Target trajectory detection method, device, equipment, medium and program product

Abstract

The application relates to the technical field of target detection, in particular to a target trajectory detection method, device, equipment, medium and program product. The method comprises the following steps: acquiring M frames of slow-time-dimension echo signals collected by a radar, each frame of slow-time-dimension echo signal comprising N groups of fast-time-dimension echo signals; performing distance transformation on the slow echo signals in the fast time dimension to obtain M groups of range images; determining a target position based on the M groups of range images; constructing M local estimation values of instantaneous frequencies based on the M frames of slow-time-dimension echo signals at the target position; constructing M confidences corresponding to the M local estimation values of instantaneous frequencies based on the amplitudes of the M frames of slow-time-dimension echo signals at the target position; and constructing and solving a global optimization function based on the M local estimation values, the M confidences and first-order and second-order continuity constraint information of the local estimation values, so as to obtain the motion trajectory of the target. In this way, the environmental perception capability of the robot can be improved.

Description

Technical Field

[0001] This application relates to the field of target detection technology, specifically to a target trajectory detection method, apparatus, equipment, medium, and program product. Background Technology

[0002] With the widespread application of robots in urban services, industrial automation, and human-computer interaction, improving their environmental perception capabilities has become an important research direction. Millimeter-wave radar, due to its insensitivity to light and its ability to operate stably in complex environments such as rain, fog, and dust, is gradually becoming a key component of multi-sensor systems for robots. Especially in tasks such as gesture interaction, close-range target perception, and micro-motion recognition, radar can directly acquire target velocity information through the Doppler effect, thus compensating to some extent for the shortcomings of vision and lidar in acquiring dynamic information.

[0003] However, current methods for environmental perception based on target speed information are complex and have low target recognition accuracy. Summary of the Invention

[0004] The purpose of this application is to provide a target trajectory detection method, apparatus, equipment, medium, and program product.

[0005] In a first aspect of this application, a target trajectory detection method is provided, applied to an electronic device. The method includes: acquiring M frames of slow-time dimension echo signals collected by a radar within an acquisition period, wherein the target detection area includes the target to be detected, and each frame of slow-time dimension echo signal includes N sets of fast-time dimension echo signals, where M and N are both integers greater than or equal to 1; performing a range transformation on each frame of slow echo signal in the fast-time dimension to obtain M sets of range profiles, each range profile representing the distance between the echo location and the radar; and determining the target based on the M sets of range profiles. The target location includes the target itself. Based on the M-frame slow-time dimension echo signal at the target location, M local estimates of instantaneous frequencies are constructed. Based on the amplitude information of the M-frame slow-time dimension echo signal at the target location, M confidence levels corresponding to the M local estimates of instantaneous frequencies are constructed. Based on the M local estimates of instantaneous frequencies, the M confidence levels, the first-order continuity constraint information of the local estimates of instantaneous frequencies, and the second-order continuity constraint information of the local estimates of instantaneous frequencies, a global optimization function is constructed. Solving the global optimization function yields the target's trajectory within the M frames.

[0006] In this embodiment, range-oriented processing of the echo signal data acquired by radar is used to separate different spatial units. Based on this, the instantaneous frequency of the slow-time signal within each range unit is directly estimated, and stable trajectory reconstruction is achieved by combining global continuity constraints. This avoids the computational complexity and resolution limitations caused by the spectrogram during construction. Furthermore, by introducing first-order and second-order continuity constraints, random fluctuations and abrupt changes in instantaneous frequency estimation are effectively suppressed, resulting in stronger adaptability to nonlinear motions such as target acceleration and deceleration, thereby improving detection robustness.

[0007] In one possible implementation of the first aspect, local estimates of M instantaneous frequencies are constructed, including: the slow-time dimension echo signal of the m-th frame at the target location. The local estimate of the instantaneous frequency of the m-th frame is calculated using the following formula: ;in, This represents the local estimate of the instantaneous frequency at the target location in the slow time dimension of the m-th frame; m represents the frame number in the slow time dimension, where m is greater than or equal to 2 and less than or equal to M. This indicates the time interval between two adjacent slow-time echo signals; This indicates the operation of taking the principal argument of the complex number within the parentheses; This represents the echo signal in the slow time dimension of the m-th frame at the target location; This represents the slow-time dimension echo signal at the target location in the (m-1)th frame. The conjugate of complex numbers.

[0008] In one possible implementation of the first aspect, constructing M confidence levels corresponding to M local estimates of instantaneous frequencies includes: calculating the confidence level corresponding to the local estimate of the instantaneous frequency in the m-th frame according to the following formula: ;in, This represents the confidence level corresponding to the local estimate of the instantaneous frequency in the slow time dimension of the m-th frame at the target location; m represents the frame number in the slow time dimension, which is greater than or equal to 2 and less than or equal to M; "||" represents the operation of taking the modulus of the complex number. This represents the echo signal in the slow time dimension of the m-th frame at the target location; α is a preset weighting adjustment coefficient.

[0009] In one possible implementation of the first aspect, the global optimization function is: ;in, This represents the instantaneous frequency at which the global optimization function reaches its minimum value in the slow time dimension of the m-th frame; m represents the frame number in the slow time dimension, where m is greater than or equal to 2 and less than or equal to M. This represents the local estimate of the instantaneous frequency at the target location in the slow time dimension of the m-th frame; This represents the confidence level corresponding to the local estimate of the instantaneous frequency in the slow time dimension of the m-th frame at the target location; For the local estimate of the instantaneous frequency of the m-th frame, there is a first-order continuity constraint term. For the local estimate of the instantaneous frequency of the m-th frame, there is a second-order continuity constraint term. These are the weighting coefficients for the first-order continuity constraint term; These are the weighting coefficients of the second-order continuity constraint term.

[0010] In this embodiment, by incorporating the amplitude information of the echo signal into the confidence calculation of instantaneous frequency estimation, the weight of each frame's local estimation value in the global optimization can be adaptively adjusted according to the intensity of the echo signal. The instantaneous frequency estimation corresponding to a stronger echo is more reliable. Through weighting, the impact of noise interference from weak echo frames on the final trajectory estimation result can be further reduced. At the same time, the global optimization function constructed in this application directly integrates confidence weighting, first-order continuity constraints, and second-order continuity constraints. By simply minimizing the global optimization function, a smooth instantaneous frequency estimation result that conforms to the motion law can be directly obtained, thereby obtaining the target's motion trajectory. There is no need to construct a high-dimensional time-spectrum map or perform complex operations such as peak search. While reducing the amount of computation, the accuracy and stability of target trajectory detection are further improved.

[0011] In one possible implementation of the first aspect, solving the global optimization function to obtain the target's trajectory within M frames includes: constructing a strip linear equation of the global optimization function; and solving the strip linear equation to obtain the target's trajectory within M frames.

[0012] In this embodiment, the target motion trajectory is solved by constructing a strip linear equation corresponding to a global optimization function. This allows for differentiated weighting of the instantaneous frequency local estimates in different frames using the aforementioned confidence level. Simultaneously, first-order and second-order continuity constraints ensure that the smoothness of the motion trajectory conforms to the physical characteristics of the target motion. Furthermore, the structure of the strip linear equation is adapted to the optimization problem form of this scheme. Compared to the general global optimization solution method, it effectively reduces the computational complexity and overhead of the solution process. It also suppresses the interference of low-confidence abnormal local estimates on the final trajectory result, improving the accuracy and robustness of the target motion trajectory estimation result within M frames.

[0013] In one possible implementation of the first aspect, solving the global optimization function to obtain the target's motion trajectory within M frames includes: solving the global optimization function through state-space filtering and smoothing algorithms to obtain the target's motion trajectory within M frames.

[0014] In one possible implementation of the first aspect, performing a range transform on each frame of slow echo signal in the fast time dimension includes: performing a Fourier transform on each frame of slow echo signal in the fast time dimension.

[0015] In a second aspect of this application, a target trajectory detection device is provided. The device includes: a data acquisition module, configured to acquire M frames of slow-time dimension echo signals collected by a radar within an acquisition period, wherein the target detection area includes a target to be detected, and each frame of slow-time dimension echo signal includes N sets of fast-time dimension echo signals, where M and N are both integers greater than or equal to 1; a data processing module, configured to perform range transformation on each frame of slow echo signal in the fast-time dimension to obtain M sets of range profiles, each range profile representing the distance between the echo location and the radar; and a target position determination module, configured to determine the target position based on the M sets of range profiles, wherein the target position includes the target. The module constructs local estimates based on the M-frame slow-time dimension echo signals at the target location, generating M local estimates of instantaneous frequencies. The module constructs confidence scores based on the amplitude information of the M-frame slow-time dimension echo signals at the target location, generating M confidence scores corresponding to the local estimates of the M instantaneous frequencies. The module constructs a global optimization function based on the local estimates of the M instantaneous frequencies, the M confidence scores, the first-order continuity constraints of the local estimates of the instantaneous frequencies, and the second-order continuity constraints of the local estimates of the instantaneous frequencies. The module determines the motion trajectory by solving the global optimization function to obtain the target's motion trajectory within the M frames.

[0016] In a third aspect of this application, an electronic device is provided, including a processor, a memory, and a communication interface. The memory is used to store instructions, and the processor is used to call and execute the instructions to implement the methods described in the first aspect and any possible implementation thereof.

[0017] In a fourth aspect of this application, a computer-readable medium is provided, on which instructions are stored, which, when executed on an electronic device, cause the electronic device to perform the methods of the first aspect and any possible implementation thereof.

[0018] In a fifth aspect of this application, a computer program product is provided, the program product including computer instructions, which, when executed by an electronic device, cause the electronic device to perform the method of the first aspect and any possible implementation of the first aspect.

[0019] The beneficial effects of the second to fifth aspects mentioned above can be found in the relevant description of the first aspect, and will not be repeated here. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings are described below.

[0021] Figure 1 According to some embodiments of this application, a flowchart of a target trajectory detection method is shown;

[0022] Figure 2 According to some embodiments of this application, a block diagram of a target trajectory detection device is shown;

[0023] Figure 3 According to some embodiments of this application, a schematic diagram of the structure of an electronic device is shown;

[0024] Figure 4 According to some embodiments of this application, a block diagram of a system-on-a-chip is shown. Detailed Implementation

[0025] The illustrative embodiments of this application include, but are not limited to, a target trajectory detection method, apparatus, device, medium, and program product.

[0026] To facilitate understanding of the solutions in the embodiments of this application by those skilled in the art, some concepts and terms involved in the embodiments of this application will be explained below.

[0027] (1) Doppler signal of radar: The essence of the Doppler effect is that the relative motion between the wave source and the observer causes a change in the frequency of the wave received by the observer. When radar emits electromagnetic waves, the frequency of the reflected electromagnetic waves changes due to the Doppler effect when the beam illuminates a moving target (such as a person waving, a pedestrian, or a vibrating machine). This echo signal component, which carries the target's velocity information and is caused by the radial motion of the target (moving towards or away from the radar), is called the Doppler signal of radar.

[0028] Radar Doppler signals, also known as echo signals reflected from moving targets and collected by radar, are data containing both fast and slow time-dimension information. Assume the radar echo signal data is in the form of x(n,m), where n=1,…,N; m=1,…,M.

[0029] (2) Fast-time signals from radar:

[0030] The first term "n" in x(n,m) (ranging from 1 to N) corresponds to the fast time dimension. It represents the data sequence obtained by the radar through high-speed sampling of the echo within one transmission cycle (e.g., within one pulse or one linear frequency modulation cycle). Its sampling interval is extremely short (typically on the order of nanoseconds), hence the term "fast" time.

[0031] The fast time dimension is directly related to the propagation time of electromagnetic waves. Since the propagation speed of electromagnetic waves is constant, differences in target distance will lead to different echo arrival times. Therefore, by analyzing the fast time signal (e.g., performing a Fourier transform), the target reflection signal at different distances can be separated, obtaining the target's range profile.

[0032] (3) Slow-time radar signals:

[0033] The second term "m" in x(n,m) (ranging from 1 to M) corresponds to the slow time dimension. It represents the sequence number of multiple consecutive transmission cycles (e.g., M pulses or M modulation cycles) of the radar. Its sampling interval is the radar's pulse repetition period or waveform repetition period, which is much slower than fast time sampling, hence the term "slow" time.

[0034] The slow-time dimension records the history of the target's motion evolution over time. When a fast-time index "n0" is fixed (i.e., a specific range cell is selected, where n0 is an integer from 1 to N), the data sequence x(n0,m) extracted along the slow-time dimension "m" (where m varies from 1 to M) constitutes the slow-time signal for that range cell. The phase of this signal changes slowly with m, and this change is caused by the target's micro-motions (such as velocity and vibration) at that range, containing rich micro-Doppler information.

[0035] As described above, effectively utilizing the target motion information contained in the Doppler information collected by radar is key to improving the robot's environmental perception capabilities.

[0036] In existing technologies, processing methods for Doppler signals from radar typically rely on time-frequency analysis techniques, such as spectrum construction based on the Short-Time Fourier Transform. Specifically, a sliding-window Fourier transform is first performed on the slow-time signal to obtain a two-dimensional time-frequency energy distribution. Then, peak detection or convolutional neural networks are used to analyze the spectrum, thereby extracting the target's motion features. However, this type of method has the following shortcomings:

[0037] First, time-frequency analysis methods inherently rely on fixed-length window functions, which inevitably involve a trade-off between time resolution and frequency resolution, making it difficult to simultaneously capture rapid motion changes and fine frequency structure. In robot environmental perception applications, the gestures or micro-motions that robots need to recognize often exhibit significant non-stationary characteristics. Traditional time-frequency representations are prone to energy diffusion and ambiguity, leading to a decrease in feature extraction accuracy.

[0038] Secondly, time-spectrum graph-based methods require energy calculation and search on a discrete frequency grid, resulting in high computational complexity and making them less suitable for embedded platforms (such as mobile robots or low-power edge devices). In real-time sensing tasks requiring high refresh rates and low latency, such methods often fail to meet system performance constraints.

[0039] Furthermore, time-spectrum plots only provide energy distribution information and lack explicit modeling of the signal's intrinsic dynamic structure. In practical applications, robots are more concerned with the continuous change of target motion over time, i.e., the instantaneous frequency trajectory. Traditional methods usually obtain this information indirectly through post-processing (such as ridge extraction), which is redundant and sensitive to noise.

[0040] In existing technologies, another type of method for instantaneous frequency is based on direct estimation of the analytic signal and phase derivative, such as constructing a complex signal using the Hilbert Transform and calculating the phase change rate. However, this type of method has poor stability in noisy environments and lacks constraints on time continuity, making it prone to frequency jumps and estimation jitter, and difficult to directly apply to continuous trajectory analysis in robotic scenarios.

[0041] In summary, existing technologies for instantaneous frequency estimation of Doppler signals in radar generally suffer from the following problems: first, reliance on time-frequency transformation leads to high computational complexity and limited resolution; second, there is a lack of unified modeling for continuous changes in instantaneous frequency; and third, there is insufficient robustness in noisy and complex dynamic scenarios.

[0042] To address this issue, this application provides a target trajectory detection method. First, the echo signal data acquired by radar is processed along the range direction to separate different spatial units, and the range unit containing the target is selected. Based on this, the instantaneous frequency of the slow-time signal within each selected range unit is directly estimated, and a global optimization function is constructed by combining global continuity constraints. By solving the global optimization function, the motion trajectory of the detected target can be obtained. In this way, a unified function with continuously changing instantaneous frequency is constructed. This function is simple to solve, avoiding the computational complexity and resolution limitations caused by constructing a spectrogram, and exhibits good performance in noisy and complex dynamic scenarios.

[0043] It is understood that the target trajectory detection method provided in this application can be applied to electronic devices, meaning that the execution entity of each step of the method is an electronic device. For the sake of simplicity, the execution entity of each step will not be described again in the following description of the method's steps.

[0044] This application does not limit the type of electronic device, which may include, but is not limited to, industrial control equipment (such as robotic arm control cabinets, automated production line controllers), and special-operation robot platforms (such as legged mobile robots, and airborne control computers for multi-legged mobile mechanisms). Furthermore, electronic devices may also include, but are not limited to, mobile phones, wearable devices (such as smartwatches), tablets, desktop, laptop, handheld computers, notebook computers, ultra-mobile personal computers (UMPCs), netbooks, as well as cellular phones, personal digital assistants (PDAs), augmented reality (AR) / virtual reality (VR) devices, etc. This application does not impose any restrictions on the specific type of electronic device.

[0045] To more clearly illustrate the technical solutions provided in the embodiments of this application, the technical solutions of this application will be described below in conjunction with the accompanying drawings.

[0046] Figure 1 A flowchart illustrating a target trajectory detection method is shown according to an embodiment of this application. Figure 1 As shown, the target trajectory detection method may include the following steps S101 to S107:

[0047] S101: Acquire M frames of slow time dimension echo signals collected by the radar in the target detection area during the acquisition period. The target detection area includes the target to be detected. Each frame of slow time dimension echo signal includes N sets of fast time dimension echo signals. M and N are both integers greater than or equal to 1.

[0048] This application embodiment can acquire raw radar sampling data from millimeter-wave radar. The data is organized as M frames of slow-time dimension echo signals, with each frame containing N sets of fast-time dimension echo signals. For example, the echo data acquired by the radar can be in the following form:

[0049] Where m represents the number of sampling points in the slow time dimension (time axis), corresponding to the number of pulses or modulation cycles emitted by the radar, used to capture the temporal changes in target motion. N represents the number of sampling points in the fast time dimension (range axis), corresponding to high-speed sampling within each pulse cycle, used to distinguish targets at different distances.

[0050] For example, assuming the radar's pulse repetition frequency (PRF) is 1 kHz and the acquisition duration is 0.5 seconds, then M = 500. Assuming each pulse has 128 sampling points, then N = 128. These specific values ​​of M and N can be flexibly set according to the radar hardware configuration and application requirements (such as refresh rate and range resolution), and this application does not impose specific limitations.

[0051] S102: For each frame of slow echo signal, perform range transformation in the fast time dimension to obtain M range profiles. Each range profile is used to characterize the distance between the echo location and the radar.

[0052] In some embodiments of this application, performing a range transform on each frame of slow echo signal in the fast time dimension may include performing a Fourier transform on each frame of slow echo signal in the fast time dimension. For example, performing a Fourier transform on each frame of slow echo signal in the fast time dimension can be implemented using the following formula: , where r represents the distance cell index.

[0053] In radar echoes, the reflected signals from targets at different distances exhibit varying time delays in the fast time dimension. Performing a Fourier transform on these fast time signals converts the time delay information into frequency (or range) information, generating a range profile. Each peak in the range profile corresponds to a strongly reflective target at a specific distance.

[0054] For example, performing a Fast Fourier Transform on a fast time series of N=128 yields 128 range cells. Assuming a radar range resolution of 0.1 meters, the peak value of the 50th range cell might correspond to a target 5 meters from the radar.

[0055] The embodiments of this application reduce the modeling complexity of subsequent processing by decomposing the original two-dimensional spatiotemporal coupled data into multiple independent one-dimensional slow-time signals (each corresponding to a distance unit).

[0056] Although the embodiments of this application use Fourier transform, the core purpose of the distance-direction transform is to map time delay to distance. Therefore, other transforms that can achieve this mapping can also be applied to the technical solutions of this application. The core of this application lies in "separation" rather than limiting the specific form of the transform.

[0057] S103: Determine the target location based on the M range profiles, where the target location includes the target itself.

[0058] For the M range images obtained by S102, the range cell (i.e., "target position") of the target can be determined according to preset detection conditions, such as finding the maximum amplitude, energy accumulation, or combining historical experience information.

[0059] It is understandable that moving targets (such as a person waving) will generate strong radar echoes, forming significant peaks in the range profile. Therefore, by analyzing the stable peak positions in multiple consecutive range profiles, the specific range cell where the target is located can be identified.

[0060] For example, in the range profile of M=500 frames, if the 75th range cell shows a peak amplitude higher than the threshold in more than 400 frames, it can be determined that the target is mainly located in that range cell (corresponding to about 7.5 meters).

[0061] It is understandable that the method for determining the target location is not limited to simple peak detection. In complex scenarios (such as multiple targets), clustering, tracking algorithms, or energy criteria can be combined to select multiple possible distance units for parallel processing.

[0062] The embodiments of this application focus global data processing on a single (or multiple) distance cell where the target is located, avoiding the processing of noisy data in irrelevant areas, thereby improving algorithm efficiency and robustness.

[0063] S104: Based on the M-frame slow-time dimension echo signal at the target location, construct local estimates of M instantaneous frequencies.

[0064] In some embodiments of this application, constructing local estimates of M instantaneous frequencies may include: the m-th frame slow-time dimension echo signal at the target location. The local estimate of the instantaneous frequency of the m-th frame is calculated using the following formula: .in, This represents the local estimate of the instantaneous frequency at the target location in the slow time dimension of the m-th frame; m represents the frame number in the slow time dimension, where m is greater than or equal to 2 and less than or equal to M. This indicates the time interval between two adjacent slow-time echo signals; This indicates the operation of taking the principal argument of the complex number within the parentheses; This represents the echo signal in the slow time dimension of the m-th frame at the target location; This represents the slow-time dimension echo signal at the target location in the (m-1)th frame. The conjugate of complex numbers.

[0065] In this embodiment, the slow-time dimension echo signal at the target location is analyzed. The formula for calculating the local estimate of the instantaneous frequency of the m-th frame is used to calculate the local estimate of the instantaneous frequency at each time step (frame), avoiding window functions or Fourier transforms and directly performing continuous estimation in the time domain. The estimation at each time step requires only one complex multiplication and phase extraction, with a complexity of O(1), which is much lower than time-frequency analysis. Direct estimation in continuous frequency space avoids the quantization error caused by discrete frequency grids in traditional methods and can capture more subtle frequency changes. It is suitable for low-latency processing on embedded platforms.

[0066] For example, suppose =0.001 seconds (PRF=1kHz), and If the phase angle of the product is 0.5 radians, then =0.5 / (2π*0.001)≈79.6Hz. This frequency corresponds to the Doppler frequency of the target.

[0067] S105: Based on the amplitude information of the M-frame slow-time dimension echo signal at the target location, construct M confidence levels corresponding to the local estimates of M instantaneous frequencies.

[0068] In some embodiments of this application, constructing M confidence levels corresponding to M local estimates of instantaneous frequencies may include: calculating the confidence level corresponding to the local estimate of the instantaneous frequency in the m-th frame according to the following formula: .in, This represents the confidence level corresponding to the local estimate of the instantaneous frequency in the slow time dimension of the m-th frame at the target location; m represents the frame number in the slow time dimension, which is greater than or equal to 2 and less than or equal to M; "||" represents the operation of taking the modulus of the complex number. This represents the echo signal in the slow time dimension of the m-th frame at the target location; α is a preset weighting adjustment coefficient.

[0069] In this embodiment of the application, the amplitude of the slow-time signal at the target location is utilized. |, according to the formula Calculate each local estimate Corresponding confidence weight .

[0070] The amplitude of the radar echo signal directly reflects the reflection intensity at the scattering point at that moment. Strong reflection (high amplitude) usually means a better signal-to-noise ratio and more reliable phase information. Confidence weighting This reliability quantification is used to differentiate the weighting of observations (local estimates) at different times in subsequent global optimization (step S106). The parameter α can then be used to adjust the contribution of the magnitude to the weights.

[0071] For example, let's set α=2. Assume a slow-time dimension echo signal for a certain frame. If the amplitude is 1.0 (strong), then =1; Assuming a slow-time dimension echo signal in a certain frame. The amplitude is 0.3 (weak, possibly affected by obstruction or noise), then =0.09. In the optimization, the influence of local estimates for weak frames will be significantly reduced.

[0072] It is understandable that the parameter α can be adjusted according to the actual scenario. α=1 represents linear weighting; α>1 will further amplify the weight difference between strong and weak signals, making it more robust to noise; α<1 will reduce the difference. Furthermore, the confidence level is not limited to power of the amplitude; other functions based on signal-to-noise ratio estimation can also be used.

[0073] This application's embodiments suppress the influence of observation data at low signal-to-noise ratio moments (weak amplitude) on the final trajectory by constructing confidence levels, thereby reducing noise interference. The physical characteristics (scattering intensity) of radar observations are incorporated into the optimization model. Furthermore, the weights dynamically change with signal energy to adapt to variations in target reflection intensity over time (such as different parts of a hand gesture).

[0074] S106: Based on the local estimates of M instantaneous frequencies, M confidence levels, first-order continuity constraint information of the local estimates of instantaneous frequencies, and second-order continuity constraint information of the local estimates of instantaneous frequencies, construct a global optimization function.

[0075] In some embodiments of this application, the global optimization function can be the following formula: .

[0076] in, This represents the instantaneous frequency at which the global optimization function reaches its minimum value in the slow time dimension of the m-th frame; m represents the frame number in the slow time dimension, where m is greater than or equal to 2 and less than or equal to M. This represents the local estimate of the instantaneous frequency at the target location in the slow time dimension of the m-th frame; This represents the confidence level corresponding to the local estimate of the instantaneous frequency in the slow time dimension of the m-th frame at the target location; For the local estimate of the instantaneous frequency of the m-th frame, there is a first-order continuity constraint term. For the local estimate of the instantaneous frequency of the m-th frame, there is a second-order continuity constraint term. These are the weighting coefficients for the first-order continuity constraint term; These are the weighting coefficients of the second-order continuity constraint term.

[0077] In this embodiment, local estimates, confidence levels, first-order continuity constraints, and second-order continuity constraints are integrated to construct a global optimization function (objective function) for solving the final smooth trajectory f(m). This function optimizes the entire trajectory {f(1),...,f(M)}, which is a global optimization problem aiming to find a path that minimizes the global optimization function. (m is an integer that takes values ​​from 1 to M in sequence), and these M numbers constitute the optimal continuous path.

[0078] The first item (observation item) This ensures that the target's trajectory f(m) is as close as possible to the weighted local observation.

[0079] The second term (first-order constraint). This ensures that the velocity changes (frequency difference) of the target's trajectory are gradual, avoiding abrupt jumps, and conforming to the continuity of the target's velocity.

[0080] The third term (second-order constraint). This ensures that the acceleration change of the trajectory (second-order frequency difference) is gradual, making the trajectory smoother and conforming to the continuity of the target acceleration.

[0081] Weighting coefficient , This balances the importance of "fitting observed data" and "trajectory smoothing." For example, in gesture recognition, where trajectory smoothing is desired to filter out jitter, a larger setting can be used. (e.g., 10) and moderate (For example, 1). For vehicle emergency braking detection, a smaller [scale] may be required. To capture sudden changes in acceleration.

[0082] In the embodiments of this application, the constraints in the global optimization function include first-order and second-order difference constraints. In some possible embodiments, higher-order constraints may also be introduced to model more complex dynamic models, and this application does not impose any restrictions on this.

[0083] In this embodiment of the application, the weighting coefficient , The specific value can be adjusted and determined through experiments or specific application scenarios (target motion characteristics). This application provides a modeling framework.

[0084] This application's embodiments unify instantaneous frequency estimation and trajectory smoothing within a single optimization framework, avoiding the accumulation of step-by-step errors inherent in the traditional "drawing the graph first, then finding the line" approach. Furthermore, first-order and second-order continuity constraints force the trajectory to conform to the general physical laws of target motion (velocity and acceleration will not change abruptly). In addition, by adjusting... , It can adapt to different types of sports: Larger sizes are better suited for uniform or smooth motion; , When smaller, they are more sensitive to rapidly changing motions (such as sudden acceleration).

[0085] S107: Solve the global optimization function to obtain the target's motion trajectory within M frames.

[0086] In some embodiments of this application, solving the global optimization function to obtain the target's motion trajectory within M frames may include: constructing a strip linear equation of the global optimization function; solving the strip linear equation to obtain the target's motion trajectory within M frames.

[0087] By transforming the quadratic global optimization function in step S106 into a system of linear equations in a band. Then, a dedicated algorithm for banded matrices (such as banded LU decomposition) is used for efficient solution.

[0088] The derivation process of the above-mentioned linear equation system is described below.

[0089] It can be understood that the above global optimization function is a function about variables. The quadratic form of the function. Finding the minimum point of this function is equivalent to solving its corresponding system of linear equations.

[0090] By taking the derivative of the optimization function and setting it to zero, we can obtain the system of linear equations that its optimal solution must satisfy. The specific derivation is as follows:

[0091] Treat the objective function as a vector The function. For each Find the partial derivatives. Combine all the conditions where the partial derivatives are equal to zero to form a system of linear equations. This system of equations can be written in matrix form as follows:

[0092] ,in:

[0093] It is a vector of local estimates. W is a diagonal matrix, and its diagonal elements are the confidence weights. ,Right now .

[0094] D1 is a matrix representing the first-order difference (frequency difference between adjacent frames). It is an (M-1)×M matrix, and each row (corresponding to the m-th difference) is in the form [0,...,0,1,-1,0,...,0], where 1 and -1 are located in the m-th and m+1-th columns, respectively. This forms an M×M matrix, where non-zero elements appear only on the main diagonal and in their adjacent positions.

[0095] D2 is a matrix representing the second-order difference (acceleration). It is an (M-2)×M matrix, and each row (corresponding to the m-th second-order difference) is in the form [0,...,0,1,-2,1,0,...,0], where 1,-2,1 are located in the m-th, m+1-th, and m+2-th columns, respectively. This also forms an M×M strip matrix.

[0096] The above system of linear equations (in , Matrix A in ) has a typical banded structure.

[0097] The non-zero elements of matrix A are concentrated only on the main diagonal and a few nearby diagonals. Specifically, because The introduced terms make matrix A have values ​​at positions (m,m) and (m,m±1); The introduced terms ensure that matrix A has values ​​at positions (m,m), (m,m±1), and (m,m±2). Therefore, matrix A is a banded matrix with a bandwidth of 5 (considering the widest case).

[0098] This banded structure reduces the computational complexity of solving this system of linear equations from O(M) to O(M) of solving a general matrix. 3 The computational cost is significantly reduced to O(M). Furthermore, algorithms specifically designed for banded matrices can be used for efficient solutions. For example, banded LU decomposition decomposes a banded matrix A into a lower triangular matrix L and an upper triangular matrix U. Since L and U also maintain a banded structure, the storage and computational costs required for decomposition and solution are significantly reduced. Another example is the Thomas Algorithm, which offers a simpler direct solution for specific forms of banded equations (such as tridiagonal systems).

[0099] Assuming M = 500, then the size of matrix A is 500 × 500. Solving a typical 500 × 500 dense matrix requires approximately O(500) time complexity. 3 The computational complexity is reduced to approximately O(5*500) for a strip matrix with a bandwidth of 5. This significantly improves computational efficiency and meets the real-time requirements of robots. Furthermore, this solution method is mature and yields stable results.

[0100] In some embodiments of this application, solving the global optimization function to obtain the target's motion trajectory within M frames may further include: solving the global optimization function through state-space filtering and smoothing algorithms to obtain the target's motion trajectory within M frames.

[0101] For example, suppose the state vector It includes not only the instantaneous frequency at the current moment. It also includes its rate of change (i.e., the first difference of frequency). This can be considered as "speed".

[0102] State transition equations (motion model): describe how the state evolves from time m to time m+1. First-order and second-order continuity constraints can be naturally incorporated into this model. A simple linear model can be represented as: Where F is the state transition matrix. To reflect smoothness, F can be designed as follows: ,in It is a slow time frame interval. It is process noise (uncaptured dynamics or model errors in the modeling), usually assumed to be zero-mean Gaussian white noise.

[0103] The observation equation describes how to obtain observations (i.e., local estimates) from the state, and it can be: .in, These are observed values. It is observation noise, and its variance can be related to the confidence weight. Related. The lower the confidence level, that is... The smaller the value, the lower the observation noise is considered to be. The larger the variance, the less reliable the observation. This achieves the equivalent role of confidence weights in the state estimation framework.

[0104] In this embodiment, the global optimization problem is reinterpreted as a state-space estimation problem. Instantaneous frequencies are considered as states, local estimates are considered as noisy observations, and first- and second-order continuity constraints correspond to the smoothness assumptions in the state transition model. Then, a Kalman filter is used for forward filtering, combined with a smoothing algorithm (such as RTS smoothing) to obtain the globally optimal trajectory.

[0105] Thus, the Kalman filter framework supports online real-time estimation, updating the trajectory estimate with each new frame of data received. The Bayesian framework provides a way to understand observation noise and model uncertainty within a probabilistic framework. Furthermore, the state-space model is easily extended (e.g., by introducing more complex motion models), enhancing robustness.

[0106] In this embodiment, the solution to the global optimization function is the continuous and smooth instantaneous frequency trajectory of the target over the entire M-frame time. The trajectory f(m) directly corresponds to the radial velocity change of the target and can be used for subsequent tasks such as gesture classification and behavior recognition. According to the formula... (λ is the radar wavelength) can be further converted into an instantaneous velocity trajectory.

[0107] In summary, the target trajectory detection method provided in this application achieves efficient, high-precision, and robust processing of radar micro-Doppler signals by constructing a technical framework of "range-axis separation, direct time-domain estimation, and global constraint optimization." This solution has the following significant advantages:

[0108] First, the significant reduction in computational complexity meets the stringent requirements of real-time robot perception. This solution abandons the highly complex process of constructing a two-dimensional time-spectrum graph and performing peak search in traditional methods, replacing it with direct frequency estimation based on phase difference and global optimization with a strip structure. This method reduces the overall computational complexity from O(M*KlogK) (where K is the number of frequency grids) or higher in traditional methods to a linear order of O(M). This enables the algorithm to achieve low-latency, high-refresh-rate real-time operation on resource-constrained embedded robot platforms (such as mobile robots and robotic arm controllers), providing a feasible technical foundation for applications with extremely high real-time requirements, such as dynamic human-robot interaction and near-field obstacle avoidance.

[0109] Secondly, this method overcomes the limitations of discrete frequency grids, achieving high-resolution instantaneous frequency estimation in the continuous domain. Traditional time-frequency analysis methods are constrained by fixed window functions and discrete frequency quantization grids, resulting in an inherent contradiction between time and frequency resolution, making it difficult to accurately characterize rapidly changing non-stationary signals. This approach estimates instantaneous frequency by directly calculating the phase change rate in the continuous time domain, avoiding quantization errors caused by frequency discretization. This allows the proposed solution to capture the fine Doppler frequency changes caused by micro-motions of targets (such as subtle finger tremors or joint flexion) with theoretically lossless resolution, thereby significantly improving the ability to characterize and recognize subtle motion features.

[0110] Furthermore, this application significantly improves the smoothness and stability of trajectory estimation through integrated modeling and physical constraints. This scheme unifies instantaneous frequency estimation and trajectory smoothing within a single global optimization model. By introducing first-order (velocity continuity) and second-order (acceleration continuity) constraints, the final output frequency trajectory conforms to the physical laws of target motion, meaning that velocity and acceleration will not undergo physically meaningless abrupt changes. This overcomes the error accumulation problem inherent in the traditional step-by-step "estimate first, then smooth" process, as well as the tendency for frequency jumps to occur under noise in pure phase derivative estimation methods. Therefore, this scheme can output a smoother, more stable, and physically meaningful motion trajectory, exhibiting stronger adaptability to nonlinear motions such as target acceleration and deceleration.

[0111] Furthermore, this application's technical solution introduces an adaptive confidence mechanism to enhance robustness in low signal-to-noise ratio (SNR) and complex environments. This application transforms the amplitude information of radar echo signals into observation confidence weights and incorporates them into the global optimization process. This enables the algorithm to adaptively distinguish between observation data at high SNR (strong scattering points) and low SNR (weak reflections or noise-dominated) times. During optimization, reliable observations with high confidence are given greater weight, while the influence of unreliable observations with low confidence is effectively suppressed. This significantly improves the algorithm's stability and anti-interference capability when facing adverse factors such as signal attenuation, partial obstruction, or environmental electromagnetic interference.

[0112] In some embodiments, this application also provides a target trajectory detection device. (See reference...) Figure 2 The diagram shown is a block diagram of a target trajectory detection device 100 according to some embodiments of this application. Figure 2 As shown, the target trajectory detection device 100 includes:

[0113] The data acquisition module 01 is used to acquire M frames of slow time dimension echo signals collected by the radar in the target detection area during the acquisition period. The target detection area includes the target to be detected. Each frame of slow time dimension echo signal includes N sets of fast time dimension echo signals. M and N are both integers greater than or equal to 1.

[0114] Data processing module 02 is used to perform range transformation on each frame of slow echo signal in the fast time dimension to obtain M sets of range profiles. Each range profile is used to characterize the distance between the echo point and the radar.

[0115] The target location determination module 03 is used to determine the target location based on M sets of range images, where the target location includes the target.

[0116] The local estimation construction module 04 is used to construct local estimates of M instantaneous frequencies based on the M-frame slow-time dimension echo signal at the target location.

[0117] The confidence construction module 05 is used to construct M confidence levels corresponding to M instantaneous frequency local estimates based on the amplitude information of the M-frame slow time dimension echo signal at the target location.

[0118] The global optimization function construction module 06 is used to construct a global optimization function based on the local estimates of M instantaneous frequencies, M confidence levels, first-order continuity constraint information of the local estimates of instantaneous frequencies, and second-order continuity constraint information of the local estimates of instantaneous frequencies.

[0119] The motion trajectory determination module 07 is used to solve the global optimization function to obtain the motion trajectory of the target within M frames.

[0120] In some embodiments, this application also provides a readable storage medium storing a program or instructions that, when executed on an electronic device, cause the electronic device to perform the target trajectory detection method described in the above embodiments.

[0121] In some embodiments, this application also provides a program product, including: a program or instructions, which, when run on an electronic device, cause the electronic device to perform the target trajectory detection method described in the above embodiments.

[0122] In some embodiments, this application also provides an electronic device, which includes: one or more processors; one or more memories; the one or more memories storing one or more programs, which, when executed by one or more processors, cause the electronic device to perform the target trajectory detection method described in the above embodiments.

[0123] Now for reference Figure 3 The diagram shows a block diagram of an electronic device 1300 according to some embodiments of this application. The electronic device 1300 may include one or more first processors 1301 coupled to a controller hub 1303. In at least one embodiment, the controller hub 1303 communicates with the first processor 1301 via a multi-branch bus such as a Front Side Bus (FSB) or a point-to-point interface such as a Quick Path Interconnect (QPI). The first processor 1301 executes instructions controlling general-type data processing operations. In one embodiment, the controller hub 1303 includes, but is not limited to, a Graphics & Memory Controller Hub (GMCH) (not shown) and an Input / Output Hub (IOH) (which may be on a separate chip) (not shown), wherein the GMCH includes memory and a graphics controller and is coupled to the IOH.

[0124] Electronic device 1300 may also include a first coprocessor 1302 and a memory 1304 coupled to a controller hub 1303. Alternatively, one or both of the memory and GMCH may be integrated within the processor (as described in this application), with memory 1304 and the first coprocessor 1302 directly coupled to the first processor 1301 and the controller hub 1303, which is located on a single chip with IOH.

[0125] Memory 1304 may be, for example, Dynamic Random Access Memory (DRAM), Phase Change Memory (PCM), or a combination of both. Memory 1304 may include one or more tangible, non-transitory computer-readable media for storing data and / or instructions. The computer-readable storage medium stores instructions, specifically, temporary and permanent copies of those instructions. These instructions may include: instructions that, when executed by at least one processor, cause electronic device 1300 to implement the target trajectory detection method provided in the embodiments of this application. When the instructions are executed on a computer, the computer performs the target trajectory detection method provided in the embodiments of this application.

[0126] In one embodiment, the first coprocessor 1302 is a dedicated processor, such as, for example, a high-throughput many integrated core (MIC) processor, a network or communication processor, a compression engine, a graphics processor, general-purpose computing on graphics processing units (GPGPU), or an embedded processor, etc. Optional properties of the first coprocessor 1302 are indicated by dashed lines. Figure 3 middle.

[0127] In one embodiment, electronic device 1300 may further include a Network Interface Controller (NIC) 1306. The network interface 1306 may include a transceiver for providing a radio interface for electronic device 1300 to communicate with any other suitable device, such as a front-end module, antenna, etc. In various embodiments, the network interface 1306 may be integrated with other components of electronic device 1300. The network interface 1306 can implement the functions of the communication unit in the above embodiments.

[0128] Electronic device 1300 may further include input / output (I / O) device 1305. Input / output device 1305 may include: a user interface designed to enable a user to interact with electronic device 1300; a peripheral component interface designed to enable peripheral components to interact with electronic device 1300; and / or sensors designed to determine environmental conditions and / or location information related to electronic device 1300.

[0129] It is worth noting that, Figure 3 This is merely an example. That is, although... Figure 3The electronic device 1300 is shown to include multiple devices such as a first processor 1301, a controller hub 1303, and a memory 1304. However, in actual applications, devices using the methods of this application may include only a portion of the devices in the electronic device 1300. For example, it may include only the first processor 1301 and the network interface 1306. Figure 3 Devices with optional properties are shown with dashed lines.

[0130] Now for reference Figure 4 The diagram shown is a block diagram of a System-on-Chip (SoC) 1400 according to some embodiments of this application. Figure 4 In the diagram, the dashed box is an optional feature for more advanced SoCs. Figure 4 In the SoC 1400, there are: an interconnect unit 1450 coupled to a second processor 1410; a system proxy unit 1480; a bus controller unit 1490; an integrated memory controller unit 1440; a group or one or more second coprocessors 1420, which may include integrated graphics logic, an image processor, an audio processor, and a video processor; and a static random access memory (SRAM) unit, i.e. Figure 4 The SRAM cell 1430 shown is a Direct Memory Access (DMA) cell, i.e. Figure 4 The DMA unit 1460 shown is illustrated. In one embodiment, the second coprocessor 1420 includes a dedicated processor, such as, for example, a network or communication processor, a compression engine, general-purpose computing on graphics processing units (GPGPU), a high-throughput MIC processor, or an embedded processor.

[0131] SRAM cell 1430 may include one or more tangible, non-transitory computer-readable media for storing data and / or instructions. The computer-readable storage medium stores instructions, specifically, temporary and permanent copies of the instructions. The instructions may include: instructions that, when executed by at least one processor, cause the SoC to implement the target trajectory detection method disclosed in embodiments of this application. When the instructions are executed on a computer, they cause the computer to perform the target trajectory detection method disclosed in embodiments of this application.

[0132] It is understood that, as used herein, the term “module” may refer to or include, or be part of, an application-specific integrated circuit (ASIC), electronic circuitry, a processor (shared, dedicated, or grouped) and / or memory that executes one or more software or firmware programs, combinational logic circuitry, and / or other suitable hardware components that provide the described functionality.

[0133] It is understood that in the various embodiments of this application, the processor may be a microprocessor, a digital signal processor, a microcontroller, etc., and / or any combination thereof. According to another aspect, the processor may be a single-core processor, a multi-core processor, etc., and / or any combination thereof.

[0134] The embodiments disclosed in this application can be implemented in hardware, software, firmware, or a combination of these implementation methods. Embodiments of this application can be implemented as computer programs or program code executable on a programmable system, the programmable system including at least one processor, a storage system (including volatile and non-volatile memory and / or storage elements), at least one input device, and at least one output device.

[0135] Program code can be applied to input instructions to execute the functions described in this application and generate output information. The output information can be applied to one or more output devices in a known manner. For the purposes of this application, the processing system includes any system having a processor such as, for example, a digital signal processor (DSP), a microcontroller, an application-specific integrated circuit (ASIC), or a microprocessor.

[0136] The program code can be implemented using a high-level procedural language or an object-oriented programming language to communicate with the processing system. Assembly language or machine language can also be used when needed. In fact, the mechanisms described in this application are not limited to any particular programming language. In either case, the language can be a compiled language or an interpreted language.

[0137] In some cases, the disclosed embodiments may be implemented in hardware, firmware, software, or any combination thereof. The disclosed embodiments may also be implemented as instructions carried or stored thereon on one or more temporary or non-temporary machine-readable (e.g., computer-readable) storage media, which may be read and executed by one or more processors. For example, the instructions may be distributed via a network or through other computer-readable media. Therefore, machine-readable media may include any mechanism for storing or transmitting information in a machine-readable (e.g., computer-readable) form, including but not limited to floppy disks, optical disks, CD-ROMs, magneto-optical disks, read-only memory (ROM), random access memory (RAM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), magnetic cards or optical cards, flash memory, or tangible machine-readable storage for transmitting information (e.g., carrier waves, infrared signals, digital signals, etc.) using the Internet in the form of electrical, optical, acoustic, or other forms of propagated signals. Therefore, machine-readable media includes any type of machine-readable medium suitable for storing or transmitting electronic instructions or information in a machine-readable (e.g., computer-readable) form.

[0138] In the accompanying drawings, some structural or methodological features may be shown in a specific arrangement and / or order. However, it should be understood that such a specific arrangement and / or order may not be necessary. Rather, in some embodiments, these features may be arranged in a manner and / or order different from that shown in the illustrative drawings. Furthermore, the inclusion of structural or methodological features in a particular figure does not imply that such features are required in all embodiments, and in some embodiments, these features may be omitted or may be combined with other features.

[0139] It should be noted that all units / modules mentioned in the device embodiments of this application are logical units / modules. Physically, a logical unit / module can be a physical unit / module, a part of a physical unit / module, or a combination of multiple physical units / modules. The physical implementation of these logical units / modules themselves is not the most important factor; the combination of functions implemented by these logical units / modules is the key to solving the technical problems proposed in this application. Furthermore, to highlight the innovative aspects of this application, the above-described device embodiments of this application have not introduced units / modules that are not closely related to solving the technical problems proposed in this application. This does not mean that the above-described device embodiments do not contain other units / modules.

[0140] It should be noted that in the examples and description of this application, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0141] Although this application has been illustrated and described with reference to certain preferred embodiments thereof, those skilled in the art will understand that various changes in form and detail may be made thereto without departing from the scope of this application.

Claims

1. A target trajectory detection method, applied to electronic devices, characterized in that, The method includes: The radar acquires M frames of slow time dimension echo signals collected from the target detection area within the acquisition period. The target detection area includes the target to be detected. Each frame of the slow time dimension echo signal includes N sets of fast time dimension echo signals, where M and N are both integers greater than or equal to 1. For each frame of the slow-time dimension echo signal, a range transformation is performed on the fast-time dimension to obtain M sets of range profiles. Each range profile is used to characterize the distance between the echo location and the radar. Based on the M sets of distance images, the target location is determined, and the target location includes the target. Based on the M-frame slow-time dimension echo signal at the target location, construct M local estimates of instantaneous frequencies; Based on the amplitude information of the M-frame slow-time dimension echo signal at the target location, M confidence levels are constructed corresponding to the local estimates of the M instantaneous frequencies. Based on the local estimates of the M instantaneous frequencies, the M confidence levels, the first-order continuity constraint information of the local estimates of the instantaneous frequencies, and the second-order continuity constraint information of the local estimates of the instantaneous frequencies, a global optimization function is constructed. Solve the global optimization function to obtain the target's motion trajectory within M frames.

2. The method according to claim 1, characterized in that, The construction of local estimates of M instantaneous frequencies includes: The slow time dimension echo signal of the m-th frame at the target location The local estimate of the instantaneous frequency of the m-th frame is calculated using the following formula: ; in, This represents the local estimate of the instantaneous frequency at the target location in the slow time dimension of the m-th frame; The frame number represents the slow time dimension, where m is greater than or equal to 2 and less than or equal to M; This represents the time interval between two adjacent frames of the slow-time dimension echo signal; This indicates the operation of taking the principal argument of the complex number within the parentheses; This represents the echo signal in the slow time dimension at the target location in the m-th frame; This represents the slow-time dimension echo signal at the target location in the (m-1)th frame. The conjugate of complex numbers.

3. The method according to claim 1 or 2, characterized in that, The construction of the M confidence levels corresponding to the local estimates of the M instantaneous frequencies includes: The confidence level corresponding to the local frequency estimate of the m-th frame is calculated using the following formula: ; in, This represents the confidence level corresponding to the local estimate of the instantaneous frequency in the slow time dimension of the m-th frame at the target location; m represents the frame number in the slow time dimension, where m is greater than or equal to 2 and less than or equal to M; "||" represents the operation of taking the modulus of a complex number; This represents the echo signal in the slow time dimension at the target location in the m-th frame; α is a preset weighting adjustment coefficient.

4. The method according to claim 1, characterized in that, The global optimization function is: ； in, Let m represent the instantaneous frequency at which the global optimization function reaches its minimum value in the slow time dimension of the m-th frame; m represents the frame number in the slow time dimension, where m is greater than 2 and less than or equal to M; This represents the local estimate of the instantaneous frequency at the target location in the slow time dimension of the m-th frame; This represents the confidence level corresponding to the local estimate of the instantaneous frequency in the slow time dimension of the m-th frame at the target location; For the local estimate of the instantaneous frequency of the m-th frame, there is a first-order continuity constraint term. For the local estimate of the instantaneous frequency of the m-th frame, there is a second-order continuity constraint term. The weighting coefficients of the first-order continuity constraint term; is the weighting coefficient of the second-order continuity constraint term.

5. The method according to claim 4, characterized in that, Solving the global optimization function to obtain the target's motion trajectory within M frames includes: Construct the banded linear equation of the global optimization function; Solving the linear equation in the band yields the trajectory of the target within M frames.

6. The method according to claim 4, characterized in that, Solving the global optimization function to obtain the target's motion trajectory within M frames includes: The global optimization function is solved by state-space filtering and smoothing algorithms to obtain the target's motion trajectory within M frames.

7. The method according to claim 1, characterized in that, For each frame of the slow-time dimension echo signal, a range transformation is performed in the fast-time dimension, including: For each frame, the slow-time dimension echo signal is subjected to a Fourier transform in the fast-time dimension.

8. A target trajectory detection device, characterized in that, The device includes: The data acquisition module is used to acquire M frames of slow time dimension echo signals collected by the radar in the target detection area during the acquisition period. The target detection area includes the target to be detected. Each frame of the slow time dimension echo signal includes N sets of fast time dimension echo signals, where M and N are both integers greater than or equal to 1. The data processing module is used to perform range transformation on the fast time dimension for each frame of the slow time dimension echo signal to obtain M sets of range profiles, each of which is used to characterize the distance between the echo location and the radar. A target location determination module is used to determine the target location based on the M sets of distance images, wherein the target location includes the target; The local estimation construction module is used to construct M instantaneous frequency local estimates based on the M-frame slow-time dimension echo signals at the target location; The confidence construction module is used to construct M confidence levels corresponding to the local estimates of the M instantaneous frequencies based on the amplitude information of the M-frame slow-time dimension echo signals at the target location. The global optimization function construction module is used to construct a global optimization function based on the local estimates of the M instantaneous frequencies, the M confidence levels, the first-order continuity constraint information of the local estimates of the instantaneous frequencies, and the second-order continuity constraint information of the local estimates of the instantaneous frequencies. The motion trajectory determination module is used to solve the global optimization function to obtain the motion trajectory of the target within M frames.

9. An electronic device, comprising a processor, a memory, and a communication interface, characterized in that, The memory is used to store instructions, and the processor is used to call and execute the instructions to implement the method as described in any one of claims 1 to 7.

10. A computer-readable medium, characterized in that, The readable medium stores instructions that, when executed on an electronic device, cause the electronic device to perform the method of any one of claims 1 to 7.

11. A computer program product, characterized in that, The program product includes computer instructions that, when executed by an electronic device, perform the method of any one of claims 1 to 7.

Images