequency electrical signal to obtain the frequency offset characteristic of the difference-frequency electrical signal in the fractional Fourier transform domain; According to the frequency offset characteristic of the difference-frequency electrical signal, the frequency modulation bandwidth, and the frequency modulation period, calculate the initial distance estimation value, and calculate the target speed according to the deviation between the corrected frequency value of the difference-frequency electrical signal and the original frequency value of the difference-frequency electrical signal; Obtain the obstacle distance and velocity detection information based on the initial distance estimate value, the target velocity, and the geometric position of the frequency-modulated continuous-wave lidar.

[0051] Specifically, in this embodiment, the original frequency value of the difference frequency electrical signal is corrected according to the ranging deviation compensation amount to obtain the compensated frequency value of the difference frequency electrical signal. The compensated frequency value of the difference frequency electrical signal is equal to the sum of the original frequency value and the ranging deviation compensation amount. This compensated frequency value of the difference frequency electrical signal takes into account the error caused by the frequency deviation. Then, this embodiment collects environmental noise data in real time. The environmental noise data reflects the external environmental noise characteristics of the radar system. This embodiment performs power spectral density estimation on the real-time collected environmental noise data to obtain the environmental noise power spectral density in the frequency domain. The power spectral density estimation can reflect the distribution characteristics of the environmental noise in the frequency domain, so that the adaptive filter can dynamically adjust the filtering parameters according to the environmental noise characteristics to minimize the mean square error. Specifically, in this embodiment, according to the environmental noise power spectral density, the parameters of the adaptive filter are designed using the least mean square error algorithm, so that the adaptive filter can dynamically adapt to the changes in the environmental noise, thereby constructing an adaptive filter. This adaptive filter can dynamically adjust the filtering parameters according to the environmental noise characteristics, thereby effectively suppressing the influence of the environmental noise on the difference frequency electrical signal and improving the signal-to-noise ratio of the signal. This embodiment uses the compensated frequency value of the difference frequency electrical signal as the input of the adaptive filter, and uses the adaptive filter to filter the compensated frequency value of the difference frequency electrical signal to suppress the influence of the environmental noise and obtain the corrected frequency value of the difference frequency electrical signal. This corrected frequency value of the difference frequency electrical signal removes the interference of the environmental noise and provides high-quality signal data for the subsequent fractional Fourier transform.

[0052] This embodiment performs fractional Fourier transform on the corrected frequency value of the difference frequency electrical signal to obtain the representation of the signal corresponding to the corrected frequency value of the difference frequency electrical signal in the fractional Fourier transform domain. In the fractional Fourier transform domain, this embodiment identifies the frequency offset characteristics of the difference frequency electrical signal. These characteristics reflect the frequency changes caused by the target movement, thereby obtaining the frequency offset characteristics of the difference frequency electrical signal in the fractional Fourier transform domain. There is a direct relationship between the frequency offset characteristics of the difference frequency electrical signal and the target distance and velocity. This embodiment can calculate the initial distance of the target through the frequency modulation bandwidth and the frequency modulation period, and can calculate the target velocity through the frequency deviation. Specifically, according to the ranging formula of the lidar, the initial distance is related to the frequency offset characteristics of the difference frequency electrical signal, the frequency modulation bandwidth, and the frequency modulation period. The frequency modulation bandwidth refers to the frequency change range of the laser signal within one frequency modulation period, and the frequency modulation period is the time taken for the frequency to complete one change. In this embodiment, the initial distance is equal to the ratio of the product of the speed of light and the frequency modulation period to twice the frequency modulation bandwidth multiplied by the frequency offset characteristics of the difference frequency electrical signal. For example, the speed of light is c, the frequency modulation period is T, the frequency modulation bandwidth is B, and the frequency offset characteristics are , the initial distance estimation value has the following calculation formula:

[0053] Thus, in this embodiment, the initial distance estimation value can be obtained. Then, the difference between the corrected frequency value and the original frequency value of the difference-frequency electrical signal is calculated to obtain the frequency change amount. At the same time, according to the Doppler effect principle, the target speed is proportional to the frequency change amount, and the target speed proportionality coefficient is determined by factors such as the laser wavelength and the frequency modulation period. In this embodiment, the target speed proportionality coefficient can be set as the ratio of the speed of light to twice the carrier frequency. Therefore, in this embodiment, by calculating the product of the frequency change amount and the target speed proportionality coefficient, the target speed is obtained. The two parameters of the initial distance estimation value and the target speed jointly describe the motion state of the target. In this embodiment, the initial distance estimation value is used as the rough distance information of the obstacle, and the target speed is used as the motion speed information of the obstacle. Finally, in this embodiment, considering the geometric position of the frequency-modulated continuous-wave lidar, the initial distance estimation value and the target speed are corrected to eliminate the systematic error caused by the radar position, and the final obstacle distance-speed detection information is obtained. The obstacle distance-speed detection information includes the distance and speed of the obstacle. For example, in this embodiment, the geometric parameters such as the position coordinates where the lidar is installed on the vehicle and the pitch angle parameter of the lidar installation are known, and the initial distance estimation value is corrected by projection in combination with these geometric parameters. In this embodiment, by calculating the ratio of the initial distance estimation value to the cosine value of the pitch angle of the lidar installation, the obstacle distance value after spatial geometric correction is obtained. At the same time, in combination with the target speed, the obstacle distance-speed detection information is obtained. The obstacle distance-speed detection information includes the precise distance between the obstacle and the lidar, the motion speed of the obstacle, and the position in space and other information.

[0054] An embodiment of the present invention provides an obstacle detection method based on a frequency-modulated continuous-wave lidar. The method includes transmitting a chaotic frequency-modulated continuous-wave signal with non-periodic characteristics into a detection area to obtain a reflected echo signal of a target object; performing decoupling calculation on the difference-frequency electrical signal to obtain a target time-frequency distribution matrix, where the difference-frequency electrical signal is extracted by performing coherent optical mixing on the reflected echo signal and the chaotic frequency-modulated continuous-wave signal; predicting the Doppler frequency shift amount of the target object at the next moment according to the target time-frequency distribution matrix; calculating a ranging deviation compensation amount based on a chaotic frequency-modulation adjustment signal, where the chaotic frequency-modulation adjustment signal is obtained by dynamically adjusting the chaotic frequency-modulation slope of the chaotic frequency-modulated continuous-wave signal with the Doppler frequency shift amount; compensating the frequency value of the difference-frequency electrical signal of the target object according to the ranging deviation compensation amount and environmental noise data to calculate the obstacle distance-velocity detection information. Compared with the prior art, this method can accurately compensate for the ranging deviation by predicting the Doppler frequency shift amount in real time and dynamically adjusting the parameters of the chaotic frequency-modulated continuous-wave signal, significantly improving the accuracy of obstacle distance-velocity detection in complex scenarios and meeting the requirements of high-dynamic scenarios such as autonomous driving.

[0055] It should be noted that the magnitudes of the sequence numbers of the above processes do not mean the order of execution. The order of execution of each process should be determined according to its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0056] In one embodiment, as Figure 2 shown, an embodiment of the present invention provides an obstacle detection system based on a frequency-modulated continuous-wave lidar. The system includes: A signal acquisition module 101, configured to transmit a chaotic frequency-modulated continuous-wave signal with non-periodic characteristics into a detection area to obtain a reflected echo signal of a target object; A difference-frequency analysis module 102, configured to perform decoupling calculation on the difference-frequency electrical signal to obtain a target time-frequency distribution matrix, where the difference-frequency electrical signal is extracted by performing coherent optical mixing on the reflected echo signal and the chaotic frequency-modulated continuous-wave signal; A frequency shift estimation module 103, configured to predict the Doppler frequency shift amount of the target object at the next moment according to the target time-frequency distribution matrix; A deviation compensation module 104, configured to calculate a ranging deviation compensation amount based on a chaotic frequency-modulation adjustment signal, where the chaotic frequency-modulation adjustment signal is obtained by dynamically adjusting the chaotic frequency-modulation slope of the chaotic frequency-modulated continuous-wave signal with the Doppler frequency shift amount; A target detection module 105, configured to compensate the frequency value of the difference-frequency electrical signal of the target object according to the ranging deviation compensation amount and environmental noise data to calculate the obstacle distance-velocity detection information.

[0057] For the specific limitations of an obstacle detection system based on a frequency-modulated continuous-wave lidar, reference may be made to the above limitations of a method for obstacle detection based on a frequency-modulated continuous-wave lidar, which will not be elaborated here. Those of ordinary skill in the art can realize that, in combination with the various modules and steps described in the embodiments disclosed in the present application, they can be implemented in hardware, software, or a combination of both. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0058] An embodiment of the present invention provides an obstacle detection system based on a frequency-modulated continuous-wave lidar. The signal acquisition module of the system emits a chaotic frequency-modulated continuous-wave signal with non-periodic characteristics to a detection area and acquires the reflected echo signal of the target object. The difference frequency analysis module performs decoupling calculations on the difference frequency electrical signal to obtain the target time-frequency distribution matrix. Among them, the difference frequency electrical signal is extracted by performing coherent optical mixing on the reflected echo signal and the chaotic frequency-modulated continuous-wave signal. The frequency shift estimation module predicts the Doppler frequency shift amount of the target object at the next moment according to the target time-frequency distribution matrix. The deviation compensation module calculates the ranging deviation compensation amount based on the chaotic frequency modulation adjustment signal. Among them, the chaotic frequency modulation adjustment signal is obtained by dynamically adjusting the chaotic frequency modulation slope of the chaotic frequency-modulated continuous-wave signal by the Doppler frequency shift amount. The target detection module compensates the frequency value of the difference frequency electrical signal of the target object according to the ranging deviation compensation amount and the environmental noise data to calculate the obstacle distance and speed detection information. Compared with the prior art, the system realizes precise compensation for ranging deviation by predicting the Doppler frequency shift amount in real time and dynamically adjusting the parameters of the chaotic frequency-modulated continuous-wave signal, significantly improving the accuracy of obstacle distance and speed detection in complex scenarios and meeting the requirements of high-dynamic scenarios such as autonomous driving.

[0059] The above embodiments only represent several preferred implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the technical field, without departing from the technical principle of the present invention, several improvements and substitutions can be made, and these improvements and substitutions should also be regarded as the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the protection scope of the claims.

Claims

1. A method for obstacle detection based on a frequency modulated continuous wave lidar, characterized in that, It includes the following steps: Transmit a chaotic frequency-modulated continuous wave signal with aperiodic characteristics to the detection area to obtain the reflected echo signal of the target object; Perform decoupling calculation on the difference-frequency electrical signal to obtain the target time-frequency distribution matrix; wherein, the difference-frequency electrical signal is extracted by performing coherent optical mixing on the reflected echo signal and the chaotic frequency-modulated continuous wave signal; Predict the Doppler frequency shift amount of the target object at the next moment according to the target time-frequency distribution matrix; Calculate the ranging deviation compensation amount based on the chaotic frequency-modulation adjustment signal; wherein, the chaotic frequency-modulation adjustment signal is obtained by dynamically adjusting the chaotic frequency-modulation slope of the chaotic frequency-modulated continuous wave signal by the Doppler frequency shift amount; Compensate the frequency value of the difference-frequency electrical signal of the target object according to the ranging deviation compensation amount and the environmental noise data to calculate the detection information of the obstacle distance and speed.

2. The obstacle detection method based on a frequency modulated continuous wave lidar according to claim 1, wherein The step of transmitting a chaotic frequency-modulated continuous wave signal with aperiodic characteristics to the detection area to obtain the reflected echo signal of the target object includes: Generate an initial chaotic sequence with aperiodic characteristics, and dynamically adjust the instantaneous frequency offset through a phase modulator to generate a chaotic frequency-modulated continuous wave signal; Modulate the chaotic frequency-modulated continuous wave signal by using a hybrid modulation method combining orthogonal frequency division multiplexing and high-order quadrature amplitude modulation to obtain a modulated chaotic frequency-modulated continuous wave signal; Load the modulated chaotic frequency-modulated continuous wave signal onto a multi-beam transducer array for space-time coding, and embed a chaotic-encrypted positioning pilot sequence in the chaotic frequency-modulated continuous wave signal loaded onto the multi-beam transducer array to obtain a chaotic frequency-modulated continuous wave transmission signal; Directly transmit the chaotic frequency-modulated continuous wave transmission signal to the detection area through the multi-beam transducer array, and receive the reflected echo signal of the target object reflected from the detection area.

3. A method for obstacle detection based on a frequency-modulated continuous-wave lidar according to claim 1, characterized in that, The specific process of obtaining the difference-frequency electrical signal is as follows: According to the timestamp information of the reflected echo signal and the chaotic frequency-modulated continuous wave signal, perform time-domain alignment on the reflected echo signal to obtain a time-domain aligned signal of the target reflected echo; Use a mixer to perform coherent optical mixing on the time-domain aligned signal of the target reflected echo and the chaotic frequency-modulated continuous wave signal to obtain an original mixed signal; Filter out the high-frequency components in the original mixed signal through a low-pass filter to extract the difference-frequency electrical signal.

4. The obstacle detection method based on a frequency modulated continuous wave lidar according to claim 1, characterized in that: The target time-frequency distribution matrix is extracted through a deep neural network, and the deep neural network includes a first-channel branch and a second-channel branch adopting a dual-channel parallel branch structure. The first-channel branch is a branch structure formed by cascading a long short-term memory network and a three-dimensional convolutional neural network, and the second-channel branch is a branch structure based on the fractional Fourier transform.

5. The obstacle detection method based on a frequency modulated continuous wave lidar according to claim 4, wherein, The step of performing decoupling calculation on the difference-frequency electrical signal to obtain the target time-frequency distribution matrix includes: According to the amplitude range of the difference-frequency electrical signal, perform standardization processing on the difference-frequency electrical signal by using the maximum-minimum normalization algorithm to obtain a standardized difference-frequency electrical signal; Capture the time-series dependence relationship in the standardized difference-frequency electrical signal through a long short-term memory network to extract the time-domain characteristics of the difference-frequency electrical signal; Perform a spatial-frequency domain convolution operation on the time-domain features of the difference-frequency electrical signal using a three-dimensional convolutional neural network to obtain a spatio-temporal frequency joint feature vector; Input the spatio-temporal frequency joint feature vector into a fully connected layer for non-linear transformation to obtain a channel state parameter vector; Based on the channel state parameter vector, use the gradient descent method to calculate the optimal fractional Fourier transform order, and perform a fractional Fourier transform on the normalized difference-frequency electrical signal according to the optimal fractional Fourier transform order to obtain a preliminary time-frequency distribution matrix; Use a morphological filtering algorithm to perform spatial domain filtering on the preliminary time-frequency distribution matrix to obtain a target time-frequency distribution matrix.

6. The obstacle detection method based on a frequency modulated continuous wave lidar according to claim 1, characterized in that, The step of predicting the Doppler frequency shift amount of the target object at the next moment according to the target time-frequency distribution matrix: Use a peak search algorithm to search for energy concentration points in the target time-frequency distribution matrix and extract the characteristic parameters of the energy concentration points; the characteristic parameters of the energy concentration points include the time center position, the frequency center value, the energy intensity, and the time-frequency distribution variance; Define the state vector of the Kalman filter, establish a state equation according to the state vector, and use the frequency center position and the time center position in the characteristic parameters of the energy concentration points as the observation variables to establish an observation equation; The state vector includes the target distance, radial velocity, and radial acceleration to be estimated; Integrate the state equation and the observation equation to form a target motion state space model; According to the characteristic parameters of the energy concentration points, use the target motion state space model to perform an initial estimation of the target motion state to obtain the initial state estimation value at the previous moment; According to the target state estimation value at the previous moment and the state equation, use the Kalman filter to predict the target state estimation value at the current moment; Calculate the radial velocity information according to the target state estimation value at the current moment, and convert the radial velocity information into a frequency shift amount using the Doppler effect principle to obtain the Doppler frequency shift amount at the next moment.

7. The obstacle detection method based on a frequency modulated continuous wave lidar according to claim 1, wherein The specific process of obtaining the chaotic frequency modulation adjustment signal is as follows: Use a spectrum analysis method to process the chaotic frequency modulation continuous wave signal and extract the initial chaotic frequency modulation slope; Calculate the chaotic frequency modulation slope adjustment amount according to the Doppler frequency shift amount, and obtain the chaotic frequency modulation slope correction value according to the chaotic frequency modulation slope adjustment amount and the initial chaotic frequency modulation slope; Generate a chaotic frequency modulation adjustment signal using the Logistic chaotic map according to the chaotic frequency modulation slope correction value and the preset signal center frequency.

8. The obstacle detection method based on a frequency-modulated continuous-wave lidar according to claim 1, wherein, The step of calculating the ranging deviation compensation amount based on the chaotic frequency modulation adjustment signal includes: Use the Hilbert transform to extract the first instantaneous frequency sequence of the chaotic frequency modulation adjustment signal and the second instantaneous frequency sequence of the preset standard chaotic frequency modulation signal; Perform time-domain alignment on the first instantaneous frequency sequence and the second instantaneous frequency sequence to obtain the corresponding first time-domain aligned instantaneous frequency sequence and second time-domain aligned instantaneous frequency sequence; Perform a point-by-point difference operation on the first time-domain aligned instantaneous frequency sequence and the second time-domain aligned instantaneous frequency sequence to obtain a frequency difference sequence; Calculate the ranging deviation at each moment according to the frequency difference sequence and the laser wavelength to obtain the ranging deviation compensation amount.

9. The obstacle detection method based on a frequency modulated continuous wave lidar according to claim 1, characterized in that, The step of compensating the frequency value of the differential frequency electrical signal of the target object according to the ranging deviation compensation amount and the environmental noise data to calculate the obstacle distance and speed detection information includes: Collect environmental noise data in real time, and perform preliminary compensation on the original frequency value of the differential frequency electrical signal according to the ranging deviation compensation amount to obtain the compensated frequency value of the differential frequency electrical signal; Perform power spectral density estimation on the environmental noise data to obtain the environmental noise power spectral density, and construct an adaptive filter using the least mean square error algorithm according to the environmental noise power spectral density; Use the adaptive filter to correct the noise of the compensated frequency value of the differential frequency electrical signal to obtain the corrected frequency value of the differential frequency electrical signal; Perform fractional Fourier transform on the corrected frequency value of the differential frequency electrical signal to obtain the frequency shift characteristic of the differential frequency electrical signal in the fractional Fourier transform domain; Calculate the initial distance estimate value according to the frequency shift characteristic of the differential frequency electrical signal, the frequency modulation bandwidth and the frequency modulation period, and calculate the target speed according to the deviation between the corrected frequency value of the differential frequency electrical signal and the original frequency value of the differential frequency electrical signal; Obtain the obstacle distance and speed detection information according to the initial distance estimate value, the target speed and the geometric position of the frequency modulated continuous wave lidar.

10. An obstacle detection system based on a frequency-modulated continuous-wave lidar, characterized in that, The system includes: A signal acquisition module, configured to transmit a chaotic frequency modulated continuous wave signal with non-periodic characteristics to a detection area and acquire a reflected echo signal of a target object; A differential frequency analysis module, configured to perform decoupling calculation on the differential frequency electrical signal to obtain a target time-frequency distribution matrix; wherein, the differential frequency electrical signal is extracted by performing coherent optical mixing on the reflected echo signal and the chaotic frequency modulated continuous wave signal; A frequency shift estimation module, configured to predict the Doppler frequency shift amount of the target object at the next moment according to the target time-frequency distribution matrix; A deviation compensation module, configured to calculate a ranging deviation compensation amount based on a chaotic frequency modulation adjustment signal; wherein, the chaotic frequency modulation adjustment signal is obtained by dynamically adjusting the chaotic frequency modulation slope of the chaotic frequency modulated continuous wave signal by the Doppler frequency shift amount; A target detection module, configured to compensate the frequency value of the differential frequency electrical signal of the target object according to the ranging deviation compensation amount and the environmental noise data to calculate the obstacle distance and speed detection information.

Images