Three-dimensional imaging method and apparatus and three-dimensional imaging device
The present application relates to a three-dimensional imaging method and apparatus, and a 3D imaging device. The method comprises generating 3D image information by capturing a 3D capture area containing a detected object using a depth camera; extracting a mask of the detected object from the 3D image information; determining an imaging area associated with the detected object based on the mask of the detected object; collecting data from a holographic data collection area containing the detected object by a holographic data collection device, generating holographic data; and performing image reconstruction on the imaging area based on the holographic data.
1 . A three-dimensional (3D) imaging method, comprising:
a 3D information obtaining step, for generating 3D image information by capturing a 3D capture area containing a detected object using a depth camera;
a mask extracting step, for extracting a mask of the detected object from the 3D image information;
an imaging area determining step, for determining an imaging area associated with the detected object based on the mask of the detected object, wherein the imaging area is an area corresponding to the mask;
a holographic data obtaining step, for collecting data from a holographic data collection area containing the detected object by a holographic data collection device, to generate holographic data; and
an image reconstructing step, for performing image reconstruction on the imaging area based on the holographic data.
2 . The 3D imaging method of claim 1 , wherein,
the mask extracting step comprises applying an image segmentation algorithm to the 3D image information to extract the mask of the detected object.
3 . The 3D imaging method of claim 2 , wherein
the image segmentation algorithm is a DeepLabV3+ segmentation algorithm based on a neural network.
4 . The 3D imaging method of claim 1 , wherein
a backpropagation imaging algorithm is applied in the image reconstructing step for the image reconstruction.
5 . The 3D imaging method of claim 4 , wherein,
the holographic data collection device comprises a multiple-input multiple-output (MIMO) antenna array, and the MIMO antenna array comprises a plurality of pairs of transmitting and receiving antennas for transmitting and receiving millimeter waves,
the backpropagation imaging algorithm comprises:
calculating, for a pixel point in the imaging area, a sum of distances from the pixel point to each pair of the transmitting and receiving antennas;
obtaining a complex reflectance of each pixel point in the imaging area by performing an operation of summing for all of the transmitting and receiving antennas, as well as transmission frequencies of the millimeter waves, based on the holographic data received by the receiving antenna and the sum of distances; and
creating a reconstructed image based on the complex reflectance calculated for all of the pixel points in the imaging area.
6 . The 3D imaging method of claim 5 , wherein,
the backpropagation imaging algorithm comprises:
calculating, by using formula (1), a sum of distances r T, R from a pixel point (x, y, z) in the imaging area to each pair of the transmitting and receiving antennas;
r T,R =√{square root over (( x−x T ) 2 +( y−y T ) 2 +z 2 )}+√{square root over (( x−x R ) 2 +( y−y R ) 2 +z 2 )} (1)
obtaining, by using formula (2), a complex reflectance Ô (x, y, z) of the pixel point (x, y, z) in the imaging area, based on the holographic data s(x T , y T , x R , t R , k) received by the receiving antenna and the sum r T, R of distances:
O
^
(
x
,
y
,
z
)
=
∑
∀
y
r
∑
∀
x
r
∑
∀
y
t
∑
∀
x
t
∑
∀
k
s
(
x
T
,
y
T
,
x
R
,
y
R
,
k
)
exp
(
jkr
T
,
R
)
(
2
)
wherein (x R , y R , 0) are coordinates of the receiving antenna, (x T , y T , 0) are coordinates of the transmitting antenna, (x, y, z) are coordinates of the pixel points in the imaging area, k is wave number, and k=2π/2, wherein 2 is wavelength; and
creating a reconstructed image based on the complex reflectance calculated for all of the pixel points in the imaging area.
7 . The 3D imaging method of claim 4 , wherein,
the holographic data collection device comprises a multiple-input multiple-output (MIMO) antenna array, and the MIMO antenna array comprises a plurality of pairs of transmitting and receiving antennas for transmitting and receiving millimeter waves,
the backpropagation imaging algorithm is a fast backpropagation imaging algorithm comprising:
calculating, for each pair of the transmitting and receiving antennas, a contribution value of the pair of the transmitting and receiving antennas to each pixel point in the imaging area;
obtaining, for all of the transmitting and receiving antennas, a complex reflectance of each pixel point in the imaging area by performing an operation of summing for all of the transmitting and receiving antennas, based on the contribution value; and
creating a reconstructed image based on the complex reflectance.
8 . The 3D imaging method of claim 7 , wherein,
calculating, for each pair of the transmitting and receiving antennas, the contribution value of the pair of the transmitting and receiving antennas to each pixel point in the imaging area comprises:
calculating, for the pair of the transmitting and receiving antennas, a sum of distances from each pixel point in the imaging area to the pair of the transmitting and receiving antennas;
calculating the contribution value based on the sum of distances and the holographic data received by the receiving antenna.
9 . The 3D imaging method of claim 8 , wherein the fast backpropagation imaging algorithm comprises:
calculating, by using formula (3), for each pair of a transmitting antenna T and a receiving antenna R, a sum r T, R of distances from a pixel point (x, y, z) in the imaging area to the pair of the transmitting and receiving antennas;
r T,R =√{square root over (( x−x T ) 2 +( y−y T ) 2 +z 2 )}+√{square root over (( x−x R ) 2 +( y−y R ) 2 +z 2 )} (3)
calculating, by using formula (4), a contribution value P T,R to the pixel point (x, y, z), based on the sum r T, R of distances and the holographic data s(x T , y T , x R , y R , k);
P T,R =Σ ∀k s ( x T ,y T ,x R ,y R ,k )exp( jkr T,R ) (4)
obtaining, by using formula (5), a complex reflectance Ô (x, y, z) of each pixel point in the imaging area, based on the contribution value P T,R :
Ô ( x,y,z )=Σ ∀y r Σ ∀x r Σ ∀y t Σ ∀x t P T,R (5)
wherein, r T, R is the sum of distances from the pixel point (x, y, z) in the imaging area to each pair of the transmitting and receiving antennas, s(x T , y T , x R , y R, , k) is the holographic data obtained through the pair of the transmitting and receiving antennas, k is wave number, and k=2π/2, wherein λ is wavelength; and
creating a reconstructed image based on the complex reflectance.
10 . The 3D imaging method of claim 7 , wherein,
calculating, for each pair of the transmitting and receiving antennas, the contribution value of the pair of the transmitting and receiving antennas to each pixel point in the imaging area comprises:
classifying the pixel points in the imaging area that have the same sum of distances to the pair of the transmitting and receiving antennas into a group; and
calculating the contribution value of the pair of transmitting and receiving antennas to any one of the pixel points in the group, as the contribution value of each pixel point in the group.
11 . The 3D imaging method of claim 7 , wherein,
calculating, for each pair of the transmitting and receiving antennas, the contribution value of the pair of the transmitting and receiving antennas to each pixel point in the imaging area comprises:
calculating a minimum value and a maximum value of the sum of distances from all of the pixel points in the imaging area to the pair of transmitting and receiving antennas;
dividing a distance between the minimum value and the maximum value into equal intervals, to obtain N equipartition values (r 1 , r 2 , . . . , r N ) of the sum of N distances;
calculating the sum of distances from any one of the pixel points in the imaging area to the pair of transmitting and receiving antennas, determining the nearest equipartition value to the sum of distances, and taking the equipartition value as the sum of distances for the pixel point;
determining the nearest equipartition value for all of the pixel points in the imaging area; and
calculating the contribution value by taking the equipartition value as the sum of the distances.
12 . The 3D imaging method of claim 1 , further comprising:
a key point determining step, for determining a core key point of the detected object based on a deep learning model; and
an imaging area adjusting step, for selecting a portion of the detected object associated with the core key point in the imaging area based on the core key point, so as to adjust the imaging area.
13 . The 3D imaging method of claim 1 , further comprising:
a key point determining step, for determining a core key point of the detected object based on a deep learning model, and
a detected object posture determining step, for determining whether a posture of the detected object is correct based on a position of the core key point, and triggering an alarm under a condition that the posture is not correct.
14 . A 3D imaging system, comprising a depth camera, a holographic data collection device, and a control device, wherein the control device comprises:
a 3D information obtaining module, for controlling the depth camera to capture a 3D capture area containing a detected object to obtain 3D image information;
a mask extracting module, for extracting a mask of the detected object from the 3D image information;
an imaging area determining module, for determining an imaging area associated with the detected object based on the mask of the detected object, wherein the imaging area is an area corresponding to the mask;
a holographic data obtaining module, for controlling the holographic data collection device to collect data from a holographic data collection area containing the detected object so as to obtain holographic data; and
an image reconstructing module, for performing image reconstruction on the imaging area based on the holographic data.
15 . The 3D imaging system of claim 14 , wherein,
the holographic data collection device comprises a multiple-input multiple-output (MIMO) antenna array, and the MIMO antenna array comprises a plurality of pairs of transmitting and receiving antennas for transmitting and receiving millimeter waves.
16 . A 3D imaging device, comprising a memory, a processor, and program instructions stored in the memory and executable by the processor, wherein the processor, when executing the program instructions, implement:
a 3D information obtaining step, for obtaining 3D image information by capturing a 3D capture area containing a detected object using a depth camera;
a mask extracting step, for extracting a mask of the detected object from the 3D image information;
an imaging area determining step, for determining an imaging area associated with the detected object based on the mask of the detected object, wherein the imaging area is an area corresponding to the mask;
a holographic data obtaining step, for obtaining holographic data collected from a holographic data collection area containing the detected object by a holographic data collection device; and
an image reconstructing step, for performing image reconstruction on the imaging area based on the holographic data.