CN112634244A - Three-dimensional complex image processing method and system for target detection - Google Patents

Three-dimensional complex image processing method and system for target detection Download PDF

Info

Publication number
CN112634244A
CN112634244A CN202011580434.5A CN202011580434A CN112634244A CN 112634244 A CN112634244 A CN 112634244A CN 202011580434 A CN202011580434 A CN 202011580434A CN 112634244 A CN112634244 A CN 112634244A
Authority
CN
China
Prior art keywords
image
dimensional
information
amplitude
phase
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202011580434.5A
Other languages
Chinese (zh)
Other versions
CN112634244B (en
Inventor
刘泽鑫
余开
唐彬
柳桃荣
孟祥新
李�诚
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Brainware Terahertz Information Technology Co ltd
Original Assignee
Brainware Terahertz Information Technology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Brainware Terahertz Information Technology Co ltd filed Critical Brainware Terahertz Information Technology Co ltd
Priority to CN202011580434.5A priority Critical patent/CN112634244B/en
Publication of CN112634244A publication Critical patent/CN112634244A/en
Application granted granted Critical
Publication of CN112634244B publication Critical patent/CN112634244B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
    • G01V3/00—Electric or magnetic prospecting or detecting; Measuring magnetic field characteristics of the earth, e.g. declination, deviation
    • G01V3/12—Electric or magnetic prospecting or detecting; Measuring magnetic field characteristics of the earth, e.g. declination, deviation operating with electromagnetic waves
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
    • G01V8/00—Prospecting or detecting by optical means
    • G01V8/005—Prospecting or detecting by optical means operating with millimetre waves, e.g. measuring the black losey radiation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/70—Determining position or orientation of objects or cameras
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning

Abstract

The invention discloses a three-dimensional complex image processing method and a system for target detection, which belong to the technical field of image processing and comprise the following steps: s1: acquiring a two-dimensional amplitude information image; s2: obtaining a target segmentation mask; s3: acquiring a two-dimensional distance information image; s4: acquiring a two-dimensional phase information image; s5: a phase unwinding operation; s6: integrating various kinds of information into the RGB image; s7: and training a target detection model. On the basis of the maximum projection of the three-dimensional complex image, the image contains more spatial information and phase information, the amplitude information, the spatial information and the phase information of the inverted three-dimensional complex image are fully used for target detection of the microwave millimeter wave image, new information input is provided for a target detection model, and the method is worthy of popularization and application.

Description

Three-dimensional complex image processing method and system for target detection
Technical Field
The invention relates to the technical field of image processing, in particular to a three-dimensional complex image processing method and a three-dimensional complex image processing system for target detection.
Background
The microwave millimeter wave three-dimensional holographic imaging security inspection system is an important technical means for active microwave millimeter wave human body security inspection, and the imaging system mainly comprises a planar mechanical scanning type imaging system, a cylindrical surface mechanical scanning type imaging system and a two-dimensional sparse array electronic scanning type imaging system. Electromagnetic waves with microwave and millimeter wave frequencies can penetrate through clothes, so that the obtained three-dimensional holographic image can represent body surface information of a human body more abundantly, dangerous goods hidden under the clothes of the human body can be effectively detected through the three-dimensional holographic image, and the method is an effective new means for human body security inspection.
The microwave millimeter wave three-dimensional holographic imaging system comprises a plurality of imaging systems, and finally, complex reflectivity images are obtained in three-dimensional space grids or space pixel bodies through an inversion algorithm. The microwave and millimeter wave holographic imaging technology measures the signal intensity, but a complex signal containing amplitude and phase information. The existing target detection method generally adopts signal amplitude to detect a target of a microwave millimeter wave image, uses a single-channel gray image as an input image of target detection, and generally obtains amplitude data of an inverted image by processing two methods, namely maximum value projection and standard deviation projection. Since the maximum value projection and the standard deviation projection are only processed according to the amplitude of the image, the spatial information and the phase information in the three-dimensional inversion image are not fully used. Therefore, a three-dimensional complex image processing method and system for target detection are provided.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: how to solve the problem that the prior art does not fully use the amplitude information, the space information and the phase information of an inverted three-dimensional complex image to detect a microwave millimeter wave image target, and provides a three-dimensional complex image processing method for target detection. On the basis of maximum projection of a three-dimensional complex image, the method enables the image to contain more space information and phase information, and provides new information input for a target detection model.
The invention solves the technical problems through the following technical scheme, and the invention comprises the following steps:
s1: obtaining two-dimensional amplitude information image
For microwave millimeter wave three-dimensional complex image O (x)i,yk,zj) X is horizontal dimension, y is vertical dimension, z is space distance dimension, and a maximum projection method is adopted to obtain a two-dimensional amplitude image, namely, any i belongs to [1, N ∈x],k∈[1,Ny]Find a j e [1, N ∈ ]z]Make O (x)i,yk,zj) Obtaining a two-dimensional amplitude information image Amp (x) with the largest amplitude valuei,yk)；
S2: obtaining a target segmentation mask
Segmenting the target and the background noise by an image segmentation method to obtain a target segmentation mask;
s3: obtaining two-dimensional distance information image
When a maximum projection method is adopted to obtain a two-dimensional amplitude information image, i.e. for any i belongs to [1, N ]x],k∈[1,Ny]Find a j e [1, N ∈ ]z]Make O (x)i,yk,zj) The maximum amplitude value is recorded, the distance position j of the maximum amplitude value is recorded, meanwhile, the effective pixel of the distance information image is determined according to the target segmentation mask, and the two-dimensional distance information image Dis (x) is obtainedi,yk)；
S4: obtaining two-dimensional phase information image
Extracting a corresponding three-dimensional complex image O (x) according to the distance position j of the maximum valuei,yk,zj) The data in (1) is arranged into two-dimensional complex image data in a plane, the phase information of the two-dimensional complex image data is calculated, and the distance is determined according to the target segmentation maskObtaining a two-dimensional phase information image Angel (x) from the effective pixels of the information imagei,yk)；
S5: phase unwrapping operation
For phase information Angel (x)i,yk) The phase unwrapping operation is performed because the phase information obtained is one at (-pi, pi)]The value of the interval, called the principal value of the phase or the winding phase, is related to the true phase by an integer multiple of 2 pi. The process of obtaining the true phase from the winding phase is called phase unwrapping;
s6: integrating various information into RGB image
Integrating the amplitude information, the distance information and the phase information into one RGB image data, and calculating to obtain the RGB image simultaneously containing the amplitude information, the distance information and the phase information by the following formula:
RGBImg(xi,yk,1)＝Amp(xi,yk)；
RGBImg(xi,yk,2)＝Amp(xi,yk)*Dis(xi,yk)；
RGBImg(xi,yk,3)＝Amp(xi,yk)*Angel(xi,yk)；
s7: training target detection model
And carrying out target detection labeling on the reconstructed RGB image data, training a preselected target detection deep learning model by combining a loss function and the edge weight, observing the loss of the training set and the verification set, finishing the training after reaching a training stopping condition, and storing and obtaining the target detection model.
Further, in the step S1, a three-dimensional complex image O (x)i,yk,zj) Grid point coordinate of (2) is xi，yk，zj，i∈[0,Nx]，j∈[0,Nz]，k∈[0,Ny]In which N isxNumber of discrete grid points divided for x dimension, NzNumber of discrete grid points divided for distance dimension z, NyDiscrete grid points divided for the y dimension.
Further, in the step S2, for each pixel of the two-dimensional amplitude image, the target segmentation mask is obtained, the background noise pixel is 0, and the target pixel is 1.
Further, in the step S3, for the divided two-dimensional image data obtained by maximum value projection, the spatial position corresponding to the maximum value projection is used as the two-dimensional distance information Dis (x)i,yk)。
Further, in the step S4, for the divided two-dimensional image data obtained by maximum value projection, the phase of the complex data corresponding to the maximum value projection is used as the two-dimensional phase information Angel (x)i,yk)。
Further, in the step S5, the specific process of the phase unwrapping operation is to change the phase information into a signal continuously changing in space by increasing or decreasing 2 pi in value and normalize to (0, 1).
Further, in the step S6, the two-dimensional amplitude image Amp (x)i,yk) In a second channel of RGB alone as a new constructed image; two-dimensional amplitude image Amp (x)i,yk) And two-dimensional distance information Dis (x)i,yk) Multiplying as a second channel in RGB of the newly constructed image; two-dimensional amplitude image Amp (x)i,yk) And two-dimensional distance information Angel (x)i,yk) The multiplication is used as a third channel in the RGB of the newly constructed image.
The invention also provides a three-dimensional complex image processing system for target detection, which adopts the processing method to process the three-dimensional complex image and comprises the following steps:
the amplitude information image module is used for acquiring a two-dimensional amplitude image by adopting a maximum projection method, namely for any i e to [1, N ]x],k∈[1,Ny]Find a j e [1, N ∈ ]z]Make O (x)i,yk,zj) Obtaining a two-dimensional amplitude information image Amp (x) with the largest amplitude valuei,yk)；
The segmentation mask module is used for segmenting the target and the background noise by an image segmentation method to obtain a target segmentation mask;
a distance information image module for recording the distance position j of the maximum value when a maximum value projection method is adopted to obtain a two-dimensional amplitude information image, and simultaneously determining the effective pixels of the distance information image according to the target segmentation mask to obtain a two-dimensional distance information image Dis (x)i,yk)；
A phase information image module for extracting corresponding three-dimensional complex image O (x) according to the distance position j of the maximum valuei,yk,zj) The data in (1) is arranged into two-dimensional complex image data in a plane, the phase information of the two-dimensional complex image data is calculated, meanwhile, the effective pixel of the distance information image is determined according to the target segmentation mask, and the two-dimensional phase information image Angel (x) is obtainedi,yk)；
A phase unwrapping module for unwrapping phase information Angel (x)i,yk) Performing phase unwrapping operation to change the phase unwrapped signal into a spatially continuous signal;
the RGB image integration module is used for integrating the amplitude information, the distance information and the phase information into one piece of RGB image data, and obtaining an RGB image simultaneously containing the amplitude information, the distance information and the phase information through formula calculation;
the target detection model training module is used for carrying out target detection labeling on the reconstructed RGB image data, training a pre-selected target detection deep learning model by combining a loss function and edge weight, observing the loss of a training set and a verification set, finishing training after a training stopping condition is reached, and storing and obtaining a target detection model;
the central processing module is used for sending instructions to other modules to complete related actions;
the amplitude information image module, the segmentation mask module, the distance information image module, the phase unwrapping module, the RGB image integration module and the target detection model training module are all electrically connected with the central processing module.
Compared with the prior art, the invention has the following advantages: according to the three-dimensional complex image processing method for target detection, on the basis of maximum projection of a three-dimensional complex image, the image contains more spatial information and phase information, target detection of a microwave millimeter wave image is performed by fully using amplitude information, the spatial information and the phase information of an inverted three-dimensional complex image, new information input is provided for a target detection model, and the method is worthy of popularization and use.
Drawings
FIG. 1 is a schematic flow chart of a three-dimensional complex image processing method for target detection according to a second embodiment of the present invention;
FIG. 2 is a diagram illustrating amplitude information, distance information, and phase information according to a second embodiment of the present invention;
fig. 3 is an exemplary diagram of an effect of integrating amplitude information, distance information, and phase information into an RGB image according to a second embodiment of the present invention.
Detailed Description
The following examples are given for the detailed implementation and specific operation of the present invention, but the scope of the present invention is not limited to the following examples.
Example one
The embodiment provides a technical scheme: a three-dimensional complex image processing method for target detection comprises the following steps:
(1) for microwave millimeter wave three-dimensional complex image O (x)i,yk,zj) X is horizontal dimension, y is vertical dimension, z is space distance dimension, firstly, a maximum projection method is adopted to obtain a two-dimensional amplitude image, namely, any i belongs to [1, N ∈x],k∈[1,Ny]Find a j e [1, N ∈ ]z]Make O (x)i,yk,zj) Obtaining a two-dimensional amplitude image Amp (x) with the largest amplitude valuei,yk)。
(2) Segmenting a target and background noise by a deep learning processing method of image segmentation, obtaining a target segmentation Mask of each pixel of the two-dimensional amplitude image, wherein the background noise pixel is 0, the target pixel is 1, and obtaining a Mask image Mask (x)i,yk)
(3) Obtaining two-dimensional amplitude image by maximum projection method, i.e. traversing coordinates in x and y directions, and finding a value j in z direction to enable O (x)i,yk,zj) The maximum amplitude value is recorded, the distance position j of the maximum amplitude value is recorded, meanwhile, the effective pixel of the distance information image is determined according to the target segmentation mask, and the two-dimensional distance information image Dis (x) is obtainedi,yk) Normalized to (0, 1).
(4) Extracting a corresponding three-dimensional complex image O (x) according to the distance position j of the maximum valuei,yk,zj) The data in (1) is arranged into two-dimensional complex image data in a plane, the phase information of the data is calculated, meanwhile, the effective pixel of the distance information image is determined according to the target segmentation mask, and a two-dimensional phase information image Angel (x) is obtainedi,yk)。
(5) Due to phase information Angel (x)i,yk) Is in the range of (-pi, pi)]The value of the interval, called the principal value of the phase or the winding phase, is related to the true phase by an integer multiple of 2 pi. Determining the variation trend of the phase, changing the jumping point in the phase information, changing the phase information into a spatially continuous signal in a mode of increasing or decreasing 2 pi on the numerical value of the phase information, and normalizing to (0, 1).
(6) Amplitude information, distance information and phase information are integrated into one piece of RGB image data, and in order to ensure that the image display effect is smaller than the actual amplitude image display difference, the RGB three-channel data are adjusted to be related to the maximum amplitude. The RGB image containing amplitude information, distance information and phase information at the same time is obtained by calculation through the following formula.
RGBImg(xi,yk,1)＝Amp(xi,yk)；
RGBImg(xi,yk,2)＝Amp(xi,yk)*Dis(xi,yk)；
RGBImg(xi,yk,3)＝Amp(xi,yk)*Angel(xi,yk)
(7) And carrying out target detection labeling on the reconstructed RGB image data, training a preselected target detection deep learning model by combining a loss function and the edge weight, observing the loss of the training set and the verification set, finishing the training after reaching a training stopping condition, and storing and obtaining the target detection model.
Example two
As shown in fig. 1, in this embodiment, the backscatter echo signal of the target obtained by the microwave millimeter wave transceiver front end is S (x, y, f), image focusing is performed through any echo reconstruction algorithm, an imaging target region is discretized, and coordinates of divided grid points are obtained as xi，yk，zjAnd obtaining the final three-dimensional complex image O (x) of the targeti,yk,zj) Where x is the horizontal dimension, y is the vertical dimension, and z is the spatial distance dimension.
Grid point coordinate xi，yk，zj，i∈[0,Nx]，j∈[0,Nz]，k∈[0,Ny],NxNumber of discrete grid points divided for x dimension, NzNumber of discrete grid points divided for distance dimension z, NyFor the number of discrete grid points divided in the y dimension, the standard of grid division should conform to: the size of the grid is
Figure BDA0002865112730000051
Number of grids
Figure BDA0002865112730000052
Wherein λ0Is the center frequency, theta, of the radio frequency signalxFor x dimension antenna beamwidth, LxIs the range covered by the x dimension. The smaller the grid size Δ x is, the finer the divided grid is, and the finer the image reconstruction result is.
And carrying out pixel-level segmentation and labeling on the two-dimensional amplitude image result, and labeling a foreground region and a background region for each two-dimensional amplitude microwave millimeter wave image to obtain classification label data. The foreground region includes a target imaging region, and the background region includes an imaging background region and a noise region. Training the selected image segmentation depth learning model by using the labeled segmentation label data, observing the loss of the training set and the verification set according to the loss function and the edge weight, finishing the training after reaching the training stopping condition, and storing and obtaining the image segmentation model;
processing the two-dimensional microwave millimeter wave amplitude image through a trained image segmentation model to segment a target from background noise, obtaining a target segmentation Mask of each pixel of the two-dimensional amplitude image, wherein the background noise pixel is 0, the target pixel is 1, and obtaining a Mask image Mask (x)i,yk)
Obtaining two-dimensional amplitude image by maximum projection method, i.e. traversing coordinates in x and y directions, and finding a value j in z direction to enable O (x)i,yk,zj) The amplitude value is maximum, the distance position j of the maximum value is recorded, meanwhile, the effective pixel of the distance information image is determined according to the target segmentation mask, if the segmentation result is 1, the area is effective, the data is j, and if the segmentation result is 0, the corresponding position data in the distance image is set to be 0. Thereby obtaining a two-dimensional distance image Dis (x)i,yk) And normalized to (0, 1).
Extracting a corresponding three-dimensional complex image O (x) according to the distance position j of the maximum valuei,yk,zj) The data in the step (1) is arranged into two-dimensional complex image data in a plane, effective pixels of a distance information image are determined according to a target segmentation mask, if the segmentation result is 1, the area is effective, the original complex image data are reserved by the data, and phase information of the complex image data is calculated; if the segmentation result is 0, the corresponding position data in the distance image is set to 0. Obtaining a two-dimensional phase image Angel (x)i,yk)。
Due to phase information Angel (x)i,yk) Is in the range of (-pi, pi)]The value of the interval, which appears spatially discontinuous, is therefore for Angel (x)i,yk) Performing phase unwrapping operation, determining the variation trend of the phase, changing the jumping point in the phase information, changing the phase information into a signal which is continuously changed on the space in a mode of increasing or decreasing 2 pi on the value of the phase information, and normalizing to (0, 1).
Amplitude information, distance information and phase information are integrated into one RGB image data in a channel retention mode, and in order to ensure that the image display effect is smaller than the actual amplitude image display difference, the RGB three-channel data are adjusted to be related to the maximum amplitude. The amplitude information corresponds to an R channel, the distance information corresponds to a G channel, the phase information corresponds to a B channel, the corresponding sequence of the information and the RGB color channels can be adjusted according to the color change required by display, and the RGB image simultaneously containing the amplitude information, the distance information and the phase information is obtained through calculation according to the following formula.
RGBImg(xi,yk,1)＝Amp(xi,yk)；
RGBImg(xi,yk,2)＝Amp(xi,yk)*Dis(xi,yk)；
RGBImg(xi,yk,3)＝Amp(xi,yk)*Angel(xi,yk)
And carrying out target detection labeling on the reconstructed RGB image data, inputting the constructed RGB image as data, transmitting the data to a training model for target detection, training a preselected target detection deep learning model by combining a loss function and edge weight, observing the loss of a training set and a verification set, finishing training after a training stopping condition is reached, and storing and obtaining the target detection model.
In summary, the three-dimensional complex image processing method for target detection in the embodiment enables the image to include more spatial information and phase information on the basis of the maximum projection of the three-dimensional complex image, fully uses the amplitude information, the spatial information and the phase information of the inverted three-dimensional complex image to perform target detection on the microwave millimeter wave image, provides new information input for the target detection model, and is worthy of being popularized and used.
Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention, and that variations, modifications, substitutions and alterations can be made to the above embodiments by those of ordinary skill in the art within the scope of the present invention.

Claims (8)

1. A three-dimensional complex image processing method for target detection is characterized by comprising the following steps:
s1: obtaining two-dimensional amplitude information image
For microwave millimeter wave three-dimensional complex image O (x)i,yk,zj) X is horizontal dimension, y is vertical dimension, z is space distance dimension, and a maximum projection method is adopted to obtain a two-dimensional amplitude image, namely, any i belongs to [1, N ∈x],k∈[1,Ny]Find a j e [1, N ∈ ]z]Make O (x)i,yk,zj) Obtaining a two-dimensional amplitude information image Amp (x) with the largest amplitude valuei,yk)；
S2: obtaining a target segmentation mask
Segmenting the target and the background noise by an image segmentation method to obtain a target segmentation mask;
s3: obtaining two-dimensional distance information image
When a maximum projection method is adopted to obtain a two-dimensional amplitude information image, i.e. for any i belongs to [1, N ]x],k∈[1,Ny]Find a j e [1, N ∈ ]z]Make O (x)i,yk,zj) The maximum amplitude value is recorded, the distance position j of the maximum amplitude value is recorded, meanwhile, the effective pixel of the distance information image is determined according to the target segmentation mask, and the two-dimensional distance information image Dis (x) is obtainedi,yk)；
S4: obtaining two-dimensional phase information image
Extracting a corresponding three-dimensional complex image O (x) according to the distance position j of the maximum valuei,yk,zj) The data in (1) is arranged into two-dimensional complex image data in a plane, the phase information of the two-dimensional complex image data is calculated, meanwhile, the effective pixel of the distance information image is determined according to the target segmentation mask, and the two-dimensional phase information image Angel (x) is obtainedi,yk)；
S5: phase unwrapping operation
For phase information Angel (x)i,yk) Performing phase unwinding operation to change the phase into a spaceA continuously varying signal;
s6: integrating various information into RGB image
Integrating the amplitude information, the distance information and the phase information into one RGB image data, and calculating to obtain the RGB image simultaneously containing the amplitude information, the distance information and the phase information by the following formula:
RGBImg(xi,yk,1)＝Amp(xi,yk)；
RGBImg(xi,yk,2)＝Amp(xi,yk)*Dis(xi,yk)；
RGBImg(xi,yk,3)＝Amp(xi,yk)*Angel(xi,yk)；
s7: training target detection model
And carrying out target detection labeling on the reconstructed RGB image data, training a preselected target detection deep learning model by combining a loss function and the edge weight, observing the loss of the training set and the verification set, finishing the training after reaching a training stopping condition, and storing and obtaining the target detection model.
2. The three-dimensional complex image processing method for object detection according to claim 1, characterized in that: in the step S1, a three-dimensional complex image O (x)i,yk,zj) Grid point coordinate of (2) is xi，yk，zj，i∈[0,Nx]，j∈[0,Nz]，k∈[0,Ny]In which N isxNumber of discrete grid points divided for x dimension, NzNumber of discrete grid points divided for distance dimension z, NyDiscrete grid points divided for the y dimension.
3. The three-dimensional complex image processing method for object detection according to claim 2, characterized in that: in step S2, for each pixel of the two-dimensional amplitude image, its target segmentation mask is obtained, the background noise pixel is 0, and the target pixel is 1.
4. A three-dimensional complex image processing method for object detection according to claim 3, characterized in that: in the above step S3, for the divided two-dimensional image data obtained by maximum value projection, the spatial position corresponding to the maximum value projection is used as the two-dimensional distance information Dis (x)i,yk)。
5. The three-dimensional complex image processing method for object detection according to claim 4, wherein: in the step S4, for the divided two-dimensional image data obtained by maximum value projection, the phase of the complex data corresponding to the maximum value projection is used as the two-dimensional phase information angell (x)i,yk)。
6. The three-dimensional complex image processing method for object detection according to claim 5, characterized in that: in step S5, the specific process of the phase unwrapping operation is to change the phase information into a spatially continuous signal by increasing or decreasing 2 pi in value and normalize to (0, 1).
7. The three-dimensional complex image processing method for object detection according to claim 6, characterized in that: in the step S6, the two-dimensional amplitude image Amp (x)i,yk) In a second channel of RGB alone as a new constructed image; two-dimensional amplitude image Amp (x)i,yk) And two-dimensional distance information Dis (x)i,yk) Multiplying as a second channel in RGB of the newly constructed image; two-dimensional amplitude image Amp (x)i,yk) And two-dimensional distance information Angel (x)i,yk) The multiplication is used as a third channel in the RGB of the newly constructed image.
8. A three-dimensional complex image processing system for object detection, wherein the three-dimensional complex image is processed by the processing method according to any one of claims 1 to 7, comprising:
the amplitude information image module is used for acquiring a two-dimensional amplitude image by adopting a maximum projection method, namely for any i e to [1, N ]x],k∈[1,Ny]Find a j e [1, N ∈ ]z]Make O (x)i,yk,zj) Obtaining a two-dimensional amplitude information image Amp (x) with the largest amplitude valuei,yk)；
The segmentation mask module is used for segmenting the target and the background noise by an image segmentation method to obtain a target segmentation mask;
a distance information image module for recording the distance position j of the maximum value when a maximum value projection method is adopted to obtain a two-dimensional amplitude information image, and simultaneously determining the effective pixels of the distance information image according to the target segmentation mask to obtain a two-dimensional distance information image Dis (x)i,yk)；
A phase information image module for extracting corresponding three-dimensional complex image O (x) according to the distance position j of the maximum valuei,yk,zj) The data in (1) is arranged into two-dimensional complex image data in a plane, the phase information of the two-dimensional complex image data is calculated, meanwhile, the effective pixel of the distance information image is determined according to the target segmentation mask, and the two-dimensional phase information image Angel (x) is obtainedi,yk)；
A phase unwrapping module for unwrapping phase information Angel (x)i,yk) Performing phase unwrapping operation to change the phase unwrapped signal into a spatially continuous signal;
the RGB image integration module is used for integrating the amplitude information, the distance information and the phase information into one piece of RGB image data, and obtaining an RGB image simultaneously containing the amplitude information, the distance information and the phase information through formula calculation;
the target detection model training module is used for carrying out target detection labeling on the reconstructed RGB image data, training a pre-selected target detection deep learning model by combining a loss function and edge weight, observing the loss of a training set and a verification set, finishing training after a training stopping condition is reached, and storing and obtaining a target detection model;
the central processing module is used for sending instructions to other modules to complete related actions;
the amplitude information image module, the segmentation mask module, the distance information image module, the phase unwrapping module, the RGB image integration module and the target detection model training module are all electrically connected with the central processing module.
CN202011580434.5A 2020-12-28 2020-12-28 Three-dimensional complex image processing method and system for target detection Active CN112634244B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011580434.5A CN112634244B (en) 2020-12-28 2020-12-28 Three-dimensional complex image processing method and system for target detection

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011580434.5A CN112634244B (en) 2020-12-28 2020-12-28 Three-dimensional complex image processing method and system for target detection

Publications (2)

Publication Number Publication Date
CN112634244A true CN112634244A (en) 2021-04-09
CN112634244B CN112634244B (en) 2022-09-30

Family

ID=75325829

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011580434.5A Active CN112634244B (en) 2020-12-28 2020-12-28 Three-dimensional complex image processing method and system for target detection

Country Status (1)

Country Link
CN (1) CN112634244B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116883998A (en) * 2023-06-20 2023-10-13 珠海微度芯创科技有限责任公司 Article labeling method and device based on millimeter wave image and electronic equipment

Citations (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5455590A (en) * 1991-08-30 1995-10-03 Battelle Memorial Institute Real-time holographic surveillance system
US20070222671A1 (en) * 2006-03-24 2007-09-27 Nec Corporation Millimeter wave image processor and millimeter wave image processing method
CN103018739A (en) * 2011-09-20 2013-04-03 中国科学院电子学研究所 Three-dimensional microwave imaging method for correcting multi-channel amplitude-phase error
CN105068126A (en) * 2015-07-02 2015-11-18 中国科学院上海微系统与信息技术研究所 Millimeter-wave holographic imaging method with function of amplitude correction
WO2018036373A1 (en) * 2016-08-23 2018-03-01 华讯方舟科技有限公司 Millimeter-wave three-dimensional holographic imaging method and system
JP2018169274A (en) * 2017-03-29 2018-11-01 キヤノン株式会社 Image processor, image processing method, and program
CN108761452A (en) * 2018-07-19 2018-11-06 山东省科学院自动化研究所 The multiple-input and multiple-output array millimeter wave three-dimensional image forming apparatus and method of compensated distance
CN109471193A (en) * 2018-11-19 2019-03-15 博微太赫兹信息科技有限公司 A kind of microwave and millimeter wave 3D hologram imaging system signal processing imaging method
CN109709621A (en) * 2019-01-31 2019-05-03 山东雷诚电子科技有限公司 A kind of method of millimeter wave human body security check system and target depth label
CN111896951A (en) * 2020-08-11 2020-11-06 中国人民解放军国防科技大学 Three-dimensional imaging and reconstructing method of millimeter wave cylindrical holographic imaging system
CN112114310A (en) * 2020-08-27 2020-12-22 博微太赫兹信息科技有限公司 Microwave millimeter wave holographic image reconstruction method based on three-dimensional decomposition

Patent Citations (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5455590A (en) * 1991-08-30 1995-10-03 Battelle Memorial Institute Real-time holographic surveillance system
US20070222671A1 (en) * 2006-03-24 2007-09-27 Nec Corporation Millimeter wave image processor and millimeter wave image processing method
CN103018739A (en) * 2011-09-20 2013-04-03 中国科学院电子学研究所 Three-dimensional microwave imaging method for correcting multi-channel amplitude-phase error
CN105068126A (en) * 2015-07-02 2015-11-18 中国科学院上海微系统与信息技术研究所 Millimeter-wave holographic imaging method with function of amplitude correction
WO2018036373A1 (en) * 2016-08-23 2018-03-01 华讯方舟科技有限公司 Millimeter-wave three-dimensional holographic imaging method and system
JP2018169274A (en) * 2017-03-29 2018-11-01 キヤノン株式会社 Image processor, image processing method, and program
CN108761452A (en) * 2018-07-19 2018-11-06 山东省科学院自动化研究所 The multiple-input and multiple-output array millimeter wave three-dimensional image forming apparatus and method of compensated distance
CN109471193A (en) * 2018-11-19 2019-03-15 博微太赫兹信息科技有限公司 A kind of microwave and millimeter wave 3D hologram imaging system signal processing imaging method
CN109709621A (en) * 2019-01-31 2019-05-03 山东雷诚电子科技有限公司 A kind of method of millimeter wave human body security check system and target depth label
CN111896951A (en) * 2020-08-11 2020-11-06 中国人民解放军国防科技大学 Three-dimensional imaging and reconstructing method of millimeter wave cylindrical holographic imaging system
CN112114310A (en) * 2020-08-27 2020-12-22 博微太赫兹信息科技有限公司 Microwave millimeter wave holographic image reconstruction method based on three-dimensional decomposition

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
LIU WEI ET AL: "FODs detection system based on millimeter wave FMCW radar", 《2013 IEEE 11TH INTERNATIONAL CONFERENCE ON ELECTRONIC MEASUREMENT & INSTRUMENTS》 *
ZHIMIAN ZHANG ET AL: "Complex-Valued Convolutional Neural Network and Its Application in Polarimetric SAR Image Classification", 《 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 》 *
孟祥新等: "一种高精度的毫米波稀疏平面阵列频域类成像算法", 《电子学报》 *
陈悦: "基于球面波的安检毫米波图像高分辨重建算法研究", 《中国优秀硕士学位论文全文数据库》 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116883998A (en) * 2023-06-20 2023-10-13 珠海微度芯创科技有限责任公司 Article labeling method and device based on millimeter wave image and electronic equipment
CN116883998B (en) * 2023-06-20 2024-04-05 珠海微度芯创科技有限责任公司 Article labeling method and device based on millimeter wave image and electronic equipment

Also Published As

Publication number Publication date
CN112634244B (en) 2022-09-30

Similar Documents

Publication Publication Date Title
Nasiri et al. Infrared small target enhancement based on variance difference
US8447383B2 (en) System and method for providing 2-dimensional computerized-tomography image corresponding to 2-dimensional ultrasound image
US8831288B1 (en) Digital processing method and system for determination of object occlusion in an image sequence
US20160350936A1 (en) Methods and Systems for Detecting Moving Objects in a Sequence of Image Frames Produced by Sensors with Inconsistent Gain, Offset, and Dead Pixels
US20120120198A1 (en) Three-dimensional size measuring system and three-dimensional size measuring method
Cheng et al. Compressive sensing radar imaging with convolutional neural networks
CN112114310B (en) Microwave millimeter wave holographic image reconstruction method based on three-dimensional decomposition
US20160213353A1 (en) Ultrasound imaging apparatus, ultrasound imaging method and ultrasound imaging program
CN106062824A (en) Edge detection device, edge detection method, and program
CN111046877A (en) Millimeter wave image suspicious article detection method and system
CN112634244B (en) Three-dimensional complex image processing method and system for target detection
CN108957448A (en) A kind of radar relevance imaging method based on broad sense total variation regularization
Song et al. Segmentation of sonar images with intensity inhomogeneity based on improved MRF
Mohsin Riaz et al. Through-wall image enhancement based on singular value decomposition
Karakuş et al. A generalized Gaussian extension to the Rician distribution for SAR image modeling
CN115698761A (en) Pattern-based array signal denoising for disturbed synthetic aperture radar
Song et al. Depth completion for kinect v2 sensor
CN112435249A (en) Dynamic small target detection method based on periodic scanning infrared search system
Li et al. Region adaptive morphological reconstruction fuzzy C-means for near-field 3-D SAR image target extraction
CN116740342A (en) Millimeter wave image target detection and identification method for improving YOLO v8
Arivazhagan et al. Optimal Gabor sub-band-based spectral kurtosis and Teager Kaiser energy for maritime target detection in SAR images
WO2018196248A1 (en) Direct wave suppression method and system for microwave imaging system
CN112099102B (en) Antenna array mechanical scanning type holographic image reconstruction method
Bivalkar et al. Development of an efficient approach for detection and measurement of crack length in ceramic tile manufacturing using millimeter-wave imaging
CN114488139A (en) Imaging method, imaging device, equipment and security check system for detection object

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant