CN114418068A - Millimeter wave radar sparse point cloud reconstruction method based on multi-step neural network structure - Google Patents

Millimeter wave radar sparse point cloud reconstruction method based on multi-step neural network structure Download PDF

Info

Publication number
CN114418068A
CN114418068A CN202210049948.0A CN202210049948A CN114418068A CN 114418068 A CN114418068 A CN 114418068A CN 202210049948 A CN202210049948 A CN 202210049948A CN 114418068 A CN114418068 A CN 114418068A
Authority
CN
China
Prior art keywords
point cloud
neural network
millimeter wave
distance
wave 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
Application number
CN202210049948.0A
Other languages
Chinese (zh)
Inventor
杨澄思
卢煜旻
朱欣恩
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanghai Silicon Microelectronics Co ltd
Original Assignee
Shanghai Silicon Microelectronics Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shanghai Silicon Microelectronics Co ltd filed Critical Shanghai Silicon Microelectronics Co ltd
Priority to CN202210049948.0A priority Critical patent/CN114418068A/en
Publication of CN114418068A publication Critical patent/CN114418068A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Molecular Biology (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Probability & Statistics with Applications (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Radar Systems Or Details Thereof (AREA)

Abstract

The invention discloses a millimeter wave radar sparse point cloud reconstruction method based on a multi-step neural network structure, which comprises the following steps of S1: first information including a speed, an angle, and a distance is obtained by a millimeter wave radar and the obtained speed and distance are subjected to fourier transform processing to take a generated speed and distance spectrogram as a first target feature. The invention discloses a millimeter wave radar sparse point cloud reconstruction method based on a multi-step Neural Network structure, which is used for directly carrying out normalization prediction on radar sparse point cloud information through a Forward transmission Neural Network (Feed Forward Neural Network) to obtain high-resolution point cloud information, and provides a novel algorithm for the problem of millimeter wave radar sparse point cloud.

Description

Millimeter wave radar sparse point cloud reconstruction method based on multi-step neural network structure
Technical Field
The invention belongs to the technical field of millimeter wave radar point cloud, and particularly relates to a millimeter wave radar sparse point cloud reconstruction method based on a multi-step neural network structure.
Background
With the rapid development of intelligent sensing technology, millimeter wave radar is widely applied to object detection and identification tasks as one of intelligent sensors. In the field of autopilot, millimeter wave radars are often used as an auxiliary sensor in cooperation with cameras and laser radars to achieve a high-order autopilot function. Compared with laser radar and camera, the millimeter wave radar has the advantages of capability of detecting object speed and distance information, low cost and suitability for various weather illumination conditions. However, millimeter wave radar has a very large defect that its resolution is low, and the point cloud obtained in the field of automatic driving is very sparse.
If the high-resolution point cloud information of the millimeter wave radar can be effectively improved, the market application prospect of the millimeter wave radar can be greatly improved, and the scale is huge. In literature, no effective method for reconstructing the millimeter wave radar sparse point cloud exists at present.
Therefore, the above problems are further improved.
Disclosure of Invention
The invention mainly aims to provide a millimeter wave radar sparse point cloud reconstruction method based on a multi-step Neural Network structure, high-resolution (high-density) point cloud information is predicted by directly carrying out normalization on radar sparse point cloud information through a Forward transfer Neural Network (Feed Forward Neural Network), and a novel algorithm is provided for the millimeter wave radar sparse point cloud problem.
In order to achieve the above purpose, the invention provides a millimeter wave radar sparse point cloud reconstruction method based on a multi-step neural network structure, which comprises the following steps:
step S1: obtaining first information including speed, angle and distance through a millimeter wave radar and performing Fourier transform processing on the obtained speed and distance to take a generated speed and distance spectrogram as a first target feature;
step S2: extracting spatial coordinate point cloud information corresponding to the speed and distance spectrogram;
step S3: clustering the spatial coordinate point cloud information to ensure that all the central point coordinates do not change any more, thereby obtaining clustering points;
step S4: the cluster points are input to a multi-step forward neural network model for regression prediction and to generate supplemental spatial coordinate point cloud information to obtain dense point cloud information.
As a more preferable mode of the above mode, in step S1, the radar frequency modulates the continuous wave signal, and the relational expression between the distance and the speed is:
Figure BDA0003473702020000021
where c is the propagation velocity of the electromagnetic wave in air, R is the distance between the object and the radar, v is the relative velocity, τ is the propagation time, and the ratio of the distance between the object and the radar is calculated to obtain R:
Figure BDA0003473702020000022
as a more preferable mode of the above-described mode, in step S1, the speed and distance are fourier-transformed by the following formulas:
Figure BDA0003473702020000023
wherein, b [ n, m, k]Refers to the nth sample point, T, of the mth chirp signal (chirp) received on the kth antennafastA sampling interval for sampling a signal within a sweep period;
performing two-dimensional discrete Fourier transform on N sampling points in a period to obtain corresponding distances, wherein the formula is as follows:
Figure BDA0003473702020000031
as a more preferable mode of the above-described mode, in step S2, the distance frequency component and the velocity frequency component having non-zero values are extracted from step S1, and the following is calculated:
x＝Rsinθ,y＝Rcosθ，
to obtain spatial coordinate point cloud information (x, y, z coordinates) of the millimeter wave radar.
As a further preferable technical solution of the above technical solution, in step S3, the K-means clustering process is performed on the spatial coordinate point cloud information, and the method specifically includes the following steps:
step S3.1: randomly selecting 5 initial point information, namely C1, C2, C3, C4 and C5;
step S3.2: finding Euclidean distances between all points of the spatial coordinate point cloud information and the initial point information, wherein the distance formula is as follows:
Figure BDA0003473702020000032
X＝(x1,x2,x3),C＝(y1,y2,y3)；
step S3.3: and allocating the point corresponding to each piece of spatial coordinate point cloud information to the center closest to the acquired distance:
Figure BDA0003473702020000033
step S3.4: recalculating the assigned points for new center points by the following formula:
Figure BDA0003473702020000034
step S3.5: step S3.2-step S3.4 are repeated to keep the coordinates of all the center points (cluster points) unchanged.
As a further preferable technical solution of the above technical solution, in step S4, the obtained clustering points are input to a multi-step forward neural network model, and the method is specifically implemented as the following steps:
step S4.1: the multi-step forward neural network model comprises an input layer, a hidden layer and an output layer, and X isLSet as the input data point of the L-th layer, the output result X of the L + 1-th layerL+1Comprises the following steps:
ZL＝F(XL)＝WL*XL+bL；
Figure BDA0003473702020000041
J＝||A-XL+1||；
wherein, WL,bLAnd A is the real labeled data of training, and the calculation formula is as follows:
Figure BDA0003473702020000042
Figure BDA0003473702020000043
step S4.2: the gradient is calculated as:
Figure BDA0003473702020000044
Figure BDA0003473702020000045
step S4.3: according to the multi-step neural network format of the differential equation, the transfer of the forward neural network is modeled as a mathematical differential equation, and the formula is as follows:
Figure BDA0003473702020000046
wherein: t is the number of layers of the neural network model, and the differential equation is expanded to obtain:
Figure BDA0003473702020000047
wherein: lambda [ alpha ]iAre coefficients.
In order to achieve the above object, the present invention further provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the millimeter wave radar sparse point cloud reconstruction method based on the multi-step neural network structure when executing the program.
To achieve the above object, the present invention further provides a non-transitory computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, implements the steps of the millimeter wave radar sparse point cloud reconstruction method based on the multi-step neural network structure.
The invention has the beneficial effects that:
1. the method is used for reconstructing the sparse millimeter wave radar point cloud by combining a clustering method with a neural network, and belongs to the algorithm-based process which is proposed for the first time.
2. A multi-step forward-transitive neural network architecture is proposed. Different from a general forward transmission network structure, the network structure is derived from strict mathematical theory and formula by adopting mathematical theory, and produces better results in practical experiments.
3. The original radar point cloud information is clustered by adopting a K-mean clustering algorithm, so that the data dimension can be effectively reduced, and the training effect is improved.
Drawings
FIG. 1 is a point cloud prediction diagram based on Kmean clustering and a multi-step forward neural network of the millimeter wave radar sparse point cloud reconstruction method based on the multi-step neural network structure.
FIG. 2 is the raw point cloud information collected by the millimeter wave radar.
FIG. 3 is a point cloud information diagram predicted by a neural network of the millimeter wave radar sparse point cloud reconstruction method based on a multi-step neural network structure.
Detailed Description
The following description is presented to disclose the invention so as to enable any person skilled in the art to practice the invention. The preferred embodiments in the following description are given by way of example only, and other obvious variations will occur to those skilled in the art. The basic principles of the invention, as defined in the following description, may be applied to other embodiments, variations, modifications, equivalents, and other technical solutions without departing from the spirit and scope of the invention.
In the preferred embodiments of the present invention, those skilled in the art should note that the electronic devices and storage media and the like to which the present invention relates may be regarded as prior art.
Preferred embodiments.
The invention discloses a millimeter wave radar sparse point cloud reconstruction method based on a multi-step neural network structure, which comprises the following steps of:
step S1: obtaining first information including speed, angle and distance through a millimeter wave radar and performing Fourier transform processing on the obtained speed and distance to take a generated speed and distance spectrogram as a first target feature;
step S2: extracting spatial coordinate point cloud information corresponding to the speed and distance spectrogram;
step S3: clustering the spatial coordinate point cloud information to ensure that all the central point coordinates do not change any more, thereby obtaining clustering points;
step S4: the cluster points are input to a multi-step forward neural network model for regression prediction and to generate supplemental spatial coordinate point cloud information to obtain dense point cloud information.
Specifically, in step S1, the radar frequency modulates the continuous wave signal, and the relationship between the distance and the speed is as follows:
Figure BDA0003473702020000061
where c is the propagation velocity of the electromagnetic wave in air, R is the distance between the object and the radar, v is the relative velocity, τ is the propagation time, and the ratio of the distance between the object and the radar is calculated to obtain R:
Figure BDA0003473702020000062
more specifically, in step S1, the fourier transform processing is performed on the velocity and the distance by the following formulas:
Figure BDA0003473702020000063
wherein, b [ n, m, k]Refers to the nth sample point, T, of the mth chirp signal (chirp) received on the kth antennafastA sampling interval for sampling a signal within a sweep period;
performing two-dimensional discrete Fourier transform on N sampling points in a period to obtain corresponding distances, wherein the formula is as follows:
Figure BDA0003473702020000071
further, in step S2, the distance frequency component and the velocity frequency component having non-zero values are extracted from step S1, and the following are calculated:
x＝Rsinθ,y＝Rcosθ，
to obtain spatial coordinate point cloud information (x, y, z coordinates) of the millimeter wave radar.
Further, in step S3, the K-means clustering process is performed on the spatial coordinate point cloud information, and the method specifically includes the following steps:
step S3.1: randomly selecting 5 initial point information, namely C1, C2, C3, C4 and C5;
step S3.2: finding Euclidean distances between all points of the spatial coordinate point cloud information and the initial point information, wherein the distance formula is as follows:
Figure BDA0003473702020000072
X＝(x1,x2,x3),C＝(y1,y2,y3)；
step S3.3: and allocating the point corresponding to each piece of spatial coordinate point cloud information to the center closest to the acquired distance:
Figure BDA0003473702020000073
step S3.4: recalculating the assigned points for new center points by the following formula:
Figure BDA0003473702020000074
step S3.5: step S3.2-step S3.4 are repeated to keep the coordinates of all the center points (cluster points) unchanged.
Preferably, in step S4, the obtained clustering points are input to the multi-step forward neural network model, and the method is implemented as the following steps:
step S4.1: the multi-step forward neural network model comprises an input layer, a hidden layer and an output layer, and X isLSet as the input data point of the L-th layer, the output result X of the L + 1-th layerL+1Comprises the following steps:
ZL＝F(XL)＝WL*XL+bL；
Figure BDA0003473702020000081
J＝||A-XL+1||；
wherein, WL,bLAnd A is the real labeled data of training, and the calculation formula is as follows:
Figure BDA0003473702020000082
Figure BDA0003473702020000083
step S4.2: the gradient is calculated as:
Figure BDA0003473702020000084
Figure BDA0003473702020000085
step S4.3: according to the multi-step neural network format of the differential equation, the transfer of the forward neural network is modeled as a mathematical differential equation, and the formula is as follows:
Figure BDA0003473702020000086
wherein: t is the number of layers of the neural network model, and the differential equation is expanded to obtain:
Figure BDA0003473702020000087
wherein: lambda [ alpha ]iAre coefficients.
The invention also discloses electronic equipment which comprises a memory, a processor and a computer program which is stored on the memory and can be run on the processor, wherein the processor realizes the steps of the millimeter wave radar sparse point cloud reconstruction method based on the multi-step neural network structure when executing the program.
The principle of the invention is as follows:
the invention provides a new algorithm aiming at the difficult problem of point cloud sparsity of millimeter wave radar. The algorithm can effectively improve the wide application range of the reconstructed millimeter wave radar high-density point cloud sensor through the output result of the existing radar.
The real-time detection algorithm comprises the following specific processes:
1. and obtaining information such as speed, angle, distance and the like through the millimeter wave radar signal.
The radar Frequency Modulation Continuous Wave (FMCW) signal is adopted, and the relation formula of the distance and the speed is as follows:
Figure BDA0003473702020000091
where c is the propagation velocity of the electromagnetic wave in air, R is the distance between the object and the radar, v is the relative velocity, and τ is the propagation time.
Figure BDA0003473702020000092
2. Fourier transform is carried out on the obtained speed and distance to generate a distance-speed spectrogram which is used as a first target characteristic,
Figure BDA0003473702020000093
b[n,m,k]refers to the nth sample point, T, of the mth chip received on the kth antennafastIs the sampling interval at which the signal is sampled during one sweep period. The adopted 2-transmitting and 4-receiving radar antenna is shown in figure 1, wherein the value of N is 256, the value of M is 128, K is 4, the frame rate is 20fps, and the corresponding distance is obtained by performing two-dimensional discrete Fourier transform on N sampling points in one period and is shown as a formula (4).
Figure BDA0003473702020000094
3. Spatial coordinate point cloud information obtained by extracting radar signals
Extracting the distance frequency component and the speed frequency component with non-zero values, and calculating according to the formula (2),
x＝Rsinθ,y＝Rcosθ (6)
namely radar space point cloud information (x, y, z coordinates).
4. The obtained point cloud (x, y, z) is subjected to K-mean (K-mean) clustering, and in order to reduce input cost and improve calculation efficiency, a specific algorithm of setting K to 5.K-mean is as follows,
(1) randomly selecting 5 initial point information, named as C1, C2, C3, C4 and C5
(2) Finding Euclidean distances d (X, C) between all points and the initial point, wherein the distance formula is as follows:
Figure BDA0003473702020000101
X＝(x1,x2,x3),C＝(y1,y2,y3)
(3) using the distances obtained above, each point is assigned to the nearest center,
Figure BDA0003473702020000102
(4) recalculating the new center point for the assigned point by the following formula:
Figure BDA0003473702020000103
(5) repeating steps 2 to 4 until all the coordinates of the center point are not changed.
5. And inputting the obtained clustering points (K is 5) into a multi-step forward neural network model, and carrying out normalization prediction to generate more point cloud information. The basic structure of the forward neural network comprises an Input layer, a Hidden layer, an output layer
The forward neural network basic structure comprises an input layer, a hidden layer and an output layer;
the mathematical formula is as follows:
an input layer: xLThe output result X of the L +1 layer is the input data point of the L layerL+1Is ZL＝F(XL)＝WL*XL+bL,
Figure BDA0003473702020000111
J＝||A-XL+1||
Wherein WL,bLThe weight and the deviation parameter of the L-th layer are obtained, A is the real labeled data of the training, and the calculation formula is
Figure BDA0003473702020000112
Figure BDA0003473702020000113
The gradient is calculated as:
Figure BDA0003473702020000114
Figure BDA0003473702020000115
one of the biggest characteristics of the invention is that based on the forward neural network theory, a multistep neural network format of a differential equation is provided, the transmission modeling of the forward neural network is shown as the mathematical differential equation below,
Figure BDA0003473702020000116
the time T is the number of layers of the neural network model. The Adams-Bashforth formula AB (Adams-Bashforth) is called AB format for short, and the differential equation is developed to obtain
Figure BDA0003473702020000117
Wherein λiFor the coefficients, a 3-step AB format was used in the experiments, with the mathematical formula,
Figure BDA0003473702020000118
6. the result is that more dense point cloud information is produced, as shown in fig. 3, and the algorithm flow of the present invention is shown in fig. 1.
The invention also discloses a non-transitory computer readable storage medium, on which a computer program is stored, which when executed by a processor implements the steps of the millimeter wave radar sparse point cloud reconstruction method based on the multi-step neural network structure. It should be noted that the technical features of the electronic device, the storage medium, and the like related to the present patent application should be regarded as the prior art, and the specific structure, the operation principle, the control mode and the spatial arrangement mode of the technical features may be selected conventionally in the field, and should not be regarded as the invention point of the present patent, and the present patent is not further specifically described in detail.
It will be apparent to those skilled in the art that modifications and equivalents may be made in the embodiments and/or portions thereof without departing from the spirit and scope of the present invention.

Claims (8)

1. A millimeter wave radar sparse point cloud reconstruction method based on a multi-step neural network structure is characterized by comprising the following steps:
step S1: obtaining first information including speed, angle and distance through a millimeter wave radar and performing Fourier transform processing on the obtained speed and distance to take a generated speed and distance spectrogram as a first target feature;
step S2: extracting spatial coordinate point cloud information corresponding to the speed and distance spectrogram;
step S3: clustering the spatial coordinate point cloud information to ensure that all the central point coordinates do not change any more, thereby obtaining clustering points;
step S4: the cluster points are input to a multi-step forward neural network model for regression prediction and to generate supplemental spatial coordinate point cloud information to obtain dense point cloud information.
2. The millimeter wave radar sparse point cloud reconstruction method based on the multi-step neural network structure as claimed in claim 1, wherein in step S1, the continuous wave signal is modulated by the radar frequency, and the relation formula of the distance and the speed is as follows:
Figure FDA0003473702010000011
where c is the propagation velocity of the electromagnetic wave in air, R is the distance between the object and the radar, v is the relative velocity, τ is the propagation time, and the ratio of the distance between the object and the radar is calculated to obtain R:
Figure FDA0003473702010000012
3. the millimeter wave radar sparse point cloud reconstruction method based on the multi-step neural network structure as claimed in claim 1, wherein in step S1, the fourier transform processing is performed on the speed and the distance by the following formula:
Figure FDA0003473702010000013
Figure FDA0003473702010000014
wherein, b [ n, m, k]Refers to the nth sampling point, T, of the mth chirp signal received on the kth antennafastA sampling interval for sampling a signal within a sweep period;
performing two-dimensional discrete Fourier transform on N sampling points in a period to obtain corresponding distances, wherein the formula is as follows:
Figure FDA0003473702010000021
Figure FDA0003473702010000022
4. the millimeter wave radar sparse point cloud reconstruction method based on the multi-step neural network structure as claimed in claim 3, wherein in step S2, the distance frequency component and the velocity frequency component with non-zero values are extracted from step S1, and the following are calculated:
x＝Rsinθ,y＝Rcosθ，
so as to obtain the spatial coordinate point cloud information of the millimeter wave radar.
5. The millimeter wave radar sparse point cloud reconstruction method based on the multi-step neural network structure as claimed in claim 4, wherein in step S3, K-means clustering processing is performed on the spatial coordinate point cloud information, and the method is specifically implemented as the following steps:
step S3.1: randomly selecting 5 initial point information, namely C1, C2, C3, C4 and C5;
step S3.2: finding Euclidean distances between all points of the spatial coordinate point cloud information and the initial point information, wherein the distance formula is as follows:
Figure FDA0003473702010000023
X＝(x1,x2,x3),C＝(y1,y2,y3)；
step S3.3: and allocating the point corresponding to each piece of spatial coordinate point cloud information to the center closest to the acquired distance:
Figure FDA0003473702010000024
step S3.4: recalculating the assigned points for new center points by the following formula:
Figure FDA0003473702010000031
step S3.5: step S3.2-step S3.4 are repeated to keep the coordinates of all the center points from changing.
6. The millimeter wave radar sparse point cloud reconstruction method based on the multi-step neural network structure as claimed in claim 5, wherein in step S4, the obtained clustering points are input into the multi-step forward neural network model, and the method is implemented as the following steps:
step S4.1: the multi-step forward neural network model comprises an input layer, a hidden layer and an output layer, and X isLSet as the input data point of the L-th layer, the output result X of the L + 1-th layerL+1Comprises the following steps:
ZL＝F(XL)＝WL*XL+bL；
Figure FDA0003473702010000032
J＝||A-XL+1||；
wherein, WL,bLAnd A is the real labeled data of training, and the calculation formula is as follows:
Figure FDA0003473702010000033
Figure FDA0003473702010000034
step S4.2: the gradient is calculated as:
Figure FDA0003473702010000035
Figure FDA0003473702010000036
step S4.3: according to the multi-step neural network format of the differential equation, the transfer of the forward neural network is modeled as a mathematical differential equation, and the formula is as follows:
Figure FDA0003473702010000037
wherein: t is the number of layers of the neural network model, and the differential equation is expanded to obtain:
Figure FDA0003473702010000041
wherein: lambda [ alpha ]iAre coefficients.
7. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the millimeter wave radar sparse point cloud reconstruction method based on the multi-step neural network structure according to any one of claims 1 to 6 when executing the program.
8. A non-transitory computer readable storage medium, on which a computer program is stored, wherein the computer program, when executed by a processor, implements the steps of the millimeter wave radar sparse point cloud reconstruction method based on the multi-step neural network structure according to any one of claims 1 to 6.
CN202210049948.0A 2022-01-17 2022-01-17 Millimeter wave radar sparse point cloud reconstruction method based on multi-step neural network structure Pending CN114418068A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210049948.0A CN114418068A (en) 2022-01-17 2022-01-17 Millimeter wave radar sparse point cloud reconstruction method based on multi-step neural network structure

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210049948.0A CN114418068A (en) 2022-01-17 2022-01-17 Millimeter wave radar sparse point cloud reconstruction method based on multi-step neural network structure

Publications (1)

Publication Number Publication Date
CN114418068A true CN114418068A (en) 2022-04-29

Family

ID=81272618

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210049948.0A Pending CN114418068A (en) 2022-01-17 2022-01-17 Millimeter wave radar sparse point cloud reconstruction method based on multi-step neural network structure

Country Status (1)

Country Link
CN (1) CN114418068A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117388836A (en) * 2023-10-27 2024-01-12 深圳承泰科技有限公司 Signal processing method of multiple wave-transmitting modes capable of increasing number of point clouds

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117388836A (en) * 2023-10-27 2024-01-12 深圳承泰科技有限公司 Signal processing method of multiple wave-transmitting modes capable of increasing number of point clouds

Similar Documents

Publication Publication Date Title
Messikommer et al. Event-based asynchronous sparse convolutional networks
CN107481264B (en) Video target tracking method with self-adaptive scale
CN110991272B (en) Multi-target vehicle track recognition method based on video tracking
CN107300698B (en) Radar target track starting method based on support vector machine
Li Research on radar signal recognition based on automatic machine learning
Liu et al. Ship detection and classification on optical remote sensing images using deep learning
CN104794730B (en) SAR image segmentation method based on super-pixel
CN116047427A (en) Small sample radar active interference identification method
CN113960587B (en) Millimeter wave radar multi-target tracking method based on category information feedback
CN114168058A (en) Method and device for recognizing handwritten characters in air by FMCW single millimeter wave radar
CN109461127B (en) SAR image sparse regularization feature enhancement method with interpretation as purpose
Zhang et al. Marine target detection for PPI images based on YOLO-SWFormer
Luo et al. Edgeactnet: Edge intelligence-enabled human activity recognition using radar point cloud
CN114418068A (en) Millimeter wave radar sparse point cloud reconstruction method based on multi-step neural network structure
CN115909086A (en) SAR target detection and identification method based on multistage enhanced network
Shaaban et al. Resonate-and-Fire Spiking Neurons for Hand Gesture Label Refinement
CN105223571B (en) The ISAR imaging method significantly paid attention to based on weighting L1 optimization with vision
CN113421281A (en) Pedestrian micromotion part separation method based on segmentation theory
Guo et al. Deep Model Based Road User Classification Using mm-Wave Radar
CN114609604B (en) Unmanned aerial vehicle cluster target detection and target contour and cluster scale estimation method
Ding et al. Ship detection in SAR images based on an improved detector with rotational boxes
CN117111013B (en) Radar target tracking track starting method, device, equipment and medium
CN117292439B (en) Human body posture recognition method and system based on indoor millimeter wave radar
Zeng et al. Follicular ultrasound image segmentation based on improved Deeplabv3
Mohamadi AI-assisted Silhouettes Generation from Sparse mmWave Sampling

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