WO2021008202A1 - Method for kernel support vector machine target classification based on millimeter-wave radar point cloud features - Google Patents
Method for kernel support vector machine target classification based on millimeter-wave radar point cloud features Download PDFInfo
- Publication number
- WO2021008202A1 WO2021008202A1 PCT/CN2020/089090 CN2020089090W WO2021008202A1 WO 2021008202 A1 WO2021008202 A1 WO 2021008202A1 CN 2020089090 W CN2020089090 W CN 2020089090W WO 2021008202 A1 WO2021008202 A1 WO 2021008202A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- point cloud
- feature
- target
- point
- target point
- Prior art date
Links
Images
Classifications
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- 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
Definitions
- the invention relates to the field of target classification, and in particular to a kernel support vector machine target classification method based on millimeter wave radar point cloud features.
- the millimeter wave radar sensor has the advantages of long detection distance, accurate ranging and speed measurement, and stable operation in extreme weather (fog, snow, etc.), making the millimeter wave radar sensor an important sensor for advanced driving assistance systems.
- the target classification application based on the millimeter wave radar sensor is still a technical problem.
- the current target classification method based on millimeter wave radar does not make full use of the azimuth and intensity information of the echo signal, resulting in low accuracy and poor practicability of the target classification method based on millimeter wave radar. Therefore, the target classification method based on millimeter wave radar cannot be promoted in practical applications. This has brought great challenges to the development and application of millimeter wave radar.
- the purpose of the present invention is to provide a kernel support vector machine target classification method based on the features of millimeter wave radar point clouds in view of the shortcomings of the prior art.
- a kernel support vector machine target classification method based on millimeter wave radar point cloud features including the following steps:
- S4 Combine the feature vectors of all target point clouds into a feature sample set, divide the feature sample set into a training set and a test set using the ten-fold cross-validation method, use the training set to train the kernel support vector machine, and use the test set to test The classification effect of the trained kernel support vector machine.
- step S2 is implemented through the following sub-steps:
- (x, y, v, I) respectively represent the x-axis coordinate, y-axis coordinate, velocity and echo intensity of any point, and w 1 and w 2 are weight values.
- S2.3 Identify all core objects. First, count the number of points in the Eps neighborhood of each point. If the number of points is greater than minpts, then the point is the core object, otherwise it is a boundary point or a noise point;
- step S3 there are 11 point cloud features in step S3, which are specifically as follows:
- Feature x 3 the width of the target point cloud rectangular frame, that is, the difference between the maximum and minimum values of all reflection points of the same target point cloud on the Y axis;
- Feature x 4 The area of the target point cloud rectangular frame, that is, the product of the length and width of the target point cloud rectangular frame;
- Feature x 5 the density of the target point cloud rectangular frame, that is, the value of the same target point divided by the area of the target frame; if the area of the target frame is 0, the density of the target frame is set to 10000;
- Feature x 7 the variance of all reflection points of the same target point cloud in the Y direction;
- Feature x 8 The average speed of all reflection points of the same target point cloud, that is, the sum of the speeds of all reflection points of the same target point cloud divided by the number of points;
- Feature x 11 the maximum value of the echo intensity of all reflection points of the same target point cloud.
- step S4 is implemented through the following sub-steps:
- S4.2 Use the ten-fold cross-validation method to randomly divide the feature sample set into ten parts, and take turns using nine of them as the training set to train the support vector machine based on the kernel function, and one as the test set to test the trained The classification accuracy of the kernel support vector machine classifier;
- the kernel function is a polynomial kernel function.
- the beneficial effect of the present invention is that the nuclear support vector machine target classification method based on the millimeter wave radar point cloud studies the characteristics of the millimeter wave radar point cloud data, extracts 11 features and fully describes the target point cloud, and realizes the goal based on millimeter wave radar.
- Target classification of radar The invention has a higher recognition accuracy rate than the traditional millimeter-wave radar-based target classification method, and has important practical significance for studying the perception ability of automatic driving.
- Figure 1 is a flowchart of a kernel support vector machine target classification method based on millimeter wave radar point cloud features.
- a kernel support vector machine target classification method based on millimeter wave radar point cloud features as shown in Figure 1, the method includes the following steps:
- (x, y, v, I) respectively represent the x-axis coordinate, y-axis coordinate, velocity and echo intensity of any point, w 1 , w 2 are weight values;
- S2.3 Identify all core objects. First, count the number of points in the Eps neighborhood of each point. If the number of points is greater than minpts, then the point is the core object, otherwise it is a boundary point or a noise point;
- Feature x 3 the width of the target point cloud rectangular frame, that is, the difference between the maximum and minimum values of all reflection points of the same target point cloud on the Y axis;
- Feature x 4 The area of the target point cloud rectangular frame, that is, the product of the length and width of the target point cloud rectangular frame;
- Feature x 5 the density of the target point cloud rectangular frame, that is, the value of the same target point divided by the area of the target frame; if the area of the target frame is 0, the density of the target frame is set to 10000;
- Feature x 7 the variance of all reflection points of the same target point cloud in the Y direction;
- Feature x 8 The average speed of all reflection points of the same target point cloud, that is, the sum of the speeds of all reflection points of the same target point cloud divided by the number of points;
- Feature x 11 the maximum value of the echo intensity of all reflection points of the same target point cloud.
- S4 Combine the feature vectors of all target point clouds into a feature sample set, divide the feature sample set into a training set and a test set using the ten-fold cross-validation method, use the training set to train the kernel support vector machine, and use the test set to test The classification effect of the trained kernel support vector machine.
- S4.2 Use the ten-fold cross-validation method to randomly divide the feature sample set into ten parts, and take turns using nine of them as the training set to train the support vector machine based on the kernel function, and one as the test set to test the trained The classification accuracy of the kernel support vector machine classifier;
- Table 1 shows the comparison of target classification results between the millimeter wave point cloud target classification method of the present invention and the traditional millimeter wave point cloud target classification method.
- the Texas Instruments IWR1642 single-chip radar sensor is used to collect 4000 pedestrian millimeter wave point clouds and 4000 automobile millimeter wave point clouds, and the target point cloud is extracted after processing in steps S1 and S2.
- the 11 features of the present invention and the traditional point cloud features are extracted respectively.
- Traditional point cloud features specifically include the number of points, distance dimension length, distance dimension upper difference, distance dimension upper standard deviation, velocity dimension length, velocity dimension upper difference, velocity dimension upper standard deviation, and radial velocity.
- support vector machines with three different kernel functions linear kernel function, Gauss kernel function and polynomial kernel function
- Table 1 that under different kernel function support vector machine classifiers, the target classification method of the present invention is more accurate than the traditional method.
- the classification effect of the support vector machine using the polynomial kernel function of the present invention is the best, and the recognition accuracy rates of pedestrians and vehicles reach 97.17% and 97.40%, respectively.
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)
- Radar Systems Or Details Thereof (AREA)
Abstract
Disclosed is a method for kernel support vector machine target classification based on millimeter-wave radar point cloud features, the method comprising: firstly preprocessing original radar point cloud data to remove point clouds outside a radar detection region; then, clustering target point clouds into one class by means of a clustering algorithm, so as to remove noise point clouds; and next, in combination with point cloud features of a target, constructing a feature vector composed of 11 features, and performing training and testing by using a kernel support vector machine classifier, thereby realizing target classification. Compared with a method for traditional target classification based on millimeter-wave radar, the present invention has a higher identification accuracy, and has an important practical significance in researching the perception capability of automatic driving.
Description
本发明涉及目标分类领域,尤其涉及一种基于毫米波雷达点云特征的核支持向量机目标分类方法。The invention relates to the field of target classification, and in particular to a kernel support vector machine target classification method based on millimeter wave radar point cloud features.
毫米波雷达传感器具有探测距离远,测距、测速精确,在极端恶劣天气(雾天、雪天等)能够稳定工作等优点,使得毫米波雷达传感器成为了高级驾驶辅助系统的重要传感器。但是由于毫米波雷达点云稀疏,包含的目标特征少等问题,基于毫米波雷达传感器的目标分类应用仍然是一个技术难题。现有的基于毫米波雷达的目标分类方法有三种,一种是基于目标在距离和速度的剖面特点实现了目标分类,第二种是从空间分布和多普勒信息中提取出五种显著的特征实现了对静止目标和行人的分类,第三种是利用目标反射的回波信号强度来实现目标分类。The millimeter wave radar sensor has the advantages of long detection distance, accurate ranging and speed measurement, and stable operation in extreme weather (fog, snow, etc.), making the millimeter wave radar sensor an important sensor for advanced driving assistance systems. However, due to the sparse point cloud of the millimeter wave radar and the few target features included, the target classification application based on the millimeter wave radar sensor is still a technical problem. There are three existing target classification methods based on millimeter wave radar. One is to achieve target classification based on the profile characteristics of the target in distance and speed, and the second is to extract five significant types from spatial distribution and Doppler information. The feature realizes the classification of stationary targets and pedestrians. The third is to use the intensity of the echo signal reflected by the target to realize the target classification.
目前基于毫米波雷达的目标分类方法没有充分利用回波信号的方位角以及强度信息,导致基于毫米波雷达的目标分类方法准确率低,实用性不好。因此无法在实际应用中推广基于毫米波雷达的目标分类方法。这给毫米波雷达的发展与应用带来了极大的挑战。The current target classification method based on millimeter wave radar does not make full use of the azimuth and intensity information of the echo signal, resulting in low accuracy and poor practicability of the target classification method based on millimeter wave radar. Therefore, the target classification method based on millimeter wave radar cannot be promoted in practical applications. This has brought great challenges to the development and application of millimeter wave radar.
发明内容Summary of the invention
本发明的目的在于针对现有技术的不足,提供一种基于毫米波雷达点云特征的核支持向量机目标分类方法。The purpose of the present invention is to provide a kernel support vector machine target classification method based on the features of millimeter wave radar point clouds in view of the shortcomings of the prior art.
本发明的目的是通过以下技术方案来实现的:一种基于毫米波雷达点云特征的核支持向量机目标分类方法,包括以下步骤:The purpose of the present invention is achieved through the following technical solutions: a kernel support vector machine target classification method based on millimeter wave radar point cloud features, including the following steps:
S1:对原始的雷达点云预处理,将雷达探测区域外的点云剔除;S1: Preprocess the original radar point cloud and remove the point cloud outside the radar detection area;
S2:通过具有噪声的基于密度的聚类方法将目标点云聚为一类,从而剔除噪声点云;S2: Clustering the target point clouds into one category through the density-based clustering method with noise, thereby eliminating the noise point clouds;
S3:提取目标的点云特征,组合成用于目标分类的特征向量;S3: Extract the point cloud features of the target and combine them into a feature vector for target classification;
S4:将所有目标点云的特征向量组合成特征样本集,采用十折交叉验证的方法将特征样本集分为训练集和测试集,采用训练集对核支持向量机进行训练,采用测试集测试训练好的核支持向量机的分类效果。S4: Combine the feature vectors of all target point clouds into a feature sample set, divide the feature sample set into a training set and a test set using the ten-fold cross-validation method, use the training set to train the kernel support vector machine, and use the test set to test The classification effect of the trained kernel support vector machine.
进一步地,所述步骤S2通过以下子步骤来实现:Further, the step S2 is implemented through the following sub-steps:
S2.1:通过下式计算预处理后的毫米波点云数据中任意两点之间的欧式距离:S2.1: Calculate the Euclidean distance between any two points in the preprocessed millimeter wave point cloud data by the following formula:
其中,(x,y,v,I)分别表示任意一点的x轴坐标、y轴的坐标、速度和回波强度,w
1、w
2为权重值。
Among them, (x, y, v, I) respectively represent the x-axis coordinate, y-axis coordinate, velocity and echo intensity of any point, and w 1 and w 2 are weight values.
S2.3:确定所有的核心对象。先统计每个点Eps邻域内的点数,如果点数大于minpts,则该点为核心对象,否则为边界点或者噪声点;S2.3: Identify all core objects. First, count the number of points in the Eps neighborhood of each point. If the number of points is greater than minpts, then the point is the core object, otherwise it is a boundary point or a noise point;
S2.4:确定核心对象之间的密度直达关系。如果点p在点q的Eps邻域内,且点p和点q均为核心对象,则点p对于点q密度直达;S2.4: Determine the direct relationship of density between core objects. If point p is in the Eps neighborhood of point q, and point p and point q are both core objects, then point p reaches the density of point q directly;
S2.5:确定核心对象之间的密度可达关系。对于任意两点p和q,如果存在关系序列X
1,X
2,…,X
N,满足p=X
1,q=X
N,并且X
i+1是X
i的密度直达,则点q对于点p密度可达,从而密度可达的核心对象以及它们邻域的点聚成一类。
S2.5: Determine the density reachability relationship between core objects. For any two points p and q, if the relationship between the presence of the sequence X 1, X 2, ..., X N, satisfies p = X 1, q = X N, and X i + 1 X i is the density of direct, to the point q The density of the point p is reachable, so that the core objects with reachable density and the points in their neighborhood are clustered into one category.
进一步地,所述步骤S3的点云特征为11个,具体如下:Further, there are 11 point cloud features in step S3, which are specifically as follows:
(1)特征x
1:目标点云的个数;
(1) Feature x 1 : the number of target point clouds;
(2)特征x
2:目标点云矩形框的长度,即同一个目标点云的所有反射点在X轴的最大值与最小值的差;
(2) Feature x 2 : the length of the rectangular frame of the target point cloud, that is, the difference between the maximum value and the minimum value of all reflection points of the same target point cloud on the X axis;
(3)特征x
3:目标点云矩形框的宽,即同一个目标点云的所有反射点在Y轴的最大值与最小值的差;
(3) Feature x 3 : the width of the target point cloud rectangular frame, that is, the difference between the maximum and minimum values of all reflection points of the same target point cloud on the Y axis;
(4)特征x
4:目标点云矩形框的面积,即目标点云矩形框的长与宽的乘积;
(4) Feature x 4 : The area of the target point cloud rectangular frame, that is, the product of the length and width of the target point cloud rectangular frame;
(5)特征x
5:目标点云矩形框的密度,即同一个目标点数除以目标框的面积的值;如果目标框面积为0,则该目标框的密度设为10000;
(5) Feature x 5 : the density of the target point cloud rectangular frame, that is, the value of the same target point divided by the area of the target frame; if the area of the target frame is 0, the density of the target frame is set to 10000;
(6)特征x
6:同一个目标点云的所有反射点在X方向的方差;
(6) Feature x 6 : The variance of all reflection points of the same target point cloud in the X direction;
(7)特征x
7:同一个目标点云的所有反射点在Y方向的方差;
(7) Feature x 7 : the variance of all reflection points of the same target point cloud in the Y direction;
(8)特征x
8:同一个目标点云的所有反射点的平均速度,即同一个目标点云的所有反射点的速度总和除以点的个数;
(8) Feature x 8 : The average speed of all reflection points of the same target point cloud, that is, the sum of the speeds of all reflection points of the same target point cloud divided by the number of points;
(9)特征x
9:同一个目标点云的所有反射点在速度维的长度,即同一个目标点云的所有反射点的速度最大值与最小值之差;
(9) Feature x 9 : The length of all reflection points of the same target point cloud in the velocity dimension, that is, the difference between the maximum and minimum speeds of all reflection points of the same target point cloud;
(10)特征x
10:同一个目标点云的所有反射点在速度维的方差;
(10) Feature x 10 : the variance of all reflection points of the same target point cloud in the velocity dimension;
(11)特征x
11:同一个目标点云的所有反射点的回波强度的最大值。
(11) Feature x 11 : the maximum value of the echo intensity of all reflection points of the same target point cloud.
进一步地,所述步骤S4通过以下子步骤来实现:Further, the step S4 is implemented through the following sub-steps:
S4.1:将所有提取完特征的目标点云数据组合成特征样本集,对每个特征进行归一化;S4.1: Combine all the feature-extracted target point cloud data into a feature sample set, and normalize each feature;
S4.2:采用十折交叉验证的方法随机将特征样本集分为十份,轮流将其中九份作训练集对基于核函数的支持向量机进行训练,一份作测试集来测试训练好的核支持向量机分类器的分类准确率;S4.2: Use the ten-fold cross-validation method to randomly divide the feature sample set into ten parts, and take turns using nine of them as the training set to train the support vector machine based on the kernel function, and one as the test set to test the trained The classification accuracy of the kernel support vector machine classifier;
S4.3:十次的测试结果的分类准确率的平均值作为对目标分类方法准确率的估计。S4.3: The average of the classification accuracy of ten test results is used as an estimate of the accuracy of the target classification method.
进一步地,所述的核函数为多项式核函数。Further, the kernel function is a polynomial kernel function.
本发明的有益效果是,基于毫米波雷达点云的核支持向量机目标分类方法,研究基于毫米波雷达点云数据的特点,提取出11个特征全面的描述了目标点云,实现基于毫米波雷达的目标分类。本发明比传统的基于毫米波雷达目标分类方法具有更高的识别准确率,对研究自动驾驶的感知能力具有重要的现实意义。The beneficial effect of the present invention is that the nuclear support vector machine target classification method based on the millimeter wave radar point cloud studies the characteristics of the millimeter wave radar point cloud data, extracts 11 features and fully describes the target point cloud, and realizes the goal based on millimeter wave radar. Target classification of radar. The invention has a higher recognition accuracy rate than the traditional millimeter-wave radar-based target classification method, and has important practical significance for studying the perception ability of automatic driving.
图1是基于毫米波雷达点云特征的核支持向量机目标分类方法流程图。Figure 1 is a flowchart of a kernel support vector machine target classification method based on millimeter wave radar point cloud features.
下面根据附图和优选实施例详细描述本发明,本发明的目的和效果将变得更加明白,应当理解,此处所描述的具体实施例仅仅用以解释本发明,并不用于限定本发明。The following describes the present invention in detail based on the drawings and preferred embodiments. The purpose and effects of the present invention will become more apparent. It should be understood that the specific embodiments described here are only used to explain the present invention and are not intended to limit the present invention.
一种基于毫米波雷达点云特征的核支持向量机目标分类方法,如图1所示,该方法包括以下步骤:A kernel support vector machine target classification method based on millimeter wave radar point cloud features, as shown in Figure 1, the method includes the following steps:
S1:对原始的雷达点云预处理,将雷达探测区域外的点云剔除;S1: Preprocess the original radar point cloud and remove the point cloud outside the radar detection area;
S2:通过具有噪声的基于密度的聚类方法将目标点云聚为一类,从而剔除噪声点云;S2: Clustering the target point clouds into one category through the density-based clustering method with noise, thereby eliminating the noise point clouds;
S2.1:通过下式计算预处理后的毫米波点云数据中任意两点之间的欧式距离:S2.1: Calculate the Euclidean distance between any two points in the preprocessed millimeter wave point cloud data by the following formula:
其中,(x,y,v,I)分别表示任意一点的x轴坐标、y轴的坐标、速度和回波强度,w
1、w
2为权重值;
Among them, (x, y, v, I) respectively represent the x-axis coordinate, y-axis coordinate, velocity and echo intensity of any point, w 1 , w 2 are weight values;
S2.3:确定所有的核心对象。先统计每个点Eps邻域内的点数,如果点数大于minpts,则该点为核心对象,否则为边界点或者噪声点;S2.3: Identify all core objects. First, count the number of points in the Eps neighborhood of each point. If the number of points is greater than minpts, then the point is the core object, otherwise it is a boundary point or a noise point;
S2.4:确定核心对象之间的密度直达关系。如果点p在点q的Eps邻域内,且点p和点q均为核心对象,则点p对于点q密度直达;S2.4: Determine the direct relationship of density between core objects. If point p is in the Eps neighborhood of point q, and point p and point q are both core objects, then point p reaches the density of point q directly;
S2.5:确定核心对象之间的密度可达关系。对于任意两点p和q,如果存在关系序列X
1,X
2,…,X
N,满足p=X
1,q=X
N,并且X
i+1是X
i的密度直达,则点q对于点p密度可达,从而密度可达的核心对象以及它们邻域的点聚成一类。
S2.5: Determine the density reachability relationship between core objects. For any two points p and q, if the relationship between the presence of the sequence X 1, X 2, ..., X N, satisfies p = X 1, q = X N, and X i + 1 X i is the density of direct, to the point q The density of the point p is reachable, so that the core objects with reachable density and the points in their neighbors are grouped together.
S3:提取目标的点云特征,组合成用于目标分类的特征向量;其中点云特征具有11个,具体如下:S3: Extract the point cloud features of the target and combine them into a feature vector for target classification; among them, there are 11 point cloud features, as follows:
(1)特征x
1:目标点云的个数;
(1) Feature x 1 : the number of target point clouds;
(2)特征x
2:目标点云矩形框的长度,即同一个目标点云的所有反射点在X轴的最大值与最小值的差;
(2) Feature x 2 : the length of the rectangular frame of the target point cloud, that is, the difference between the maximum value and the minimum value of all reflection points of the same target point cloud on the X axis;
(3)特征x
3:目标点云矩形框的宽,即同一个目标点云的所有反射点在Y轴的最大值与最小值的差;
(3) Feature x 3 : the width of the target point cloud rectangular frame, that is, the difference between the maximum and minimum values of all reflection points of the same target point cloud on the Y axis;
(4)特征x
4:目标点云矩形框的面积,即目标点云矩形框的长与宽的乘积;
(4) Feature x 4 : The area of the target point cloud rectangular frame, that is, the product of the length and width of the target point cloud rectangular frame;
(5)特征x
5:目标点云矩形框的密度,即同一个目标点数除以目标框的面积的值;如果目标框面积为0,则该目标框的密度设为10000;
(5) Feature x 5 : the density of the target point cloud rectangular frame, that is, the value of the same target point divided by the area of the target frame; if the area of the target frame is 0, the density of the target frame is set to 10000;
(6)特征x
6:同一个目标点云的所有反射点在X方向的方差;
(6) Feature x 6 : The variance of all reflection points of the same target point cloud in the X direction;
(7)特征x
7:同一个目标点云的所有反射点在Y方向的方差;
(7) Feature x 7 : the variance of all reflection points of the same target point cloud in the Y direction;
(8)特征x
8:同一个目标点云的所有反射点的平均速度,即同一个目标点云的所有反射点的速度总和除以点的个数;
(8) Feature x 8 : The average speed of all reflection points of the same target point cloud, that is, the sum of the speeds of all reflection points of the same target point cloud divided by the number of points;
(9)特征x
9:同一个目标点云的所有反射点在速度维的长度,即同一个目标点云的所有反射点的速度最大值与最小值之差;
(9) Feature x 9 : The length of all reflection points of the same target point cloud in the velocity dimension, that is, the difference between the maximum and minimum speeds of all reflection points of the same target point cloud;
(10)特征x
10:同一个目标点云的所有反射点在速度维的方差;
(10) Feature x 10 : the variance of all reflection points of the same target point cloud in the velocity dimension;
(11)特征x
11:同一个目标点云的所有反射点的回波强度的最大值。
(11) Feature x 11 : the maximum value of the echo intensity of all reflection points of the same target point cloud.
S4:将所有目标点云的特征向量组合成特征样本集,采用十折交叉验证的方法将特征样本集分为训练集和测试集,采用训练集对核支持向量机进行训练,采用测试集测试训练好的核支持向量机的分类效果。S4: Combine the feature vectors of all target point clouds into a feature sample set, divide the feature sample set into a training set and a test set using the ten-fold cross-validation method, use the training set to train the kernel support vector machine, and use the test set to test The classification effect of the trained kernel support vector machine.
S4.1:将所有提取完特征的目标点云数据组合成特征样本集,对每个特征进行归一化;S4.1: Combine all the feature-extracted target point cloud data into a feature sample set, and normalize each feature;
S4.2:采用十折交叉验证的方法随机将特征样本集分为十份,轮流将其中九份作训练集对基于核函数的支持向量机进行训练,一份作测试集来测试训练好的核支持向量机分类器的分类准确率;S4.2: Use the ten-fold cross-validation method to randomly divide the feature sample set into ten parts, and take turns using nine of them as the training set to train the support vector machine based on the kernel function, and one as the test set to test the trained The classification accuracy of the kernel support vector machine classifier;
S4.3:十次的测试结果的分类准确率的平均值作为对目标分类方法准确率的估计。S4.3: The average of the classification accuracy of ten test results is used as an estimate of the accuracy of the target classification method.
本发明的毫米波点云目标分类方法与传统的毫米波点云目标分类方法的目标分类结果对比如表1所示。Table 1 shows the comparison of target classification results between the millimeter wave point cloud target classification method of the present invention and the traditional millimeter wave point cloud target classification method.
表1本发明的目标分类方法与传统的毫米波点云目标分类方法的结果对比Table 1 Comparison of results between the target classification method of the present invention and the traditional millimeter wave point cloud target classification method
本实施例中采用德州仪器公司IWR1642单芯片雷达传感器采集了4000个行人毫米波点云与4000个汽车毫米波点云,经过步骤S1和S2处理后提取目标点云。分别提取本发明的11种特征和传统的点云特征。传统点云特征具体包含点数、距离维长度、距离维上方差、距离维上标准差、速度维长度、速度维上方差、速度维上标准差和径向速度。最后利用三种不同核函数(线性核函数、高斯核函数和多项式核函数)的支持向量机进行训练和测试。从表1中可以看出,在不同的核函数支持向量机分类器下,本发明的目标分类方法均比传统方法更加准确。其中本发明在使用多项式核函数的支持向量机分类效果最好,行人和车辆的识别准确率分别达到97.17%和97.40%。In this embodiment, the Texas Instruments IWR1642 single-chip radar sensor is used to collect 4000 pedestrian millimeter wave point clouds and 4000 automobile millimeter wave point clouds, and the target point cloud is extracted after processing in steps S1 and S2. The 11 features of the present invention and the traditional point cloud features are extracted respectively. Traditional point cloud features specifically include the number of points, distance dimension length, distance dimension upper difference, distance dimension upper standard deviation, velocity dimension length, velocity dimension upper difference, velocity dimension upper standard deviation, and radial velocity. Finally, support vector machines with three different kernel functions (linear kernel function, Gauss kernel function and polynomial kernel function) are used for training and testing. It can be seen from Table 1 that under different kernel function support vector machine classifiers, the target classification method of the present invention is more accurate than the traditional method. Among them, the classification effect of the support vector machine using the polynomial kernel function of the present invention is the best, and the recognition accuracy rates of pedestrians and vehicles reach 97.17% and 97.40%, respectively.
本领域普通技术人员可以理解,以上所述仅为发明的优选实例而已,并不用于限制发明,尽管参照前述实例对发明进行了详细的说明,对于本领域的技术人员来说,其依然可以对前述各实例记载的技术方案进行修改,或者对其中部分技术特征进行等同替换。凡在发明的精神和原则之内,所做的修改、等同替换等均应包含在发明的保护范围之内。Those of ordinary skill in the art can understand that the above descriptions are only preferred examples of the invention and are not intended to limit the invention. Although the invention has been described in detail with reference to the foregoing examples, those skilled in the art can still The technical solutions recorded in the foregoing examples are modified, or some of the technical features are equivalently replaced. All modifications and equivalent substitutions made within the spirit and principle of the invention shall be included in the protection scope of the invention.
Claims (5)
- 一种基于毫米波雷达点云特征的核支持向量机目标分类方法,其特征在于,包括以下步骤:A kernel support vector machine target classification method based on millimeter wave radar point cloud features is characterized by including the following steps:S1:对原始的雷达点云预处理,将雷达探测区域外的点云剔除;S1: Preprocess the original radar point cloud and remove the point cloud outside the radar detection area;S2:通过具有噪声的基于密度的聚类方法将目标点云聚为一类,从而剔除噪声点云;S2: Clustering the target point clouds into one category through the density-based clustering method with noise, thereby eliminating the noise point clouds;S3:提取目标的点云特征,组合成用于目标分类的特征向量;S3: Extract the point cloud features of the target and combine them into a feature vector for target classification;S4:将所有目标点云的特征向量组合成特征样本集,采用十折交叉验证的方法将特征样本集分为训练集和测试集,采用训练集对核支持向量机进行训练,采用测试集测试训练好的核支持向量机的分类效果。S4: Combine the feature vectors of all target point clouds into a feature sample set, divide the feature sample set into a training set and a test set using the ten-fold cross-validation method, use the training set to train the kernel support vector machine, and use the test set to test The classification effect of the trained kernel support vector machine.
- 根据权利要求1所述的目标分类方法,其特征在于,所述步骤S2通过以下子步骤来实现:The target classification method according to claim 1, wherein the step S2 is implemented by the following sub-steps:S2.1:通过下式计算预处理后的毫米波点云数据中任意两点之间的欧式距离:S2.1: Calculate the Euclidean distance between any two points in the preprocessed millimeter wave point cloud data by the following formula:其中,(x,y,v,I)分别表示任意一点的x轴坐标、y轴的坐标、速度和回波强度,w 1、w 2为权重值。 Among them, (x, y, v, I) respectively represent the x-axis coordinate, y-axis coordinate, velocity and echo intensity of any point, and w 1 and w 2 are weight values.S2.3:确定所有的核心对象。先统计每个点Eps邻域内的点数,如果点数大于minpts,则该点为核心对象,否则为边界点或者噪声点;S2.3: Identify all core objects. First, count the number of points in the Eps neighborhood of each point. If the number of points is greater than minpts, then the point is the core object, otherwise it is a boundary point or a noise point;S2.4:确定核心对象之间的密度直达关系。如果点p在点q的Eps邻域内,且点p和点q均为核心对象,则点p对于点q密度直达;S2.4: Determine the direct relationship of density between core objects. If point p is in the Eps neighborhood of point q, and point p and point q are both core objects, then point p reaches the density of point q directly;S2.5:确定核心对象之间的密度可达关系。对于任意两点p和q,如果存在关系序列X 1,X 2,…,X N,满足p=X 1,q=X N,并且X i+1是X i的密度直达,则点q对于点p密度可达,从而密度可达的核心对象以及它们邻域的点聚成一类。 S2.5: Determine the density reachability relationship between core objects. For any two points p and q, if the relationship between the presence of the sequence X 1, X 2, ..., X N, satisfies p = X 1, q = X N, and X i + 1 X i is the density of direct, to the point q The density of the point p is reachable, so that the core objects with reachable density and the points in their neighborhood are clustered into one category.
- 根据权利要求1所述的目标分类方法,其特征在于,所述步骤S3的点云特征为11个,具体如下:The target classification method according to claim 1, wherein there are 11 point cloud features in step S3, which are specifically as follows:(1)特征x 1:目标点云的个数; (1) Feature x 1 : the number of target point clouds;(2)特征x 2:目标点云矩形框的长度,即同一个目标点云的所有反射点在X轴的最大值与最小值的差; (2) Feature x 2 : the length of the rectangular frame of the target point cloud, that is, the difference between the maximum value and the minimum value of all reflection points of the same target point cloud on the X axis;(3)特征x 3:目标点云矩形框的宽,即同一个目标点云的所有反射点在Y轴的最大值与最小值的差; (3) Feature x 3 : the width of the target point cloud rectangular frame, that is, the difference between the maximum and minimum values of all reflection points of the same target point cloud on the Y axis;(4)特征x 4:目标点云矩形框的面积,即目标点云矩形框的长与宽的乘积; (4) Feature x 4 : The area of the target point cloud rectangular frame, that is, the product of the length and width of the target point cloud rectangular frame;(5)特征x 5:目标点云矩形框的密度,即同一个目标点数除以目标框的面积的值;如果目标框面积为0,则该目标框的密度设为10000; (5) Feature x 5 : the density of the target point cloud rectangular frame, that is, the value of the same target point divided by the area of the target frame; if the area of the target frame is 0, the density of the target frame is set to 10000;(6)特征x 6:同一个目标点云的所有反射点在X方向的方差; (6) Feature x 6 : The variance of all reflection points of the same target point cloud in the X direction;(7)特征x 7:同一个目标点云的所有反射点在Y方向的方差; (7) Feature x 7 : the variance of all reflection points of the same target point cloud in the Y direction;(8)特征x 8:同一个目标点云的所有反射点的平均速度,即同一个目标点云的所有反射点的速度总和除以点的个数; (8) Feature x 8 : The average speed of all reflection points of the same target point cloud, that is, the sum of the speeds of all reflection points of the same target point cloud divided by the number of points;(9)特征x 9:同一个目标点云的所有反射点在速度维的长度,即同一个目标点云的所有反射点的速度最大值与最小值之差; (9) Feature x 9 : The length of all reflection points of the same target point cloud in the velocity dimension, that is, the difference between the maximum and minimum speeds of all reflection points of the same target point cloud;(10)特征x 10:同一个目标点云的所有反射点在速度维的方差; (10) Feature x 10 : the variance of all reflection points of the same target point cloud in the velocity dimension;(11)特征x 11:同一个目标点云的所有反射点的回波强度的最大值。 (11) Feature x 11 : the maximum value of the echo intensity of all reflection points of the same target point cloud.
- 根据权利要求1所述的目标分类方法,其特征在于,所述步骤S4通过以下子步骤来实现:The target classification method according to claim 1, wherein the step S4 is implemented by the following sub-steps:S4.1:将所有提取完特征的目标点云数据组合成特征样本集,对每个特征进行归一化;S4.1: Combine all the feature-extracted target point cloud data into a feature sample set, and normalize each feature;S4.2:采用十折交叉验证的方法随机将特征样本集分为十份,轮流将其中九份作训练集对基于核函数的支持向量机进行训练,一份作测试集来测试训练好的核支持向量机分类器的分类准确率;S4.2: Use the ten-fold cross-validation method to randomly divide the feature sample set into ten parts, and take turns using nine of them as the training set to train the support vector machine based on the kernel function, and one as the test set to test the trained The classification accuracy of the kernel support vector machine classifier;S4.3:十次的测试结果的分类准确率的平均值作为对目标分类方法准确率的估计。S4.3: The average of the classification accuracy of ten test results is used as an estimate of the accuracy of the target classification method.
- 根据权利要求4所述的目标分类方法,其特征在于,所述的核函数为多项式核函数。The target classification method according to claim 4, wherein the kernel function is a polynomial kernel function.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910640745.7 | 2019-07-16 | ||
CN201910640745.7A CN110427986B (en) | 2019-07-16 | 2019-07-16 | Target classification method of kernel support vector machine based on millimeter wave radar point cloud characteristics |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2021008202A1 true WO2021008202A1 (en) | 2021-01-21 |
Family
ID=68410715
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2020/089090 WO2021008202A1 (en) | 2019-07-16 | 2020-05-08 | Method for kernel support vector machine target classification based on millimeter-wave radar point cloud features |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN110427986B (en) |
WO (1) | WO2021008202A1 (en) |
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112767552A (en) * | 2021-01-25 | 2021-05-07 | 绍兴文理学院 | Support vector machine-based scattered point cloud triangularization method |
CN113537316A (en) * | 2021-06-30 | 2021-10-22 | 南京理工大学 | Vehicle detection method based on 4D millimeter wave radar point cloud |
CN113567953A (en) * | 2021-07-28 | 2021-10-29 | 哈尔滨工业大学 | Full-waveform laser echo signal classification method based on SIFT visual word bag |
CN113610044A (en) * | 2021-08-19 | 2021-11-05 | 清华大学 | 4D millimeter wave three-dimensional target detection method and system based on self-attention mechanism |
CN113671481A (en) * | 2021-07-21 | 2021-11-19 | 西安电子科技大学 | 3D multi-target tracking processing method based on millimeter wave radar |
CN113705617A (en) * | 2021-07-30 | 2021-11-26 | 北京万集科技股份有限公司 | Point cloud data processing method and device, computer equipment and storage medium |
CN113723365A (en) * | 2021-09-29 | 2021-11-30 | 西安电子科技大学 | Target feature extraction and classification method based on millimeter wave radar point cloud data |
CN114236488A (en) * | 2021-11-18 | 2022-03-25 | 深圳成谷科技有限公司 | Object classification method, object classification device, terminal device and storage medium |
CN115131594A (en) * | 2021-03-26 | 2022-09-30 | 航天科工深圳(集团)有限公司 | Millimeter wave radar data point classification method and device based on ensemble learning |
CN115236674A (en) * | 2022-06-15 | 2022-10-25 | 北京踏歌智行科技有限公司 | Mining area environment sensing method based on 4D millimeter wave radar |
CN115236627A (en) * | 2022-09-21 | 2022-10-25 | 深圳安智杰科技有限公司 | Millimeter wave radar data clustering method based on multi-frame Doppler velocity dimension expansion |
CN115271096A (en) * | 2022-07-27 | 2022-11-01 | 阿波罗智能技术(北京)有限公司 | Point cloud processing and machine learning model training method and device and automatic driving vehicle |
CN116824877A (en) * | 2023-08-29 | 2023-09-29 | 湖南纳雷科技有限公司 | Vehicle detection method, medium and system for traffic flow millimeter wave radar |
Families Citing this family (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110427986B (en) * | 2019-07-16 | 2022-02-01 | 浙江大学 | Target classification method of kernel support vector machine based on millimeter wave radar point cloud characteristics |
CN111199555B (en) * | 2019-12-13 | 2023-10-13 | 意诺科技有限公司 | Millimeter wave radar target identification method |
CN111142085B (en) * | 2020-01-15 | 2021-12-03 | 武汉大学 | External radiation source radar target classification and identification method based on track feature extraction |
CN113296086A (en) * | 2020-02-21 | 2021-08-24 | 华为技术有限公司 | Target identification method and device |
CN111832665B (en) * | 2020-09-10 | 2021-01-05 | 福瑞泰克智能系统有限公司 | Fence detection method and device, computer equipment and storage medium |
CN112150501A (en) * | 2020-09-18 | 2020-12-29 | 浙江吉利控股集团有限公司 | Target detection method, device and equipment based on laser radar and storage medium |
CN113516052B (en) * | 2021-05-21 | 2023-04-18 | 同济大学 | Imaging millimeter wave radar point cloud target classification method based on machine learning |
CN113671849A (en) * | 2021-09-03 | 2021-11-19 | 珠海格力电器股份有限公司 | Intelligent household equipment control method and device |
CN113820682B (en) * | 2021-09-26 | 2023-08-18 | 华南农业大学 | Millimeter wave radar-based target detection method and device |
CN114460582B (en) * | 2021-12-14 | 2023-04-14 | 江苏航天大为科技股份有限公司 | Millimeter wave radar cart identification method based on point cloud speed |
CN114633782B (en) * | 2022-03-30 | 2024-02-27 | 南京慧尔视智能科技有限公司 | Train arrival early warning method, device, equipment and medium for railway level crossing |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103093235A (en) * | 2012-12-30 | 2013-05-08 | 北京工业大学 | Handwriting digital recognition method based on improved distance core principal component analysis |
US20180136332A1 (en) * | 2016-11-15 | 2018-05-17 | Wheego Electric Cars, Inc. | Method and system to annotate objects and determine distances to objects in an image |
CN109061600A (en) * | 2018-09-28 | 2018-12-21 | 上海市刑事科学技术研究院 | A kind of target identification method based on millimetre-wave radar data |
CN109581361A (en) * | 2018-11-22 | 2019-04-05 | 九牧厨卫股份有限公司 | A kind of detection method, detection device, terminal and detection system |
CN110427986A (en) * | 2019-07-16 | 2019-11-08 | 浙江大学 | A kind of kernel support vectors machine objective classification method based on millimetre-wave radar point cloud feature |
Family Cites Families (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102270301B (en) * | 2011-06-07 | 2013-05-08 | 南京理工大学 | Method for detecting unstructured road boundary by combining support vector machine (SVM) and laser radar |
CN106022259B (en) * | 2016-05-20 | 2019-04-12 | 江苏得得空间信息科技有限公司 | A kind of mountain road extracting method based on laser point cloud three-dimensional feature descriptive model |
CN108732584B (en) * | 2017-04-17 | 2020-06-30 | 百度在线网络技术(北京)有限公司 | Method and device for updating map |
CN107292335A (en) * | 2017-06-06 | 2017-10-24 | 云南电网有限责任公司信息中心 | A kind of transmission line of electricity cloud data automatic classification method based on Random Forest model |
CN109145680B (en) * | 2017-06-16 | 2022-05-27 | 阿波罗智能技术(北京)有限公司 | Method, device and equipment for acquiring obstacle information and computer storage medium |
CN108226883B (en) * | 2017-11-28 | 2020-04-28 | 深圳市易成自动驾驶技术有限公司 | Method and device for testing millimeter wave radar performance and computer readable storage medium |
CN108509918B (en) * | 2018-04-03 | 2021-01-08 | 中国人民解放军国防科技大学 | Target detection and tracking method fusing laser point cloud and image |
CN109934230A (en) * | 2018-09-05 | 2019-06-25 | 浙江大学 | A kind of radar points cloud dividing method of view-based access control model auxiliary |
CN109581312B (en) * | 2018-11-22 | 2023-07-14 | 西安电子科技大学昆山创新研究院 | High-resolution millimeter wave radar multi-target clustering method |
CN109753874A (en) * | 2018-11-28 | 2019-05-14 | 南京航空航天大学 | A kind of low slow small classification of radar targets method based on machine learning |
CN109901193A (en) * | 2018-12-03 | 2019-06-18 | 财团法人车辆研究测试中心 | The light of short distance barrier reaches arrangement for detecting and its method |
-
2019
- 2019-07-16 CN CN201910640745.7A patent/CN110427986B/en active Active
-
2020
- 2020-05-08 WO PCT/CN2020/089090 patent/WO2021008202A1/en active Application Filing
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103093235A (en) * | 2012-12-30 | 2013-05-08 | 北京工业大学 | Handwriting digital recognition method based on improved distance core principal component analysis |
US20180136332A1 (en) * | 2016-11-15 | 2018-05-17 | Wheego Electric Cars, Inc. | Method and system to annotate objects and determine distances to objects in an image |
CN109061600A (en) * | 2018-09-28 | 2018-12-21 | 上海市刑事科学技术研究院 | A kind of target identification method based on millimetre-wave radar data |
CN109581361A (en) * | 2018-11-22 | 2019-04-05 | 九牧厨卫股份有限公司 | A kind of detection method, detection device, terminal and detection system |
CN110427986A (en) * | 2019-07-16 | 2019-11-08 | 浙江大学 | A kind of kernel support vectors machine objective classification method based on millimetre-wave radar point cloud feature |
Cited By (22)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112767552A (en) * | 2021-01-25 | 2021-05-07 | 绍兴文理学院 | Support vector machine-based scattered point cloud triangularization method |
CN115131594A (en) * | 2021-03-26 | 2022-09-30 | 航天科工深圳(集团)有限公司 | Millimeter wave radar data point classification method and device based on ensemble learning |
CN113537316A (en) * | 2021-06-30 | 2021-10-22 | 南京理工大学 | Vehicle detection method based on 4D millimeter wave radar point cloud |
CN113537316B (en) * | 2021-06-30 | 2024-04-09 | 南京理工大学 | Vehicle detection method based on 4D millimeter wave radar point cloud |
CN113671481B (en) * | 2021-07-21 | 2024-04-09 | 西安电子科技大学 | 3D multi-target tracking processing method based on millimeter wave radar |
CN113671481A (en) * | 2021-07-21 | 2021-11-19 | 西安电子科技大学 | 3D multi-target tracking processing method based on millimeter wave radar |
CN113567953A (en) * | 2021-07-28 | 2021-10-29 | 哈尔滨工业大学 | Full-waveform laser echo signal classification method based on SIFT visual word bag |
CN113705617A (en) * | 2021-07-30 | 2021-11-26 | 北京万集科技股份有限公司 | Point cloud data processing method and device, computer equipment and storage medium |
CN113610044A (en) * | 2021-08-19 | 2021-11-05 | 清华大学 | 4D millimeter wave three-dimensional target detection method and system based on self-attention mechanism |
CN113610044B (en) * | 2021-08-19 | 2022-02-15 | 清华大学 | 4D millimeter wave three-dimensional target detection method and system based on self-attention mechanism |
CN113723365B (en) * | 2021-09-29 | 2024-05-14 | 西安电子科技大学 | Millimeter wave Lei Dadian cloud data-based target feature extraction and classification method |
CN113723365A (en) * | 2021-09-29 | 2021-11-30 | 西安电子科技大学 | Target feature extraction and classification method based on millimeter wave radar point cloud data |
CN114236488A (en) * | 2021-11-18 | 2022-03-25 | 深圳成谷科技有限公司 | Object classification method, object classification device, terminal device and storage medium |
CN114236488B (en) * | 2021-11-18 | 2024-05-14 | 深圳成谷科技有限公司 | Target classification method, target classification device, terminal device and storage medium |
CN115236674A (en) * | 2022-06-15 | 2022-10-25 | 北京踏歌智行科技有限公司 | Mining area environment sensing method based on 4D millimeter wave radar |
CN115236674B (en) * | 2022-06-15 | 2024-06-04 | 北京踏歌智行科技有限公司 | Mining area environment sensing method based on 4D millimeter wave radar |
CN115271096B (en) * | 2022-07-27 | 2023-08-18 | 阿波罗智能技术(北京)有限公司 | Point cloud processing and machine learning model training method and device and automatic driving vehicle |
CN115271096A (en) * | 2022-07-27 | 2022-11-01 | 阿波罗智能技术(北京)有限公司 | Point cloud processing and machine learning model training method and device and automatic driving vehicle |
CN115236627B (en) * | 2022-09-21 | 2022-12-16 | 深圳安智杰科技有限公司 | Millimeter wave radar data clustering method based on multi-frame Doppler velocity dimension expansion |
CN115236627A (en) * | 2022-09-21 | 2022-10-25 | 深圳安智杰科技有限公司 | Millimeter wave radar data clustering method based on multi-frame Doppler velocity dimension expansion |
CN116824877A (en) * | 2023-08-29 | 2023-09-29 | 湖南纳雷科技有限公司 | Vehicle detection method, medium and system for traffic flow millimeter wave radar |
CN116824877B (en) * | 2023-08-29 | 2023-11-24 | 湖南纳雷科技有限公司 | Vehicle detection method, medium and system for traffic flow millimeter wave radar |
Also Published As
Publication number | Publication date |
---|---|
CN110427986B (en) | 2022-02-01 |
CN110427986A (en) | 2019-11-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2021008202A1 (en) | Method for kernel support vector machine target classification based on millimeter-wave radar point cloud features | |
US20230099113A1 (en) | Training method and apparatus for a target detection model, target detection method and apparatus, and medium | |
JP6571545B2 (en) | Object detection apparatus and object detection method | |
Breitenstein et al. | Robust tracking-by-detection using a detector confidence particle filter | |
CN105976376B (en) | High-resolution SAR image target detection method based on component model | |
CN110895333B (en) | Rapid 77G vehicle-mounted radar data clustering method based on Doppler frequency | |
CN113486961A (en) | Radar RD image target detection method and system based on deep learning under low signal-to-noise ratio and computer equipment | |
Pandey et al. | Classification of automotive targets using inverse synthetic aperture radar images | |
Ye et al. | Overhead ground wire detection by fusion global and local features and supervised learning method for a cable inspection robot | |
Yuan et al. | A graph-based vehicle proposal location and detection algorithm | |
Qing et al. | A novel particle filter implementation for a multiple-vehicle detection and tracking system using tail light segmentation | |
Li et al. | Pedestrian liveness detection based on mmwave radar and camera fusion | |
JP7418476B2 (en) | Method and apparatus for determining operable area information | |
Guo et al. | Feature‐based detection and classification of moving objects using LiDAR sensor | |
JP7484492B2 (en) | Radar-based attitude recognition device, method and electronic device | |
JP6657934B2 (en) | Object detection device | |
CN111353481A (en) | Road obstacle identification method based on laser point cloud and video image | |
Yang et al. | Learn to model and filter point cloud noise for a near-infrared ToF LiDAR in adverse weather | |
CN116310552A (en) | Three-dimensional target detection method based on multi-scale feature fusion | |
Qing et al. | Localization and tracking of same color vehicle under occlusion problem | |
Bai et al. | Multi-objective classification of three-dimensional imaging radar point clouds: support vector machine and PointNet | |
JP2013008310A (en) | Three-dimensional map creation device and its window area detection device | |
Fazekas et al. | Performance metrics and validation methods for vehicle position estimators | |
Lu et al. | A vision-based system for the prevention of car collisions at night | |
CN115128571A (en) | Multi-person and non-motor vehicle identification method based on millimeter wave radar |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 20840976 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 20840976 Country of ref document: EP Kind code of ref document: A1 |