CN112099014A - Road millimeter wave noise model detection and estimation method based on deep learning - Google Patents

Road millimeter wave noise model detection and estimation method based on deep learning Download PDF

Info

Publication number
CN112099014A
CN112099014A CN202010859256.3A CN202010859256A CN112099014A CN 112099014 A CN112099014 A CN 112099014A CN 202010859256 A CN202010859256 A CN 202010859256A CN 112099014 A CN112099014 A CN 112099014A
Authority
CN
China
Prior art keywords
noise
road
layer
characteristic
feature
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.)
Granted
Application number
CN202010859256.3A
Other languages
Chinese (zh)
Other versions
CN112099014B (en
Inventor
刘震宇
陈泽伟
梁进杰
严远鹏
张鑫
刘昊明
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.)
Gree Iot Technology Shenzhen Co ltd
Original Assignee
Guangdong University of Technology
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 Guangdong University of Technology filed Critical Guangdong University of Technology
Priority to CN202010859256.3A priority Critical patent/CN112099014B/en
Publication of CN112099014A publication Critical patent/CN112099014A/en
Application granted granted Critical
Publication of CN112099014B publication Critical patent/CN112099014B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
    • G01S13/88—Radar or analogous systems specially adapted for specific applications
    • G01S13/93—Radar or analogous systems specially adapted for specific applications for anti-collision purposes
    • G01S13/931—Radar or analogous systems specially adapted for specific applications for anti-collision purposes of land vehicles
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
    • G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
    • G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
    • G01S7/415—Identification of targets based on measurements of movement associated with the target
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
    • G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
    • G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
    • G01S7/418—Theoretical aspects
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation

Landscapes

  • Engineering & Computer Science (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Physics & Mathematics (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • General Physics & Mathematics (AREA)
  • Electromagnetism (AREA)
  • Radar Systems Or Details Thereof (AREA)

Abstract

The invention relates to a road millimeter wave noise model detection and estimation method based on deep learning, which comprises the following steps: s1: preprocessing millimeter wave radar echo signals of R road sections; s2: performing time-frequency conversion on the digitized intermediate-frequency signal, and extracting the noise clutter characteristic of the road; s3: processing the noise clutter characteristic to generate a normalized noise clutter characteristic data set; s4: adding road coordinate information and uploading the road coordinate information to a road end server; s5: and the road end server processes the data. And S6, the road end server deeply learns the network training road millimeter wave noise model parameters. The invention can improve the sensitivity and accuracy of the vehicle-mounted radar in detecting the target and enhance the safety in the driving process.

Description

Road millimeter wave noise model detection and estimation method based on deep learning
Technical Field
The invention relates to the technical field of millimeter wave radar application, in particular to a road millimeter wave noise model detection and estimation method based on deep learning.
Background
With the advent of the 5G era, the information era will have seen a great progress. In the field of transportation, advances in information technology can greatly enhance the safety performance of vehicles. ADAS, an advanced driving assistance system, is a research hotspot in the field of automobile unmanned driving at present. The signal processing of the automobile radar is one of the key technologies of the ADAS, and common automobile radars include an ultrasonic radar, a laser radar and a millimeter wave radar. The effective frequency spectrum bandwidth of the millimeter wave radar is 30GHz to 300GHz, the millimeter wave radar has the advantages of short wavelength, small volume, light weight, high precision, capability of working in all weather and small influence of the weather environment.
In the field of radar research, the most important is accurate detection of target signals, including target object information such as target distance, speed, angle and the like, as well as vehicle-mounted millimeter wave radars. At present, a great deal of research work has been carried out on millimeter wave radar signal processing methods such as noise removal, interference suppression and the like. And in the aspect of sensing the whole road environment by adopting the millimeter wave radar signal, corresponding documents and technical data are less. The existing research idea is that a millimeter wave radar receives a radar echo signal, and then the signal is analyzed and processed to extract useful target information.
According to the traditional signal processing thought, when the noise clutter of the road is large, the accuracy of the millimeter wave radar for detecting the target object can be greatly reduced, and the false alarm probability of radar detection can be improved along with the increase of interference noise. In the road with high environmental noise, the automatic driving system is influenced by noise clutter of the road environment, if the radar of the vehicle finds a front target vehicle later, the braking distance is short, severe braking in a short time can influence the riding comfort of a driver and people in the vehicle, and even serious traffic accidents can be caused. In order to avoid casualties and property loss caused by traffic accidents, more advanced methods must be adopted to deal with the problem of road environmental noise.
Disclosure of Invention
The invention provides a road millimeter wave noise model detection and estimation method based on deep learning, aiming at overcoming the defects of low sensitivity and accuracy of a vehicle-mounted radar detection target caused by environmental noise in the prior art.
The method comprises the following steps:
s1: preprocessing millimeter wave radar echo signals of R road sections; the preprocessing process comprises the steps of mixing a received signal with a radar local oscillator signal to obtain an intermediate frequency signal, filtering the intermediate frequency signal, and digitizing the filtered intermediate frequency signal through AD conversion.
S2: and performing time-frequency conversion on the digitized intermediate-frequency signal, and extracting the road noise clutter characteristic.
S3: and processing the noise clutter characteristic to generate a normalized noise clutter characteristic data set.
S4: and acquiring road information data, and integrating the road information data and the noise characteristic data.
S5: the road end server constructs a road millimeter wave noise deep learning network model and performs data processing on road information data and noise characteristic data;
s6: deep learning and training a road millimeter wave noise deep learning network model to obtain optimal network model parameters, wherein the noise model is represented by the trained model parameters, and detection and estimation of the noise model are realized according to the model parameters.
Preferably, S2 includes the steps of:
s2.1: taking radar echo intermediate frequency signals of an R road section from the preprocessed millimeter wave radar echo signals of the R road sections;
s2.2, taking the nth Chirp signal, wherein the value range of N is 1-N, and N represents the number of all Chirp signals in the radar echo intermediate frequency signal of the r road section;
s2.3: performing time-frequency conversion on the nth Chirp radar echo intermediate-frequency signal; obtaining a frequency domain signal;
s2.4: extracting road noise clutter characteristics from the obtained frequency domain signals; and storing the noise clutter characteristic data in an array
Figure BDA0002647575900000021
The array subscript R represents the R-th road section, and the value of R is 1 to R; the value of the array superscript i is 1 to 9, and 9 noise clutter characteristic data are represented;
wherein, the road noise clutter characteristic data are respectively: noise clutter characteristic I; noise clutter characteristic II; noise clutter characteristic III;
s2.5: judging whether N is larger than or equal to N, if not, enabling N to be N +1, and returning to S2.2; if so, step S2.6 is performed.
S2.6: frequency domain information noise clutter characteristic array for outputting N Chirps
Figure BDA0002647575900000025
Preferably, S2.4 comprises the steps of:
s2.4.1: calculating noise clutter characteristic I data from the frequency domain signal data:
calculating the mean value of the frequency domain signals:
Figure BDA0002647575900000022
stored in arrays
Figure BDA0002647575900000023
Calculating the polar difference of the frequency domain signals: r ═ Xmax-XminStored in an array
Figure BDA0002647575900000024
Calculating the median of the frequency domain signal: z ═ mean (X)i) (ii) a Stored in arrays
Figure BDA0002647575900000031
Step S2.4.2: calculating noise clutter characteristic II data from the frequency domain signal and the noise clutter characteristic I data:
calculating the variance of the frequency domain signal:
Figure BDA0002647575900000032
stored in arrays
Figure BDA0002647575900000033
Calculating the standard deviation of the frequency domain signals:
Figure BDA0002647575900000034
stored in arrays
Figure BDA0002647575900000035
Calculating the average deviation of the frequency domain signals:
Figure BDA0002647575900000036
stored in arrays
Figure BDA0002647575900000037
Step S2.4.3: calculating noise clutter characteristic III data from the frequency domain signal, the noise clutter characteristic I data and the noise clutter characteristic II data:
calculating the relative average deviation of the frequency domain signals:
Figure BDA0002647575900000038
stored in arrays
Figure BDA0002647575900000039
Calculating the relative standard deviation of the frequency domain signals:
Figure BDA00026475759000000310
stored in arrays
Figure BDA00026475759000000311
Calculating the absolute deviation of the median of the frequency domain signal: MAD meani(|Xi-medianj(Xj) | stored in the array
Figure BDA00026475759000000312
Preferably, S3 includes the steps of:
S3.1：array from S2
Figure BDA00026475759000000313
In the specification, get
Figure BDA00026475759000000314
Normalization processing is carried out to obtain normalized noise clutter characteristic data
Figure BDA00026475759000000315
S3.2: judging whether i is greater than or equal to 9, if not, enabling i to be i +1, and returning to S3.1; if yes, executing S3.3;
s3.3: from normalized noise clutter characteristic data
Figure BDA00026475759000000316
Form a 3 × 3 matrix mn(ii) a Wherein n represents the nth Chirp signal;
s3.4: judging whether N is larger than or equal to N, if not, enabling N to be N +1, and returning to S3.3; if yes, executing step S3.5; this step is performed to obtain N matrices mn；
S3.5: by matrix mnForming a noise clutter characteristic matrix M of length sxsr(ii) a Wherein s has a value of
Figure BDA00026475759000000317
MrThe subscript r of (a) denotes the r-th road segment;
s3.6: judging whether R is greater than or equal to R, if not, making R equal to R +1, and returning to S3.1; if yes, the noise characteristic matrix M is outputrI.e. a noise clutter characteristic data set.
Preferably, the first and second electrodes are formed of a metal,
Figure BDA00026475759000000318
the normalization process comprises the following steps:
s3.1.1: extracting maxima in noise clutter characteristic data
Figure BDA00026475759000000319
S3.1.2: let n be 1, start normalization from the 1 st Chirp signal.
S3.1.3: taking noise clutter characteristic data
Figure BDA00026475759000000320
S3.1.4: according to the normalization formula:
Figure BDA0002647575900000041
calculating a t value;
s3.1.5: the value of t is rounded down and stored in an array
Figure BDA0002647575900000042
Performing the following steps;
s3.1.6: judging whether N is larger than or equal to N, if not, enabling N to be N +1, and returning to S3.1.3; if yes, all are output
Figure BDA0002647575900000043
Preferably, S4 includes the steps of:
s4.1: acquiring road information data;
s4.2: extracting road name rn{ G, S, T }, wherein G represents national road, S represents provincial road, and T represents town road;
extracting vehicle specific position information cl{ ak ± b }, where ak represents the center position of the road segment length responsible for the road end server, and b represents the current position of the vehicle from the road end server;
extracting vehicle direction of travel cd{ E, S, W, N }, where E, S, W, N represents south-east-west-north;
s4.3: integrating road information data Tagr{rn,cl,cd}。
S4.4: integrating road information and noise characteristic data M { Tag }r,Mr}。
S4.5: and uploading the M to a road end server.
Preferably, S5 includes the steps of:
s5.1: fetch data M { Tagr,Mr}。
S5.2: constructing a road millimeter wave noise model deep learning network model; taking a noise characteristic matrix MrAnd a noise characteristic matrix M is formedrInputting the millimeter wave noise model into a deep learning network model of a road for model training;
s5.3: get road information data Tagr{rn,cl,cdAnd the road information data Tagr{rn,cl,cdIs stored in the server.
Preferably, the road millimeter wave noise model deep learning network model in S5.2 is composed of L layers of subnetworks, and each layer of subnetwork includes two parts, namely a deep feature extraction subnetwork and a deep feature mapping subnetwork.
Preferably, the deep feature extraction sub-network comprises a four-layer network structure:
the depth feature extraction subnet first sub-layer is used for shallow feature coding; the depth feature extraction subnet second sublayer is used for shallow feature sampling; the third sub-layer of the depth feature extraction sub-network is used for deep feature coding; the fourth sub-layer of the depth feature extraction sub-network is used for sampling the deep features;
the depth feature mapping subnet comprises a five-layer network structure:
a depth feature mapping subnet first sub-layer for initial deep feature reconstruction; the depth feature mapping subnet second sub-layer is used for primary deep feature recovery; the depth feature mapping subnet third sublayer is used for secondary deep feature reconstruction; the depth feature mapping subnet sub-network sub-layer four is used for secondary deep feature recovery; a fifth sublayer of the depth feature mapping subnetwork for mapping the deep features to output data.
Preferably, S6 includes the steps of:
s6.1: initializing deep network parameters: setting the number L of subnet layers and the noise characteristic data training iteration times J of a road millimeter wave noise characteristic deep learning network model; initializing weight parameter w0And a bias parameter b0；
S6.2: matrix noise characteristicsMrInputting the millimeter wave noise characteristic deep learning network model into a road to perform one-time iteration unsupervised training;
outputting deep noise characteristics after finishing training of the j-th subnet
Figure BDA0002647575900000051
Wherein j ∈ [1, L ]]Calculating the weight w of each layer of subnetjAnd bias bjTaking the data as the input data of the j +1 th sub-network, and continuing to train the j +1 th sub-network of the next layer;
s6.3: noise characteristic matrix MrInputting the depth feature to extract a first subnet;
the depth feature extraction first sub-network comprises a first feature coding layer, a first sampling layer, a second feature coding layer and a second sampling layer;
first eigen coding layer to noise characteristic matrix MrPerforming primary feature extraction to obtain a shallow noise feature map
Figure BDA0002647575900000052
First sampling layer to shallow noise characteristic diagram
Figure BDA0002647575900000053
Carrying out feature sampling to enhance noise features; the second characteristic coding layer carries out secondary characteristic extraction on the shallow noise characteristic diagram to obtain a deep noise characteristic diagram
Figure BDA0002647575900000054
The second sampling layer performs secondary characteristic sampling on the deep noise characteristic diagram and outputs the deep noise characteristic diagram with enhanced noise characteristics
Figure BDA0002647575900000055
S6.4: deep noise signature
Figure BDA0002647575900000056
Inputting the depth characteristics of the first subnet to map the first subnet;
the first subnet depth feature mapping first subnet comprises a feature first reconstruction layer, a first up-sampling layer, a feature second reconstruction layer, a second up-sampling layer and a feature mapping layer;
feature first reconstruction layer and feature second reconstruction layer versus deep noise feature map
Figure BDA0002647575900000057
And performing feature reconstruction twice, performing feature recovery twice on the depth noise feature map by the first up-sampling layer and the second up-sampling layer, and alternately and repeatedly performing feature reconstruction and feature recovery.
S6.5: the feature mapping layer reconstructs the recovered deep noise features
Figure BDA0002647575900000058
According to the set size of the feature map, mapping the deep noise features output by the jth sub-network
Figure BDA0002647575900000059
Calculating the weight w of the first subnet1And bias b1Outputting to the next layer of sub-network;
s6.6: judging whether j is greater than or equal to L, if so, executing step S6.8, otherwise, executing step S67;
s6.7: inputting data into a sub-network layer j +1 and returning to S6.3;
s6.8: taking noise characteristic data stored in a road end server as a label to perform supervised training, and adjusting each layer of subnet network parameters; error e is recovered according to noise characteristics of each layer of sub-networkjDifference, adjusting weight w of each layer of sub-networkjAnd bias bj；
S6.9: determining a reconstruction error ejWhether or not it is less than or equal to the minimum reconstruction error ejminIf yes, executing S6.10, if no, executing S6.3;
s6.10: determining an optimal network layer structure and optimal parameters of each layer; complete training and output noise characteristics
Figure BDA0002647575900000061
Compared with the prior art, the technical scheme of the invention has the beneficial effects that:
according to the method, the noise characteristic parameters under different road conditions are obtained and uploaded to the road end server, and the complex noise model is obtained in the road end server by applying deep learning.
The complex road millimeter wave noise model acquired by the road-side server can be provided for vehicles passing the road, the vehicles acquire road millimeter wave noise model parameters in advance, the vehicle-mounted radar detection algorithm is updated, the sensitivity and the accuracy of the vehicle-mounted radar detection target are improved, and the safety in the driving process is further enhanced.
By applying the method, the vehicle-mounted radar can more accurately detect the position, the speed and the angle of the vehicle on the road ahead, and meanwhile, the vehicle-mounted system can remind a driver of slowly decelerating in advance, so that the riding comfort of passengers in the vehicle is improved.
Drawings
Fig. 1 is a flowchart of a deep learning-based road millimeter wave noise model detection and estimation method according to embodiment 1.
Fig. 2 is a flowchart of a method of extracting road clutter characteristics.
Fig. 3 is a flow chart for generating a normalized clutter characteristic data set.
Fig. 4 is a flow chart of a noise clutter characteristic normalization algorithm.
Fig. 5 is a flow chart of adding a road information mark.
Fig. 6 is a flow chart of the way-end server processing data.
FIG. 7 is a diagram of a road millimeter wave noise feature deep learning network model structure.
FIG. 8 is a flow chart of training a road millimeter wave noise model.
Detailed Description
The drawings are for illustrative purposes only and are not to be construed as limiting the patent;
for the purpose of better illustrating the embodiments, certain features of the drawings may be omitted, enlarged or reduced, and do not represent the size of an actual product;
it will be understood by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted.
The technical solution of the present invention is further described below with reference to the accompanying drawings and examples.
Example 1:
example 1
The embodiment provides a deep learning-based road millimeter wave noise model detection and estimation method, as shown in fig. 1, the method includes the following steps:
s1: preprocessing millimeter wave radar echo signals of R road sections; the preprocessing process comprises the steps of mixing a received signal with a radar local oscillator signal to obtain an intermediate frequency signal, filtering the intermediate frequency signal, and digitizing the filtered intermediate frequency signal through AD conversion;
s2: performing time-frequency conversion on the digitized intermediate-frequency signal, and extracting the noise clutter characteristic of the road;
s3: processing the noise clutter characteristic to generate a normalized noise clutter characteristic data set;
s4: acquiring road information data, and integrating the road information data and noise characteristic data;
s5: constructing a road millimeter wave noise deep learning network model, and performing data processing on road information data and noise characteristic data;
s6: deep learning and training of a road millimeter wave noise deep learning network model to obtain optimal network model parameters, and detection and estimation of the noise model are achieved according to the model parameters.
According to the embodiment, the millimeter wave noise model of the road is extracted, the vehicle-mounted radar obtains the environmental noise parameters of the current road in advance, and corresponding algorithm processing is carried out before the vehicle enters the specific road condition, so that the vehicle target detection sensitivity is improved, the driving safety performance of the vehicle is improved, and the comfort of passengers in the driving process is improved on the basis of reducing the occurrence of traffic accidents.
Example 2:
the method for detecting and estimating the millimeter wave noise model of the road based on deep learning provided by the embodiment is consistent with embodiment 1, and only the steps are further limited.
Step S1: and preprocessing the millimeter wave radar echo signals of the R road sections.
Step S2: and performing time-frequency conversion on the digitized intermediate-frequency signal, and extracting the road noise clutter characteristic.
Step S3: and processing the noise clutter characteristic to generate a normalized noise clutter characteristic data set.
Step S4: and adding the road coordinate information and uploading the road coordinate information to a road end server.
Step S5: and the road end server processes the data.
Step S6: and the road end server deeply learns the network training road millimeter wave noise model parameters.
As shown in fig. 2, the step 2 of extracting the road noise clutter characteristic specifically includes the following steps:
step S21: and taking N Chirp signals in total of radar echo intermediate frequency signals of the r-th road section.
And step S22, taking the nth Chirp signal, wherein the value range of N is 1 to N, and the N represents a total of N Chirp signals.
Step S23: and performing time-frequency conversion on the nth Chirp radar echo intermediate-frequency signal.
Step S24: extracting road noise clutter characteristics from the obtained frequency domain signals; the road noise clutter characteristic data are respectively as follows: noise clutter characteristic i: mean, range, median; noise clutter characteristic ii: variance, standard deviation, mean deviation; noise clutter characteristic iii: relative mean deviation, relative standard deviation, median absolute deviation.
Step S25: storing and array of noise clutter characteristic data
Figure BDA0002647575900000081
The array subscript R represents the R-th road section, and the value of R is 1 to R, which represents that the number of the R road sections is total; the value of the array superscript i is 1 to 9, and the data represent 9 noise clutter characteristic data.
Step S26: judging whether N is greater than or equal to N, if not, executing step S27; if yes, go to step S28.
Step S27: the index value n is n + 1.
Step S28: frequency domain information noise clutter characteristic array for outputting N Chirps
Figure BDA0002647575900000082
The step S24 extracts the frequency domain signal noise clutter characteristic data and the step S25 stores the noise clutter characteristic in the array
Figure BDA0002647575900000083
The method comprises the following steps:
step S241: calculating noise clutter characteristic I data from frequency domain signal data, and averaging:
Figure BDA0002647575900000084
extremely poor: r ═ Xmax-Xmin(ii) a Median: z ═ mean (X)i) (ii) a Are respectively stored in an array
Figure BDA0002647575900000085
Figure BDA0002647575900000086
Step S242: calculating noise clutter characteristic II data from the frequency domain signal and the noise clutter characteristic I data, wherein the variance is as follows:
Figure BDA0002647575900000087
standard deviation:
Figure BDA0002647575900000088
average deviation:
Figure BDA0002647575900000089
are respectively stored in an array
Figure BDA00026475759000000810
Step S243: calculating noise clutter from frequency domain signal and noise clutter characteristic I data and noise clutter characteristic II dataWave characteristics iii data, relative mean deviation:
Figure BDA00026475759000000811
relative standard deviation:
Figure BDA00026475759000000812
median absolute deviation: MAD meani(|Xi-medianj(Xj) I)); are respectively stored in an array
Figure BDA00026475759000000813
As shown in fig. 3, the noise clutter characteristic is further processed in step S3 to generate a normalized noise clutter characteristic data set. The method comprises the following steps:
step S31: taking the array obtained in the last step
Figure BDA0002647575900000091
Step S32: get
Figure BDA0002647575900000092
And (6) carrying out normalization processing.
Step S33: judging whether i is greater than or equal to 9, if not, executing step S311; if yes, go to step S34. Step S34: from normalized noise clutter characteristic data
Figure BDA0002647575900000093
Form a 3 × 3 matrix mn. Where n denotes the nth Chirp signal.
Step S35: judging whether N is greater than or equal to N, if not, executing step S310; if yes, go to step S36. This step is performed to obtain N matrices mn。
Step S36: by matrix mnForming a noise clutter characteristic matrix M of length sxsr. Wherein s has a value of
Figure BDA0002647575900000094
MrUnder (2) isThe symbol r indicates the r-th road segment.
Step S37: judging whether R is greater than or equal to R, if not, executing step S39; if yes, go to step S38.
Step S38: output noise characteristic matrix Mr。
Step S39: the index value r is r + 1.
Step S310: the index value n is n + 1.
Step S311: the index value i ═ i + 1.
As shown in fig. 4, the step S32 is implemented by
Figure BDA0002647575900000095
And (6) carrying out normalization processing. The method comprises the following steps:
step S321: extracting a maximum value in noise clutter characteristic data
Figure BDA0002647575900000096
Step S322: the index value is assigned n to 1, and normalization is started from the 1 st Chirp signal.
Step S323: taking noise clutter characteristic data
Figure BDA0002647575900000097
Step S324: according to the normalization formula:
Figure BDA0002647575900000098
and calculating the value of t.
Step S325: the value of t is rounded down and stored in an array
Figure BDA0002647575900000099
In (1).
Step S326: judging whether N is greater than or equal to N, if not, executing step S328 to obtain an index value N which is N + 1; if yes, go to step S327.
Step S327: output all of
Figure BDA00026475759000000910
Step S328: the index value n is n + 1.
As shown in fig. 5, the step S4 of adding the road coordinate information and uploading the road coordinate information to the road end server includes the following steps:
at step S41, road information data is acquired.
Step S42: extracting road name rn{ G, S, T }, where G represents national road, S represents provincial road, and T represents town road. Extracting vehicle specific position information cl{ ak ± b }, where ak represents the center position of the road end server responsible for the length of the road segment, and b represents the current position of the vehicle from the road end server. Extracting vehicle direction of travel cd{ E, S, W, N }, where E, S, W, N represents the south-east-west-north.
Step S43 integrating road information data Tagr{rn,cl,cd}。
Step S44 of integrating road information and noise characteristic data M { Tag }r,Mr}。
And step S45, uploading the M to a road end server.
As shown in fig. 6, the step S5 of the way-end server processing data includes the following steps:
step S51, data M { Tag) uploaded by the vehicle is fetchedr,Mr}。
Step S52, taking noise characteristic matrix Mr。
Step S53, getting road information data Tagr{rn,cl,cd}。
Step S54 noise characteristic matrix MrInputting the data into a deep learning network for model training.
Step S55 road information data Tagr{rn,cl,cdIs stored in the server.
As shown in fig. 7, the noise characteristic matrix M of step S54rInputting the data into a deep learning network for model training. The structure of the road millimeter wave noise model feature deep learning network is as follows: the road millimeter wave noise feature deep learning network model is composed of L layers of subnetworks, wherein each layer of subnetwork comprises a deep feature extraction subnetwork and a deep feature mapping subnetworkAnd (4) partial.
The depth feature extraction subnet comprises four-layer network structures:
the depth feature extraction subnet first sub-layer has the main function of shallow feature coding; the depth feature extraction subnet second sublayer, the main function is shallow feature sampling; the third sub-layer of the depth feature extraction sub-network has the main function of deep feature coding. The fourth sub-layer of the depth feature extraction sub-network has the main function of deep feature sampling. For example, the depth feature extraction subnet may be set to: the convolution filter comprises a first convolution active layer, a first sampling layer, a second convolution active layer and a second sampling layer.
The depth feature mapping subnet comprises a five-layer network structure:
the depth feature maps a first sub-layer of the subnet, and the main function is primary deep feature reconstruction; the depth feature maps a second sub-layer of the subnet, and the main function is primary deep feature recovery; the depth feature maps a third sub-layer of the subnet, and the main function is secondary deep feature reconstruction; the depth feature maps the fourth sub-layer of the subnet, and the main function is secondary deep feature recovery; the fifth sub-layer of the depth feature mapping sub-network has the main function of mapping deep features into output data. For example, the depth feature mapping subnet may be set to: the system comprises a first deconvolution active layer, a first up-sampling layer, a second deconvolution active layer, a second up-sampling layer and a third deconvolution layer.
As shown in fig. 8, in the step S6, the road millimeter wave noise feature deep learning network model trains road millimeter wave noise model parameters. The method comprises the following steps:
step S61, initializing deep network parameters. Setting the number L of subnet layers and the noise characteristic data training iteration times J of a road millimeter wave noise characteristic deep learning network model; initializing weight parameter w0And a bias parameter b0。
Step S62 noise characteristic matrix MrAnd inputting the data into a road millimeter wave noise characteristic deep learning network model to perform one-time iteration unsupervised training. Outputting a deep noise characteristic diagram after the training of the j-th subnet in the middle is finished
Figure BDA00026475759000001120
Wherein j ∈ [1, L ]]Calculating the weight w of each layer of subnetjAnd bias bjAnd as the input data of the j +1 th sub-network, continuing to train the j +1 th sub-network of the next layer.
Step S63 noise characteristic matrix Mr(28 × 28 × 1) is input to the depth feature extraction first subnet. Convolution active layer 1 vs. noise feature matrix MrPerforming primary feature extraction (28 × 28 × 1) to obtain a shallow noise feature map
Figure BDA0002647575900000111
(28X 16). First sampling layer to shallow noise characteristic diagram
Figure BDA0002647575900000112
(28 x 16) feature sampling to enhance shallow noise features
Figure BDA0002647575900000113
(14X 16). The convolution activation layer 2 carries out feature extraction again on the shallow noise feature map to obtain a deep noise feature map
Figure BDA0002647575900000114
(14X 8). The second sampling layer performs secondary characteristic sampling on the deep noise characteristic diagram and outputs the deep noise characteristic diagram with enhanced noise characteristics
Figure BDA0002647575900000115
(7X 8). Wherein the convolutional layer output can be expressed by the following formula:
Figure BDA0002647575900000116
w is a convolution kernel matrix which is successfully initialized, b is a bias parameter matrix matched with the convolution kernel, BN (-) is a batch normalization function, sigma (-) is an activation function, and h (M)r) For the convolution kernel w with the input MrAnd (4) outputting the feature map after convolution.
Step S64 deep noise feature map
Figure BDA0002647575900000117
(7 x 8) the input to the first subnet depth feature maps the first subnet. Deconvolution active layer 1 vs. deep noise signature
Figure BDA0002647575900000118
Performing primary reconstruction to obtain a deep noise characteristic reconstruction picture (7 multiplied by 8)
Figure BDA0002647575900000119
(7 × 7 × 8), the first upsampling layer reconstructs the deep noise features
Figure BDA00026475759000001110
(7 x 8) performing feature recovery to obtain a deep noise feature recovery map
Figure BDA00026475759000001111
(14X 8). Deconvolution active layer 2 to deep noise feature recovery map
Figure BDA00026475759000001112
Reconstructing again (14 × 14 × 8) to obtain a deep noise feature reconstruction image
Figure BDA00026475759000001113
(14 x 16), the second upsampling layer reconstructs the deep noise features
Figure BDA00026475759000001114
(14 × 14 × 16) performing feature recovery to obtain deep noise feature recovery map
Figure BDA00026475759000001115
(28×28×16)。
Step S65, the feature mapping layer restores the deep noise features to the graph
Figure BDA00026475759000001116
(28 x 16) mapping deep noise features of a single channel
Figure BDA00026475759000001117
(28X 1). Wherein, it is reversedThe convolutional layer can be represented by the following formula:
Figure BDA00026475759000001118
Figure BDA00026475759000001119
an image is reconstructed for the feature map. Calculating the weight w of the first subnet1And bias b1And outputting to the next layer of subnet.
And step S66, judging whether j is greater than or equal to L, if so, executing step S68, and if not, executing step S67.
Step S67 data is input into the sub-net layer j + 1.
Step S68: and taking noise characteristic data stored in the route end server as a label to perform supervised training, and finely adjusting each layer of subnet network parameters. Error e is recovered according to noise characteristics of each layer of sub-networkjWeight w of each layer of subnetwork is differentially adjustedjAnd bias bj。
Step S69, judging the reconstruction error ejWhether or not it is equal to or greater than the minimum reconstruction error ejmjnIf yes, step S610 is executed, and if no, step S63 is executed.
Step S610: and determining an optimal network layer structure and optimal parameters of each layer. Complete training and output noise characteristics
Figure BDA0002647575900000121
The present embodiment can detect noise model parameters of various road environments. Vehicles passing through the road section can download road millimeter wave noise model parameters from a road end server, and the road environment of the current driving road and the noise level of the current driving road are sensed through algorithm processing. The vehicle-mounted system updates the vehicle-mounted radar detection algorithm according to the road millimeter wave noise model, so that the overall anti-interference performance of the vehicle-mounted radar system is improved, the missing rate is reduced, the identification precision of radar detection targets is improved, and traffic accidents are reduced. This embodiment can let the driver make sufficient prejudgement to the place ahead road condition under the prerequisite that improves vehicle security of traveling, alleviates the control speed of a motor vehicle, improves passenger's driving comfort in the car.
The terms describing positional relationships in the drawings are for illustrative purposes only and are not to be construed as limiting the patent;
it should be understood that the above-described embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.

Claims (10)

1. A road millimeter wave noise model detection and estimation method based on deep learning is characterized by comprising the following steps:
s1: preprocessing millimeter wave radar echo signals of R road sections; the preprocessing process comprises the steps of mixing a received signal with a radar local oscillator signal to obtain an intermediate frequency signal, filtering the intermediate frequency signal, and digitizing the filtered intermediate frequency signal through AD conversion;
s2: performing time-frequency conversion on the digitized intermediate-frequency signal, and extracting the noise clutter characteristic of the road;
s3: processing the noise clutter characteristic to generate a normalized noise clutter characteristic data set;
s4: acquiring road information data, and integrating the road information data and noise characteristic data;
s5: the road end server constructs a road millimeter wave noise deep learning network model and performs data processing on road information data and noise characteristic data;
s6: deep learning and training of a road millimeter wave noise deep learning network model to obtain optimal network model parameters, and detection and estimation of the noise model are achieved according to the model parameters.
2. The deep learning-based road millimeter wave noise model detection and estimation method according to claim 1, wherein S2 includes the steps of:
s2.1: taking radar echo intermediate frequency signals of an R road section from the preprocessed millimeter wave radar echo signals of the R road sections;
s2.2, taking the nth Chirp signal, wherein the value range of N is 1-N, and N represents the number of all Chirp signals in the radar echo intermediate frequency signal of the r road section;
s2.3: performing time-frequency conversion on the nth Chirp radar echo intermediate-frequency signal; obtaining a frequency domain signal;
s2.4: extracting road noise clutter characteristics from the obtained frequency domain signals; and storing the noise clutter characteristic data in an array
Figure FDA0002647575890000011
The array subscript R represents the R-th road section, and the value of R is 1 to R; the value of the array superscript i is 1 to 9, and 9 noise clutter characteristic data are represented;
wherein, the road noise clutter characteristic data are respectively: noise clutter characteristic I; noise clutter characteristic II; noise clutter characteristic III;
s2.5: judging whether N is larger than or equal to N, if not, enabling N to be N +1, and returning to S2.2; if yes, executing step S2.6;
s2.6: frequency domain information noise clutter characteristic array for outputting N Chirps
Figure FDA0002647575890000012
3. The deep learning-based road millimeter wave noise model detection and estimation method according to claim 2, wherein S2.4 comprises the following steps:
s2.4.1: calculating noise clutter characteristic I data from the frequency domain signal data:
calculating the mean value of the frequency domain signals:
Figure FDA0002647575890000021
stored in arrays
Figure FDA0002647575890000022
Calculating the polar difference of the frequency domain signals: r ═ Xmax-XminStored in an array
Figure FDA0002647575890000023
Calculating the median of the frequency domain signal: z ═ mean (X)i) (ii) a Stored in arrays
Figure FDA0002647575890000024
Step S2.4.2: calculating noise clutter characteristic II data from the frequency domain signal and the noise clutter characteristic I data:
calculating the variance of the frequency domain signal:
Figure FDA0002647575890000025
stored in arrays
Figure FDA0002647575890000026
Calculating the standard deviation of the frequency domain signals:
Figure FDA0002647575890000027
stored in arrays
Figure FDA0002647575890000028
Calculating the average deviation of the frequency domain signals:
Figure FDA0002647575890000029
stored in arrays
Figure FDA00026475758900000210
Step S2.4.3: calculating noise clutter characteristic III data from the frequency domain signal, the noise clutter characteristic I data and the noise clutter characteristic II data:
calculating the relative average deviation of the frequency domain signals:
Figure FDA00026475758900000211
stored in arrays
Figure FDA00026475758900000212
Calculating the relative standard deviation of the frequency domain signals:
Figure FDA00026475758900000213
stored in arrays
Figure FDA00026475758900000214
Calculating the absolute deviation of the median of the frequency domain signal: MAD meani(|Xi-medianj(Xj) | stored in the array
Figure FDA00026475758900000215
4. The deep learning-based road millimeter wave noise model detection and estimation method according to claim 2 or 3, characterized in that S3 includes the steps of:
s3.1: array from S2
Figure FDA00026475758900000216
In the specification, get
Figure FDA00026475758900000217
Normalization processing is carried out to obtain normalized noise clutter characteristic data
Figure FDA00026475758900000218
S3.2: judging whether i is greater than or equal to 9, if not, enabling i to be i +1, and returning to S3.1; if yes, executing S3.3;
s3.3: by normalizing noiseClutter characteristic data
Figure FDA00026475758900000219
Form a 3 × 3 matrix mn(ii) a Wherein n represents the nth Chirp signal;
s3.4: judging whether N is larger than or equal to N, if not, enabling N to be N +1, and returning to S3.3; if yes, executing step S3.5; this step is performed to obtain N matrices mn；
S3.5: by matrix mnForming a noise clutter characteristic matrix M of length sxsr(ii) a Wherein s has a value of
Figure FDA00026475758900000220
MrThe subscript r of (a) denotes the r-th road segment;
s3.6: judging whether R is greater than or equal to R, if not, making R equal to R +1, and returning to S3.1; if yes, the noise characteristic matrix M is outputrI.e. a noise clutter characteristic data set.
5. The deep learning-based road millimeter wave noise model detection and estimation method according to claim 4, characterized in that,
Figure FDA0002647575890000031
the normalization process comprises the following steps:
s3.1.1: extracting maxima in noise clutter characteristic data
Figure FDA0002647575890000032
S3.1.2: let n be 1, start normalization from the 1 st Chirp signal;
s3.1.3: taking noise clutter characteristic data
Figure FDA0002647575890000033
S3.1.4: according to the normalization formula:
Figure FDA0002647575890000034
calculating a t value;
s3.1.5: the value of t is rounded down and stored in an array
Figure FDA0002647575890000035
Performing the following steps;
s3.1.6: judging whether N is larger than or equal to N, if not, enabling N to be N +1, and returning to S3.1.3; if yes, all are output
Figure FDA0002647575890000036
6. The deep learning-based road millimeter wave noise model detection and estimation method according to claim 5, wherein S4 includes the steps of:
s4.1: acquiring road information data;
s4.2: extracting road name rn{ G, S, T }, wherein G represents national road, S represents provincial road, and T represents town road;
extracting vehicle specific position information cl{ ak ± b }, where ak represents the center position of the road segment length responsible for the road end server, and b represents the current position of the vehicle from the road end server;
extracting vehicle direction of travel cd{ E, S, W, N }, where E, S, W, N represents south-east-west-north;
s4.3: integrating road information data Tagr{rn,cl,cd}；
S4.4: integrating road information and noise characteristic data M { Tag }r,Mr}；
S4.5: and uploading the M to a road end server.
7. The deep learning-based road millimeter wave noise model detection and estimation method according to claim 6, wherein S5 includes the steps of:
s5.1: data M { Tag of getting trackr,Mr}；
S5.2: engineering of road millimeter wave noise model depthLearning a network model; taking a noise characteristic matrix MrAnd a noise characteristic matrix M is formedrInputting the millimeter wave noise model into a deep learning network model of a road for model training;
s5.3: get road information data Tagr{rn,cl,cdAnd the road information data Tagr{rn,cl,cdIs stored in the server.
8. The deep learning-based road millimeter wave noise model detection and estimation method according to claim 7, wherein S5.2 the road millimeter wave noise model deep learning network model is composed of L layers of subnetworks, and each layer of subnetwork includes two parts, namely a depth feature extraction subnetwork and a depth feature mapping subnetwork.
9. The deep learning-based road millimeter wave noise model detection and estimation method according to claim 8, wherein the deep feature extraction sub-network comprises a four-layer network structure:
the depth feature extraction subnet first sub-layer is used for shallow feature coding; the depth feature extraction subnet second sublayer is used for shallow feature sampling; the third sub-layer of the depth feature extraction sub-network is used for deep feature coding; the fourth sub-layer of the depth feature extraction sub-network is used for sampling the deep features;
the depth feature mapping subnet comprises a five-layer network structure:
a depth feature mapping subnet first sub-layer for initial deep feature reconstruction; the depth feature mapping subnet second sub-layer is used for primary deep feature recovery; the depth feature mapping subnet third sublayer is used for secondary deep feature reconstruction; the depth feature mapping subnet sub-network sub-layer four is used for secondary deep feature recovery; a fifth sublayer of the depth feature mapping subnetwork for mapping the deep features to output data.
10. The deep learning-based road millimeter wave noise model detection and estimation method according to claim 9, wherein S6 includes the steps of:
s6.1: initializing deep network parameters: setting the number L of subnet layers and the noise characteristic data training iteration times J of a road millimeter wave noise characteristic deep learning network model; initializing weight parameter w0And a bias parameter b0；
S6.2: matrix M of noise characteristicsrInputting the millimeter wave noise characteristic deep learning network model into a road to perform one-time iteration unsupervised training;
outputting deep noise characteristics after finishing training of the j-th subnet
Figure FDA0002647575890000041
Wherein j ∈ [1, L ]]Calculating the weight w of each layer of subnetjAnd bias bjTaking the data as the input data of the j +1 th sub-network, and continuing to train the j +1 th sub-network of the next layer;
s6.3: noise characteristic matrix MrInputting the depth feature to extract a first subnet;
the depth feature extraction first sub-network comprises a first feature coding layer, a first sampling layer, a second feature coding layer and a second sampling layer;
first eigen coding layer to noise characteristic matrix MrPerforming primary feature extraction to obtain a shallow noise feature map
Figure FDA0002647575890000042
First sampling layer to shallow noise characteristic diagram
Figure FDA0002647575890000043
Carrying out feature sampling to enhance noise features; the second characteristic coding layer carries out secondary characteristic extraction on the shallow noise characteristic diagram to obtain a deep noise characteristic diagram
Figure FDA0002647575890000044
The second sampling layer performs secondary characteristic sampling on the deep noise characteristic diagram and outputs the deep noise characteristic diagram with enhanced noise characteristics
Figure FDA0002647575890000045
S6.4: deep noise signature
Figure FDA0002647575890000046
Inputting the depth characteristics of the first subnet to map the first subnet;
the first subnet depth feature mapping first subnet comprises a feature first reconstruction layer, a first up-sampling layer, a feature second reconstruction layer, a second up-sampling layer and a feature mapping layer;
feature first reconstruction layer and feature second reconstruction layer versus deep noise feature map
Figure FDA0002647575890000051
Performing feature reconstruction twice, performing feature recovery twice on the depth noise feature map by the first upper sampling layer and the second upper sampling layer, and alternately and repeatedly performing feature reconstruction and feature recovery;
s6.5: the feature mapping layer reconstructs the recovered deep noise features
Figure FDA0002647575890000052
According to the set size of the feature map, mapping the deep noise features output by the jth sub-network
Figure FDA0002647575890000053
Calculating the weight w of the first subnet1And bias b1Outputting to the next layer of sub-network;
s6.6: judging whether j is greater than or equal to L, if so, executing step S6.8, otherwise, executing step S67;
s6.7: inputting data into a sub-network layer j +1 and returning to S6.3;
s6.8: taking noise characteristic data stored in a road end server as a label to perform supervised training, and adjusting each layer of subnet network parameters; error e is recovered according to noise characteristics of each layer of sub-networkjDifference, adjusting weight w of each layer of sub-networkjAnd bias bj；
S6.9: determining reconstruction errorsejWhether or not it is less than or equal to the minimum reconstruction error ejminIf yes, executing S6.10, if no, executing S6.3;
s6.10: determining an optimal network layer structure and optimal parameters of each layer; complete training and output noise characteristics
Figure FDA0002647575890000054
CN202010859256.3A 2020-08-24 2020-08-24 Road millimeter wave noise model detection estimation method based on deep learning Active CN112099014B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010859256.3A CN112099014B (en) 2020-08-24 2020-08-24 Road millimeter wave noise model detection estimation method based on deep learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010859256.3A CN112099014B (en) 2020-08-24 2020-08-24 Road millimeter wave noise model detection estimation method based on deep learning

Publications (2)

Publication Number Publication Date
CN112099014A true CN112099014A (en) 2020-12-18
CN112099014B CN112099014B (en) 2023-08-22

Family

ID=73754366

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010859256.3A Active CN112099014B (en) 2020-08-24 2020-08-24 Road millimeter wave noise model detection estimation method based on deep learning

Country Status (1)

Country Link
CN (1) CN112099014B (en)

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108169745A (en) * 2017-12-18 2018-06-15 电子科技大学 A kind of borehole radar target identification method based on convolutional neural networks
CN108226892A (en) * 2018-03-27 2018-06-29 天津大学 A kind of radar signal restoration methods under complicated noise based on deep learning
CN108229404A (en) * 2018-01-09 2018-06-29 东南大学 A kind of radar echo signal target identification method based on deep learning
CN108599765A (en) * 2018-04-14 2018-09-28 上海交通大学 The device and method of the noise suppressed distortion correction of analog-digital converter based on deep learning
CN109389058A (en) * 2018-09-25 2019-02-26 中国人民解放军海军航空大学 Sea clutter and noise signal classification method and system
DE102018219255A1 (en) * 2018-11-12 2020-05-14 Zf Friedrichshafen Ag Training system, data set, training method, evaluation device and deployment system for a road vehicle for recording and classifying traffic noise
WO2020139355A1 (en) * 2018-12-27 2020-07-02 Didi Research America, Llc System for automated lane marking
CN111507233A (en) * 2020-04-13 2020-08-07 吉林大学 Multi-mode information fusion intelligent vehicle pavement type identification method

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108169745A (en) * 2017-12-18 2018-06-15 电子科技大学 A kind of borehole radar target identification method based on convolutional neural networks
CN108229404A (en) * 2018-01-09 2018-06-29 东南大学 A kind of radar echo signal target identification method based on deep learning
CN108226892A (en) * 2018-03-27 2018-06-29 天津大学 A kind of radar signal restoration methods under complicated noise based on deep learning
CN108599765A (en) * 2018-04-14 2018-09-28 上海交通大学 The device and method of the noise suppressed distortion correction of analog-digital converter based on deep learning
CN109389058A (en) * 2018-09-25 2019-02-26 中国人民解放军海军航空大学 Sea clutter and noise signal classification method and system
DE102018219255A1 (en) * 2018-11-12 2020-05-14 Zf Friedrichshafen Ag Training system, data set, training method, evaluation device and deployment system for a road vehicle for recording and classifying traffic noise
WO2020139355A1 (en) * 2018-12-27 2020-07-02 Didi Research America, Llc System for automated lane marking
CN111507233A (en) * 2020-04-13 2020-08-07 吉林大学 Multi-mode information fusion intelligent vehicle pavement type identification method

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
C.J.OLIVER等: "基于噪声模型和神经网络的雷达杂波分类", 《现代雷达》, no. 6, pages 51 - 61 *
GUANGYAO ZHAI等: "Millimeter Wave Radar Target Tracking Based on Adaptive Kalman Filter road millimeter wave radar noise-model deep learning", 《2018 IEEE INTELLIGENT VEHICLES SYMPOSIUM (IV)》, pages 453 - 458 *

Also Published As

Publication number Publication date
CN112099014B (en) 2023-08-22

Similar Documents

Publication Publication Date Title
US6420997B1 (en) Track map generator
Bilik et al. GMM-based target classification for ground surveillance Doppler radar
CN110588653A (en) Control system, control method and controller for autonomous vehicle
CN106407931A (en) Novel deep convolution neural network moving vehicle detection method
CN113495251B (en) Object classification using low level radar data
CN110596696B (en) Apparatus and method for improved radar beamforming
DE102018118220B4 (en) Method for estimating the localization quality in the self-localization of a vehicle, device for carrying out method steps of the method, vehicle and computer program
DE102019134351A1 (en) SYSTEM AND METHOD FOR SENSOR DIAGNOSIS
CN110703272B (en) Surrounding target vehicle state estimation method based on vehicle-to-vehicle communication and GMPHD filtering
CN113064193B (en) Combined positioning system based on vehicle road cloud cooperation
Zhang et al. A framework for turning behavior classification at intersections using 3D LIDAR
US20220335279A1 (en) Radar System Using a Machine-Learned Model for Stationary Object Detection
CN106778907A (en) A kind of intelligent travelling crane early warning system based on multi-sensor information fusion
DE102018000517A1 (en) Method for radar-based measurement and / or classification of objects in a vehicle environment
CN115775378A (en) Vehicle-road cooperative target detection method based on multi-sensor fusion
DE112020003963T5 (en) INFORMATION PROCESSING DEVICE, INFORMATION PROCESSING METHOD AND PROGRAM
CN112099014A (en) Road millimeter wave noise model detection and estimation method based on deep learning
DE102020113418B4 (en) AUTONOMOUS DRIVING DEVICE AND METHOD
DE102018216999A1 (en) Process, computer program and measuring system for the evaluation of movement information
CN116224232A (en) Radar-based object height estimation
Susithra et al. Deep learning-based activity monitoring for smart environment using radar
CN114839628A (en) Object detection with multiple distances and resolutions
DE112021005777T5 (en) Systems and methods for range rate dealiasing using position consistency
CN112230216A (en) Vehicle-mounted millimeter wave radar multi-target detection method for cloud control intelligent chassis
EP4174523A1 (en) Electronic device, method for controlling electronic device, and program

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
GR01 Patent grant
GR01 Patent grant
TR01 Transfer of patent right
TR01 Transfer of patent right

Effective date of registration: 20240410

Address after: Room 201, Building A, No. 318 Outer Ring West Road, University City, Panyu District, Guangzhou City, Guangdong Province, 510000

Patentee after: Guangzhou University Town (Guangong) Science and Technology Achievement Transformation Center

Country or region after: China

Address before: 510062 Dongfeng East Road, Yuexiu District, Guangzhou, Guangdong 729

Patentee before: GUANGDONG University OF TECHNOLOGY

Country or region before: China

TR01 Transfer of patent right
TR01 Transfer of patent right

Effective date of registration: 20240425

Address after: 518000 Room 201, building 6, Shenzhen Software Park, No.2, second Gaoxin Road, Maling community, Yuehai street, Nanshan District, Shenzhen City, Guangdong Province

Patentee after: Gree IoT Technology (Shenzhen) Co.,Ltd.

Country or region after: China

Address before: Room 201, Building A, No. 318 Outer Ring West Road, University City, Panyu District, Guangzhou City, Guangdong Province, 510000

Patentee before: Guangzhou University Town (Guangong) Science and Technology Achievement Transformation Center

Country or region before: China