WO2022068142A1 - Road surface terrain recognition method and apparatus based on suspension vibration signal, and storage medium - Google Patents

Abstract

Disclosed are a road surface terrain recognition method and apparatus based on a suspension vibration signal, and a storage medium. The method comprises: acquiring vibration signal data of a vehicle suspension that is currently collected by a suspension sensor during the traveling process of a vehicle; and when the sequence length of vibration signal data that is accumulated this time and is to be subjected to feature recognition reaches a preset data interception width of a sliding window, taking the vibration signal data that is accumulated this time and is to be subjected to feature recognition as an input quantity, and inputting same into a preset road surface terrain recognition model, so as to recognize a road surface terrain feature, thereby obtaining a recognition result of a road surface terrain. The present invention can effectively solve the problems in the prior art of a road surface terrain recognition result not being accurate enough, and too many computing resources being occupied.

Description

Road terrain recognition method, device and storage medium based on suspension vibration signal technical field

The present invention relates to the field of computer technology, and in particular, to a road terrain identification method, device and storage medium based on suspension vibration signals.

Background technique

At present, the road terrain can be recognized by the image recognition algorithm, specifically: when the vehicle is driving on the road, the vehicle camera captures the road image, and the road image is input into the road terrain recognition model for terrain recognition. However, the existing image-based road terrain recognition method is easily affected by the image shooting environment (such as dark light environment), resulting in an inaccurate result of the road terrain recognition, and the computational resources are occupied due to the large amount of algorithm calculation. excessive.

SUMMARY OF THE INVENTION

Embodiments of the present invention provide a road terrain identification method, device and storage medium based on suspension vibration signals, which can effectively solve the problems of the prior art that the results of road terrain identification are not accurate enough and occupy too many computing resources.

An embodiment of the present invention provides a road terrain identification method based on a vehicle suspension vibration signal, which includes:

Obtain the vibration signal data of the vehicle suspension currently collected by the suspension sensor during the driving process of the vehicle;

When the sequence length of the accumulated vibration signal data to be characterized by identification reaches the preset data interception width of the sliding window, the accumulated vibration signal data to be characterized by identification is used as the input quantity and input to the preset road terrain Identifying the road topography features in the recognition model, thereby obtaining the road topography recognition result; wherein the road surface topography recognition model is pre-trained according to the vibration signal data samples.

As an improvement of the above solution, after the vibration signal data of the vehicle suspension is acquired, and before the feature identification is performed on the vibration signal data, the method further includes:

Data processing is performed on the acquired vibration signal data to obtain vibration signal data after data processing; the data processing includes at least one of the following: data screening, data cleaning, and deletion of vacancies.

As an improvement of the above solution, there are at least two suspension sensors, which are distributed in different places of the vehicle suspension;

Then, after the vibration signal data of the vehicle suspension is acquired, and before the feature identification is performed on the vibration signal data, the acquisition method of the vibration signal data sample is as follows:

The acquired vibration signal data is stored in the form of a data matrix according to the order of time, so as to obtain a vibration signal data matrix for which data feature extraction is to be performed.

As an improvement of the above scheme, when the sequence length of the vibration signal data to be characterized by the feature recognition accumulated this time reaches the data interception width preset by the sliding window, the vibration signal data to be characterized by the feature recognition accumulated this time is used as the input input into the preset pavement terrain recognition model for identification of pavement terrain features, including:

When the sequence length of the data in the vibration signal data matrix to be feature identification accumulated this time reaches the data interception width preset by the sliding window, the vibration signal data matrix is subjected to PCA dimension reduction processing to obtain the vibration signal after dimension reduction. signal data;

The dimensionality-reduced vibration signal data is input into a preset road terrain recognition model as an input to identify road terrain features.

As an improvement of the above scheme, the road topography recognition model is a deep neural network model for identifying road topography, then the vibration signal data matrix is obtained after PCA dimensionality reduction processing is performed on the vibration signal data matrix. .

As an improvement of the above solution, the road terrain recognition model is an XGBoost model for recognizing road terrain, and the vibration signal data vector is obtained after PCA dimension reduction processing is performed on the vibration signal data matrix.

As an improvement of the above scheme, the data interception width w' of the sliding window corresponds to the current speed v of the vehicle, and its calculation formula is:

Among them, a is the preset window deviation value, and _wi is the initial data interception width.

As an improvement of the above scheme, the method for acquiring the vibration signal data sample includes:

Obtain the road surface image sequence collected by the vehicle camera when the vehicle is driving on the test road, and obtain the vibration signal time series data of the vehicle suspension collected by the suspension sensor when the vehicle is driving on the test road; the test road The road topography of the working condition includes the target road topography;

Performing data sliding interception on the vibration signal time series data with a sliding window of a preset data interception width, to obtain multi-segment vibration signal interception data;

Perform similarity calculation on the plurality of pieces of the vibration signal interception data and the preset vibration signal data template corresponding to the target road terrain to obtain the corresponding similarity value;

Judging whether the similarity value is greater than a preset similarity threshold;

If so, mark the image in the road surface image sequence with the vibration signal interception data corresponding to the similarity value at the same time stamp as the target road surface terrain image;

The vibration signal interception data confirmed by the user based on the target road topographic image is used as the vibration signal data sample.

As an improvement of the above solution, after the acquisition of the vibration signal time series data of the vehicle suspension collected by the suspension sensor when the vehicle is driving on the test road, after the acquisition by the suspension sensor when the vehicle is driving on the test road Before the time series data of the vibration signal of the vehicle suspension collected under the working conditions, the method further includes:

Normalize the time series data of the vibration signal.

As an improvement to the above solution, if the vehicle repeatedly drives under the test road conditions for many times, and the number of vibration signal time series data is multiple, then the vibration signal time series data is subjected to a sliding window with a preset data interception width. Sliding interception of data to obtain multi-segment vibration signal interception data, including:

Obtaining the median vibration signal time series data in the multiple different vibration signal time series data;

Perform data sliding interception on the vibration signal time series data whose median is a sliding window with a preset data interception width, so as to obtain multi-segment vibration signal interception data.

As an improvement of the above scheme, the data interception width of the sliding window corresponds to the speed of the vehicle, the faster the vehicle speed, the smaller the data interception width; the sliding step size of the sliding window is consistent with the sampling frequency of the suspension sensor.

As an improvement of the above scheme, the similarity calculation is performed on the multiple pieces of the vibration signal intercepted data corresponding to the target road terrain and is a preset vibration signal data template, and the corresponding similarity value is obtained, including:

Based on the DTW algorithm, the similarity between the curve segment formed by each segment of the vibration signal interception data and the curve template corresponding to the target road terrain and which is the vibration signal data is calculated, and the corresponding similarity value is obtained.

As an improvement of the above scheme, the method for acquiring the vibration signal data sample includes:

Obtain the vibration signal data stream generated by the vehicle suspension sensor under the current road terrain in the vehicle driving domain;

Performing data sliding interception on the vibration signal data stream with a sliding window of a preset data interception width, to intercept and obtain a corresponding vibration signal data sequence and store;

Perform data splicing one by one with the currently intercepted vibration signal data sequence and the previously intercepted vibration signal data sequence stored locally;

Perform data feature analysis on the vibration signal data sequence obtained by splicing, and determine the data feature pattern sequence in the vibration signal data sequence obtained by splicing according to the data feature analysis result; the data feature pattern sequence is used as the vibration signal data sample.

As an improvement of the above solution, before the data splicing of the currently intercepted vibration signal data sequence and the locally stored previously intercepted vibration signal data sequence is performed one by one, the sliding movement with the preset data interception width is performed. The window performs sliding interception of data on the vibration signal data stream, so as to intercept and obtain the corresponding vibration signal data sequence and store it, the method further includes:

Carrying out similarity calculation between the currently intercepted vibration signal data sequence and each vibration signal data sequence previously intercepted in the local storage to obtain a corresponding similarity comparison result;

Then, after performing data feature analysis on the vibration signal data sequence obtained by splicing, and determining the data feature pattern sequence in the vibration signal data sequence obtained by splicing according to the data feature analysis result, the method further includes:

Perform data fusion between the data characteristic pattern sequence and the previously intercepted vibration signal data sequence with the highest similarity result to obtain a fused data characteristic pattern sequence.

As an improvement of the above scheme, after obtaining the fused data feature pattern sequence, the method further includes:

uploading the fused data feature pattern sequence and the current road topography to the server, so that the server performs data analysis according to all acquired data feature pattern sequences corresponding to the same road topography to obtain a data analysis result; wherein , the data analysis results include road surface terrain recognition results.

As an improvement of the above-mentioned solution, the sliding window of the preset data interception width is used to perform data sliding interception on the vibration signal data stream, so as to intercept and obtain the corresponding vibration signal data sequence and store, including:

Perform data sliding interception on the vibration signal data stream with a preset sliding window with a data interception width corresponding to the current vehicle speed, so as to intercept and obtain a corresponding vibration signal data sequence;

The intercepted vibration signal data sequence is normalized and cached.

As an improvement of the above scheme, the data feature analysis is performed on the vibration signal data sequence obtained by splicing, and according to the data feature analysis result, the data feature pattern sequence in the vibration signal data sequence obtained by splicing is determined, including:

The vibration signal data sequence obtained by splicing is subjected to data characteristic analysis by the Matrix Profile algorithm, and according to the data characteristic analysis result, the data characteristic pattern sequence in the vibration signal data sequence obtained by the splicing is determined.

As an improvement of the above solution, the data interception width dynamic_wsize of the sliding window is:

Or, the data interception width dynamic_wsize of the sliding window is:

Wherein, max_speed is the preset maximum vehicle speed threshold, base_wsize is the preset data interception width reference value, and smooth(vspeed) is the current average vehicle speed.

As an improvement of the above scheme, if the sliding window is to jointly extract multiple associated vibration signals on the data stream, the data interception width dynamic_wsize of the sliding window is:

dynamic_wsize _i =dynamic_wsize _{i -1} +Δi

Wherein, t _sample is a preset time sampling length, and σ(s ₁ , s ₂ , s ₃ , s _i ) is the standard deviation at time ti of multiple scalar data streams.

Another embodiment of the present invention correspondingly provides a road terrain recognition device based on a vehicle suspension vibration signal, which includes:

The data acquisition module is used to acquire the vibration signal data of the vehicle suspension currently collected by the suspension sensor during the driving process of the vehicle;

The terrain identification module is used for inputting the accumulated vibration signal data to be identified as input when the sequence length of the accumulated vibration signal data to be identified by the feature reaches the preset data interception width of the sliding window. The road terrain features are identified in the preset road terrain recognition model, so as to obtain the recognition result of the road terrain; wherein, the road terrain recognition model is pre-trained according to the vibration signal data samples.

As an improvement of the above scheme, the device also includes:

The acquisition module is used to acquire the road surface image sequence collected by the vehicle camera when the vehicle is driving on the test road, and obtain the vibration signal time series data of the vehicle suspension collected by the suspension sensor when the vehicle is driving on the test road ; the pavement terrain of the test road condition includes the target pavement terrain;

a data interception module, configured to perform data sliding interception on the vibration signal time series data with a sliding window with a preset data interception width to obtain multiple segments of vibration signal interception data;

Similarity calculation module, for carrying out similarity calculation with the vibration signal data template corresponding to the multi-section described vibration signal interception data and target road topography, and obtains the corresponding similarity value;

a judgment module, configured to judge whether the similarity value is greater than a preset similarity threshold;

a data labeling module, configured to, if so, mark an image with the same time stamp of the vibration signal interception data corresponding to the similarity value in the road surface image sequence as a target road surface terrain image;

The data confirmation module is configured to use the vibration signal interception data confirmed by the user based on the target road terrain image as the vibration signal data sample.

As an improvement of the above scheme, the device also includes:

The data stream acquisition module is used to acquire the vibration signal data stream generated by the vehicle suspension sensor under the current road terrain in the vehicle driving domain;

a data sequence interception module, used for performing sliding interception of data on the vibration signal data stream with a sliding window of a preset data interception width, so as to intercept and obtain a corresponding vibration signal data sequence and store;

A data splicing module, for performing data splicing one by one with the currently intercepted vibration signal data sequence and the previously intercepted vibration signal data sequence stored locally;

The data feature analysis module is used to perform data feature analysis on the vibration signal data sequence obtained by splicing, and according to the data feature analysis result, determine the data feature pattern sequence in the vibration signal data sequence obtained by splicing; the data feature pattern sequence used as the vibration signal data sample.

Another embodiment of the present invention provides a road terrain recognition device based on a vehicle suspension vibration signal, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, the When the processor executes the computer program, the method for recognizing the road surface terrain based on the vibration signal of the vehicle suspension according to the above embodiments of the invention is implemented.

Another embodiment of the present invention provides a storage medium, where the computer-readable storage medium includes a stored computer program, wherein when the computer program runs, a device on which the computer-readable storage medium is located is controlled to execute the foregoing embodiments of the invention The road terrain recognition method based on the vibration signal of the vehicle suspension.

Compared with the prior art, one of the foregoing inventive embodiments has the following advantages:

By acquiring the vibration signal data of the vehicle suspension currently collected by the suspension sensor during the vehicle driving process; when the sequence length of the vibration signal data to be characterized identification accumulated this time reaches the data interception width preset by the sliding window, the The vibration signal data to be characterized this time accumulated this time is input into a preset road topography recognition model as an input to recognize road topography features, thereby obtaining a road topography recognition result. It can be seen that the vibration of the vehicle suspension can accurately reflect the terrain condition of the driving road of the vehicle, and the embodiment of the present invention realizes the real-time perception of the vibration of the vehicle suspension and analyzes the vibration signal of the vehicle suspension. Therefore, the embodiment of the present invention can improve the accuracy of the recognition result of the road surface terrain, and compared with the image-based road terrain recognition method, the algorithm calculation amount of the embodiment of the present invention is greatly reduced, so as to avoid occupation more computing resources. Of course, it is not necessary for any product embodying the present invention to achieve all of the above-described advantages simultaneously.

Description of drawings

FIG. 1 is a schematic flowchart of a road terrain recognition method based on a vehicle suspension vibration signal provided by an embodiment of the present invention;

2 is an application schematic diagram of a road terrain identification method based on vehicle suspension vibration signals provided by an embodiment of the present invention;

Fig. 3 shows that sliding window carries out data sequence interception to vibration signal data stream with preset data interception width;

Fig. 4 shows the process that the Matrix Profile algorithm performs feature analysis on the same data sequence as two dimensions of the matrix;

5 is a signal diagram containing a plurality of associated vibration signals in an embodiment of the present invention;

6 is a schematic structural diagram of a road terrain recognition device based on a vehicle suspension vibration signal provided by an embodiment of the present invention;

7 is a schematic structural diagram of a road terrain recognition device based on a vehicle suspension vibration signal provided by an embodiment of the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

Referring to FIG. 1 , it is a schematic flowchart of a road terrain identification method based on a vehicle suspension vibration signal provided by an embodiment of the present invention. The method is executed by a device of a road terrain recognition method based on a vehicle suspension vibration signal, such as a main control device of the vehicle (eg, a body domain controller). The method includes:

S10, acquiring vibration signal data of the vehicle suspension currently collected by the suspension sensor during the running of the vehicle;

Wherein, during the driving process of the vehicle, the driving road surface of the vehicle includes various road terrains, such as pothole terrain, flat terrain, roadblock terrain, and the like. The inventors of the present invention found that when the vehicle travels on different road terrains, the amplitude and frequency of the vibration of the suspension of the vehicle are different. Therefore, road topography recognition can be performed based on feature extraction and analysis of the vibration signal of the vehicle suspension.

S11, when the sequence length of the vibration signal data to be characterized by the feature recognition accumulated this time reaches the preset data interception width of the sliding window, the vibration signal data to be characterized by the feature recognition accumulated this time is input as an input to the preset data interception width. In the road topography recognition model, the road topography features are recognized, so as to obtain the road topography recognition result; wherein, the road surface topography recognition model is pre-trained according to the vibration signal data samples.

Wherein, the data interception width of the sliding window is preset and can be set as: the setting of the data interception width enables the vibration signal data of the corresponding sequence length to be used to identify the road topography. That is, the setting of the data interception width enables the accumulated vibration signal data to be characterized this time to be used for road terrain features. Therefore, the data interception width cannot be too narrow, which will make the accumulated vibration signal data for feature recognition to be too small to be used for road terrain recognition. In addition, the data interception width does not need to be set to be too wide. Too wide will cause too much vibration signal data to be used for feature identification accumulated this time and increase the difficulty of road terrain identification.

In addition, the road terrain recognition model is pre-trained according to the vibration signal data samples, and the training method may be the existing model training method, which will not be repeated here.

To sum up, since the vibration of the vehicle suspension can accurately reflect the terrain condition of the driving road of the vehicle, the embodiment of the present invention is to perceive the vibration of the vehicle suspension in real time, and analyze the vibration signal of the vehicle suspension to obtain the results. Therefore, the embodiment of the present invention can improve the accuracy of the recognition result of the road surface terrain, and compared with the image-based road terrain recognition method, the algorithm calculation amount of the embodiment of the present invention is greatly reduced, so as to avoid occupation more computing resources.

In the embodiment of the present invention, further, after the vibration signal data of the vehicle suspension is acquired, and before the feature identification is performed on the vibration signal data, the method further includes:

Data processing is performed on the acquired vibration signal data to obtain vibration signal data after data processing; the data processing includes at least one of the following: data screening, data cleaning, and deletion of vacancies.

In the embodiment of the present invention, by performing data processing on the acquired vibration signal data, abnormal vibration signal data can be eliminated, thereby facilitating subsequent data analysis and ultimately improving the identification result of road topography.

Exemplarily, there are at least two suspension sensors, which are distributed in different places on the vehicle suspension. For example, there are four suspension sensors, which are respectively distributed on the left side of the suspension of the front wheel of the vehicle (marked as FrntLelv1), the right side of the suspension of the front wheel of the vehicle (marked as FrntRilv1), and the suspension of the rear wheel of the vehicle. the left side of the vehicle (marked ReLelv1 ) and the right side of the suspension of the rear wheel of the vehicle (marked ReRilv1 ).

Further, after acquiring the vibration signal data of the vehicle suspension, before performing feature identification on the vibration signal data, the method further includes:

The acquired vibration signal data is stored in the form of a data matrix according to the order of time, so as to obtain a vibration signal data matrix for which data feature extraction is to be performed.

Specifically, when the sequence length of the vibration signal data to be characterized by the feature recognition accumulated this time reaches the data interception width preset by the sliding window, the vibration signal data to be characterized by the feature recognition accumulated this time is input as an input to the Recognition of pavement terrain features in the preset pavement terrain recognition model, including:

When the sequence length of the data in the vibration signal data matrix to be feature identification accumulated this time reaches the data interception width preset by the sliding window, the vibration signal data matrix is subjected to PCA dimension reduction processing to obtain the vibration signal after dimension reduction. signal data;

The dimensionality-reduced vibration signal data is input into a preset road terrain recognition model as an input to identify road terrain features.

In this embodiment, by performing PCA dimensionality reduction processing on the vibration signal data matrix, high-dimensional variables that may have correlation can be synthesized into linearly independent low-dimensional variables, which is beneficial to the subsequent vibration signal data matrix. Feature extraction and processing Recognition analysis, and then improve the accuracy of road terrain recognition.

As a specific example, referring to FIG. 2 , the road terrain recognition model is a deep neural network model for recognizing road terrain, then the vibration signal data matrix is subjected to PCA dimensionality reduction processing and obtained after PCA dimensionality reduction The vibration signal data matrix.

Among them, the deep neural network model is pre-trained by a large number of vibration signal data samples, and the specific training method can refer to the existing deep neural network model training method.

As another specific example, referring to FIG. 2 , the road terrain recognition model is an XGBoost model used to identify road terrain, then the vibration signal data matrix is subjected to PCA dimensionality reduction processing to obtain the PCA dimensionality reduction vibration Signal data vector.

Specifically, the objective function formula of the XGBoost model is:

Among them, ∑ _k Ω(f _t ) is expressed as the summation of all k-particle complexities in XGBoost, which is used as the regularization term of the objective function to prevent the model from overfitting. y _i ′ represents the residual value of the previous binary tree, y _i represents the predicted value, l is a constant, and i is the tree depth. It is understandable that the objective function formula of the XGBoost model is also pre-trained through a large number of vibration signal data samples, and the specific training method can refer to the existing XGBoost model training method.

As an improvement of the above scheme, the data interception width w' of the sliding window corresponds to the current speed v of the vehicle, and its calculation formula is:

Among them, a is the preset window deviation value, _wi is the initial data interception width, and n is the number of samples.

Specifically, the data interception width of the sliding window is adjusted according to the type of road terrain to be identified. Before determining the window function, it is necessary to collect the data interception width of the window that passes through the terrain at different speeds; the data interception width of the sliding window is a variation, which can be determined according to Adjust the vehicle speed v. In addition, the sliding step S of the sliding window can be selected according to the sampling frequency of the suspension sensor. For example, the sliding step S can be set to be consistent with the sampling frequency of the suspension sensor, both being 10ms.

Exemplarily, the method for acquiring vibration signal data samples for training the road terrain recognition model includes steps S20 to S25:

S20, acquiring the road surface image sequence collected by the vehicle camera when the vehicle is driving on the test road, and acquiring the vibration signal time series data of the vehicle suspension collected by the suspension sensor when the vehicle is driving on the test road; the The pavement terrain of the test road condition includes the target pavement terrain.

Among them, when the vehicle is running under the test road conditions, the suspension sensor will collect the vibration of the vehicle suspension in real time and generate the vibration signal, and according to the collected vibration signal, the time series data of the vibration signal can be generated according to the order of collection time. Among them, the road topography of the test road conditions can be various, such as pothole road topography, road block road topography, and stone road section road topography. And the pavement terrain of the test road condition includes the target pavement terrain.

S21, performing sliding data interception on the vibration signal time series data by using a sliding window with a preset data interception width to obtain multiple segments of vibration signal interception data.

S22, performing similarity calculation on a plurality of pieces of the vibration signal interception data and a preset vibration signal data template corresponding to the target road terrain to obtain a corresponding similarity value;

S23: Determine whether the similarity value is greater than a preset similarity threshold.

Among them, by setting an appropriate similarity threshold, the vibration signal interception data of non-target terrain can be filtered out, and only the vibration signal interception data corresponding to the target terrain can be retained.

S24, if yes, mark the image with the same time stamp of the vibration signal interception data corresponding to the similarity value in the road surface image sequence as the target road surface topographic image;

S25: Use the vibration signal interception data confirmed by the user based on the target road topography image as the vibration signal data sample.

After marking the target road terrain images, the user can verify these target road terrain images to confirm whether the road surface captured by the target road terrain images is the target road terrain. The vibration signal interception data corresponding to the target road topography image whose result is confirmed as yes by the user is taken as the vibration signal data sample, which is used for the training of the road surface topography recognition model.

The embodiment of the present invention can automatically mark the road surface image as the target road terrain by analyzing the vibration signal of the vehicle suspension, and use the corresponding vibration signal interception data as the vibration according to the user's confirmation result based on the target road terrain image. Signal data samples, which can improve the efficiency and accuracy of data sample acquisition.

In the above-mentioned embodiment of the invention, further, after the acquisition of the vibration signal time series data of the vehicle suspension collected by the suspension sensor when the vehicle is running on the test road, after the acquisition of the vibration signal of the vehicle suspension by the suspension sensor Before driving the vibration signal time series data of the vehicle suspension collected under test road conditions, the method further includes:

Normalize the time series data of the vibration signal.

Wherein, by normalizing the time series data of the vibration signal, it is more beneficial to the subsequent data analysis, thereby making the data analysis more accurate.

Specifically, the normalization method is as follows

Wherein, X is the current vibration signal data to be analyzed, X _min is the minimum value in the vibration signal time series data, and X _max is the maximum value in the vibration signal time series data.

In the embodiment of the present invention, exemplarily, if the vehicle repeatedly drives under the test road conditions for many times, and the number of vibration signal time series data is multiple, the sliding window with the preset data interception width will Sliding interception of the vibration signal time series data to obtain multi-segment vibration signal interception data, including:

Acquiring vibration signal time series data that is median in a plurality of different described vibration signal time series data;

Perform data sliding interception on the vibration signal time series data whose median is a sliding window with a preset data interception width, so as to obtain multi-segment vibration signal interception data.

In the embodiment of the present invention, a plurality of vibration signal time series data are obtained through repeated experiments, and the median vibration signal time series data is used for analysis, which can make the data analysis more accurate.

In the above embodiment, specifically, the data interception width of the sliding window corresponds to the speed of the vehicle, and the faster the vehicle speed, the smaller the data interception width; the sliding step size of the sliding window corresponds to the sampling of the suspension sensor. the same frequency.

Among them, the faster the vehicle speed is, the smaller the data interception width can be, so as to avoid the sampled vibration signal from containing too many road terrain conditions, ideally, it can only contain one road terrain condition. When the speed of the vehicle is slower, the data interception width can be set larger, so that the road terrain can be sampled more comprehensively.

In addition, as an example, the sampling frequency of the suspension sensor is 10ms, and then the sliding step size of the sliding window is 10ms each time.

In the above-mentioned embodiment, specifically, the similarity calculation is performed on the plurality of pieces of the vibration signal intercepted data and the preset vibration signal data template corresponding to the target road terrain, and the corresponding similarity value is obtained, including:

Based on the DTW algorithm, the similarity between the curve segment formed by each segment of the vibration signal interception data and the curve template corresponding to the target road terrain and which is the vibration signal data is calculated, and the corresponding similarity value is obtained. Among them, the similarity calculation formula of DTW algorithm is:

Exemplarily, after the vibration signal data of the vehicle suspension is acquired, in order to automatically perform feature extraction on the vibration signal data of the vehicle suspension, so as to perform feature analysis on the extracted feature data of the vibration signal, so as to automatically perform feature extraction on the vibration signal data of the vehicle suspension. Identifying new vibration signal data features corresponding to each road topography, and using the new vibration signal data features as the vibration signal data samples for training the road surface topography recognition model, so that the road surface topography can be recognized The model can identify the road topography more accurately, and can identify the road topography under more vehicle operating conditions. In order to achieve the above object, an embodiment of the present invention also provides a method for automatic feature extraction of vibration signal data of a vehicle suspension sensor, which includes:

S30, acquiring the vibration signal data stream generated by the vehicle suspension sensor under the current road terrain in the vehicle driving domain;

S31, referring to FIG. 3, slidingly intercepting the vibration signal data stream with the sliding window of the preset data interception width Δt, to intercept and obtain the corresponding vibration signal data sequence and store;

S32, carry out data splicing one by one with the previously intercepted vibration signal data sequence of the currently intercepted vibration signal data sequence and the locally stored vibration signal data sequence;

S33, perform data characteristic analysis on the vibration signal data sequence obtained by splicing, and determine the data characteristic pattern sequence in the vibration signal data sequence obtained by splicing according to the data characteristic analysis result; the data characteristic pattern sequence is used as the Vibration signal data sample.

In the embodiment of the present invention, the embodiment of the present invention performs data splicing one by one between the vibration signal data sequence obtained by the current sliding interception of the vibration signal data stream and the previously intercepted vibration signal data sequence; The vibration signal data sequence is analyzed by data feature, and according to the data feature analysis result, the data feature pattern sequence in the vibration signal data sequence obtained by splicing is determined, so that the data feature pattern sequence is new and can be used to represent the current road terrain. vibration data characteristics. It can be seen that the embodiment of the present invention can automatically identify the new vibration signal data features corresponding to each road terrain, and use the new vibration signal data features as the vibration signal data samples for training the road terrain recognition Therefore, the road terrain recognition model can identify the road terrain more accurately, and can identify more road terrains under different vehicle operating conditions.

As an improvement of the above solution, before the data splicing of the currently intercepted vibration signal data sequence and the locally stored previously intercepted vibration signal data sequence is performed one by one, the sliding movement with the preset data interception width is performed. The window performs sliding interception of data on the vibration signal data stream, so as to intercept and obtain the corresponding vibration signal data sequence and store it, the method further includes:

Carrying out similarity calculation between the currently intercepted vibration signal data sequence and each vibration signal data sequence previously intercepted in the local storage to obtain a corresponding similarity comparison result;

Then, after performing data feature analysis on the vibration signal data sequence obtained by splicing, and determining the data feature pattern sequence in the vibration signal data sequence obtained by splicing according to the data feature analysis result, the method further includes:

Perform data fusion between the data characteristic pattern sequence and the previously intercepted vibration signal data sequence with the highest similarity result to obtain a fused data characteristic pattern sequence.

Exemplarily, the calculation formula of the similarity may be the correlation coefficient calculation formula:

Alternatively, the formula can be calculated for the similarity distance:

Among them, x and y in the formula are the sample sequences of the two groups of data patterns to be calculated, μ _x is the sample mean of x, σ _x is the standard deviation, E(x) is the expectation of the x sample, and m is the sample length.

Exemplarily, the data fusion algorithm is:

Among them, c _old is the preset original data pattern correction coefficient, p _old is the original data pattern matching degree, c _new is the preset new data pattern correction coefficient, p _new is the original data pattern matching degree, motif_old(t _i ) is the previously intercepted vibration signal data sequence with the highest similarity, and motifs_new(t _i ) is the current data feature pattern sequence.

In this embodiment, specifically, data fusion is to merge the characteristic pattern sequence in the newly discovered data stream and the vibration signal sequence with the greatest matching degree in the cache according to the cache index i of the sequence. The purpose of data fusion is that the new data feature pattern sequence identified by the algorithm is random, and after a large number of data feature pattern sequences are fused and accumulated, a stable feature pattern can be generated to avoid accidental errors in the identified data feature pattern sequence, resulting in The final data feature extraction result is wrong.

As an improvement of the above scheme, after obtaining the fused data feature pattern sequence, the method further includes:

uploading the fused data feature pattern sequence and the current road topography to the server, so that the server performs data analysis according to all acquired data feature pattern sequences corresponding to the same road topography, and obtains a data analysis result; wherein , the data analysis results include road surface terrain recognition results.

As an improvement of the above-mentioned solution, the sliding window of the preset data interception width is used to perform data sliding interception on the vibration signal data stream, so as to intercept and obtain the corresponding vibration signal data sequence and store, including:

Perform data sliding interception on the vibration signal data stream with a preset sliding window with a data interception width corresponding to the current vehicle speed, so as to intercept and obtain a corresponding vibration signal data sequence;

The intercepted vibration signal data sequence is normalized and cached.

In this embodiment, normalization processing is performed on the intercepted vibration signal data sequence, which facilitates subsequent feature analysis of the vibration signal data.

Exemplarily, the normalization processing algorithm is:

Among them, □t is the data interception width of the sliding window, x _Δt is the intercepted vibration signal data sequence, σ is the standard deviation,

is the mean value of the vibration signal data series.

As the improvement of the above scheme, the described vibration signal data sequence obtained by splicing is carried out data feature analysis, and according to the data feature analysis result, determine the data feature pattern sequence in the described vibration signal data sequence obtained by splicing, including:

The vibration signal data sequence obtained by splicing is subjected to data characteristic analysis by the Matrix Profile algorithm, and according to the data characteristic analysis result, the data characteristic pattern sequence in the vibration signal data sequence obtained by the splicing is determined.

Specifically, the Matrix Profile algorithm is based on the principle as shown in Figure 4. Among them, a segment is taken from the time series data and slid along the rest of the time series, and how much it is similar to the overlap of the time series segment at each new position is calculated. More specifically, the Euclidean distance between a subsequence and each time series segment of the same length can be calculated, thus establishing a so-called distance pattern for time series segments. If the subsequence repeats itself in the data, there will be at least one exact match and the minimum Euclidean distance will be zero. Whether the time series data contains similar patterns can be obtained by calculating the minimum value of the matrix profile in the time series data window. The same time series data are used as the two dimensions of the matrix, and the similarity of each corresponding point in the matrix is calculated in a one-to-one correspondence. Distance di, dj, and then find the minimum value in the column direction to get Pi, in the sequence from P1 to Pn-m+1, the index corresponding to the minimum value is the starting point MPI (MatrixProfile Index) that generates similar data patterns, and take MPI as the starting point to take a specific The window length data is the data characteristic pattern sequence in the data sequence obtained by splicing.

Specifically, the calculation formula of Matrix Profile is:

The formula for finding the minimum value of similarity distance by column: MP(t) _i =min(d _{(i, j)} );

The formula for finding the minimum value of the calculated Pi sequence: MPI=min(MP(Δt)).

As an improvement of the above solution, the data interception width dynamic_wsize of the sliding window is:

Or, the data interception width dynamic_wsize of the sliding window is:

Wherein, max_speed is the preset maximum vehicle speed threshold, base_wsize is the preset data interception width reference value, and smooth(vspeed) is the current average vehicle speed.

As an improvement of the above scheme, if the sliding window is to jointly extract multiple associated vibration signals on the data stream, the data interception width dynamic_wsize of the sliding window is:

dynamic_wsize _i =dynamic_wsize _{i -1} +Δi

Wherein, t _sample is a preset time sampling length, and σ(s ₁ , s ₂ , s ₃ , s _i ) is the standard deviation at time ti of multiple scalar data streams.

Wherein, the joint extraction of multiple associated vibration signals means that the characteristic patterns shown in the data appear in multiple vibration signals (these vibration signals are simultaneously collected by different suspension sensors). As shown in Figure 5, the data patterns jointly exhibited by the changes of the four correlated vibration signals over time.

As an improvement of the above solution, the sliding step size of the sliding window corresponds to the sampling frequency of the vehicle suspension sensor.

Referring to FIG. 6, it is a schematic structural diagram of a road terrain recognition device based on a vehicle suspension vibration signal provided by an embodiment of the present invention, and the device includes:

The data acquisition module 10 is used for acquiring the vibration signal data of the vehicle suspension currently collected by the suspension sensor during the driving process of the vehicle;

The terrain identification module 11 is used to use the accumulated vibration signal data to be identified this time as the input amount when the sequence length of the vibration signal data to be characterized by the current accumulation reaches the preset data interception width of the sliding window Input into a preset road topography recognition model to identify road topography features, thereby obtaining a road topography recognition result; wherein the road surface topography recognition model is pre-trained according to vibration signal data samples.

To sum up, since the vibration of the vehicle suspension can accurately reflect the terrain condition of the driving road of the vehicle, the embodiment of the present invention is to perceive the vibration of the vehicle suspension in real time, and analyze the vibration signal of the vehicle suspension to obtain the results. Therefore, the embodiment of the present invention can improve the accuracy of the recognition result of the road surface terrain, and compared with the image-based road terrain recognition method, the algorithm calculation amount of the embodiment of the present invention is greatly reduced, so as to avoid occupation more computing resources.

As an improvement of the above solution, there are at least two suspension sensors, which are distributed in different places of the vehicle suspension; then the device further includes:

The data storage module is used for storing the acquired vibration signal data in the form of a data matrix according to the order of time, so as to obtain a vibration signal data matrix for data feature extraction.

As an improvement of the above scheme, the terrain recognition module includes:

A data dimensionality reduction unit, used to perform PCA dimensionality reduction processing on the vibration signal data matrix when the sequence length of the data in the vibration signal data matrix to be characterized identification accumulated this time reaches the data interception width preset by the sliding window , obtain the vibration signal data after dimension reduction;

The terrain recognition unit is used for inputting the vibration signal data after dimensionality reduction as an input into a preset road terrain recognition model to identify road terrain features.

As an improvement of the above scheme, the device also includes:

Data processing is performed on the acquired vibration signal data to obtain vibration signal data after data processing; the data processing includes at least one of the following: data screening, data cleaning, and deletion of vacancies.

As an improvement of the above scheme, the road topography recognition model is a deep neural network model for identifying road topography, then the vibration signal data matrix is obtained after PCA dimensionality reduction processing is performed on the vibration signal data matrix. .

As an improvement of the above solution, the road terrain recognition model is an XGBoost model for recognizing road terrain, and the vibration signal data vector is obtained after PCA dimension reduction processing is performed on the vibration signal data matrix.

As an improvement of the above solution, the data interception width w' of the sliding window corresponds to the current speed v of the vehicle, and its calculation formula is:

Among them, a is the preset window deviation value, and _wi is the initial data interception width.

Exemplarily, an embodiment of the present invention further provides a device for marking road terrain data based on a vehicle suspension vibration signal, the device comprising:

The acquisition module is used to acquire the road surface image sequence collected by the vehicle camera when the vehicle is driving on the test road, and obtain the vibration signal time series data of the vehicle suspension collected by the suspension sensor when the vehicle is driving on the test road ; the pavement terrain of the test road condition includes the target pavement terrain;

a data interception module, configured to perform data sliding interception on the vibration signal time series data with a sliding window with a preset data interception width to obtain multiple segments of vibration signal interception data;

a similarity calculation module, configured to perform similarity calculation on a plurality of pieces of the vibration signal intercepted data and a preset vibration signal data template corresponding to the target road terrain to obtain a corresponding similarity value;

a judgment module, configured to judge whether the similarity value is greater than a preset similarity threshold;

a data labeling module, configured to, if so, mark an image with the same time stamp of the vibration signal interception data corresponding to the similarity value in the road surface image sequence as a target road surface terrain image;

The data confirmation module is configured to use the vibration signal interception data confirmed by the user based on the target road terrain image as the vibration signal data sample.

The embodiment of the present invention can automatically mark the road surface image as the target road terrain by analyzing the vibration signal of the vehicle suspension, and use the corresponding vibration signal interception data as the vibration according to the user's confirmation result based on the target road terrain image. Signal data samples, which can improve the efficiency and accuracy of data sample acquisition.

As an improvement of the above scheme, the device also includes:

The normalization module is used for normalizing the time series data of the vibration signal.

As an improvement of the above scheme, the similarity calculation module is specifically used for:

Based on the DTW algorithm, the similarity between the curve segment formed by each segment of the vibration signal interception data and the curve template corresponding to the target road terrain and which is the vibration signal data is calculated, and the corresponding similarity value is obtained.

As an improvement of the above scheme, if the vehicle repeatedly drives under the test road conditions for many times, and the number of vibration signal time series data is multiple, the data interception module is specifically used for:

Obtaining the median vibration signal time series data in the multiple different vibration signal time series data;

Perform data sliding interception on the vibration signal time series data whose median is a sliding window with a preset data interception width, so as to obtain multi-segment vibration signal interception data.

As an improvement of the above solution, the data interception width of the sliding window corresponds to the speed of the vehicle. The faster the vehicle speed is, the smaller the data interception width is; the sliding step size of the sliding window is consistent with the sampling frequency of the suspension sensor.

Exemplarily, after the vibration signal data of the vehicle suspension is acquired, in order to automatically perform feature extraction on the vibration signal data of the vehicle suspension, so as to perform feature analysis on the extracted feature data of the vibration signal, so as to automatically perform feature extraction on the vibration signal data of the vehicle suspension. Identifying new vibration signal data features corresponding to each road topography, and using the new vibration signal data features as the vibration signal data samples for training the road surface topography recognition model, so that the road surface topography can be recognized The model can identify the road topography more accurately, and can identify the road topography under more vehicle operating conditions. In order to achieve the above object, an embodiment of the present invention also provides an automatic feature extraction device for vibration signal data of a vehicle suspension sensor, which includes:

The data stream acquisition module is used to acquire the vibration signal data stream generated by the vehicle suspension sensor under the current road terrain in the vehicle driving domain;

a data sequence interception module, used for performing sliding interception of data on the vibration signal data stream with a sliding window of a preset data interception width, so as to intercept and obtain a corresponding vibration signal data sequence and store;

A data splicing module, for performing data splicing one by one with the currently intercepted vibration signal data sequence and the previously intercepted vibration signal data sequence stored locally;

The data feature analysis module is used to perform data feature analysis on the vibration signal data sequence obtained by splicing, and according to the data feature analysis result, determine the data feature pattern sequence in the vibration signal data sequence obtained by splicing; the data feature pattern sequence used as the vibration signal data sample.

In the embodiment of the present invention, the embodiment of the present invention performs data splicing one by one between the vibration signal data sequence obtained by the current sliding interception of the vibration signal data stream and the previously intercepted vibration signal data sequence; The vibration signal data sequence is analyzed by data feature, and according to the data feature analysis result, the data feature pattern sequence in the vibration signal data sequence obtained by splicing is determined, so that the data feature pattern sequence is new and can be used to represent the current road terrain. vibration data characteristics. It can be seen that the embodiment of the present invention can automatically identify the new vibration signal data features corresponding to each road terrain, and use the new vibration signal data features as the vibration signal data samples for training the road terrain recognition Therefore, the road terrain recognition model can identify the road terrain more accurately, and can identify more road terrains under different vehicle operating conditions.

As an improvement of the above scheme, the device also includes:

A similarity calculation module, used to calculate the similarity between the currently intercepted vibration signal data sequence and each vibration signal data sequence previously intercepted in the local storage to obtain a corresponding similarity comparison result;

Then, the device also includes:

The data fusion module is used for data fusion between the data characteristic pattern sequence and the previously intercepted vibration signal data sequence with the highest similarity result to obtain the fused data characteristic pattern sequence.

Exemplarily, the calculation formula of the similarity may be the correlation coefficient calculation formula:

Alternatively, the formula can be calculated for the similarity distance:

Among them, x and y in the formula are the sample sequences of the two groups of data patterns to be calculated, μ _x is the sample mean of x, σ _x is the standard deviation, E(x) is the expectation of the x sample, and m is the sample length.

Exemplarily, the data fusion algorithm is:

Among them, c _old is the preset original data pattern correction coefficient, p _old is the original data pattern matching degree, c _new is the preset new data pattern correction coefficient, p _new is the original data pattern matching degree, motif_old(t _i ) is the previously intercepted vibration signal data sequence with the highest similarity, and motifs_new(t _i ) is the current data feature pattern sequence.

In this embodiment, specifically, data fusion is to merge the characteristic pattern sequence in the newly discovered data stream and the vibration signal sequence with the greatest matching degree in the cache according to the cache index i of the sequence. The purpose of data fusion is that the new data feature pattern sequence identified by the algorithm is random, and after a large number of data feature pattern sequences are fused and accumulated, a stable feature pattern can be generated to avoid accidental errors in the identified data feature pattern sequence, resulting in The final data feature extraction result is wrong.

As an improvement of the above scheme, the device also includes:

The data uploading module is used for uploading the fused data characteristic pattern sequence and the current road terrain to the server, so that the server performs data analysis according to all the acquired data characteristic pattern sequences corresponding to the same road terrain, and A data analysis result is obtained; wherein, the data analysis result includes a road surface terrain recognition result.

As an improvement of the above scheme, the data sequence interception module includes:

a data sequence interception unit, configured to perform sliding interception of the vibration signal data stream with a sliding window of a preset data interception width corresponding to the current vehicle speed, so as to intercept and obtain a corresponding vibration signal data sequence;

The data normalization unit is used for normalizing and buffering the intercepted vibration signal data sequence.

In this embodiment, normalization processing is performed on the intercepted vibration signal data sequence, which facilitates subsequent feature analysis of the vibration signal data.

Exemplarily, the normalization processing algorithm is:

Among them, □t is the data interception width of the sliding window, x _Δt is the intercepted vibration signal data sequence, σ is the standard deviation,

is the mean value of the vibration signal data series.

As an improvement of the above solution, the data feature analysis module is specifically used for:

The vibration signal data sequence obtained by splicing is subjected to data characteristic analysis by the Matrix Profile algorithm, and according to the data characteristic analysis result, the data characteristic pattern sequence in the vibration signal data sequence obtained by the splicing is determined.

Specifically, the Matrix Profile algorithm is based on the principle as shown in the figure. Among them, a segment is taken from the time series data and slid along the rest of the time series, and how much it is similar to the overlap of the time series segment at each new position is calculated. More specifically, the Euclidean distance between a subsequence and each time series segment of the same length can be calculated, thereby establishing a so-called distance pattern for time series segments. If the subsequence repeats itself in the data, there will be at least one exact match and the minimum Euclidean distance will be zero. Whether the time series data contains similar patterns can be obtained by calculating the minimum value of the matrix profile in the time series data window. The same time series data are used as the two dimensions of the matrix, and the similarity of each corresponding point in the matrix is calculated in a one-to-one correspondence. Distance di, dj, and then find the minimum value in the column direction to obtain Pi, in the sequence from P1 to Pn-m+1, the index corresponding to the minimum value is the starting point MPI (Matrix Profile Index) for generating similar data patterns, taking MPI as the starting point to take The specific window length data is the data characteristic pattern sequence in the data sequence obtained by splicing.

Specifically, the calculation formula of Matrix Profile is:

The formula for finding the minimum value of similarity distance by column: MP(t) _i =min(d _{(i, j)} );

The formula for finding the minimum value of the calculated Pi sequence: MPI=min(MP(Δt)).

As an improvement of the above solution, the data interception width dynamic_wsize of the sliding window is:

Or, the data interception width dynamic_wsize of the sliding window is:

Wherein, max_speed is the preset maximum vehicle speed threshold, base_wsize is the preset data interception width reference value, and smooth(vspeed) is the current average vehicle speed.

As an improvement of the above scheme, if the sliding window is to jointly extract multiple associated vibration signals on the data stream, the data interception width dynamic_wsize of the sliding window is:

dynamic_wsize _i =dynamic_wsize _{i -1} +Δi

Wherein, t _sample is a preset time sampling length, and σ(s ₁ , s ₂ , s ₃ , s _i ) is the standard deviation at time ti of multiple scalar data streams.

As an improvement of the above solution, the sliding step size of the sliding window corresponds to the sampling frequency of the vehicle suspension sensor.

Referring to FIG. 7 , it is a schematic diagram of a road terrain recognition device based on a vehicle suspension vibration signal provided by an embodiment of the present invention. The apparatus for recognizing road terrain based on vehicle suspension vibration signals of this embodiment includes: a processor 1, a memory 2, and a computer program stored in the memory and executable on the processor, for example, based on the vehicle suspension vibration signals Pavement Topography Recognition Program. When the processor executes the computer program, the steps in each of the above embodiments of the road terrain recognition method based on the vehicle suspension vibration signal are implemented. Alternatively, when the processor executes the computer program, the functions of the modules/units in the foregoing device embodiments are implemented.

Exemplarily, the computer program may be divided into one or more modules/units, and the one or more modules/units are stored in the memory and executed by the processor to accomplish the present invention. The one or more modules/units may be a series of computer program instruction segments capable of accomplishing specific functions, and the instruction segments are used to describe the execution of the computer program in the road terrain recognition device based on vehicle suspension vibration signals process.

The road terrain recognition device based on the vibration signal of the vehicle suspension may be a main control device of the vehicle, such as a body domain controller. The device for identifying road terrain based on the vibration signal of the vehicle suspension may include, but is not limited to, a processor and a memory. Those skilled in the art can understand that the schematic diagram is only an example of a road terrain recognition device based on the vibration signal of the vehicle suspension, and does not constitute a limitation on the road terrain recognition device based on the vibration signal of the vehicle suspension. More or less components, or a combination of some components, or different components, for example, the road terrain identification device based on the vibration signal of the vehicle suspension may also include input and output devices, network access devices, buses, and the like.

The so-called processor can be a central processing unit (Central Processing Unit, CPU), or other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), off-the-shelf processors. Field-Programmable Gate Array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The processor is the control center of the road terrain recognition device based on the vehicle suspension vibration signal, using various interfaces and circuits. Connect the various parts of the whole road terrain recognition device based on the vibration signal of the vehicle suspension.

The memory can be used to store the computer program and/or module, and the processor implements the based on by running or executing the computer program and/or module stored in the memory and calling the data stored in the memory. Various functions of road terrain recognition devices for vehicle suspension vibration signals. The memory may mainly include a stored program area and a stored data area, wherein the stored program area may store an operating system, an application program required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the storage data area may store Data (such as audio data, phonebook, etc.) created according to the usage of the mobile phone, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory such as hard disk, internal memory, plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card , a flash card (Flash Card), at least one magnetic disk storage device, flash memory device, or other volatile solid-state storage device.

Wherein, if the integrated module/unit of the road terrain recognition device based on the vehicle suspension vibration signal is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention can implement all or part of the processes in the methods of the above embodiments, and can also be completed by instructing relevant hardware through a computer program, and the computer program can be stored in a computer-readable storage medium. When the program is executed by the processor, the steps of the foregoing method embodiments can be implemented. Wherein, the computer program includes computer program code, and the computer program code may be in the form of source code, object code, executable file or some intermediate form, and the like. The computer-readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM, Read-Only Memory) , Random Access Memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable media may be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction, for example, in some jurisdictions, according to legislation and patent practice, the computer-readable media Electric carrier signals and telecommunication signals are not included.

It should be noted that the device embodiments described above are only schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical unit, that is, it can be located in one place, or it can be distributed over multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution in this embodiment. In addition, in the drawings of the apparatus embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. Those of ordinary skill in the art can understand and implement it without creative effort.

The above are the preferred embodiments of the present invention. It should be pointed out that for those skilled in the art, without departing from the principles of the present invention, several improvements and modifications can be made, and these improvements and modifications may also be regarded as It is the protection scope of the present invention.

Claims (16)

1. A road terrain identification method based on vehicle suspension vibration signals, characterized in that it includes:

   Obtain the vibration signal data of the vehicle suspension currently collected by the suspension sensor during the driving process of the vehicle;

   When the sequence length of the accumulated vibration signal data to be characterized by identification reaches the preset data interception width of the sliding window, the accumulated vibration signal data to be characterized by identification is used as the input quantity and input to the preset road terrain Identifying the road topography features in the recognition model, thereby obtaining the road topography recognition result; wherein the road surface topography recognition model is pre-trained according to the vibration signal data samples.
2. The road terrain identification method based on the vibration signal of the vehicle suspension according to claim 1, characterized in that, after acquiring the vibration signal data of the vehicle suspension, before performing feature identification on the vibration signal data, the method Also includes:

   Data processing is performed on the acquired vibration signal data to obtain vibration signal data after data processing; the data processing includes at least one of the following: data screening, data cleaning, and deletion of vacancies.
3. The road terrain identification method based on the vibration signal of the vehicle suspension according to claim 1, wherein there are at least two suspension sensors, which are distributed in different places of the vehicle suspension;

   Then, after the vibration signal data of the vehicle suspension is acquired, and before the feature identification is performed on the vibration signal data, the method further includes:

   The acquired vibration signal data is stored in the form of a data matrix according to the order of time, so as to obtain a vibration signal data matrix for which data feature extraction is to be performed.
4. The road terrain identification method based on the vibration signal of the vehicle suspension according to claim 3, characterized in that, when the sequence length of the vibration signal data to be characterized identification accumulated this time reaches the data interception width preset by the sliding window , the accumulated vibration signal data to be identified this time is input into the preset identification model of pavement terrain as the input to identify the terrain features of the pavement, including:

   When the sequence length of the data in the vibration signal data matrix to be feature identification accumulated this time reaches the data interception width preset by the sliding window, the vibration signal data matrix is subjected to PCA dimension reduction processing to obtain the vibration signal after dimension reduction. signal data;

   The dimensionality-reduced vibration signal data is input into a preset road terrain recognition model as an input to identify road terrain features.
5. The road topography identification method based on vehicle suspension vibration signals according to claim 4, wherein the road topography identification model is a deep neural network model for identifying road topography, and the vibration signal data matrix is After PCA dimensionality reduction processing, the vibration signal data matrix obtained by PCA dimensionality reduction is obtained.
6. The road terrain identification method based on vehicle suspension vibration signals according to claim 4, wherein the road terrain identification model is an XGBoost model for identifying road terrain, and the vibration signal data matrix is subjected to PCA reduction. After dimension processing, the vibration signal data vector obtained by PCA dimension reduction is obtained.
7. The road terrain identification method based on the vibration signal of the vehicle suspension according to claim 1, wherein the data interception width w' of the sliding window corresponds to the current speed v of the vehicle, and the calculation formula is:

   

   Among them, a is the preset window deviation value, and _wi is the initial data interception width.
8. The road terrain identification method based on the vibration signal of the vehicle suspension according to claim 1, wherein the method for acquiring the vibration signal data sample comprises:

   Obtain the road surface image sequence collected by the vehicle camera when the vehicle is driving on the test road, and obtain the vibration signal time series data of the vehicle suspension collected by the suspension sensor when the vehicle is driving on the test road; the test road The road topography of the working condition includes the target road topography;

   Performing data sliding interception on the vibration signal time series data with a sliding window of a preset data interception width, to obtain multi-stage vibration signal interception data;

   Perform similarity calculation on the plurality of pieces of the vibration signal interception data and the preset vibration signal data template corresponding to the target road terrain to obtain the corresponding similarity value;

   Judging whether the similarity value is greater than a preset similarity threshold;

   If so, mark the image in the road surface image sequence with the vibration signal interception data corresponding to the similarity value at the same time stamp as the target road surface terrain image;

   The vibration signal interception data confirmed by the user based on the target road topographic image is used as the vibration signal data sample.
9. The road terrain identification method based on the vibration signal of the vehicle suspension according to claim 8, characterized in that, in the acquisition of the time sequence of the vibration signal of the vehicle suspension collected by the suspension sensor when the vehicle is running on a test road After the data, before the acquisition of the vibration signal time series data of the vehicle suspension collected by the suspension sensor when the vehicle is running on the test road, the method further includes:

   Normalize the time series data of the vibration signal.
10. The road terrain identification method based on the vibration signal of the vehicle suspension according to claim 8, characterized in that, the step of performing a plurality of pieces of the vibration signal interception data corresponding to the target road terrain and is a preset vibration signal data template. Calculate the similarity to get the corresponding similarity value, including:

    Based on the DTW algorithm, the similarity between the curve segment formed by each segment of the vibration signal interception data and the curve template corresponding to the target road terrain and which is the vibration signal data is calculated, and the corresponding similarity value is obtained.
11. The road terrain identification method based on the vibration signal of the vehicle suspension according to claim 1, wherein the method for acquiring the vibration signal data sample comprises:

    Obtain the vibration signal data stream generated by the vehicle suspension sensor under the current road terrain in the vehicle driving domain;

    Performing data sliding interception on the vibration signal data stream with a sliding window of a preset data interception width, to intercept and obtain a corresponding vibration signal data sequence and store;

    Perform data splicing one by one with the currently intercepted vibration signal data sequence and the previously intercepted vibration signal data sequence stored locally;

    Perform data feature analysis on the vibration signal data sequence obtained by splicing, and determine the data feature pattern sequence in the vibration signal data sequence obtained by splicing according to the data feature analysis result; the data feature pattern sequence is used as the vibration signal data sample.
12. The road terrain identification method based on the vibration signal of the vehicle suspension according to claim 11, characterized in that, in the current intercepted vibration signal data sequence and the locally stored previously intercepted vibration signal data sequence one by one Before performing data splicing, the sliding window of the vibration signal data stream with the preset data interception width is used to perform sliding interception of data, so as to intercept and obtain the corresponding vibration signal data sequence and store it, the method further includes:

    Carrying out similarity calculation between the currently intercepted vibration signal data sequence and each vibration signal data sequence previously intercepted in the local storage to obtain a corresponding similarity comparison result;

    Then, after performing data feature analysis on the vibration signal data sequence obtained by splicing, and determining the data feature pattern sequence in the vibration signal data sequence obtained by splicing according to the data feature analysis result, the method further includes:

    Perform data fusion between the data characteristic pattern sequence and the previously intercepted vibration signal data sequence with the highest similarity result to obtain a fused data characteristic pattern sequence.
13. The road terrain identification method based on the vibration signal of the vehicle suspension according to claim 11, wherein after obtaining the fused data characteristic pattern sequence, the method further comprises:

    uploading the fused data feature pattern sequence and the current road topography to the server, so that the server performs data analysis according to all acquired data feature pattern sequences corresponding to the same road topography to obtain a data analysis result; wherein , the data analysis results include road surface terrain recognition results.
14. A road terrain identification device based on a vehicle suspension vibration signal, characterized in that it includes:

    The data acquisition module is used to acquire the vibration signal data of the vehicle suspension currently collected by the suspension sensor during the driving process of the vehicle;

    The terrain recognition module is used for inputting the accumulated vibration signal data to be characterized as input when the sequence length of the vibration signal data to be characterized by the current accumulation reaches the preset data interception width of the sliding window. The road terrain features are identified in the preset road terrain recognition model, so as to obtain the recognition result of the road terrain; wherein, the road terrain recognition model is pre-trained according to the vibration signal data samples.
15. A road terrain recognition device based on a vehicle suspension vibration signal, characterized by comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, the processor executing the When the computer program is implemented, the road terrain identification method based on the vibration signal of the vehicle suspension according to any one of claims 1 to 13 is realized.
16. A computer-readable storage medium, characterized in that the computer-readable storage medium comprises a stored computer program, wherein, when the computer program runs, the device where the computer-readable storage medium is located is controlled to perform the steps as claimed in claims 1 to 1. The road terrain recognition method based on the vibration signal of the vehicle suspension according to any one of 13.

Images