CN110674853A - Ultrasonic data processing method and device and vehicle - Google Patents
Ultrasonic data processing method and device and vehicle Download PDFInfo
- Publication number
- CN110674853A CN110674853A CN201910848772.3A CN201910848772A CN110674853A CN 110674853 A CN110674853 A CN 110674853A CN 201910848772 A CN201910848772 A CN 201910848772A CN 110674853 A CN110674853 A CN 110674853A
- Authority
- CN
- China
- Prior art keywords
- ultrasonic
- data
- classification model
- obstacle
- radar
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Traffic Control Systems (AREA)
Abstract
Ultrasonic data processing method, device and vehicle, the method includes: acquiring ultrasonic sample data of the object class marked out; training an initial classification model by using the ultrasonic sample data to obtain a trained ultrasonic data classification model; wherein the initial classification model is a support vector machine; acquiring ultrasonic data to be identified; inputting the ultrasonic data to be identified into an ultrasonic data classification model, and determining the object type of the ultrasonic data to be identified based on the output result of the ultrasonic data classification model. By implementing the embodiment of the invention, the data acquired by the ultrasonic sensor can be classified without depending on the characteristics of manual design, and the classification accuracy can be improved.
Description
Technical Field
The invention relates to the technical field of ultrasonic data processing, in particular to an ultrasonic data processing method, an ultrasonic data processing device and a vehicle.
Background
Ultrasonic sensors are widely used in automotive solutions. However, based on the data detected by the ultrasonic sensor, only the distance between the vehicle and the obstacle can be recognized, and it is not possible to recognize which kind of object is the obstacle around the vehicle. That is, the data collected by the ultrasonic sensor cannot be classified.
Disclosure of Invention
The embodiment of the invention discloses an ultrasonic data processing method, an ultrasonic data processing device and a vehicle, which can classify data acquired by an ultrasonic sensor.
The first aspect of the embodiments of the present invention discloses an ultrasonic data processing method, including:
acquiring ultrasonic sample data of the object class marked out;
training an initial classification model by using the ultrasonic sample data to obtain a trained ultrasonic data classification model; wherein the initial classification model is a support vector machine;
acquiring ultrasonic data to be identified;
inputting the ultrasonic data to be identified into an ultrasonic data classification model, and determining the object type of the ultrasonic data to be identified based on the output result of the ultrasonic data classification model.
As an optional implementation manner, in the first aspect of the embodiment of the present invention, the training the initial classification model by using the ultrasonic sample data to obtain a trained ultrasonic data classification model includes:
inputting the ultrasonic sample data into an initial classification model to obtain a classification result output by the initial classification model;
comparing the classification result with the marked object class to obtain a comparison result;
adjusting parameters in the initial classification model according to the comparison result so that the parameters w in the initial classification model satisfy the following conditions:
minimizing a distance d of a decision boundary of the initial classification model to 2/| w |,
wherein N is the total number of the ultrasonic sample data, xiRepresents the ith ultrasonic sample data, yiFor the marked object class to which the ith ultrasonic sample data belongs, λ | w |)2Is a regular term;
and determining the initial classification model when the parameter w meets the condition as a trained ultrasonic data classification model.
As an optional implementation manner, in the first aspect of the embodiment of the present invention, the determining, based on an output result of the ultrasound data classification model, an object class to which the ultrasound data to be identified belongs includes:
solving for f (x) ═ w'Tx + b; wherein, x is the ultrasonic data to be identified, w' is a parameter in the trained ultrasonic data classification model, and b is a calibration parameter;
if f (x) is more than or equal to 1, determining the ultrasonic data to be identified as one type; if f (x) ≦ 1, the ultrasound data to be identified is determined as another class.
As an optional implementation manner, in the first aspect of the embodiment of the present invention, the acquiring ultrasound sample data marked as belonging to a category includes:
acquiring laser radar data acquired by a laser radar and ultrasonic radar data acquired by an ultrasonic radar;
identifying ultrasonic radar data belonging to the obstacle in the ultrasonic radar data by taking the laser radar data as a true value of the obstacle;
and marking the object type to which the ultrasonic radar data belonging to the obstacle belongs as the object type of the obstacle, and taking the marked ultrasonic radar data belonging to the obstacle as ultrasonic sample data.
As an optional implementation manner, in the first aspect of the embodiments of the present invention, the identifying, by using the lidar data as a true value of an obstacle, the ultrasonic radar data belonging to the obstacle in the ultrasonic radar data includes:
calculating the position coordinate of a laser reflection point for reflecting laser according to the installation position of the laser radar on the vehicle and the laser radar data;
calculating the position coordinates of ultrasonic reflection points of reflected ultrasonic waves according to the ultrasonic radar data at the installation position of the ultrasonic radar on the vehicle; the position coordinates of the laser reflection points and the position coordinates of the ultrasonic reflection points are in the same coordinate system;
determining the range occupied by the barrier according to the position coordinates of the laser reflection points;
and determining the ultrasonic radar data corresponding to the ultrasonic reflection points falling within the range occupied by the obstacle as the ultrasonic radar data belonging to the obstacle.
A second aspect of the embodiments of the present invention discloses an ultrasonic data processing apparatus, including:
the first acquisition unit is used for acquiring ultrasonic sample data of the object class marked by the first acquisition unit;
the training unit is used for training the initial classification model by using the ultrasonic sample data to obtain a trained ultrasonic data classification model; wherein the initial classification model is a support vector machine;
the second acquisition unit is also used for acquiring ultrasonic data to be identified;
and the identification unit is used for inputting the ultrasonic data to be identified into an ultrasonic data classification model and determining the object class to which the ultrasonic data to be identified belongs based on the output result of the ultrasonic data classification model.
As an optional implementation manner, in a second aspect of the embodiment of the present invention, the training unit includes:
the input subunit inputs the ultrasonic sample data into an initial classification model to obtain a classification result output by the initial classification model;
the comparison subunit is used for comparing the classification result with the marked object class to obtain a comparison result;
a parameter adjusting subunit, configured to adjust a parameter in the initial classification model according to the comparison result, so that the parameter w in the initial classification model satisfies the following condition:
minimizing a distance d of a decision boundary of the initial classification model to 2/| w |,
wherein N is the total number of the ultrasonic sample data, xiRepresents the ith ultrasonic sample data, yiFor the marked object class to which the ith ultrasonic sample data belongs, λ | w |)2Is a regular term;
a determining subunit 4024, configured to determine the initial classification model when the parameter w satisfies the above condition as a trained ultrasound data classification model.
As an optional implementation manner, in the second aspect of the embodiment of the present invention, the manner that the identification unit is configured to determine the object class to which the ultrasonic data to be identified belongs based on the output result of the ultrasonic data classification model is specifically:
the identification unit is used for solving f (x) wTx + b; wherein, x is the ultrasonic data to be identified, w is a parameter in the ultrasonic data classification model, and b is a calibration parameter; if f (x) is more than or equal to 1, determining the ultrasonic data to be identified as one type; if f (x) ≦ 1, the ultrasound data to be identified is determined as another class.
As an optional implementation manner, in a second aspect of the embodiment of the present invention, the first obtaining unit includes:
the acquisition subunit is used for acquiring laser radar data acquired by a laser radar and ultrasonic radar data acquired by an ultrasonic radar;
the identification subunit is used for identifying the ultrasonic radar data belonging to the obstacle in the ultrasonic radar data by taking the laser radar data as a true value of the obstacle;
and a marking subunit, configured to mark the object class to which the ultrasonic radar data belonging to the obstacle belongs as the object class of the obstacle, and use the marked ultrasonic radar data belonging to the obstacle as ultrasonic sample data.
As an alternative implementation, in the second aspect of the embodiments of the present invention, the identifier unit includes:
the calculation module is used for calculating the position coordinates of laser reflection points of reflected laser according to the installation position of the laser radar on the vehicle and the laser radar data; calculating the position coordinates of ultrasonic reflection points of reflected ultrasonic waves according to the ultrasonic radar data at the installation position of the ultrasonic radar on the vehicle; the position coordinates of the laser reflection points and the position coordinates of the ultrasonic reflection points are in the same coordinate system;
the first determining module is used for determining the range occupied by the barrier according to the position coordinates of the laser reflection points;
and the second determination module is used for determining the ultrasonic radar data corresponding to the ultrasonic reflection points falling into the range occupied by the obstacle as the ultrasonic radar data belonging to the obstacle.
A third aspect of the embodiments of the present invention discloses a vehicle including any one of the ultrasonic data processing apparatuses disclosed in the second aspect of the embodiments of the present invention.
A fourth aspect of the present invention discloses a computer-readable storage medium storing a computer program, wherein the computer program causes a computer to execute any one of the methods disclosed in the first aspect of the embodiments of the present invention.
A fifth aspect of the embodiments of the present invention discloses a computer program product, which, when running on a computer, causes the computer to execute any one of the methods disclosed in the first aspect of the embodiments of the present invention.
Compared with the prior art, the embodiment of the invention has the following beneficial effects:
and training the initial classification model by using the ultrasonic sample data of the object class to which the label belongs, and identifying the object class to which the ultrasonic data belongs by using the ultrasonic data classification model obtained after training. Therefore, the ultrasonic data to be recognized is input into the ultrasonic data classification model, and the classification result returned by the ultrasonic data classification model can be obtained, so that the object class to which the ultrasonic data to be recognized belongs is recognized, and the data collected by the ultrasonic sensor is classified. Because the classification model is a support vector machine, the characteristics of ultrasonic data belonging to different object types can be learned through ultrasonic sample data, the classification accuracy can be improved without depending on manually designed characteristics, different driving control operations of the vehicle aiming at different types of obstacles can be favorably executed, and the realization of automatic driving is greatly promoted.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a schematic flow chart of a method for processing ultrasonic data according to an embodiment of the present invention;
FIG. 2 is a schematic flow chart diagram illustrating one embodiment of step 102 of FIG. 1;
FIG. 3 is a schematic flow chart of another ultrasonic data processing method disclosed in the embodiments of the present invention;
FIG. 4 is a schematic structural diagram of an ultrasonic data processing apparatus according to an embodiment of the present invention;
fig. 5 is a schematic structural diagram of another ultrasonic data processing apparatus disclosed in the embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It is to be noted that the terms "comprises" and "comprising" and any variations thereof in the embodiments and drawings of the present invention are intended to cover non-exclusive inclusions. For example, a process, method, system, article, or apparatus that comprises a list of steps or elements is not limited to only those steps or elements listed, but may alternatively include other steps or elements not listed, or inherent to such process, method, article, or apparatus.
The embodiment of the invention discloses an ultrasonic data processing method which can classify data acquired by an ultrasonic sensor. The following are detailed below.
Example one
Referring to fig. 1, fig. 1 is a schematic flow chart of an ultrasonic data processing method according to an embodiment of the present invention. As shown in fig. 1, the ultrasonic data processing method may include the steps of:
101. and acquiring ultrasonic sample data of the object class marked with the mark.
In the embodiment of the present invention, the ultrasonic sample data may be represented by a feature vector of the ultrasonic wave. The elements constituting the feature vector may include at least the echo intensity of the ultrasonic wave and the measured echo distance. The measured echo distance is a distance between the ultrasonic radar and an object that reflects the ultrasonic wave, which is measured by receiving the echo of the ultrasonic wave. Further, the echo intensity may include a primary echo intensity, a secondary echo intensity; the echo distance may include a primary echo distance and a secondary echo distance.
102. Training the initial classification model by using ultrasonic sample data to obtain a trained ultrasonic data classification model; wherein, the initial classification model is a support vector machine.
In an embodiment of the present invention, a Support Vector Machine (SVM) is a linear classifier. When classification is performed by using SVM, it can be regarded as a binary classification problem. For example, when ultrasonic data belonging to a vehicle is recognized by the SVM, the SVM returns a result indicating whether the object type to which the input ultrasonic data belongs is a vehicle or not.
Generally, positive samples belonging to a certain object class and negative samples not belonging to the object class may be included in the ultrasonic sample data. Optionally, before the initial classification model is trained, a part of the ultrasonic sample data may be selected as a training set, and another part of the ultrasonic sample data may be selected as a test set. Both the training set and the test set may also contain positive and negative examples.
The process of training the initial classification model is described below. Optionally, referring to fig. 2, a specific training method may be as shown in the steps of fig. 2:
1021. and inputting the ultrasonic sample data into the initial classification model to obtain a classification result output by the initial classification model.
1022. And comparing the classification result with the marked object class to obtain a comparison result.
1023. And adjusting parameters in the initial classification model according to the comparison result so that the parameters w in the initial classification model satisfy the following conditions:
the distance d that minimizes the decision boundary of the initial classification model is 2/| w |,
wherein N is the total number of ultrasonic sample data, xiRepresents the ith ultrasonic sample data, yiFor the marked object class to which the ith ultrasonic sample data belongs, λ | w |)2Is a regular term.
1024. And determining the initial classification model when the parameter w meets the conditions as a trained ultrasonic data classification model.
In the embodiment of the invention, the training target of the SVM classifier is to find a hyperplane w in the n-dimensional data spaceTX + b is 0, the hyperplane may separate the data points in the data point set X into two classes and maximize the separation between the data point closest to the hyperplane (i.e., the support vector) and the hyperplane. The above training targets may be converted to the following conditions: min d | w |, constrained to (subject to, s.t.)
Assume that the set Y is a label of the ultrasonic sample data, and one element in Y corresponds to an object class to which one feature vector belongs. Object type y to which ith ultrasonic sample data belongsiRepresent different object classes. Such as yi＝1、yi＝2、yiThree different object classes are represented by 3.
If the SVM is used to solve the binary problem, y is used for calculation convenienceiThe value of (b) may be-1 or 1. When y isiWhen the value is 1, the corresponding characteristic vector belongs to one class, and when y isiWhen the value is-1, the corresponding feature vector belongs to another class. For example, when yiWhen 1, the representative belongs to this class (such as a vehicle); when y isiWhen-1, the representation does not belong to this class (e.g., is not a vehicle). It will be understood that y is particularly preferrediWhether 1 represents belonging to this class or yiThe term-1 is used to designate the category, and may be set artificially.
The process of solving the above training target is actually to iterate the parameter w in the modelAnd (5) updating. The initial classification model may generally be trained using feature vectors in a training set as input data: inputting the characteristic vectors in the training set into an initial classification model, and judging whether a classification result output by the initial classification model is consistent with the object class marked by the characteristic vectors or not; if not, returning the value of the adjusting parameter w; repeating the above steps until the parameter w satisfies min d 2/| w |,
i.e. to meet the training objectives. When the initial classification model meets the training target and the preliminarily trained ultrasonic data classification model is obtained, the feature vectors in the test set can be used as input data to test whether the accuracy of the preliminarily trained ultrasonic data classification model meets a preset accuracy threshold value.
If the parameter w meets the training target, or further, the accuracy of the preliminarily trained ultrasound data classification model meets a preset accuracy threshold, the initial classification model when the parameter w meets can be determined as the trained ultrasound data classification model.
In some cases, when the data in the training set is not uniformly distributed or the initial classification model is over-trained, an over-fitting condition occurs, which results in low accuracy of the initially trained ultrasound data classification model on the test set. In the embodiment of the invention, the regularization term λ | w | can be obtained through the regularization term2And the occurrence of overfitting is reduced, so that the accuracy of the trained ultrasonic data classification model is improved.
103. Ultrasonic data to be identified is acquired.
104. And inputting the ultrasonic data to be recognized into the ultrasonic data classification model, and determining the object type of the ultrasonic data to be recognized based on the output result of the ultrasonic data classification model.
In the embodiment of the present invention, the ultrasonic data to be identified may also be represented by a feature vector, and the dimension of the feature vector of the ultrasonic data to be identified is consistent with the dimension of the feature vector of the ultrasonic sample data.
If the initial classification model is trained based on the steps shown in fig. 2 to obtain an ultrasound data classification model, as an alternative embodiment, the specific implementation of step 104 may be:
solving for f (x) ═ w'Tx + b; wherein x is ultrasonic data to be identified, w' is a parameter in a trained ultrasonic data classification model, and b is a calibration parameter;
if f (x) is more than or equal to 1, determining the ultrasonic data to be identified as one type; if f (x) ≦ 1, the ultrasound data to be identified is determined as another class.
Specifically, whether f (x) is more than or equal to 1 is judged as belonging to the class or f (x) is more than or equal to-1 is judged as belonging to the class, and the judgment is consistent with the setting when the SVM classifier is designed.
Further, it can be appreciated that when solving the multi-classification problem of ultrasound data using an SVM classifier, the multi-classification problem can be converted into a one-to-many two-classification problem.
It can be seen that, in the method described in fig. 1, after the SVM classifier is trained, the object class to which the ultrasonic data to be recognized belongs can be recognized by using the SVM classifier, so as to implement classification of the ultrasonic data. The method of machine learning is independent of the characteristics of artificial design, and the accuracy of classification can be improved.
Example two
Referring to fig. 3, fig. 3 is a schematic flow chart of another ultrasonic data processing method according to the embodiment of the present invention. As shown in fig. 3, the ultrasonic data processing method may include the steps of:
301. and acquiring laser radar data acquired by the laser radar and ultrasonic radar data acquired by the ultrasonic radar.
In the embodiment of the present invention, the collected lidar data may at least include a laser ranging distance, and the collected ultrasonic radar data may at least include an ultrasonic echo intensity and an ultrasonic echo distance.
302. And identifying the ultrasonic radar data belonging to the obstacle in the ultrasonic radar data by taking the laser radar data as a true value (group truth) of the obstacle.
Optionally, the specific implementation of step 302 may include the following steps:
3021. calculating the position coordinate of a laser reflection point for reflecting laser according to the installation position of the laser radar on the vehicle and the data of the laser radar;
3022. calculating the position coordinates of the ultrasonic reflection points of the reflected ultrasonic waves according to the ultrasonic radar data of the installation position of the ultrasonic radar on the vehicle; the position coordinates of the laser reflection points and the position coordinates of the ultrasonic reflection points are in the same coordinate system;
in an embodiment of the invention, the lidar and the ultrasonic radar may be mounted on the same vehicle. Optionally, the laser radar data and the ultrasonic radar data may be unified to the vehicle coordinate system for representation according to the installation position of the laser radar on the vehicle and the installation position of the ultrasonic radar on the vehicle, so as to calculate the position coordinates of the laser reflection point and the position coordinates of the ultrasonic reflection point in the same coordinate system.
3023. Determining the range occupied by the barrier according to the position coordinates of the laser reflection points;
in the embodiment of the invention, because the beam angle of the laser radar is relatively smaller, the range of the obstacle detected by the laser radar is more accurate. Therefore, step 3023 determines the boundary of the obstacle defined by the position coordinates of the plurality of laser reflection points using the laser reflection points as the true value. The obstacle boundary and the range enclosed by the obstacle boundary may be considered as the range occupied by the obstacle.
3024. And determining the ultrasonic radar data corresponding to the ultrasonic reflection point falling in the range occupied by the obstacle as the ultrasonic radar data belonging to the obstacle.
In the embodiment of the invention, the laser radar is used for benchmarking, and the ultrasonic radar data belonging to the obstacle can be more accurately identified.
Further, in determining the range occupied by an obstacle, some error may be allowed in the vicinity of the obstacle boundary. For example, the range occupied by the obstacle may include the boundary of the obstacle, the range enclosed by the boundary of the obstacle, and the area where the boundary of the obstacle extends outward for a certain distance.
Since the ultrasound radar data is greatly affected by the beam angle, there may be some redundant ultrasound reflection points, which are generally generated due to the edge of the ultrasound beam angle being reflected by an obstacle. However, when calculating the position coordinates of the ultrasonic reflection points, it is impossible to distinguish whether the edge of the ultrasonic beam angle is reflected by an obstacle or the center of the ultrasonic beam angle is reflected by an obstacle, and therefore the calculated position coordinates of the reflection points may not fall within the obstacle boundary or the range enclosed by the obstacle boundary.
Therefore, it is possible to allow a certain error in the vicinity of the boundary of the obstacle so that the redundant ultrasonic wave reflection point corresponding to the ultrasonic radar data can be correctly identified as belonging to the obstacle.
As an optional implementation manner, after the step 3022 is executed to unify the laser reflection point and the ultrasonic reflection point in the same coordinate system, the coordinate positions of the laser reflection point and the ultrasonic reflection point in the same coordinate system may be visually represented in the human-computer interaction interface, so that the data processing personnel can perform targeting on the laser reflection point and the ultrasonic reflection point. Further, after step 3023 is executed, the area occupied by the obstacle can be visually represented in the human-computer interaction interface.
303. The object type to which the ultrasonic radar data belonging to the obstacle belongs is marked as the object type of the obstacle, and the marked ultrasonic radar data belonging to the obstacle is used as ultrasonic sample data.
Before the initial classification model is trained, marking (i.e., labeling) the ultrasonic data is an important link, and the accuracy of the marking affects the accuracy of the trained ultrasonic data classification model. By executing the steps 301 to 303, the object type to which the ultrasonic radar data belongs can be accurately marked by using the laser radar data as a group.
In addition, the sampling frequency of the ultrasonic radar and the laser radar is high and can reach the level of kilohertz. Therefore, the number of the ultrasonic radar data and the laser radar data collected is large. By executing the above steps 301 to 303, by determining whether the ultrasonic wave reflection point falls within the range occupied by the obstacle, the ultrasonic radar data belonging to the obstacle can be quickly identified, and the marking efficiency can be improved.
304. Training the initial classification model by using ultrasonic sample data to obtain a trained ultrasonic data classification model; wherein, the initial classification model is a support vector machine.
In the embodiment of the present invention, a specific implementation of step 304 may be as shown in the step in fig. 2, and details are not described below.
305. Ultrasonic data to be identified is acquired.
306. And inputting the ultrasonic data to be recognized into the ultrasonic data classification model, and determining the object type of the ultrasonic data to be recognized based on the output result of the ultrasonic data classification model.
In the embodiment of the present invention, the specific implementation manner of step 104 may be:
solving for f (x) ═ w'Tx + b; wherein x is ultrasonic data to be identified, w' is a parameter in a trained ultrasonic data classification model, and b is a calibration parameter;
if f (x) is more than or equal to 1, determining the ultrasonic data to be identified as one type; if f (x) ≦ 1, the ultrasound data to be identified is determined as another class.
Specifically, whether f (x) is more than or equal to 1 is judged as belonging to the class or f (x) is more than or equal to-1 is judged as belonging to the class, and the judgment is consistent with the setting when the SVM classifier is designed.
It can be seen that in the method described in fig. 3, an SVM classifier capable of recognizing the object class to which the ultrasonic data belongs may be trained, and the ultrasonic data to be recognized is classified by using the SVM classifier. Furthermore, when ultrasonic sample data used in training is marked, the data of the laser radar is used as a group route, the object type to which the ultrasonic radar data belongs can be accurately marked, and meanwhile, the marking efficiency can be improved.
EXAMPLE III
Referring to fig. 4, fig. 4 is a schematic structural diagram of an ultrasonic data processing apparatus according to an embodiment of the present invention. As shown in fig. 4, the ultrasonic data processing apparatus may include:
a first acquisition unit 401, configured to acquire ultrasonic sample data of an object class to which the object class has been marked;
a training unit 402, configured to train the initial classification model by using ultrasonic sample data to obtain a trained ultrasonic data classification model; wherein, the initial classification model is a support vector machine;
a second acquiring unit 403, further configured to acquire ultrasonic data to be identified;
the identification unit 404 is configured to input the ultrasonic data to be identified into the ultrasonic data classification model, and determine the object class to which the ultrasonic data to be identified belongs based on the output result of the ultrasonic data classification model.
Optionally, the training unit 402 may specifically include:
the input subunit 4021, inputs the ultrasonic sample data into the initial classification model, and obtains a classification result output by the initial classification model;
a comparison subunit 4022, configured to compare the classification result with the marked object class to obtain a comparison result;
a parameter adjusting subunit 4023, configured to adjust a parameter in the initial classification model according to the comparison result, so that the parameter w in the initial classification model satisfies the following condition:
the distance d that minimizes the decision boundary of the initial classification model is 2/| w |,
wherein the content of the first and second substances,n is the total number of ultrasonic sample data, xiRepresents the ith ultrasonic sample data, yiFor the marked object class to which the ith ultrasonic sample data belongs, λ | w |)2Is a regular term;
a determining subunit 4024, configured to determine the initial classification model when the parameter w satisfies the above condition as a trained ultrasound data classification model.
Accordingly, the above-mentioned manner for determining the object class to which the ultrasonic data to be identified belongs based on the output result of the ultrasonic data classification model by the identification unit 404 may specifically be:
an identification unit 404 for solving for f (x) wTx + b; wherein x is ultrasonic data to be identified, w is a parameter in the ultrasonic data classification model, and b is a calibration parameter; if f (x) is more than or equal to 1, determining the ultrasonic data to be identified as one type; if f (x) ≦ 1, the ultrasound data to be identified is determined as another class.
It can be seen that, by implementing the ultrasonic data processing apparatus shown in fig. 4, an SVM classifier capable of identifying the object class to which the ultrasonic data belongs can be trained, and the ultrasonic data to be identified can be classified by using the SVM classifier. The trained SVM classifier does not depend on the characteristics of manual design, and the classification accuracy can be improved.
Example four
Referring to fig. 5, fig. 5 is a schematic structural diagram of another ultrasonic data processing apparatus according to an embodiment of the present disclosure. The ultrasonic data processing apparatus shown in fig. 5 is optimized by the ultrasonic data processing apparatus shown in fig. 4. In the ultrasonic data processing apparatus shown in fig. 5:
the first obtaining unit 401 may specifically include:
the acquiring sub-unit 4011 is configured to acquire laser radar data acquired by a laser radar and ultrasonic radar data acquired by an ultrasonic radar;
the identifying sub-unit 4012 is configured to identify, using the lidar data as a true value of the obstacle, the lidar data belonging to the obstacle in the lidar data;
a labeling subunit 4013 configured to label an object class to which the ultrasonic radar data belonging to the obstacle belongs as an object class of the obstacle, and use the labeled ultrasonic radar data belonging to the obstacle as ultrasonic sample data.
Further optionally, the identifier unit 4012 may specifically include:
the calculation module 40121 is configured to calculate a position coordinate of a laser reflection point that reflects laser light according to an installation position of the laser radar on the vehicle and the laser radar data; calculating the position coordinates of the ultrasonic reflection points of the reflected ultrasonic waves according to the ultrasonic radar data of the installation position of the ultrasonic radar on the vehicle; the position coordinates of the laser reflection points and the position coordinates of the ultrasonic reflection points are in the same coordinate system;
optionally, the calculation module 40121 may also perform visual representation on coordinate positions of the laser reflection point and the ultrasonic reflection point in the same coordinate system in the human-computer interaction interface, so that a data processing person performs alignment on the laser reflection point and the ultrasonic reflection point;
the first determining module 40122 is configured to determine a range occupied by the obstacle according to the position coordinates of the laser reflection point;
optionally, the range occupied by the obstacle determined by the first determining module 40122 may include the obstacle boundary, the range enclosed by the obstacle boundary, and the area where the obstacle boundary extends outward by a certain distance, so that the redundant ultrasound reflection point corresponding to the ultrasound radar data can be correctly identified as belonging to the obstacle;
in addition, the first determination module 40122 may also visually represent the range occupied by the obstacle in the human-computer interaction interface;
a second determination module 40123, configured to determine the ultrasonic radar data corresponding to the ultrasonic wave reflection point falling within the range occupied by the obstacle as the ultrasonic radar data belonging to the obstacle.
It can be seen that, by implementing the ultrasonic data processing apparatus shown in fig. 5, an SVM classifier capable of identifying the object class to which the ultrasonic data belongs can be trained, and the ultrasonic data to be identified can be classified by using the SVM classifier. Further, when the ultrasonic sample data used in training is marked, the ultrasonic data processing apparatus shown in fig. 5 uses the data of the laser radar as a group route, and can accurately mark the object type to which the ultrasonic radar data belongs, and can also improve the marking efficiency.
In addition, the embodiment of the invention discloses a vehicle which comprises any one ultrasonic data processing device shown in fig. 4 or fig. 5.
The embodiment of the invention discloses a computer-readable storage medium which stores a computer program, wherein the computer program enables a computer to execute any one of the ultrasonic data processing methods shown in fig. 1 or fig. 3.
An embodiment of the invention discloses a computer program product, which comprises a non-transitory computer readable storage medium storing a computer program, and the computer program is operable to make a computer execute any one of the ultrasonic data processing methods shown in fig. 1 or fig. 3.
It should be appreciated that reference throughout this specification to "one embodiment" or "an embodiment" means that a particular feature, structure or characteristic described in connection with the embodiment is included in at least one embodiment of the present invention. Thus, the appearances of the phrases "in one embodiment" or "in an embodiment" in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. Those skilled in the art should also appreciate that the embodiments described in this specification are exemplary and alternative embodiments, and that the acts and modules illustrated are not required in order to practice the invention.
In various embodiments of the present invention, it should be understood that the sequence numbers of the above-mentioned processes do not imply an inevitable order of execution, and the execution order of the processes should be determined by their functions and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated units, if implemented as software functional units and sold or used as a stand-alone product, may be stored in a computer accessible memory. Based on such understanding, the technical solution of the present invention, which is a part of or contributes to the prior art in essence, or all or part of the technical solution, can be embodied in the form of a software product, which is stored in a memory and includes several requests for causing a computer device (which may be a personal computer, a server, a network device, or the like, and may specifically be a processor in the computer device) to execute part or all of the steps of the above-described method of each embodiment of the present invention.
It will be understood by those skilled in the art that all or part of the steps in the methods of the embodiments described above may be implemented by instructions associated with a program, which may be stored in a computer-readable storage medium, where the storage medium includes Read-Only Memory (ROM), Random Access Memory (RAM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), One-time Programmable Read-Only Memory (OTPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), compact disc-Read-Only Memory (CD-ROM), or other Memory, magnetic disk, magnetic tape, or magnetic tape, Or any other medium which can be used to carry or store data and which can be read by a computer.
The above detailed description of the ultrasonic data processing method, the ultrasonic data processing apparatus and the vehicle according to the embodiments of the present invention has been provided, and the principles and embodiments of the present invention are explained herein by applying specific examples, and the above description of the embodiments is only provided to help understanding the method and the core idea of the present invention. Meanwhile, for a person skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.
Claims (10)
1. An ultrasonic data processing method, characterized in that the method comprises:
acquiring ultrasonic sample data of the object class marked out;
training an initial classification model by using the ultrasonic sample data to obtain a trained ultrasonic data classification model; wherein the initial classification model is a support vector machine;
acquiring ultrasonic data to be identified;
inputting the ultrasonic data to be identified into an ultrasonic data classification model, and determining the object type of the ultrasonic data to be identified based on the output result of the ultrasonic data classification model.
2. The method of claim 1, wherein training an initial classification model using the ultrasound sample data to obtain a trained ultrasound data classification model comprises:
inputting the ultrasonic sample data into an initial classification model to obtain a classification result output by the initial classification model;
comparing the classification result with the marked object class to obtain a comparison result;
adjusting parameters in the initial classification model according to the comparison result so that the parameters w in the initial classification model satisfy the following conditions:
minimizing a distance d of a decision boundary of the initial classification model to 2/| w |,
wherein N is the total number of the ultrasonic sample data, xiRepresents the ith ultrasonic sample data, yiFor the marked object class to which the ith ultrasonic sample data belongs, λ | w |)2Is a regular term;
and determining the initial classification model when the parameter w meets the condition as a trained ultrasonic data classification model.
3. The method according to claim 2, wherein the determining the object class to which the ultrasonic data to be recognized belongs based on the output result of the ultrasonic data classification model includes:
solving for f (x) ═ w'Tx + b; wherein, x is the ultrasonic data to be identified, w' is a parameter in the trained ultrasonic data classification model, and b is a calibration parameter;
if f (x) is more than or equal to 1, determining the ultrasonic data to be identified as one type; if f (x) ≦ 1, the ultrasound data to be identified is determined as another class.
4. The method according to any one of claims 1 to 3, wherein the acquiring the ultrasonic sample data marked as belonging to the category comprises:
acquiring laser radar data acquired by a laser radar and ultrasonic radar data acquired by an ultrasonic radar;
identifying ultrasonic radar data belonging to the obstacle in the ultrasonic radar data by taking the laser radar data as a true value of the obstacle;
and marking the object type to which the ultrasonic radar data belonging to the obstacle belongs as the object type of the obstacle, and taking the marked ultrasonic radar data belonging to the obstacle as ultrasonic sample data.
5. The method of claim 4, wherein the identifying of the sodar data belonging to the obstacle from the sodar data as a true value of the obstacle comprises:
calculating the position coordinate of a laser reflection point for reflecting laser according to the installation position of the laser radar on the vehicle and the laser radar data;
calculating the position coordinates of ultrasonic reflection points of reflected ultrasonic waves according to the ultrasonic radar data at the installation position of the ultrasonic radar on the vehicle; the position coordinates of the laser reflection points and the position coordinates of the ultrasonic reflection points are in the same coordinate system;
determining the range occupied by the barrier according to the position coordinates of the laser reflection points;
and determining the ultrasonic radar data corresponding to the ultrasonic reflection points falling within the range occupied by the obstacle as the ultrasonic radar data belonging to the obstacle.
6. An ultrasonic data processing apparatus, characterized in that the apparatus comprises:
the first acquisition unit is used for acquiring ultrasonic sample data of the object class marked by the first acquisition unit;
the training unit is used for training the initial classification model by using the ultrasonic sample data to obtain a trained ultrasonic data classification model; wherein the initial classification model is a support vector machine;
the second acquisition unit is also used for acquiring ultrasonic data to be identified;
and the identification unit is used for inputting the ultrasonic data to be identified into an ultrasonic data classification model and determining the object class to which the ultrasonic data to be identified belongs based on the output result of the ultrasonic data classification model.
7. The apparatus of claim 6, wherein the first obtaining unit comprises:
the acquisition subunit is used for acquiring laser radar data acquired by a laser radar and ultrasonic radar data acquired by an ultrasonic radar;
the identification subunit is used for identifying the ultrasonic radar data belonging to the obstacle in the ultrasonic radar data by taking the laser radar data as a true value of the obstacle;
and a marking subunit, configured to mark the object class to which the ultrasonic radar data belonging to the obstacle belongs as the object class of the obstacle, and use the marked ultrasonic radar data belonging to the obstacle as ultrasonic sample data.
8. The apparatus of claim 7, wherein the identifier subunit comprises:
the calculation module is used for calculating the position coordinates of laser reflection points of reflected laser according to the installation position of the laser radar on the vehicle and the laser radar data; calculating the position coordinates of ultrasonic reflection points of reflected ultrasonic waves according to the ultrasonic radar data at the installation position of the ultrasonic radar on the vehicle; the position coordinates of the laser reflection points and the position coordinates of the ultrasonic reflection points are in the same coordinate system;
the first determining module is used for determining the range occupied by the barrier according to the position coordinates of the laser reflection points;
and the second determination module is used for determining the ultrasonic radar data corresponding to the ultrasonic reflection points falling into the range occupied by the obstacle as the ultrasonic radar data belonging to the obstacle.
9. A vehicle, characterized in that the vehicle comprises an ultrasonic data processing device according to any one of claims 6 to 8.
10. A computer-readable storage medium characterized by storing a computer program, wherein the computer program causes a computer to execute the ultrasonic data processing method according to any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910848772.3A CN110674853A (en) | 2019-09-09 | 2019-09-09 | Ultrasonic data processing method and device and vehicle |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910848772.3A CN110674853A (en) | 2019-09-09 | 2019-09-09 | Ultrasonic data processing method and device and vehicle |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110674853A true CN110674853A (en) | 2020-01-10 |
Family
ID=69076688
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910848772.3A Pending CN110674853A (en) | 2019-09-09 | 2019-09-09 | Ultrasonic data processing method and device and vehicle |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110674853A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112462353A (en) * | 2020-11-26 | 2021-03-09 | 中国第一汽车股份有限公司 | Test method, device, system and equipment of ultrasonic radar |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105911554A (en) * | 2016-06-06 | 2016-08-31 | 清华大学 | Target identification method based on ultrasonic sensor array |
| US20170369051A1 (en) * | 2016-06-28 | 2017-12-28 | Toyota Motor Engineering & Manufacturing North America, Inc. | Occluded obstacle classification for vehicles |
| CN107784151A (en) * | 2016-08-26 | 2018-03-09 | 福特全球技术公司 | The physical modeling of radar and sonac |
| CN107870322A (en) * | 2017-11-03 | 2018-04-03 | 北京清瑞维航技术发展有限公司 | Target identification method, apparatus and system based on single band radar |
| US20180306924A1 (en) * | 2017-04-24 | 2018-10-25 | Toyota Research Institute, Inc. | Systems and methods for surface property identification using waveform classification |
| CN109622416A (en) * | 2019-01-21 | 2019-04-16 | 哈尔滨工业大学（深圳） | Refuse classification method and its system based on ultrasonic sensor array |
| CN109814112A (en) * | 2019-01-15 | 2019-05-28 | 北京百度网讯科技有限公司 | A kind of ultrasonic radar and laser radar information fusion method and system |
| CN109870698A (en) * | 2019-01-15 | 2019-06-11 | 北京百度网讯科技有限公司 | A kind of ultrasonic array detection of obstacles result processing method and system |
-
2019
- 2019-09-09 CN CN201910848772.3A patent/CN110674853A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105911554A (en) * | 2016-06-06 | 2016-08-31 | 清华大学 | Target identification method based on ultrasonic sensor array |
| US20170369051A1 (en) * | 2016-06-28 | 2017-12-28 | Toyota Motor Engineering & Manufacturing North America, Inc. | Occluded obstacle classification for vehicles |
| CN107784151A (en) * | 2016-08-26 | 2018-03-09 | 福特全球技术公司 | The physical modeling of radar and sonac |
| US20180306924A1 (en) * | 2017-04-24 | 2018-10-25 | Toyota Research Institute, Inc. | Systems and methods for surface property identification using waveform classification |
| CN107870322A (en) * | 2017-11-03 | 2018-04-03 | 北京清瑞维航技术发展有限公司 | Target identification method, apparatus and system based on single band radar |
| CN109814112A (en) * | 2019-01-15 | 2019-05-28 | 北京百度网讯科技有限公司 | A kind of ultrasonic radar and laser radar information fusion method and system |
| CN109870698A (en) * | 2019-01-15 | 2019-06-11 | 北京百度网讯科技有限公司 | A kind of ultrasonic array detection of obstacles result processing method and system |
| CN109622416A (en) * | 2019-01-21 | 2019-04-16 | 哈尔滨工业大学（深圳） | Refuse classification method and its system based on ultrasonic sensor array |
Non-Patent Citations (2)
| Title |
|---|
| JAMES BOND: "SVM中寻找最大间隔的理解", 《知乎专栏-HTTPS://ZHUANLAN.ZHIHU.COM/P/30487021》 * |
| 康琦 等: "《机器学习中的不平衡分类方法》", 31 October 2017, 上海 同济大学出版社 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112462353A (en) * | 2020-11-26 | 2021-03-09 | 中国第一汽车股份有限公司 | Test method, device, system and equipment of ultrasonic radar |
| CN112462353B (en) * | 2020-11-26 | 2023-08-15 | 中国第一汽车股份有限公司 | Ultrasonic radar testing method, device, system and equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113269073B (en) | Ship multi-target tracking method based on YOLO V5 algorithm | |
| EP3506158B1 (en) | Method and apparatus for determining lane line on road | |
| US11783568B2 (en) | Object classification using extra-regional context | |
| KR102371317B1 (en) | Rare Instance Classifiers | |
| US20210334651A1 (en) | Learning point cloud augmentation policies | |
| JP5908193B1 (en) | Radar signal processing device | |
| WO2020132102A2 (en) | Neural networks for coarse- and fine-object classifications | |
| EP3875905B1 (en) | Method, device and medium for detecting environmental change | |
| US11391825B2 (en) | Sensor calibration parameter sensitivity analysis | |
| JP2017207348A (en) | Radar device and sensor fusion device using the same | |
| CN106371106A (en) | People shape targets recognizing method, apparatus and vehicle using single-line laser radar | |
| CN111723724A (en) | Method and related device for identifying road surface obstacle | |
| CN116964588A (en) | Target detection method, target detection model training method and device | |
| CN114460598A (en) | Target identification method, device, equipment and storage medium | |
| CN115797736A (en) | Method, device, equipment and medium for training target detection model and target detection | |
| CN110674853A (en) | Ultrasonic data processing method and device and vehicle | |
| CN114819135A (en) | Training method of detection model, target detection method, device and storage medium | |
| CN110824478B (en) | Automatic classification method and device for precipitation cloud types based on diversified 3D radar echo characteristics | |
| CN115792945B (en) | Floating obstacle detection method and device, electronic equipment and storage medium | |
| CN115049996B (en) | Double-sensor target detection fusion method and system based on evidence reasoning rule | |
| WO2023193923A1 (en) | Maritime traffic management | |
| CN112508970B (en) | Point cloud data segmentation method and device | |
| CN116964472A (en) | Method for detecting at least one object of an environment by means of a reflected signal of a radar sensor system | |
| CN112861708A (en) | Semantic segmentation method and device for radar image and storage medium | |
| CN114638947A (en) | Data labeling method and device, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20201225 Address after: Room 46, room 406, No.1, Yichuang street, Zhongxin knowledge city, Huangpu District, Guangzhou City, Guangdong Province 510000 Applicant after: Guangzhou Xiaopeng Automatic Driving Technology Co.,Ltd. Address before: Room 245, No. 333, jiufo Jianshe Road, Zhongxin Guangzhou Knowledge City, Guangzhou City, Guangdong Province Applicant before: GUANGZHOU XIAOPENG MOTORS TECHNOLOGY Co.,Ltd. |
|
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200110 |