A vehicle trajectory reconstruction method based on rotation projection

Abstract

This invention discloses a vehicle trajectory reconstruction method based on rotational projection, relating to the field of vehicle trajectory detection and reconstruction. The method includes: acquiring the original distributed fiber acoustic sensor data matrix and preprocessing it to obtain a preprocessed matrix; rotating the preprocessed matrix by multiple preset angles, performing column-direction integration on each rotated matrix to obtain the corresponding projection vector, generating a multi-angle projection matrix; performing signal-to-noise ratio enhancement processing to generate a spatial signal enhancement function; determining the peak position of each candidate vehicle trajectory in the spatial dimension and its corresponding optimal rotation angle through a two-stage peak-finding strategy; and reconstructing the rotated trajectory lines to the original spatiotemporal domain through coordinate inverse mapping based on the peak position and its corresponding optimal rotation angle, obtaining the vehicle trajectory equation. This method transforms two-dimensional trajectory detection into one-dimensional peak detection through rotational projection, achieving vehicle trajectory reconstruction that is training-free, robust, and lightweight.

Description

Technical Field

[0001] This invention relates to the field of vehicle trajectory detection and reconstruction, and more specifically to a vehicle trajectory reconstruction method based on rotational projection. Background Technology

[0002] With the development of smart cities and intelligent transportation systems, distributed fiber optic acoustic sensing technology has been widely used in vehicle trajectory perception in scenarios such as highways, tunnels and cross-sea bridges because it can utilize existing communication optical cables to achieve continuous monitoring over all weather, long distances and with strong interference resistance. In the spatiotemporal data map of this technology, the vehicle driving trajectory is characterized by discontinuous linear features with a specific slope. Accurately extracting these trajectories from a strong noise background is a key prerequisite for realizing traffic flow statistics and vehicle speed monitoring.

[0003] In existing methods, the Hough transform is a common means of extracting straight-line trajectories. However, it is highly dependent on the quality of preprocessing steps such as edge detection. In actual distributed fiber acoustic sensing data, vehicle trajectories are generally severely discontinuous due to occlusion, weak signals, or multipath interference, resulting in a large number of fragmented edges in edge detection. This makes the votes of the Hough transform dispersed in the parameter space, making it difficult to form effective peaks. As a result, the detection effect of the Hough transform is poor and the false negative rate is high. At the same time, the detection performance of the Hough transform is extremely sensitive to the preprocessing parameters, while the noise distribution in real-world scenarios is complex and non-stationary, making stable deployment difficult.

[0004] Another technical approach is to use convolutional neural networks combined with Hough transform for trajectory extraction. This type of method first converts the sensor data into fine line images through the neural network, and then outputs the trajectory through Hough transform. However, the neural network model contains a large number of hyperparameters, and the parameter tuning process is complex and dependent on experience. Its performance is highly limited by the scale and distribution of the training data. When the application scenario changes, such as the fiber optic laying method, road surface material, and vehicle type, the model's generalization ability drops significantly, and the accuracy is difficult to maintain. In addition, the reasoning process of deep neural networks lacks interpretability. When an incorrect trajectory is output, the cause cannot be traced, which increases the difficulty of maintenance on the engineering site.

[0005] In addition to the common problems of complex parameter tuning and insufficient generalization of the above methods, existing technologies also face the constraint of deployment cost. The Hough transform has a large computational overhead when processing large-scale data, while neural network solutions usually require dedicated hardware such as graphics processors, making it difficult to integrate into the edge computing nodes of distributed fiber acoustic sensing systems. This limits the real-time operation capability of vehicle trajectory detection algorithms on low-cost, low-power devices in actual roadside or tunnel environments.

[0006] Therefore, how to design a vehicle trajectory reconstruction method based on rotation projection to achieve automatic vehicle trajectory reconstruction that requires no training data, is robust to discontinuous trajectory noise, and is lightweight and deployable is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0007] In view of this, the present invention provides a vehicle trajectory reconstruction method based on rotational projection, which aims to overcome the technical defects of existing Hough transform, such as poor detection effect on discontinuous trajectory, weak noise robustness, and difficulty in parameter tuning, as well as convolutional neural network schemes, such as reliance on a large amount of training data, poor generalization, and high deployment cost. It only requires adjustment of angle interval and peak finding threshold parameters, can run stably in a central processing unit environment, and the detection process is fully interpretable. It can accurately extract vehicle driving trajectories in both positive and negative directions from distributed fiber acoustic sensing data.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] A vehicle trajectory reconstruction method based on rotation projection includes the following steps: S1. Obtain the raw distributed fiber acoustic sensor data matrix and preprocess it to obtain the preprocessed matrix; S2. Rotate the preprocessed matrix by multiple preset angles, perform column integration on each rotated matrix to obtain the corresponding projection vector, and generate a multi-angle projection matrix. S3. Perform signal-to-noise ratio enhancement processing on the multi-angle projection matrix to generate a spatial signal enhancement function; S4. Based on the spatial signal enhancement function, determine the peak position of each candidate vehicle trajectory in the spatial dimension and its corresponding optimal rotation angle through a two-level peak-finding strategy. S5. Based on the peak position and its corresponding optimal rotation angle, the rotated trajectory line is reconstructed to the original spatiotemporal domain through coordinate inverse mapping to obtain the vehicle trajectory equation.

[0010] Preferably, in step S1, the preprocessing includes: S11. Perform Hilbert transform on each spatial channel sequence of the original distributed fiber acoustic sensor data matrix to extract the instantaneous amplitude envelope and generate the envelope energy matrix. S12. Based on the envelope energy matrix, calculate the statistical mean of the response of all spatial channels at each time step, and generate a binary mask matrix according to the preset noise multiplier coefficient. S13. Perform morphological dilation on the binary mask matrix, and perform spatiotemporal scale normalization resampling on the dilated matrix to obtain a preprocessed matrix.

[0011] Preferably, S11 specifically includes: For the original spacetime matrix Each spatial sequence Its analytic signal is constructed through Hilbert transform. And calculate the instantaneous amplitude envelope. The envelope energy matrix E is obtained; Where t is the time sampling point index, x is the spatial channel index, j is the imaginary unit, and H{ } represents the Hilbert transform.

[0012] Preferably, S12 specifically includes: For each time t, calculate the statistical mean of the responses of all spatial channels. ; Define a binary mask matrix based on the preset noise multiplier K. :

[0013] in, Let x be the instantaneous amplitude envelope, and x be the spatial channel index. This represents the total number of spatial channels.

[0014] Preferably, S13 specifically includes: Define a rectangular window with structuring element S, and perform morphological dilation on the binary mask matrix M to obtain the dilated matrix. ; The dilated matrix is ​​then subjected to bilinear interpolation. Resampling to target temporal resolution The preprocessing matrix is ​​obtained. ; in, The total number of spatial channels. Bilinear interpolation function.

[0015] Preferably, S2 includes: Define rotation operator For the preprocessed matrix Perform a rotation operation to obtain the rotated matrix. ; For the rotated matrix Perform column integration to obtain the projection vector. :

[0016] Where t is the index of the time sampling point, x represents the total number of time sampling points, and x represents the spatial channel index. Traverse the preset angle range Given all discrete angles within the range, stack all projection vectors according to angles to construct a multi-angle projection matrix P.

[0017] Preferably, the preset angle range Includes a positive angle range and a negative angle range ,in >0, <0, used to detect the trajectories of vehicles traveling in the forward and reverse directions, respectively.

[0018] Preferably, S3 includes: For each spatial location x, the maximum value is obtained by taking the maximum value in the angular dimension of the multi-angle projection matrix P. :

[0019] in, For the preset angle range, The rotation angle; For each spatial location x, the arithmetic mean is taken along the angular dimension to obtain the background noise response. :

[0020] in, The number of discrete angles within the angle interval; To suppress background noise, a spatial signal enhancement function is defined. The vehicle trajectory is transformed into an enhancement function. The local extreme point.

[0021] Preferably, in step S4, the two-stage peak-finding strategy includes: Primary peak finding: in spatial signal enhancement functions Above, extract salience greater than the first preset threshold. The positions of all candidate peaks are denoted as set. The prominence is defined as the vertical distance between the peak and the higher of its two nearest troughs. Secondary screening: For each candidate peak position Determine the optimal angle that maximizes the projection value from the multi-angle projection matrix P. :

[0022] in, The angle range is preset, where θ is the rotation angle; The projection curve corresponding to the optimal angle Above, cut off Centered on, with radius neighborhood window ; If there is a prominence greater than the second preset threshold within the neighborhood window If a peak is identified, the candidate peak is confirmed to correspond to a real vehicle trajectory, and its precise peak position is recorded. and the corresponding optimal angle .

[0023] Preferably, S5 includes: Define the inverse rotation operator Mapping the points in the rotated image back to the original spatiotemporal coordinate system, the mapping relationship is as follows:

[0024] in, For the optimal rotation angle, t and x are the time and space coordinates in the original spacetime coordinate system, respectively. and These are the spacetime coordinates of the center of rotation. For the time index in the rotated image, For precise peak location; Traverse all The values ​​are taken, the corresponding original spatiotemporal coordinates (t, x) are calculated, the discrete trajectory point set is obtained, and the vehicle trajectory straight line equation is fitted. ,in, b represents the vehicle speed, and b is the intercept.

[0025] As can be seen from the above technical solution, compared with the prior art, the technical solution of the present invention has the following beneficial effects: 1. This method transforms the two-dimensional trajectory detection problem into a peak search problem of a one-dimensional spatial curve by performing multi-angle rotation and column integration on the preprocessed spatiotemporal matrix. Due to the adoption of the full column integration accumulation strategy, even if there are discontinuities in the vehicle trajectory in the original data due to occlusion or weak signals, a significant projected peak can be formed at the correct rotation angle. Therefore, it does not need to rely on preprocessing methods that are susceptible to noise interference, such as edge detection, and can still stably extract the trajectory in complex signal-to-noise ratio environments.

[0026] 2. It does not rely on any neural network model or pre-labeled training dataset. It achieves trajectory reconstruction entirely based on the deterministic computation process of rotation projection and two-level peak finding. Therefore, when the application scenario changes such as fiber optic laying environment, road type, and vehicle type distribution, there is no need to retrain or adjust the personalized model. Only the core parameters such as angle range and peak finding threshold need to be adjusted for application. It has strong generalization ability across scenarios.

[0027] 3. All operations in this method involve basic matrix operations such as matrix rotation, column integration, peak detection, and coordinate inverse mapping. It can run efficiently in a central processing unit environment without a graphics processor, making it easy to integrate into the edge computing nodes of existing distributed fiber optic acoustic sensing systems. At the same time, from the multi-angle projection matrix, the spatial response maximum envelope, the background noise curve to the two-level peak finding results, each intermediate step has a clear physical meaning and a visualized output product. When the detection results are abnormal, the location can be traced step by step, reducing the cost of engineering maintenance and debugging. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0029] Figure 1 A flowchart of a vehicle trajectory reconstruction method based on rotation projection is provided in an embodiment of the present invention; Figure 2 The original DAS vibration phase spatiotemporal diagram provided in the embodiments of the present invention; Figure 3 This is a preprocessed vibration signal diagram provided in an embodiment of the present invention; Figure 4 The matrix image after rotation by -6.6° is provided in an embodiment of the present invention; Figure 5 The projection vector curve and maximum value envelope diagram provided for embodiments of the present invention; Figure 6 A spatial signal enhancement function curve provided for an embodiment of the present invention; Figure 7 This is a superimposed image of the reconstructed trajectory and the original signal provided in an embodiment of the present invention. Detailed Implementation

[0030] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0031] like Figure 1 As shown, this embodiment provides a vehicle trajectory reconstruction method based on rotational projection, including... The following steps: S1. Obtain the raw distributed fiber acoustic sensor data matrix and preprocess it to obtain the preprocessed matrix; S2. Rotate the preprocessed matrix by multiple preset angles, perform column integration on each rotated matrix to obtain the corresponding projection vector, and generate a multi-angle projection matrix. S3. Perform signal-to-noise ratio enhancement processing on the multi-angle projection matrix to generate a spatial signal enhancement function; S4. Based on the spatial signal enhancement function, determine the peak position of each candidate vehicle trajectory in the spatial dimension and its corresponding optimal rotation angle through a two-level peak-finding strategy. S5. Based on the peak position and its corresponding optimal rotation angle, the rotated trajectory line is reconstructed to the original spatiotemporal domain through coordinate inverse mapping to obtain the vehicle trajectory equation.

[0032] It transforms the vehicle trajectory detection problem in a two-dimensional spatiotemporal map into a peak search problem in a one-dimensional projection curve by performing multi-angle rotation and column integration on the preprocessed distributed fiber acoustic sensor data matrix. This process does not rely on any neural network model or pre-labeled training data, and can be completed in a central processing unit environment through basic matrix operations, exhibiting good cross-scenario generalization ability. At the same time, each intermediate step in the entire process has a clear physical meaning and visual output, which facilitates engineering debugging and error tracing, and can effectively support practical applications such as real-time traffic flow monitoring, vehicle speed statistics, and abnormal event early warning in intelligent transportation systems.

[0033] The following provides a further explanation of each step and related features in the above method; In this implementation, S1, the raw distributed fiber acoustic sensor data matrix is ​​acquired and preprocessed to obtain a preprocessed matrix. In practical engineering applications, the original distributed fiber optic acoustic sensing data matrix can be obtained as follows: Standard single-mode communication optical fibers deployed along highways serve as the sensing medium. Distributed fiber optic acoustic sensing devices continuously collect the phase changes of backscattered Rayleigh light at fixed spatial sampling intervals and temporal sampling rates, thereby constructing a matrix such as... Figure 2 The two-dimensional spatiotemporal data matrix shown represents the time sampling points on the horizontal axis and the spatial channel positions along the optical fiber on the vertical axis, corresponding to the actual mileage; for example... Figure 2 As shown, the vehicle trajectory appears as a discontinuous bright line with a certain slope. However, the strong vibration noise in the background, such as bridge joints and uneven road surfaces, is significant, and direct trajectory extraction will face a high rate of missed detection.

[0034] Further preprocessing includes: S11. Perform Hilbert transform on each spatial channel sequence of the original distributed fiber acoustic sensor data matrix to extract the instantaneous amplitude envelope and generate the envelope energy matrix. S12. Based on the envelope energy matrix, calculate the statistical mean of the response of all spatial channels at each time step, and generate a binary mask matrix according to the preset noise multiplier coefficient. S13. Perform morphological dilation on the binary mask matrix, and perform spatiotemporal scale normalization resampling on the dilated matrix to obtain a preprocessed matrix.

[0035] Furthermore, S11 specifically includes: For the original spacetime matrix Each spatial sequence Its analytic signal is constructed through Hilbert transform. And calculate the instantaneous amplitude envelope. The envelope energy matrix E is obtained; Where t is the time sampling point index, x is the spatial channel index, j is the imaginary unit, and H{ } represents the Hilbert transform.

[0036] Furthermore, S12 specifically includes: For each time t, calculate the statistical mean of the responses of all spatial channels. ; Define a binary mask matrix based on the preset noise multiplier K. :

[0037] in, Let x be the instantaneous amplitude envelope, and x be the spatial channel index. This represents the total number of spatial channels.

[0038] Furthermore, S13 specifically includes: Define a rectangular window with structuring element S, and perform morphological dilation on the binary mask matrix M to obtain the dilated matrix. ; The dilated matrix is ​​then subjected to bilinear interpolation. Resampling to target temporal resolution The preprocessing matrix is ​​obtained. ; in, The total number of spatial channels. Bilinear interpolation function.

[0039] To suppress the aforementioned noise and enhance trajectory continuity, this embodiment employs a three-stage preprocessing approach. First, the instantaneous amplitude envelope of each spatial channel sequence is extracted using Hilbert transform to generate an envelope energy matrix. This operation effectively smooths out high-frequency oscillation components in the original phase data while preserving the energy envelope characteristics caused by vehicle passage. Second, adaptive threshold dynamic denoising is used: the statistical mean of the responses of all spatial channels at each moment is calculated, and a noise multiplier K (K=2.0 in this embodiment) is set to filter out noise with amplitudes lower than a certain threshold. Points above the threshold are treated as background noise and set to zero, while points above the threshold are retained as candidate trajectory points, thus obtaining a binary mask matrix; Finally, a morphological dilation operation is performed on the binary mask matrix, using a rectangular structuring element S with a width of 5 spatial channels and a length of 20 time sampling points to connect the trajectory breakpoints caused by the instantaneous fading of the signal; then, the time resolution is resampled to the target size through bilinear interpolation. Balancing computational efficiency with trajectory resolution capability; such as Figure 3 As shown, after the above preprocessing, the signal-to-noise ratio of the vehicle trajectory is significantly improved, and the discontinuous line segments are effectively connected, providing a high-quality preprocessing matrix for subsequent rotational projection detection.

[0040] In this implementation S2, the preprocessed matrix is ​​rotated by multiple preset angles, and column integration is performed on each rotated matrix to obtain the corresponding projection vector, thereby generating a multi-angle projection matrix; including: Define rotation operator For the preprocessed matrix Perform a rotation operation to obtain the rotated matrix. ; For the rotated matrix Perform column integration to obtain the projection vector. :

[0041] Where t is the index of the time sampling point, x represents the total number of time sampling points, and x represents the spatial channel index. Traverse the preset angle range Given all discrete angles within the range, stack all projection vectors according to angles to construct a multi-angle projection matrix P.

[0042] Among them, the preset angle range Includes a positive angle range and a negative angle range ,in >0, <0, used to detect the trajectories of vehicles traveling in the forward and reverse directions, respectively.

[0043] This step transforms the problem of detecting oblique line trajectories in a two-dimensional spatiotemporal graph into the problem of peak detection in a one-dimensional projected curve through rotation and projection transformation; in specific implementation, the preprocessing matrix... Rotate it counterclockwise by θ degrees around its geometric center to obtain the rotated matrix. When the rotation angle θ is equal to the actual tilt angle of a vehicle trajectory, the trajectory becomes a vertical straight line in the rotated image; for example... Figure 4 As shown, when the rotation angle At that time, the originally diagonal vehicle trajectory becomes vertical after rotation. Then, column integration is performed on the rotated matrix, that is, the absolute values ​​of the amplitudes at all time points in each column (spatial channel) are summed to obtain the projection vector. The peak position in this vector corresponds to the spatial channel where the vertical line is located; Since there are traffic flows in both forward and reverse directions on actual roads, with their trajectory slopes being positive and negative respectively, this embodiment divides the road into two preset angle intervals: the positive angle interval... Used to detect vehicles traveling in the forward direction, negative angle range. To detect vehicles traveling in the opposite direction, a multi-angle projection matrix P is constructed by traversing all discrete angles with a fixed step size within each angle interval; for example... Figure 5 As shown, the gray line represents a single angle ( The projection curve under ( ), each peak corresponds to Figure 4 The vertical trajectory line is shown in the figure; while the blue and orange lines represent the maximum envelope curves obtained by taking the maximum value of all angle projections at each spatial location in the negative and positive angle intervals, respectively, providing basic data for subsequent signal-to-noise ratio enhancement.

[0044] In this implementation S3, the signal-to-noise ratio (SNR) enhancement process is performed on the multi-angle projection matrix to generate a spatial signal enhancement function; including: For each spatial location x, the maximum value is obtained by taking the maximum value in the angular dimension of the multi-angle projection matrix P. :

[0045] in, For the preset angle range, The rotation angle; For each spatial location x, the arithmetic mean is taken along the angular dimension to obtain the background noise response. :

[0046] in, The number of discrete angles within the angle interval; To suppress background noise, a spatial signal enhancement function is defined. The vehicle trajectory is transformed into an enhancement function. The local extreme point.

[0047] In the multi-angle projection matrix P, each spatial position x corresponds to a set of projection values ​​that vary with the rotation angle θ. For the spatial position where the actual vehicle trajectory is located, a significant peak will appear when its corresponding optimal angle participates in the projection. For non-trajectory positions or noise-dominated positions, the projection values ​​under each angle are relatively average. This step transforms the vehicle trajectory into a local extreme point in the spatial dimension through signal-to-noise ratio enhancement processing. Specifically, for each spatial location x, the maximum value is first taken among all the angular projection values ​​corresponding to that location to obtain the maximum spatial response. This value reflects the intensity of the strongest signal response at that location; secondly, the arithmetic mean of all angular projection values ​​is calculated to obtain the background noise response. This value represents the typical noise level at that location when there is no matching trajectory; for example... Figure 6 As shown, after subtracting the background noise curve from the maximum value curve, the spatial signal enhancement function s(x) is obtained. The orange curve in the figure corresponds to the enhancement result of the positive angle interval (forward traffic flow), and the blue curve corresponds to the enhancement result of the negative angle interval (reverse traffic flow). It can be seen that the trajectory position that was originally submerged in the background presents as a prominent local extreme peak on s(x), while the amplitude of the noise region is effectively suppressed, providing a high-quality signal for subsequent peak search.

[0048] In this implementation S4, based on the spatial signal enhancement function, a two-stage peak-finding strategy is used to determine the peak position of each candidate vehicle trajectory in the spatial dimension and its corresponding optimal rotation angle; wherein, the two-stage peak-finding strategy includes: Primary peak finding: in spatial signal enhancement functions Above, extract salience greater than the first preset threshold. The positions of all candidate peaks are denoted as set. The prominence is defined as the vertical distance between the peak and the higher of its two nearest troughs. Secondary screening: For each candidate peak position Determine the optimal angle that maximizes the projection value from the multi-angle projection matrix P. :

[0049] in, The angle range is preset, where θ is the rotation angle; The projection curve corresponding to the optimal angle Above, cut off Centered on, with radius neighborhood window ; If there is a prominence greater than the second preset threshold within the neighborhood window If a peak is identified, the candidate peak is confirmed to correspond to a real vehicle trajectory, and its precise peak position is recorded. and the corresponding optimal angle ; The first threshold at this location With the second threshold All are preset positive real numbers, and the second threshold Greater than the first threshold It is used to confirm the true trajectory with higher standards in the secondary screening.

[0050] After obtaining the spatial signal enhancement function s(x), S4 employs a two-stage peak-finding strategy to accurately locate the spatial position of each vehicle trajectory and its corresponding optimal rotation angle. The first stage is primary peak-finding: extracting all s(x) whose prominence is greater than a first preset threshold. The candidate peak position is defined by prominence, which is the vertical distance between the peak and the higher of its two nearest troughs. This indicator is more effective than the peak amplitude in distinguishing true peaks from random fluctuations. In this embodiment... It can be set to 0.15×max(s(x)) to recall all possible target locations, allowing for a certain number of false positives; The second level is fine-tuning: for each candidate peak position, the optimal angle that maximizes the projection value is backtracked from the multi-angle projection matrix P. Subsequently, the projection curve corresponding to this optimal angle Above, cut off A neighborhood window centered on a point with a radius of Δw (in this embodiment, Δw corresponds to a spatial range of 150 meters) is defined, and within this window, a more precise peak finding is performed again using a higher standard; for example... Figure 6 As indicated by the arrows, after secondary screening, only peaks that simultaneously meet the primary recall criteria and the local fine-grained verification criteria are confirmed as genuine vehicle trajectories, and their precise peak positions are recorded. and the corresponding optimal angle This two-stage mechanism effectively eliminates false positives that may be introduced during the initial peak search, while improving positioning accuracy through local window search.

[0051] In this implementation S5, based on the peak position and its corresponding optimal rotation angle, the rotated trajectory line is reconstructed to the original spatiotemporal domain through coordinate inverse mapping to obtain the vehicle trajectory equation; including: Define the inverse rotation operator Mapping the points in the rotated image back to the original spatiotemporal coordinate system, the mapping relationship is as follows:

[0052] in, For the optimal rotation angle, t and x are the time and space coordinates in the original spacetime coordinate system, respectively. and These are the spacetime coordinates of the center of rotation. For the time index in the rotated image, For precise peak location; Traverse all The values ​​are taken, the corresponding original spatiotemporal coordinates (t, x) are calculated, the discrete trajectory point set is obtained, and the vehicle trajectory straight line equation is fitted. ,in, b represents the vehicle speed, and b is the intercept.

[0053] In practice, the precise peak position and optimal rotation angle determined by S4 are used, combined with the preset rotation center ( , For each time index in the rotated image Apply the inverse rotation transformation to calculate its corresponding coordinates (t, x) in the original space-time domain; by traversing all This yields a series of discrete original spatiotemporal points, which can be directly fitted into a linear equation. ; like Figure 7 As shown, the reconstructed trajectory straight line (black dashed line) is superimposed on the original DAS signal for display. In this embodiment, the trajectory found by the vehicle trajectory reconstruction method based on rotation projection is highly consistent with the actual vehicle driving path, completing the complete detection process from the original DAS data matrix to the vehicle trajectory equation. The obtained trajectory data can be further used in the following specific scenarios: In highway operation and management, the instantaneous speed of vehicles can be directly calculated based on the slope of each trajectory, and the average vehicle speed and speed distribution of the road segment can be statistically analyzed; the number of trajectories at different spatial locations can be used to estimate traffic density and traffic flow in real time, assisting in dynamic lane control and ramp adjustment; at the same time, when abnormal sudden stops, lateral deviations, or sudden changes in speed are detected in the trajectory, abnormal event warnings for special road sections such as tunnels and bridges can be triggered in a timely manner, providing data support for traffic management departments.

[0054] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0055] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A vehicle trajectory reconstruction method based on rotational projection, characterized in that, Includes the following steps: S1. Obtain the raw distributed fiber acoustic sensor data matrix and preprocess it to obtain the preprocessed matrix; S2. Rotate the preprocessed matrix by multiple preset angles, perform column integration on each rotated matrix to obtain the corresponding projection vector, and generate a multi-angle projection matrix. S3. Perform signal-to-noise ratio enhancement processing on the multi-angle projection matrix to generate a spatial signal enhancement function; S4. Based on the spatial signal enhancement function, determine the peak position of each candidate vehicle trajectory in the spatial dimension and its corresponding optimal rotation angle through a two-level peak-finding strategy. S5. Based on the peak position and its corresponding optimal rotation angle, the rotated trajectory line is reconstructed to the original spatiotemporal domain through coordinate inverse mapping to obtain the vehicle trajectory equation.

2. The vehicle trajectory reconstruction method based on rotational projection according to claim 1, characterized in that, In step S1, the preprocessing includes: S11. Perform Hilbert transform on each spatial channel sequence of the original distributed fiber acoustic sensor data matrix to extract the instantaneous amplitude envelope and generate the envelope energy matrix. S12. Based on the envelope energy matrix, calculate the statistical mean of the response of all spatial channels at each time step, and generate a binary mask matrix according to the preset noise multiplier coefficient. S13. Perform morphological dilation on the binary mask matrix, and perform spatiotemporal scale normalization resampling on the dilated matrix to obtain a preprocessed matrix.

3. The vehicle trajectory reconstruction method based on rotational projection according to claim 2, characterized in that, S11 specifically includes: For the original spacetime matrix Each spatial sequence Its analytic signal is constructed through Hilbert transform. And calculate the instantaneous amplitude envelope. The envelope energy matrix E is obtained; Where t is the time sampling point index, x is the spatial channel index, j is the imaginary unit, and H{ } represents the Hilbert transform.

4. The vehicle trajectory reconstruction method based on rotational projection according to claim 2, characterized in that, S12 specifically includes: For each time t, calculate the statistical mean of the responses of all spatial channels. ; Define a binary mask matrix based on the preset noise multiplier K. : in, Let x be the instantaneous amplitude envelope, and x be the spatial channel index. This represents the total number of spatial channels.

5. The vehicle trajectory reconstruction method based on rotational projection according to claim 2, characterized in that, S13 specifically includes: Define a rectangular window with structuring element S, and perform morphological dilation on the binary mask matrix M to obtain the dilated matrix. ; The dilated matrix is ​​then subjected to bilinear interpolation. Resampling to target temporal resolution The preprocessing matrix is ​​obtained. ; in, The total number of spatial channels. Bilinear interpolation function.

6. The vehicle trajectory reconstruction method based on rotational projection according to claim 1, characterized in that, S2 includes: Define rotation operator For the preprocessed matrix Perform a rotation operation to obtain the rotated matrix. ; For the rotated matrix Perform column-direction integration to obtain the projection vector. : Where t is the index of the time sampling point, x represents the total number of time sampling points, and x represents the spatial channel index. Traverse the preset angle range Given all discrete angles within the range, stack all projection vectors according to angles to construct a multi-angle projection matrix P.

7. The vehicle trajectory reconstruction method based on rotational projection according to claim 6, characterized in that, The preset angle range Includes a positive angle range and a negative angle range ,in >0, <0, used to detect the trajectories of vehicles traveling in the forward and reverse directions, respectively.

8. The vehicle trajectory reconstruction method based on rotational projection according to claim 1, characterized in that, S3 includes: For each spatial location x, the maximum value is obtained by taking the maximum value in the angular dimension of the multi-angle projection matrix P. : in, For the preset angle range, The rotation angle; For each spatial location x, the arithmetic mean is taken along the angular dimension to obtain the background noise response. : in, The number of discrete angles within the angle interval; To suppress background noise, a spatial signal enhancement function is defined. The vehicle trajectory is transformed into an enhancement function. The local extreme point.

9. The vehicle trajectory reconstruction method based on rotational projection according to claim 1, characterized in that, In S4, the two-stage peak-finding strategy includes: Primary peak finding: in spatial signal enhancement functions Above, extract salience greater than the first preset threshold. The positions of all candidate peaks are denoted as set. The prominence is defined as the vertical distance between the peak and the higher of its two nearest troughs. Secondary screening: For each candidate peak position Determine the optimal angle that maximizes the projection value from the multi-angle projection matrix P. : in, The angle range is preset, where θ is the rotation angle; The projection curve corresponding to the optimal angle Above, cut off Centered on, with radius neighborhood window ; If there is a prominence greater than the second preset threshold within the neighborhood window If a peak is identified, the candidate peak is confirmed to correspond to a real vehicle trajectory, and its precise peak position is recorded. and the corresponding optimal angle .

10. The vehicle trajectory reconstruction method based on rotational projection according to claim 1, characterized in that, S5 includes: Define the inverse rotation operator Mapping the points in the rotated image back to the original spatiotemporal coordinate system, the mapping relationship is as follows: in, For the optimal rotation angle, t and x are the time and space coordinates in the original spacetime coordinate system, respectively. and These are the spacetime coordinates of the center of rotation. For the time index in the rotated image, For precise peak location; Traverse all The values ​​are taken, the corresponding original spatiotemporal coordinates (t, x) are calculated, the discrete trajectory point set is obtained, and the vehicle trajectory straight line equation is fitted. ,in, b represents the vehicle speed, and b is the intercept.

Images