Automobile electric tail gate control method and system

Abstract

The application provides a kind of automobile electric tailgate control method and system, periodically based on UWB radar acquisition automobile tailgate N frame echo data;For each frame of echo data, the echo data is processed to obtain the range-doppler plot of echo data;The pre-set speed-distance threshold range of kicking is used to cut each frame of echo data range-doppler plot;From the range-doppler plot of each frame of echo data after cutting, select each amplitude greater than the preset amplitude threshold as each target amplitude of each frame of echo data;According to each target amplitude of each frame of echo data, determine the kicking feature of each frame of echo data;Through the pre-trained kicking action recognition model, the kicking feature of each frame of echo data is recognized to obtain the corresponding recognition result;If the recognition result is kicking action, control the opening or closing of automobile tailgate.

Description

Technical Field

[0001] The present invention relates to the field of automobile technology, and more particularly to a method and system for controlling an automobile electric tailgate. Background Art

[0002] At present, cars have become an indispensable means of transportation for people to travel. In order to bring greater convenience to people, many manufacturers have added electric tailgates to their cars.

[0003] At present, the opening methods of electric tailgates of automobiles can be roughly divided into two types. One is a control structure based on the traditional manual button-type opening or closing. This control structure requires the operator to manually press the button to open or close. When the operator's hands are not free, this control method is not applicable and is not convenient enough; the other is a capacitive control structure based on the capacitive electric field. The structure is generally placed under the rear bumper of the vehicle. An emitting electrode of the capacitive electric field is placed at two positions according to a strict installation position and spacing. When an action occurs in the detection area, the capacitance will produce a specific change. According to a certain change rule, it can be determined whether the action is a kicking action, and then the signal quantity can be used to achieve the purpose of controlling the tailgate of the automobile. However, this method is easily affected by factors such as temperature, humidity or dust, resulting in insensitive control and false triggering. Summary of the Invention

[0004] In view of this, the present invention provides a method and system for controlling an electric tailgate of an automobile, with the purpose of improving the control accuracy of the electric tailgate of an automobile.

[0005] A first aspect of the present application provides a method for controlling an electric tailgate of an automobile, the method comprising:

[0006] Periodically collect N frames of echo data from the car's tailgate based on UWB radar;

[0007] For each frame of the echo data, the echo data is processed to obtain a range-Doppler map of the echo data;

[0008] Using a preset kicking speed-distance threshold range, the range-Doppler graph of each frame of echo data is clipped; wherein the clipped range-Doppler graph includes multiple amplitudes;

[0009] Selecting, from the multiple amplitudes of the range-Doppler graph of each frame of echo data after clipping, each amplitude that is greater than a preset amplitude threshold as each target amplitude of the echo data of each frame;

[0010] determining a kicking feature of each frame of echo data according to each target amplitude of each frame of echo data;

[0011] A pre-trained kicking action recognition model is used to identify the kicking features of the echo data of each frame to obtain a corresponding recognition result; wherein the kicking action recognition model is obtained by training a neural network using real kicking training data and non-real kicking training data;

[0012] If the recognition result is a kicking action, the tailgate of the car is controlled to be opened or closed.

[0013] Optionally, the preset kicking speed-distance threshold range includes a kicking speed threshold range and a kicking distance threshold range;

[0014] The step of clipping the range-Doppler graph of each frame of echo data using a preset kicking speed-distance threshold range comprises:

[0015] Determining a row index number of a range-Doppler graph of each frame of the echo data according to the kick distance threshold range;

[0016] Determining a column index number of a range-Doppler graph of each frame of the echo data according to the kicking speed threshold range;

[0017] For each frame of the range-Doppler map of the echo data, the range-Doppler map of the echo data is cropped according to the row index number and the column index number of the range-Doppler map of the echo data.

[0018] Optionally, determining the row index number of the range-Doppler map of each frame of the echo data according to the kick distance threshold range includes:

[0019] For each frame of the range-Doppler map of the echo data, determining a maximum value and a minimum value of the kicking action distance from the range-Doppler map of the echo data according to the kicking distance threshold range;

[0020] determining a maximum row number of a characteristic value in a distance dimension of the range-Doppler map according to a maximum value of the kicking action distance and a distance resolution;

[0021] determining a minimum row number of characteristic values ​​in the distance dimension of the range-Doppler map according to the minimum value of the kicking action distance and the distance resolution;

[0022] A row index number of the range-Doppler map is determined according to the maximum row number and the minimum row number.

[0023] Optionally, the kicking speed threshold range includes a positive threshold range of the speed dimension and a negative threshold range of the speed dimension;

[0024] Determining the column index number of the range-Doppler graph of each frame of the echo data according to the kicking speed threshold range includes:

[0025] For each frame of the range-Doppler map of the echo data, the range-Doppler map of the echo data is divided into two equal columns, with the vertical axis as the y-axis and the rows as the x-axis to construct a two-dimensional array in a spatial xy coordinate system;

[0026] The two-dimensional array is processed according to the positive threshold range of the velocity dimension and the negative threshold range of the velocity dimension to obtain a column index number of the range-Doppler map of the echo data.

[0027] Optionally, the processing the two-dimensional array according to the positive threshold range of the velocity dimension and the negative threshold range of the velocity dimension to obtain the column index number of the range-Doppler map includes:

[0028] determining a maximum value of the first kicking action speed and a minimum value of the first kicking action speed from the two-dimensional array according to the positive threshold range of the speed dimension;

[0029] determining a maximum value and a minimum value of the second kicking action speed from the two-dimensional array according to the negative threshold range of the speed dimension;

[0030] Determining a first maximum column number of eigenvalues ​​within a positive threshold range of a velocity dimension of the two-dimensional array based on a maximum value of the first kicking motion velocity and a velocity resolution, and determining a first minimum column number of eigenvalues ​​within a positive threshold range of the velocity dimension of the two-dimensional array based on a minimum value of the first kicking motion velocity and a velocity resolution;

[0031] Determining a second maximum column number of eigenvalues ​​within a negative threshold range of the velocity dimension of the two-dimensional array based on a maximum value of the second kicking motion velocity and a velocity resolution, and determining a second minimum column number of eigenvalues ​​within a negative threshold range of the velocity dimension of the two-dimensional array based on a minimum value of the second kicking motion velocity and a velocity resolution;

[0032] A column index number of the range-Doppler map is determined according to the first maximum column number, the second maximum column number, the first minimum column number, and the second minimum column number.

[0033] Optionally, determining the kicking feature of each frame of echo data according to each target amplitude of each frame of echo data includes:

[0034] Calculating the weighted Doppler, weighted positive Doppler, weighted negative Doppler and weighted distance of each frame of echo data according to each target amplitude of each frame of echo data;

[0035] For each frame of echo data, the weighted Doppler, weighted positive Doppler, weighted negative Doppler and weighted distance of the echo data are spliced ​​to obtain the kicking feature of the echo data.

[0036] Optionally, the step of training a neural network using real kicking training data and non-real kicking training data to obtain a kicking action recognition model includes:

[0037] Acquire real kicking training data and non-real kicking training data; wherein the real kicking training data includes N frames of first historical echo data with kicking action; the non-real kicking training data includes N frames of second historical echo data without kicking action;

[0038] Processing each frame of the first historical echo data to obtain a first historical range-Doppler map of each frame of the first historical echo data, and processing each frame of the second historical echo data to obtain a second historical range-Doppler map of each frame of the second historical echo data;

[0039] cropping a first historical range-Doppler graph of each frame of the first historical echo data and a second historical range-Doppler graph of each frame of the second historical echo data using a preset kicking speed-distance threshold range; wherein the cropped first historical range-Doppler graph includes a plurality of first historical amplitudes, and the cropped second range-Doppler graph includes a plurality of second historical amplitudes;

[0040] From the plurality of first historical amplitudes of the first historical range-Doppler of each frame of the trimmed first historical echo data, selecting each first historical amplitude greater than a preset threshold as each target first historical amplitude of each frame of the first historical echo data;

[0041] From the plurality of second historical amplitudes of the second historical range-Doppler of each frame of the cropped second historical echo data, selecting each second historical amplitude greater than the preset threshold as each target second historical amplitude of each frame of the second historical echo data;

[0042] determining a first historical kick feature of each frame of the first historical echo data according to each target first historical amplitude of each frame of the first historical echo data;

[0043] determining a second historical kick feature of each frame of the second historical echo data according to each target second historical amplitude of each frame of the second historical echo data;

[0044] A neural network is trained using the first historical kicking features of each frame of the first historical echo data and the second historical kicking features of each frame of the second historical echo data to obtain a kicking action recognition model.

[0045] A second aspect of the present application provides a vehicle tailgate control system, the system comprising:

[0046] A periodic acquisition unit, used to periodically acquire N frames of echo data from the tailgate of a car based on the UWB radar;

[0047] a first processing unit, configured to process the echo data for each frame of the echo data to obtain a range-Doppler map of the echo data;

[0048] A first cropping unit is configured to crop the range-Doppler graph of each frame of the echo data using a preset kicking speed-distance threshold range; wherein the cropped range-Doppler graph includes a plurality of amplitudes;

[0049] A first selection unit is configured to select, from a plurality of amplitudes of the range-Doppler map of each frame of echo data after clipping, respective amplitudes that are greater than a preset amplitude threshold as respective target amplitudes of the echo data of each frame;

[0050] a first determining unit, configured to determine a kicking feature of each frame of echo data according to each target amplitude of each frame of echo data;

[0051] a recognition unit configured to recognize kicking features of the echo data of each frame using a pre-trained kicking action recognition model to obtain a corresponding recognition result; wherein the kicking action recognition model is obtained by training a neural network using real kicking training data and non-real kicking training data by the training unit;

[0052] A control unit is used to control the opening or closing of the car tailgate if the recognition result is a kicking action.

[0053] Optionally, the preset kicking speed-distance threshold range includes a kicking speed threshold range and a kicking distance threshold range, and the first clipping unit includes:

[0054] a second determining unit, configured to determine a row index number of the range-Doppler graph of each frame of the echo data according to the kick distance threshold range;

[0055] a third determining unit, configured to determine a column index number of the range-Doppler map of each frame of the echo data according to the kicking speed threshold range;

[0056] The first cropping subunit is configured to crop the range-Doppler map of each frame of the echo data according to the row index number and the column index number of the range-Doppler map of the echo data.

[0057] Optionally, the second determining unit includes:

[0058] a fourth determining unit, configured to determine, for each frame of the range-Doppler map of the echo data, a maximum value and a minimum value of the kicking action distance from the range-Doppler map of the echo data according to the kicking distance threshold range;

[0059] a fifth determining unit, configured to determine a maximum row number of characteristic values ​​in the distance dimension of the range-Doppler map according to the maximum value of the kicking action distance and the distance resolution;

[0060] a sixth determining unit, configured to determine a minimum row number of characteristic values ​​in the distance dimension of the range-Doppler map according to the minimum value of the kicking action distance and the distance resolution;

[0061] A seventh determining unit is configured to determine a row index number of the range-Doppler map according to the maximum row number and the minimum row number.

[0062] The present invention provides a method and system for controlling an electric tailgate of an automobile. A kick recognition model is obtained by training a neural network using real kick training data and non-real kick training data. The system then processes each frame of echo data collected from the tailgate of an automobile using a UWB radar to obtain a range-Doppler map of each frame of echo data. The model then clips the range-Doppler map of each frame of echo data according to a preset kick speed-distance threshold range to remove unnecessary background information from the range-Doppler map of each frame of echo data. Information is collected and redundancy in the distance-Doppler map is reduced. If a kicking action really exists at present, the features of the kicking action in the distance-Doppler map can also be highlighted. The kicking features are constructed using the cropped distance-Doppler map of each frame of echo data, and the pre-trained kicking action recognition model is used to identify whether there is a kicking action at present according to the constructed kicking features. It can be seen that the present invention does not need to place a capacitive electric field emitting electrode at two positions according to a strict installation position and spacing, and can ensure the accuracy of kicking action recognition in various environments, and has strong anti-interference ability. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0064] Figure 1 A schematic flow chart of a method for controlling an electric tailgate of an automobile provided by an embodiment of the present invention;

[0065] Figure 2 A flowchart of a method for training a kicking action recognition model provided by an embodiment of the present invention;

[0066] Figure 3 The present invention provides a schematic structural diagram of an electric tailgate control system for an automobile. DETAILED DESCRIPTION

[0067] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0068] In this application, relational terms such as first and second, etc. are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus comprising the element.

[0069] See also Figure 1 , shows a flow chart of a method for controlling an electric tailgate of an automobile provided by an embodiment of the present invention, the method specifically comprising the following steps:

[0070] S101: Periodically collect N frames of echo data from the car based on the UWB radar.

[0071] In the embodiment of the present invention, the period of a complete kicking action is generally longer than one frame of echo data. Therefore, an ultra-wideband (UWB) radar can periodically collect N frames of echo data and their timestamps from the tailgate of the vehicle, where N is an integer greater than 1.

[0072] It should be noted that a complete kicking action may be: the UBW radar kicks from left to right, or from right to left, or from directly below the radar to directly above the radar, which is not limited in this embodiment of the present invention.

[0073] S102: Processing the echo data for each frame to obtain a range-Doppler map of the echo data.

[0074] In the specific process of executing step S102 , for each frame of first echo data, a range-Doppler fast Fourier transform may be performed on the first echo data to obtain a first range-Doppler map of the echo data.

[0075] It should be noted that the data form corresponding to the range-Doppler diagram of each frame of echo data can be mathematically expressed as an m*n two-dimensional matrix. Each value in the matrix represents the amplitude of a "grid", m represents the total number of "grids" in the distance dimension, n represents the total number of "grids" in the velocity dimension, and one "grid" represents a specific resolution size.

[0076] It is worth noting that, given a certain waveform and various hardware and software components of the UBW radar, the range and velocity resolutions of the echo data collected by the radar are also certain. For example, if the velocity resolution is determined to be DopplerRes, then n×DopplerRes can represent its maximum detection velocity. Similarly, if the range resolution is determined to be RangeRes, then the maximum detection range can be expressed as n×RangeRes.

[0077] S103: cropping the range-Doppler graph of each frame of echo data using a preset kicking speed-distance threshold range; wherein the cropped range-Doppler graph includes multiple amplitudes.

[0078] In an embodiment of the present invention, the applicant has discovered through research that a kicking action has a certain movement trend, and a kicking action is usually within a certain speed and distance range. Therefore, a corresponding kicking speed-distance threshold range can be pre-set, wherein the kicking speed-distance threshold range can include a kicking speed threshold range and a kicking distance threshold range; the pre-set kicking speed threshold range and kicking distance threshold range can be used to crop the distance-Doppler map of each frame of echo data to remove unnecessary background information in the distance-Doppler map of each frame of echo data.

[0079] It should be noted that unnecessary background information can be interpreted as signal data outside the range of the kicking action. Since the UWB radar can detect speed and distance within a certain range, the range-Doppler map may include some signal data outside the range of the kicking action. For example, the UWB radar here can detect a maximum speed of 10m / s and a distance of 10m. The echo data collected by the UWB radar after 2DFFT is a 10x10 matrix. The 10 grids in the row direction represent 10m, that is, one grid represents 1m. However, the kicking action may only be within 1-2m of the radar (the kicking threshold range). Therefore, the range-Doppler map can be cropped according to the kicking speed threshold range and the kicking distance threshold range to remove unnecessary background information in the range-Doppler map.

[0080] Optionally, the process of cropping the distance-Doppler graph of each frame of echo data using a pre-set kicking speed-distance threshold range can be specifically as follows: determining the row index number of the distance-Doppler graph of each frame of echo data according to the kicking distance threshold range; determining the column index number of the distance-Doppler graph of each frame of echo data according to the kicking speed threshold range; and for the distance-Doppler graph of each frame of echo data, cropping the distance-Doppler graph of the echo data according to the row index number and column index number of the distance-Doppler graph of the echo data.

[0081] As an implementation method of an embodiment of the present invention, the process of determining the row index number of the distance-Doppler map of each frame of the echo data according to the kick distance threshold range can be specifically as follows: for each frame of the distance-Doppler map of the echo data, according to the kick distance threshold range, determining the maximum value of the kicking action distance and the minimum value of the kicking action distance from the distance-Doppler map of the echo data; determining the maximum row number of the eigenvalue in the distance dimension of the distance-Doppler map according to the maximum value of the kicking action distance and the distance resolution; determining the minimum row number of the eigenvalue in the distance dimension of the distance-Doppler map according to the minimum value of the kicking action distance and the distance resolution; and determining the row index number of the distance-Doppler map according to the maximum row number and the minimum row number.

[0082] In actual application, the corresponding range resolution RangeRes can be pre-set according to the waveform and software and hardware of the UBW radar; if the pre-set kick distance threshold range is [RangeThresMin, RangeThresMax], the value corresponding to RangeThresMax in the range-Doppler map can be determined as the maximum value of the kick action distance, and the value corresponding to RangeThresMin in the range-Doppler map can be determined as the minimum value of the kick action distance; the value obtained by dividing the minimum value of the kick action distance by the range resolution is rounded down to obtain the minimum row number of the eigenvalue in the distance dimension of the range-Doppler map, as shown in formula (1); the value obtained by dividing the maximum value of the kick action distance by the range resolution is rounded down to obtain the maximum row number of the eigenvalue in the distance dimension of the range-Doppler map, as shown in formula (2).

[0083]

[0084]

[0085] Among them, RangeThresMin1 is the minimum value of the kicking action distance, RangeThresMax1 is the maximum value of the kicking action distance, MinRow is the minimum row number, MaxRow is the maximum row number, and floor() is rounded down.

[0086] It should be noted that after determining the row index number of the range-Doppler map, the row index number can be used to crop the range-Doppler map from the row of the range-Doppler map, as shown in formula (3); it is worth noting that the range-Doppler map can be imagined as a two-dimensional array (a matrix of m*n rows), and then a small matrix can be cropped from this two-dimensional array (a matrix of m*n rows) according to the row index number.

[0087] The cut matrix is ​​the one in formula (3)

[0088]

[0089] in, is the range-Doppler map after clipping using the row index number.

[0090] Optionally, the kick speed threshold range includes a positive threshold range of the velocity dimension and a negative threshold range of the velocity dimension, and the process of determining the column index number of the range-Doppler map of each frame of the echo data according to the kick speed threshold range can be specifically: for the range-Doppler map of each frame of the echo data, the range-Doppler map is constructed into a two-dimensional array in a spatial xy coordinate system in a column 2 equal division manner with the vertical axis as the y axis and the horizontal axis as the x axis; the two-dimensional array is processed according to the positive threshold range of the velocity dimension and the negative threshold range of the velocity dimension to obtain the column index number of the range-Doppler map of the echo data.

[0091] In the embodiment of the application, the applicant has found that there is a positive and negative distinction in the velocity dimension of the range-Doppler map, so the range-Doppler map can be constructed into a two-dimensional array in a spatial xy coordinate system in a column 2 equal division manner with the vertical axis as the y axis and the horizontal axis as the x axis, so as to determine the maximum value of the first kick action speed and the minimum value of the first kick action speed from the two-dimensional array according to the positive threshold range of the velocity dimension, determine the maximum value of the second kick action speed and the minimum value of the second kick action speed from the two-dimensional array according to the negative threshold range of the velocity dimension, determine the first maximum column number of the characteristic value in the positive threshold range of the velocity dimension of the two-dimensional array according to the maximum value of the first kick action speed and the velocity resolution, and determine the first minimum column number of the characteristic value in the positive threshold range of the velocity dimension of the two-dimensional array according to the minimum value of the first kick action speed and the velocity resolution, determine the second maximum column number of the characteristic value in the negative threshold range of the velocity dimension of the two-dimensional array according to the maximum value of the second kick action speed and the velocity resolution, and determine the second minimum column number of the characteristic value in the negative threshold range of the velocity dimension of the two-dimensional array according to the minimum value of the second kick action speed and the velocity resolution, and determine the column index number of the range-Doppler map according to the first maximum column number, the second maximum column number, the first minimum column number and the second minimum column number.

[0092] In actual application, the distance-Doppler graph is divided into two equal parts by columns, and a two-dimensional array in a spatial xy coordinate system is constructed with the vertical axis as the y-axis and the rows as the x-axis; if the positive threshold range of the velocity dimension is [DopplerThresMin, DopplerThresMax], since the maximum positive and negative Doppler velocities of the kick are roughly equal, the negative threshold range of the velocity dimension can be [-DopplerThresMin, -DopplerThresMax]; the value corresponding to DopplerThresMax in the two-dimensional array can be determined as the maximum value of the first kicking action speed, and the value corresponding to DopplerThresMin in the two-dimensional array can be determined as the minimum value of the first kicking action speed; the value corresponding to -DopplerThresMax in the two-dimensional array can be determined as the maximum value of the second kicking action speed, and the value corresponding to -DopplerThresMin in the two-dimensional array can be determined as the minimum value of the second kicking action speed.

[0093] The minimum value of the first kicking action speed is divided by the speed resolution, and the value obtained is rounded down to obtain the first minimum row number of the eigenvalue in the speed dimension of the two-dimensional array. The specific calculation formula can be found in formula (4). The maximum value of the first kicking action speed is divided by the speed resolution, and the value obtained is rounded down to obtain the first maximum row number of the eigenvalue in the speed dimension of the two-dimensional array. The specific calculation formula can be found in formula (5). The minimum value of the second kicking action speed is divided by the speed resolution, and the value obtained is rounded down to obtain the second minimum row number of the eigenvalue in the speed dimension of the two-dimensional array. The specific calculation formula can be found in formula (4). The maximum value of the second kicking action speed is divided by the speed resolution, and the value obtained is rounded down to obtain the second maximum row number of the eigenvalue in the speed dimension of the two-dimensional array. The specific calculation formula can be found in formula (5). Finally, [first maximum column number, first minimum column number] and [second maximum column number, second minimum column number] are concatenated to obtain the column index number of the range-Doppler map.

[0094]

[0095] Among them, DopplerMin1 is the minimum value of the first kicking action speed / the minimum value of the second kicking action speed, DopplerMax1 is the maximum value of the first kicking action speed / the maximum value of the second kicking action speed, Mincolumn is the first minimum column number / the second minimum column number, and Maxcolumn is the first maximum column number / the second maximum column number.

[0096] It should be noted that after the column index number of the range-Doppler map is determined, the range-Doppler map after being cropped by the row index number can be cropped by the column index number, and the specific cropping manner can be referred to formula (6); it should be noted that the range-Doppler map after being cropped by the row index number can be imagined as a two-dimensional array (matrix) of rows, and then a small matrix is cropped from the two-dimensional array (matrix) of rows according to the row index number, and the cropped matrix is in formula (6)

[0097]

[0098] It should be further noted that since the Doppler velocity has positive and negative, the range-Doppler map f n =matrix m×n The vertical axis equally divided according to column 2 is the y axis, and the x axis is constructed by taking the maximum row axis as the x axis to construct a two-dimensional array in the space xy coordinate system, so as to obtain the range-Doppler map after being cropped according to a certain speed-distance threshold range according to formula (4)-(6)

[0099] S104: Select each amplitude greater than the preset amplitude threshold value from the multiple amplitudes of the range-Doppler map of each frame of the cropped echo data as each target amplitude of each frame of the echo data.

[0100] In the embodiment of the application, the applicant has found that when the UWB radar detects a kicking action, according to the detection principle of the radar, there must be a strong amplitude change in the "lattice" greater than a certain speed and distance dimension, so different amplitude threshold values can be set according to different UWB radar hardware and software properties, so that after the range-Doppler map of each frame of echo data is cropped, a preset amplitude threshold value matched with the currently used UWB radar can be determined from the preset amplitude threshold values, and each target amplitude of each frame of the echo data can be selected from the multiple amplitudes of the range-Doppler map of each frame of the cropped echo data.

[0101] As an implementation manner of the embodiment of the application, the preset amplitude threshold value can be a topk threshold value, and then each amplitude greater than the topk threshold value can be selected from the multiple amplitudes of the range-Doppler map of the cropped echo data, and the selected amplitudes are sorted from large to small, and finally the top K amplitudes are arranged in the front as the K target amplitudes of the range-Doppler map of the echo data, and the two-dimensional coordinates of the selected K target amplitudes are C = {(x 1， y1),(x 2， ​​​y2),...,(x k， y k )}, where K is greater than 2.

[0102] S105: Determine the kicking feature of each frame of echo data according to each target amplitude of each frame of echo data.

[0103] In the specific process of executing step S105, after selecting the target amplitudes of each frame of echo data, for each frame of echo data, multiple characteristic values ​​of the echo data can be calculated based on the target amplitudes of the echo data, and the kicking characteristics of the echo data can be determined based on the characteristic values ​​of the echo data.

[0104] Optionally, the process of determining the kicking feature of each frame of echo data based on the target amplitudes of each frame of echo data can be specifically as follows: according to the target amplitudes of each frame of echo data, calculating the weighted Doppler (weighted doppler), weighted positive Doppler (weighted positive doppler), weighted negative Doppler (weighted negative doppler) and weighted distance of each frame of echo data; for each frame of echo data, splicing the weighted Doppler, weighted positive Doppler, weighted negative Doppler and weighted distance of the echo data to obtain the kicking feature of the echo data.

[0105] In some embodiments, for each frame of echo data, the x-coordinate, positive x-coordinate, negative x-coordinate, y-coordinate, positive y-coordinate and negative y-coordinate of each target amplitude of the echo data on the distance-Doppler diagram of the corresponding cropped echo data can be determined, and the weighted Doppler, weighted positive Doppler, weighted negative Doppler and weighted distance of the echo data can be calculated based on the respective target amplitudes of the echo data and their x-coordinate, positive x-coordinate, negative x-coordinate, y-coordinate, positive y-coordinate and negative y-coordinate. The specific calculation method is shown in Formula (7) to Formula (10).

[0106]

[0107]

[0108] Among them, WeightedDoppler, WeightedPosDoppler, WeightNegDoppler, and WeightedRange represent weighted Doppler, weighted positive Doppler, weighted negative Doppler, and weighted range, respectively. i y i ) indicates that according to the coordinate (x i ,y i ) in the cropped range-Doppler map The kth amplitude value taken out from i 、xposi 、x negi 、y i 、y posi and y negi Indicates that the i-th amplitude is The x coordinate under the coordinates, positive x coordinate, negative x coordinate and the y coordinate, positive y coordinate, negative y coordinate.

[0109] In actual application, after obtaining the WeightedDoppler, WeightedPosDoppler, WeightNegDoppler, and WeightedRange of each frame of echo data, the WeightedDoppler, WeightedPosDoppler, WeightNegDoppler, and WeightedRange of each frame of echo data can be used to generate the topk feature point set F N =(f1, f2, ..., f N ), where f N =(WeightedDoppler, WeightedPosDoppler, WeightNegDoppler, WeightedRange), and then f can be N The WeightedDoppler, WeightedPosDoppler, WeightNegDoppler, and WeightedRange in the function are concatenated into a feature vector of length N*4 to represent f. N A complete kick feature.

[0110] S106: Recognize the kicking features of the echo data of each frame through a pre-trained kicking action recognition model to obtain a corresponding recognition result; wherein, the kicking action recognition model is obtained by training a neural network using real kicking training data and non-real kicking training data.

[0111] In the specific process of executing step S106, after obtaining the kicking features of each frame of echo data, the kicking features of each frame of echo data can be input into the pre-trained kicking action recognition model, so that the kicking action recognition model uses the input kicking features of each frame of echo data to perform kicking action recognition and obtain corresponding recognition results; wherein, the recognition result can be a kicking action or a non-kicking action.

[0112] In an embodiment of the present invention, the kicking action recognition model is obtained by training a neural network using real kicking training data and non-real kicking training data.

[0113] See also Figure 2, which shows a flow chart of a method for training a kicking action recognition model provided by an embodiment of the present invention. The method for controlling an electric tailgate of an automobile specifically includes the following steps:

[0114] S201: Acquire real kicking training data and non-real kicking training data; wherein, the real kicking training data includes N frames of first historical echo data with kicking action; the non-real kicking training data includes N frames of second historical echo data without kicking action.

[0115] In an embodiment of the present invention, a kick recognition data set for machine learning can be produced, wherein the kick recognition data set includes real kick training data and non-real kick training data; the real kick training data is N frames of first historical echo data and their timestamps with kicking actions collected based on UBW radar; the non-real kick training data is N frames of second historical echo data with no kicking actions collected based on UBW.

[0116] It should be noted that the non-real kick training data may be interference data. For example, the non-real kick training data may be walking data near a car, which is not limited in this embodiment of the present invention.

[0117] S202: Process each frame of the first historical echo data to obtain a first historical range-Doppler map of each frame of the first historical echo data, and process each frame of the second historical echo data to obtain a second historical range-Doppler map of each frame of the second historical echo data.

[0118] In the specific process of executing step S202, for each frame of first historical echo data, a range-dimensional-Doppler-dimensional fast Fourier transform can be performed on the first historical echo data to obtain a first historical range-Doppler map of the first historical echo data; for each frame of second historical echo data, a range-dimensional-Doppler-dimensional fast Fourier transform can be performed on the second historical echo data to obtain a second historical range-Doppler map of the second historical echo data.

[0119] It should be noted that, no matter whether it is the first historical distance-Doppler map or the second historical distance-Doppler map, the corresponding data form can be specifically expressed mathematically as an m*n two-dimensional matrix.

[0120] S203: Using a preset kicking speed-distance threshold range, cropping the first historical range-Doppler graph of each frame of the first historical echo data and the second historical range-Doppler graph of each frame of the second historical echo data; wherein the cropped first historical range-Doppler graph includes multiple first historical amplitudes, and the cropped second range-Doppler graph includes multiple second historical amplitudes.

[0121] In an embodiment of the present invention, the kicking speed-distance threshold range may include a kicking speed threshold range and a kicking distance threshold range, and the pre-set kicking speed threshold range and kicking distance threshold range can be used to crop the first historical range-Doppler map of each frame of the first historical echo data and the second historical range-Doppler map of each frame of the second historical echo data to remove unnecessary background information in each first historical range-Doppler map and each second historical range-Doppler map.

[0122] Optionally, the process of cropping the historical range-Doppler map of each frame of historical echo data using a preset kicking speed-distance threshold range may specifically include: for each frame of the historical range-Doppler map of the historical echo data, determining a historical row index number of the historical range-Doppler map of each frame of the historical echo data based on the kicking distance threshold range, and determining a historical column index number of the historical range-Doppler map of each frame of the historical echo data based on the kicking speed threshold range. The historical range-Doppler map may be the first historical range-Doppler map or the second historical range-Doppler map.

[0123] As an implementation method of an embodiment of the present invention, the process of determining the historical row index number of the historical distance-Doppler graph based on the kick distance threshold range can be specifically as follows: according to the kick distance threshold range, determining the maximum value of the historical kicking action distance and the minimum value of the historical kicking action distance from the historical distance-Doppler graph; according to the maximum value of the historical kicking action distance and the distance resolution, determining the historical maximum row number of the characteristic value on the distance dimension of the historical distance-Doppler graph; according to the minimum value of the kicking action distance and the distance resolution, determining the historical minimum row number of the characteristic value on the distance dimension of the historical distance-Doppler graph; according to the historical maximum row number and the historical minimum row number, determining the historical row index number of the historical distance-Doppler graph.

[0124] In actual application, the corresponding distance resolution can be pre-set according to the waveform and software and hardware of the UBW radar; if the pre-set kick distance threshold range is [RangeThresMin,RangeThresMax], the value corresponding to RangeThresMax in the historical range-Doppler map can be determined as the maximum value of the historical kick action distance, and the value corresponding to RangeThresMin in the historical range-Doppler map can be determined as the minimum value of the historical kick action distance; the value obtained by dividing the maximum value of the historical kick action distance by the distance resolution is rounded down to obtain the historical maximum row number of the characteristic value in the distance dimension of the historical range-Doppler map. The specific calculation method can be found in formula (2) and will not be repeated here; the value obtained by dividing the minimum value of the historical kick action distance by the distance resolution is rounded down to obtain the historical minimum row number of the characteristic value in the distance dimension of the historical range-Doppler map. The specific calculation method can be found in formula (1) and will not be repeated here.

[0125] It should be noted that after determining the historical row index number of the historical distance-Doppler map, the historical row index number can be used to crop the historical distance-Doppler map from the row of the historical distance-Doppler map. The specific cropping method can be referred to formula (3), which will not be described in detail here. It is worth noting that the historical distance-Doppler map can be imagined as a two-dimensional array (a matrix of m*n rows), and then a small matrix can be cropped from this two-dimensional array (a matrix of m*n rows) according to the historical row index number.

[0126] As an implementation method of an embodiment of the present invention, the kicking speed threshold range includes a positive threshold range of the speed dimension and a negative threshold range of the speed dimension. The process of determining the historical column index number of the historical distance-Doppler graph based on the kicking speed threshold range can be specifically as follows: the historical distance-Doppler graph is divided into two equal columns, and the historical distance-Doppler graph is constructed with the vertical axis as the y-axis and the rows as the x-axis in a spatial xy coordinate system; the historical two-dimensional array is processed according to the positive threshold range of the speed dimension and the negative threshold range of the speed dimension to obtain the historical column index number of the historical distance-Doppler graph.

[0127] Optionally, the maximum value of the first historical kicking action speed and the minimum value of the first historical kicking action speed can be determined from the historical two-dimensional array according to the positive threshold range of the speed dimension; the maximum value of the second historical kicking action speed and the minimum value of the second historical kicking action speed can be determined from the historical two-dimensional array according to the negative threshold range of the speed dimension; the first historical maximum column number of the characteristic value on the positive threshold range of the speed dimension of the historical two-dimensional array can be determined according to the maximum value and speed resolution of the first historical kicking action speed, and the first historical maximum column number of the characteristic value on the historical two-dimensional array can be determined according to the minimum value and speed resolution of the first historical kicking action speed. The first historical minimum column number of the eigenvalue on the positive threshold range of the velocity dimension of the group; according to the maximum value and velocity resolution of the second historical kicking action velocity, the second historical maximum column number of the eigenvalue on the negative threshold range of the velocity dimension of the historical two-dimensional array is determined, and according to the minimum value and velocity resolution of the second historical kicking action velocity, the second historical minimum column number of the eigenvalue on the negative threshold range of the velocity dimension of the historical two-dimensional array is determined; according to the first historical maximum column number, the second historical maximum column number, the first historical minimum column number and the second historical minimum column number, the historical column index number of the historical distance-Doppler map is determined.

[0128] In actual application, the positive threshold range of the speed dimension is [DopplerThresMin, DopplerThresMax], and the negative threshold range of the speed dimension is [-DopplerThresMin, -DopplerThresMax]. The value corresponding to DopplerThresMax in the historical two-dimensional array is determined as the maximum value of the first historical kicking movement speed, and the value corresponding to DopplerThresMin in the historical two-dimensional array is determined as the minimum value of the first historical kicking movement speed; the value corresponding to -DopplerThresMax in the historical two-dimensional array is determined as the maximum value of the second historical kicking movement speed, and the value corresponding to -DopplerThresMin in the historical two-dimensional array is determined as the minimum value of the second historical kicking movement speed.

[0129] The value obtained by dividing the maximum value of the first historical kicking action speed by the speed resolution is rounded down to obtain the first historical maximum row number of the eigenvalue on the speed dimension of the two-dimensional array. The specific calculation formula can be found in formula (5), which will not be repeated here. The value obtained by dividing the minimum value of the first historical kicking action speed by the speed resolution is rounded down to obtain the first historical minimum row number of the eigenvalue on the speed dimension of the historical two-dimensional array. The specific calculation formula can be found in formula (4), which will not be repeated here. The value obtained by dividing the maximum value of the second historical kicking action speed by the speed resolution is rounded down to obtain the second historical maximum row number of the eigenvalue on the speed dimension of the historical two-dimensional array. The specific calculation formula can be found in formula (5), which will not be repeated here. The value obtained by dividing the minimum value of the second historical kicking action speed by the speed resolution is rounded down to obtain the second historical minimum row number of the eigenvalue on the speed dimension of the historical two-dimensional array. The specific calculation formula can be found in formula (4), which will not be repeated here. Finally, [the first historical maximum column number, the first historical minimum column number] and [the second historical maximum column number, the second historical minimum column number] are spliced ​​together to obtain the column index number of the historical distance-Doppler map.

[0130] It should be noted that after determining the historical column index number of the historical distance-Doppler map, the historical column index number can be used to crop the historical distance-Doppler map that has been cropped using the historical row index number. The specific cropping method can be found in formula (6), which will not be described in detail here.

[0131] S204: From the multiple first historical amplitudes of the first historical range-Doppler of each frame of the cropped first historical echo data, select each first historical amplitude that is greater than the preset threshold as each target first historical amplitude of each frame of the first historical echo data, and determine the first historical kick feature of each frame of the first historical echo data based on each target first historical amplitude of each frame of the first historical echo data.

[0132] In an embodiment of the present invention, for each frame of first historical echo data, after cropping the range-Doppler plot of the first historical echo data, a preset amplitude threshold matching the UBW radar that collected the first historical echo data can be determined from among various preset amplitude thresholds. First historical amplitudes greater than the preset amplitude threshold are selected from the plurality of first historical amplitudes in the range-Doppler plot of the cropped first historical echo data as the respective target first historical amplitudes of the first historical echo data. Based on the respective target first historical amplitudes of the first historical echo data, multiple feature values ​​of the first historical echo data are calculated, and based on the feature values ​​of the first historical echo data, a first historical kick feature of the first historical echo data is determined.

[0133] Optionally, the process of determining the first historical kick feature of each frame of the first historical echo data based on the first historical amplitudes of each target of each frame of the first historical echo data can be specifically as follows: according to the first historical amplitudes of each target of each frame of the first historical echo data, the first historical weighted Doppler, the first historical weighted positive Doppler, the first historical weighted negative Doppler and the first historical weighted distance of each frame of the first historical echo data are calculated; for each frame of the first historical echo data, the first historical weighted Doppler, the first historical weighted positive Doppler, the first historical weighted negative Doppler and the first historical weighted distance of the first historical echo data are spliced ​​to obtain the first historical kick feature of the first historical echo data.

[0134] It should be noted that, according to the first historical amplitudes of each target in each frame of the first historical echo data, the specific calculation method of calculating the first historical weighted Doppler, the first historical weighted positive Doppler, the first historical weighted negative Doppler and the first historical weighted distance of each frame of the first historical echo data can be referred to Formula (7) to Formula (10), which will not be repeated here.

[0135] S205: From the multiple second historical amplitudes of the second historical range-Doppler of each frame of the cropped second historical echo data, select each second historical amplitude that is greater than the preset threshold as each target second historical amplitude of each frame of the second historical echo data, and determine the second historical kick feature of each frame of the second historical echo data based on each target second historical amplitude of each frame of the second historical echo data.

[0136] In an embodiment of the present invention, for each frame of second historical echo data, after cropping the range-Doppler plot of the second historical echo data, a preset amplitude threshold matching the UBW radar that collected the second historical echo data can be determined from among various preset amplitude thresholds. Second historical amplitudes greater than the preset amplitude threshold are selected from the plurality of second historical amplitudes in the range-Doppler plot of the cropped second historical echo data as the respective target second historical amplitudes of the second historical echo data. Based on the respective target second historical amplitudes of the second historical echo data, multiple feature values ​​of the second historical echo data are calculated, and based on the respective feature values ​​of the second historical echo data, a second historical kick feature of the second historical echo data is determined.

[0137] Optionally, the process of determining the second historical kicking feature of each frame of the second historical echo data according to each target second historical amplitude of each frame of the second historical echo data can be specifically: calculating the second historical weighted Doppler, the second historical weighted positive Doppler, the second historical weighted negative Doppler and the second historical weighted distance of each frame of the second historical echo data according to each target second historical amplitude of each frame of the second historical echo data; for each frame of the second historical echo data, splicing the second historical weighted Doppler, the second historical weighted positive Doppler, the first historical weighted negative Doppler and the second historical weighted distance of the second historical echo data to obtain the second historical kicking feature of the second historical echo data.

[0138] It should be noted that the specific calculation method of calculating the second historical weighted Doppler, the second historical weighted positive Doppler, the second historical weighted negative Doppler and the second historical weighted distance of each frame of the first historical echo data according to each target second historical amplitude of each frame of the second historical echo data can refer to formula (7)-formula (10), which will not be repeated here.

[0139] S206: training the neural network using the first historical kicking feature of each frame of the first historical echo data and the second historical kicking feature of each frame of the second historical echo data to obtain a kicking action recognition model.

[0140] In the specific execution process of step S206, after obtaining the first historical kicking feature of each frame of the first historical echo data and the second historical kicking feature of each frame of the second historical echo data, the first historical kicking feature of each frame of the first historical echo data and the second historical kicking feature of each frame of the second historical echo data can be input into the neural network to be trained, so that the neural network is trained using the input first historical kicking feature of each frame of the first historical echo data and the second historical kicking feature of each frame of the second historical echo data, until the neural network converges, and a kicking action recognition model capable of identifying the kicking action and the non-kicking action is obtained.

[0141] S107: determining whether the recognition result is a kicking action; if the recognition result is a kicking action, executing step S108; if the recognition result is not a kicking action, returning to step S101.

[0142] In the specific execution process of step S107, after identifying the corresponding recognition result according to the kicking feature of each frame of the echo data by the pre-trained kicking action recognition model, it can be further determined whether the obtained recognition result is a kicking action; if so, step S108 can be executed, and in the case that the current automobile tailgate is opened, the automobile tailgate is controlled to be closed; in the case that the current automobile tailgate is closed, the automobile tailgate is controlled to be opened.

[0143] S108: controlling the automobile tail gate to open or close.

[0144] The embodiment of the application is a kind of automobile electric tailgate control method, the neural network is trained through real kick training data and non-real kick training data, obtain a kick identification model that can identify kick action, so as to when based on UWB radar collects N frame echo data of automobile tailgate, each frame echo data is processed, the range-doppler map of each frame echo data is obtained, and each frame echo data is cropped according to the pre-set kick speed-distance threshold range, so as to remove unnecessary background information in the range-doppler map of each frame echo data, reduce the redundancy in the range-doppler map, if there is really a kick action, it can also highlight the characteristics of the kick action in the range-doppler map;The range-doppler map of each frame echo data after cropping is used to construct kick features, and whether there is a kick action at present is identified according to the constructed various kick features by the pre-trained kick action identification model, so it can be seen that the application does not need to place one capacitive electric field transmitting electrode in two positions according to strict installation position and spacing, can ensure the accuracy of kick action recognition in various environments, and has strong anti-interference ability.

[0145] Based on the automobile tailgate control method provided by the above embodiment of the application, correspondingly, the application also provides an automobile tailgate control system, as shown in Figure 3 The system comprises:

[0146] Periodic acquisition unit 31, for periodically collecting N frame echo data of automobile tailgate based on UWB radar;

[0147] The first processing unit 32 is used for processing the echo data for each frame, and obtaining the range-doppler map of the echo data;

[0148] The first clipping unit 33 is used for respectively clipping the range-doppler map of each frame echo data using the pre-set kick speed-distance threshold range;Wherein, the range-doppler map after clipping includes a plurality of amplitude values;

[0149] The first selection unit 34 is used for selecting each amplitude value greater than the preset amplitude threshold value from the plurality of amplitude values of the range-doppler map of each frame echo data after clipping as each target amplitude value of each frame echo data;

[0150] The first determination unit 35 is used for determining the kick feature of each frame echo data according to each target amplitude value of each frame echo data;

[0151] The recognition unit 36 is configured to recognize the kick features of each frame of the echo data according to a pre-trained kick action recognition model to obtain a corresponding recognition result.

[0152] The control unit 37 is configured to control the opening or closing of the automobile tail door if the recognition result is a kick action.

[0153] The above-mentioned embodiments of the present application disclose the specific principles and execution processes of each unit in the automobile tail door control system, which are the same as the automobile tail door control method disclosed in the above-mentioned embodiments of the present application. For details, refer to the corresponding parts in the automobile tail door control method disclosed in the above-mentioned embodiments of the present application, which will not be described here again.

[0154] The embodiment of the present application discloses an automobile electric tail door control system. The neural network is trained by using real kick training data and non-real kick training data to obtain a kick recognition model capable of recognizing kick actions. When N frames of echo data of the automobile tail door are collected based on the UWB radar, each frame of the collected echo data is processed to obtain a range-Doppler graph of each frame of the echo data. The range-Doppler graph of each frame of the echo data is cropped according to a pre-set kick speed-distance threshold range to remove unnecessary background information in the range-Doppler graph of each frame of the echo data and reduce the redundancy in the range-Doppler graph. If there is really a kick action, the characteristics of the kick action in the range-Doppler graph can be highlighted. The range-Doppler graph of each frame of the echo data after the cropping is used to construct a kick feature, and a pre-trained kick action recognition model is used to recognize whether there is a kick action according to each constructed kick feature. As can be seen, the present application does not need to place one capacitive electric field transmitting electrode at each of two positions according to strict installation positions and intervals, and can ensure the accuracy of kick action recognition in various environments and has strong anti-interference ability.

[0155] Optionally, the pre-set kick speed-distance threshold range includes a kick speed threshold range and a kick distance threshold range. The first cropping unit includes:

[0156] The second determination unit is configured to determine a row index number of the range-Doppler graph of each frame of the echo data according to the kick distance threshold range.

[0157] The third determination unit is configured to determine a column index number of the range-Doppler graph of each frame of the echo data according to the kick speed threshold range.

[0158] The first cropping subunit is configured to crop the range-Doppler map of each frame of the echo data according to the row index number and the column index number of the range-Doppler map of the echo data.

[0159] Optionally, the second determining unit includes:

[0160] a fourth determining unit, configured to determine, for each frame of the range-Doppler map of the echo data, a maximum value and a minimum value of the kicking action distance from the range-Doppler map of the echo data according to the kicking distance threshold range;

[0161] a fifth determining unit, configured to determine a maximum row number of characteristic values ​​in the distance dimension of the range-Doppler map according to the maximum value of the kicking action distance and the distance resolution;

[0162] a sixth determining unit, configured to determine a minimum row number of characteristic values ​​in the distance dimension of the range-Doppler map according to the minimum value of the kicking action distance and the distance resolution;

[0163] A seventh determining unit is configured to determine a row index number of the range-Doppler map according to the maximum row number and the minimum row number.

[0164] Optionally, the third determining unit includes:

[0165] a constructing unit, configured to construct a two-dimensional array in a spatial xy coordinate system by dividing the range-Doppler map of each frame of echo data into two equal columns, with the vertical axis being the y-axis and the rows being the x-axis;

[0166] The second processing unit is configured to process the two-dimensional array according to the positive threshold range of the velocity dimension and the negative threshold range of the velocity dimension to obtain a column index number of the range-Doppler map of the echo data.

[0167] Optionally, the second processing unit includes:

[0168] an eighth determining unit, configured to determine a maximum value of the first kicking action speed and a minimum value of the first kicking action speed from the two-dimensional array according to the positive threshold range of the speed dimension;

[0169] a ninth determining unit, configured to determine, from the two-dimensional array, a maximum value of the second kicking action speed and a minimum value of the second kicking action speed according to the negative threshold range of the speed dimension;

[0170] a tenth determining unit, configured to determine a first maximum column number of the eigenvalues ​​within a positive threshold range of the velocity dimension of the two-dimensional array based on a maximum value of the first kicking motion velocity and a velocity resolution, and to determine a first minimum column number of the eigenvalues ​​within the positive threshold range of the velocity dimension of the two-dimensional array based on a minimum value of the first kicking motion velocity and a velocity resolution;

[0171] an eleventh determining unit, configured to determine a second maximum column number of eigenvalues ​​within a negative threshold range of the velocity dimension of the two-dimensional array based on a maximum value of the second kicking motion velocity and a velocity resolution, and to determine a second minimum column number of eigenvalues ​​within a negative threshold range of the velocity dimension of the two-dimensional array based on a minimum value of the second kicking motion velocity and a velocity resolution;

[0172] A twelfth determining unit is configured to determine a column index number of the range-Doppler map according to the first maximum column number, the second maximum column number, the first minimum column number, and the second minimum column number.

[0173] Optionally, the first determining unit includes:

[0174] a first calculation unit, configured to calculate a weighted Doppler, a weighted positive Doppler, a weighted negative Doppler, and a weighted distance of each frame of echo data according to each target amplitude of the echo data of each frame;

[0175] The splicing unit is used to splice the weighted Doppler, weighted positive Doppler, weighted negative Doppler and weighted distance of the echo data for each frame of the echo data to obtain the kick feature of the echo data.

[0176] Optional training sessions including:

[0177] an acquisition unit, configured to acquire real kicking training data and non-real kicking training data; wherein the real kicking training data includes N frames of first historical echo data in which a kicking action is present; and the non-real kicking training data includes N frames of second historical echo data in which a kicking action is not present;

[0178] a third processing unit, configured to process each frame of the first historical echo data to obtain a first historical range-Doppler map of each frame of the first historical echo data, and to process each frame of the second historical echo data to obtain a second historical range-Doppler map of each frame of the second historical echo data;

[0179] a second cropping unit, configured to crop a first historical range-Doppler map of each frame of the first historical echo data and a second historical range-Doppler map of each frame of the second historical echo data using a preset kicking speed-distance threshold range; wherein the cropped first historical range-Doppler map includes a plurality of first historical amplitudes, and the cropped second range-Doppler map includes a plurality of second historical amplitudes;

[0180] a third clipping unit, configured to select, from the plurality of first historical amplitudes of the first historical range-Doppler of each frame of the clipped first historical echo data, each first historical amplitude greater than the preset threshold as each target first historical amplitude of each frame of the first historical echo data;

[0181] A second selection unit is configured to select, from the plurality of second historical amplitudes of the second historical range-Doppler of each frame of the cropped second historical echo data, each second historical amplitude greater than the preset threshold as each target second historical amplitude of each frame of the second historical echo data;

[0182] a thirteenth determining unit, configured to determine a first historical kick feature of each frame of the first historical echo data according to each target first historical amplitude of each frame of the first historical echo data;

[0183] a fourteenth determining unit, configured to determine a second historical kick feature of each frame of the second historical echo data according to the second historical amplitude of each target in each frame of the second historical echo data;

[0184] The training subunit is used to train a neural network using the first historical kicking features of each frame of the first historical echo data and the second historical kicking features of each frame of the second historical echo data to obtain a kicking action recognition model.

[0185] Each embodiment in this specification is described in a progressive manner. The same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments. In particular, for system or system embodiments, since they are basically similar to method embodiments, the description is relatively simple. For relevant parts, refer to the partial description of the method embodiment. The system and system embodiments described above are merely schematic. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. A person of ordinary skill in the art can understand and implement it without making any creative efforts.

[0186] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the components and steps of each example according to their functions. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present invention.

[0187] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

[0188] The above are only preferred embodiments of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for controlling an electric tailgate of an automobile, characterized in that: The method comprises: Periodically collect N frames of echo data from the car's tailgate based on UWB radar; For each frame of the echo data, the echo data is processed to obtain a range-Doppler map of the echo data; Using a preset kicking speed-distance threshold range, the range-Doppler graph of each frame of echo data is clipped; wherein the clipped range-Doppler graph includes multiple amplitudes; Selecting, from the multiple amplitudes of the range-Doppler graph of each frame of echo data after clipping, each amplitude that is greater than a preset amplitude threshold as each target amplitude of the echo data of each frame; determining a kicking feature of each frame of echo data according to each target amplitude of each frame of echo data; A pre-trained kicking action recognition model is used to identify the kicking features of the echo data of each frame to obtain a corresponding recognition result; wherein the kicking action recognition model is obtained by training a neural network using real kicking training data and non-real kicking training data; If the recognition result is a kicking action, controlling the tailgate of the vehicle to open or close; The step of determining the kicking feature of each frame of echo data according to each target amplitude of each frame of echo data includes: Calculating the weighted Doppler, weighted positive Doppler, weighted negative Doppler and weighted distance of each frame of echo data according to each target amplitude of each frame of echo data; For each frame of echo data, the weighted Doppler, weighted positive Doppler, weighted negative Doppler and weighted distance of the echo data are spliced ​​to obtain the kicking feature of the echo data.

2. The method according to claim 1, characterized in that The preset kicking speed-distance threshold range includes a kicking speed threshold range and a kicking distance threshold range; The step of clipping the range-Doppler graph of each frame of echo data using a preset kicking speed-distance threshold range comprises: Determining a row index number of a range-Doppler graph of each frame of the echo data according to the kick distance threshold range; Determining a column index number of a range-Doppler graph of each frame of the echo data according to the kicking speed threshold range; For each frame of the range-Doppler map of the echo data, the range-Doppler map of the echo data is cropped according to the row index number and the column index number of the range-Doppler map of the echo data.

3. The method according to claim 2, characterized in that Determining the row index number of the range-Doppler graph of each frame of the echo data according to the kick distance threshold range includes: For each frame of the range-Doppler map of the echo data, determining a maximum value and a minimum value of the kicking action distance from the range-Doppler map of the echo data according to the kicking distance threshold range; determining a maximum row number of a characteristic value in a distance dimension of the range-Doppler map according to a maximum value of the kicking action distance and a distance resolution; determining a minimum row number of characteristic values ​​in the distance dimension of the range-Doppler map according to the minimum value of the kicking action distance and the distance resolution; A row index number of the range-Doppler map is determined according to the maximum row number and the minimum row number.

4. The method according to claim 2, characterized in that The kicking speed threshold range includes a positive threshold range of the speed dimension and a negative threshold range of the speed dimension; Determining the column index number of the range-Doppler graph of each frame of the echo data according to the kicking speed threshold range includes: For each frame of the range-Doppler map of the echo data, the range-Doppler map of the echo data is divided into two equal columns, with the vertical axis as the y-axis and the rows as the x-axis to construct a two-dimensional array in a spatial xy coordinate system; The two-dimensional array is processed according to the positive threshold range of the velocity dimension and the negative threshold range of the velocity dimension to obtain a column index number of the range-Doppler map of the echo data.

5. The method according to claim 4, characterized in that The processing of the two-dimensional array according to the positive threshold range of the velocity dimension and the negative threshold range of the velocity dimension to obtain the column index number of the range-Doppler map includes: determining a maximum value of the first kicking action speed and a minimum value of the first kicking action speed from the two-dimensional array according to the positive threshold range of the speed dimension; determining a maximum value and a minimum value of the second kicking action speed from the two-dimensional array according to the negative threshold range of the speed dimension; Determining a first maximum column number of eigenvalues ​​within a positive threshold range of a velocity dimension of the two-dimensional array based on a maximum value of the first kicking motion velocity and a velocity resolution, and determining a first minimum column number of eigenvalues ​​within a positive threshold range of the velocity dimension of the two-dimensional array based on a minimum value of the first kicking motion velocity and a velocity resolution; Determining a second maximum column number of eigenvalues ​​within a negative threshold range of the velocity dimension of the two-dimensional array based on a maximum value of the second kicking motion velocity and a velocity resolution, and determining a second minimum column number of eigenvalues ​​within a negative threshold range of the velocity dimension of the two-dimensional array based on a minimum value of the second kicking motion velocity and a velocity resolution; A column index number of the range-Doppler map is determined according to the first maximum column number, the second maximum column number, the first minimum column number, and the second minimum column number.

6. The method according to claim 1, characterized in that The method of training a neural network using real kicking training data and non-real kicking training data to obtain a kicking action recognition model includes: Acquire real kicking training data and non-real kicking training data; wherein the real kicking training data includes N frames of first historical echo data with kicking action; the non-real kicking training data includes N frames of second historical echo data without kicking action; Processing each frame of the first historical echo data to obtain a first historical range-Doppler map of each frame of the first historical echo data, and processing each frame of the second historical echo data to obtain a second historical range-Doppler map of each frame of the second historical echo data; cropping a first historical range-Doppler graph of each frame of the first historical echo data and a second historical range-Doppler graph of each frame of the second historical echo data using a preset kicking speed-distance threshold range; wherein the cropped first historical range-Doppler graph includes a plurality of first historical amplitudes, and the cropped second range-Doppler graph includes a plurality of second historical amplitudes; From the plurality of first historical amplitudes of the first historical range-Doppler of each frame of the trimmed first historical echo data, selecting each first historical amplitude greater than a preset threshold as each target first historical amplitude of each frame of the first historical echo data; From the plurality of second historical amplitudes of the second historical range-Doppler of each frame of the cropped second historical echo data, selecting each second historical amplitude greater than the preset threshold as each target second historical amplitude of each frame of the second historical echo data; determining a first historical kick feature of each frame of the first historical echo data according to each target first historical amplitude of each frame of the first historical echo data; determining a second historical kick feature of each frame of the second historical echo data according to each target second historical amplitude of each frame of the second historical echo data; A neural network is trained using the first historical kicking features of each frame of the first historical echo data and the second historical kicking features of each frame of the second historical echo data to obtain a kicking action recognition model.

7. A car tailgate control system, characterized in that: The system comprises: A periodic acquisition unit, used to periodically acquire N frames of echo data from the tailgate of a car based on the UWB radar; a first processing unit, configured to process the echo data for each frame of the echo data to obtain a range-Doppler map of the echo data; A first cropping unit is configured to crop the range-Doppler graph of each frame of the echo data using a preset kicking speed-distance threshold range; wherein the cropped range-Doppler graph includes a plurality of amplitudes; A first selection unit is configured to select, from a plurality of amplitudes of the range-Doppler map of each frame of echo data after clipping, respective amplitudes that are greater than a preset amplitude threshold as respective target amplitudes of the echo data of each frame; a first determining unit, configured to determine a kicking feature of each frame of echo data according to each target amplitude of each frame of echo data; a recognition unit configured to recognize kicking features of the echo data of each frame using a pre-trained kicking action recognition model to obtain a corresponding recognition result; wherein the kicking action recognition model is obtained by training a neural network using real kicking training data and non-real kicking training data by the training unit; A control unit, configured to control the tailgate of the vehicle to open or close if the recognition result is a kicking action; The first determining unit includes: a first calculation unit, configured to calculate a weighted Doppler, a weighted positive Doppler, a weighted negative Doppler, and a weighted distance of each frame of echo data according to each target amplitude of the echo data of each frame; The splicing unit is used to splice the weighted Doppler, weighted positive Doppler, weighted negative Doppler and weighted distance of the echo data for each frame of the echo data to obtain the kick feature of the echo data.

8. The system according to claim 7, characterized in that The preset kicking speed-distance threshold range includes a kicking speed threshold range and a kicking distance threshold range, and the first clipping unit includes: a second determining unit, configured to determine a row index number of the range-Doppler graph of each frame of the echo data according to the kick distance threshold range; a third determining unit, configured to determine a column index number of the range-Doppler map of each frame of the echo data according to the kicking speed threshold range; The first cropping subunit is configured to crop the range-Doppler map of each frame of the echo data according to the row index number and the column index number of the range-Doppler map of the echo data.

9. The system according to claim 8, characterized in that The second determining unit includes: a fourth determining unit, configured to determine, for each frame of the range-Doppler map of the echo data, a maximum value and a minimum value of the kicking action distance from the range-Doppler map of the echo data according to the kicking distance threshold range; a fifth determining unit, configured to determine a maximum row number of characteristic values ​​in the distance dimension of the range-Doppler map according to the maximum value of the kicking action distance and the distance resolution; a sixth determining unit, configured to determine a minimum row number of characteristic values ​​in the distance dimension of the range-Doppler map according to the minimum value of the kicking action distance and the distance resolution; A seventh determining unit is configured to determine a row index number of the range-Doppler map according to the maximum row number and the minimum row number.

Images