A terahertz continuous wave compressed sensing imaging method
By optimizing the combination of Hadamar matrix and sparse representation base, high-speed imaging and high-quality image reconstruction of terahertz continuous wave imaging at low sampling rates are achieved, solving the problems of slow imaging speed and low image quality in the prior art.
Patent Information
- Application Number
- CN202411132452.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-19
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2044-08-19
AI Technical Summary
The existing terahertz continuous wave imaging method has a long scanning time and a large amount of sampled data, resulting in slow imaging speed and low image reconstruction quality at low sampling rates.
Design the optimized Hadamma matrix and perform connection degree rearrangement and under-determined sampling processing to build a measurement matrix. Using a compression perception scheme based on spatial light modulation, the terahertz continuous wave signal is encoded in the airspace amplitude, and the sparse representation base is selected, and the convexity is used. Optimize the problem and solve the sparse reconstruction signal, and finally perform two-dimensional graphics processing to obtain compressed sensing imaging results.
The terahertz continuous wave imaging speed is greatly improved under low sampling rate conditions and obtains higher image reconstruction quality in a very short time.
Smart Images

Figure CN118657848B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of imaging technology, and in particular relates to a terahertz continuous wave compressed sensing imaging method. Background Art
[0002] The frequency of terahertz (THz) is between 0.1 and 10 THz, and the wavelength is between the millimeter wave and the infrared band. Due to its rich spectrum resources and unique physical properties, terahertz waves play an important role in terahertz imaging, terahertz high-resolution radar, terahertz high-speed wireless communication and other fields. Terahertz waves are penetrable to non-polar materials and dielectric materials, and can penetrate clothing, packaging materials, etc. In addition, terahertz wave photons have low energy and non-ionizing radiation, which will not cause harm to the human body or the samples to be tested. Therefore, terahertz imaging has broad application prospects in security inspection, biomedical imaging, non-contact and non-destructive imaging, etc. Terahertz continuous wave imaging has large bandwidth and high imaging resolution, and has a higher signal-to-noise ratio and a simpler system than pulsed terahertz imaging systems. On the other hand, traditional imaging solutions based on grating scanning and cameras face the problems of being too slow and too expensive, which greatly limits the development of terahertz imaging. Therefore, in recent years, imaging solutions based on compressed sensing have gradually entered the field of vision of researchers. Terahertz compressed sensing imaging uses only a single terahertz detector for reception and achieves high-speed imaging by quickly switching masks and downsampling. The continuous-wave terahertz compressed sensing imaging system can make the imaging speed not limited by the pulse repetition rate, but only depend on the mask switching speed and data acquisition speed, further improving the upper limit of the system's imaging speed.
[0003] According to literature search, Hajun Song et al. published a paper titled "Continuous-wave THz vector imaging system utilizing two-tone signal generation and selfmixing detection" in the journal Optic Express in 2017. The paper proposes a photonics-based continuous-wave terahertz imaging method. The method uses the optical signal emitted by a suppressed carrier double-sideband modulated laser. The modulated signal and the optical signal emitted by another laser are beat in a single-line carrier photodiode to generate a dual-frequency terahertz signal. After the dual-frequency terahertz signal passes through the sample, it is mixed in the Schottky barrier diode at the receiving end. The low-frequency signal obtained by mixing is amplified and output by a phase-locked amplifier. The paper uses a point-by-point scanning method for single-pixel imaging. The sample is placed on an electric workbench and moves in a plane at a fixed speed. The continuous-wave terahertz imaging scheme proposed in the paper can achieve imaging without the local oscillator signal required for heterodyne or homodyne detection, and has a high signal-to-noise ratio. However, the imaging method based on point-by-point scanning severely limits the imaging speed.
[0004] A literature search revealed that Rayko Ivanov Stantchev et al. published a paper titled "Real-time terahertz imaging with a single-pixel detector" in the journal Nature Communication in 2020. The paper proposed and demonstrated a terahertz imaging method based on compressed sensing. The method uses laser reflected by a digital micromirror to regulate the photosensitive semiconductor, and then uses the photosensitive semiconductor to modulate the terahertz pulse incident on the sample to achieve spatial light encoding of the terahertz pulse. The method proposed in the paper realizes 32*32 pixels and 6 frames per second of compressed sensing real-time imaging. However, since the time for the carriers in the photosensitive semiconductor to reach saturation is much longer than the mask switching period, the modulation depth is limited, which affects the imaging quality. It is necessary to introduce a fitting algorithm for the response amplitude of the photosensitive semiconductor in post-processing to obtain an estimate closer to the actual modulation depth to improve the imaging quality, which increases the algorithm complexity and processing time. In addition, the document selects the Paley I-type Hadamard matrix construction method to generate the measurement matrix. Each mask matrix generated by this method can be obtained by cyclic translation of the previous mask. Therefore, each mask matrix has the same frequency component. During downsampling, the mask matrix containing more image information is not selected, which affects the image reconstruction quality. Summary of the invention
[0005] In view of the above, the purpose of the present invention is to provide a terahertz continuous wave compressed sensing imaging method to solve the problems of long scanning time and large amount of sampling data in the existing terahertz continuous wave imaging method, and adopt an optimized Hadamard matrix to improve the image reconstruction quality under low sampling rate conditions.
[0006] To achieve the above-mentioned purpose of the invention, an embodiment provides a terahertz continuous wave compressed sensing imaging method, comprising the following steps:
[0007] Design the Hadamard matrix according to the imaging requirements, and perform rearrangement optimization and underdetermined sampling on the Hadamard matrix based on connectivity to obtain the measurement matrix;
[0008] The terahertz continuous wave signal is spatially amplitude encoded using a measurement matrix, and a transmission signal formed after the encoded terahertz continuous wave signal is transmitted through the sample is measured to obtain a measurement signal;
[0009] Selecting a sparse representation basis, and obtaining a sparse representation of the signal according to the measurement matrix, the sparse representation basis, and the measurement signal;
[0010] According to the sparse expression of the signal, a sparse reconstructed signal is obtained by solving a convex optimization problem;
[0011] The compressed sensing imaging result is obtained by performing two-dimensional graphics processing based on the sparsely reconstructed signal.
[0012] Preferably, the Hadamard matrix is designed according to the imaging requirements, including:
[0013] Determine the number of image pixels based on the spatial light modulator resolution, imaging accuracy requirements, and data acquisition requirements , according to the number of image pixels Determine the Hadamard matrix Dimensions ,get dimensional Hadamard matrix .
[0014] Preferably, the Hadamard matrix is subjected to connectivity-based rearrangement optimization and underdetermined sampling processing to obtain a measurement matrix, including:
[0015] Connectivity-based rearrangement optimization of the Hadamard matrix: No. The connectivity of a row is defined as The rows are split and concatenated into Dimensional matrix ,matrix Each local maximum rectangular block composed of the same elements contributes a connectivity value of 1; the Hadamard matrix is sorted in order of connectivity from small to large Rearrange each row of to obtain the Cake-cutting Hadamard matrix ;
[0016] Cake-cutting Hadamard Matrix Perform underdetermined sampling processing: Determine the measurement matrix based on the sampling rate The row dimension , and extract the Cake-cutting Hadamard matrix Before Row Formation Dimensional measurement matrix .
[0017] Preferably, the terahertz continuous wave signal is subjected to spatial domain amplitude encoding using a measurement matrix, comprising:
[0018] Constructing a positive measurement matrix for differential measurement based on the measurement matrix and the negative measurement matrix ,in, Middle position Positive measurement at , Middle position Negative measurements at Calculated as:
[0019] ;
[0020] ;
[0021] Align the measurement matrix separately Matrix each row in to get individual dimensional mask matrix ; For the negative measurement matrix Matrix each row in to get indivual dimensional negative mask matrix , is the mask matrix index; using the spatial light modulator The mask matrix and Negative mask matrix Play them sequentially to realize spatial amplitude coding of terahertz continuous wave signals.
[0022] Preferably, measuring a transmission signal formed after the encoded terahertz continuous wave signal is transmitted through the sample to obtain a measurement signal comprises:
[0023] ;
[0024] in, To measure the signal, is the signal to be reconstructed.
[0025] Preferably, a one-dimensional discrete cosine transform basis is selected as the sparse representation basis of the signal , which can be expressed as:
[0026] ;
[0027] in, Indicates the number of signal points to be reconstructed.
[0028] Preferably, obtaining a signal sparse expression according to the measurement matrix, the sparse representation basis, and the measurement signal includes:
[0029] Based on the measurement matrix and sparse representation basis Get the matrix ;
[0030] Based on the sparse representation and the signal to be reconstructed Get sparse signal ;
[0031] The signal is sparsely expressed as: the measured signal .
[0032] Preferably, obtaining a sparse reconstructed signal by solving a convex optimization problem according to the sparse expression of the signal includes:
[0033] Based on the sparse expression of signals Norm minimization problem: ,in, For sparsely reconstructed signals, represents the variable value that minimizes the objective function, express of norm;
[0034] Will be based on The convex relaxation of the norm minimization problem is equivalent to Norm minimization problem: , express of norm;
[0035] Solve using the basis pursuit method based on linear programming The norm minimization problem is to obtain the sparse reconstructed signal .
[0036] Preferably, performing two-dimensional graphics processing on the sparsely reconstructed signal to obtain a compressed sensing imaging result includes:
[0037] The sparsely reconstructed signal With sparse representation basis Multiply and perform sparse inverse transform to obtain the reconstructed signal ;
[0038] The reconstructed signal The matrix is dimensional matrix, the The dimensional matrix is the compressed sensing imaging result.
[0039] Compared with the prior art, the present invention has the following beneficial effects:
[0040] The present invention constructs different measurement matrices by performing connectivity-based rearrangement optimization and underdetermined sampling processing on the Hadamard matrix, and uses a compressed sensing scheme based on spatial light modulation based on the measurement matrix to image the sample under low sampling rate conditions and in an extremely short time. This greatly improves the terahertz continuous wave imaging speed and obtains higher image reconstruction quality under low sampling rate conditions. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0042] Figure 1 It is a flow chart of the terahertz continuous wave compressed sensing imaging method provided by the present invention;
[0043] Figure 2 It is a schematic diagram of the terahertz continuous wave compressed sensing imaging experimental system implemented by the present invention;
[0044] Figure 3 It is a sample image reconstructed based on the general Hadamard matrix with a sampling rate of 20%;
[0045] Figure 4 is a sample image reconstructed when the sampling rate is 20% in a specific embodiment;
[0046] Figure 5 is a sample image reconstructed when the sampling rate is 30% in a specific embodiment;
[0047] Figure 6 is a sample image reconstructed when the sampling rate is 40% in a specific embodiment;
[0048] Figure 7 It is a sample image reconstructed when the sampling rate is 50% in a specific embodiment. DETAILED DESCRIPTION
[0049] To make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific implementation methods described herein are only used to explain the present invention and do not limit the scope of protection of the present invention.
[0050] The inventive concept of the present invention is: in response to the technical problems of long scanning time, large amount of sampling data and slow imaging speed in the existing terahertz continuous wave imaging method, the present invention provides a terahertz continuous wave compressed sensing imaging method. The method of the present invention can be based on a terahertz continuous wave compressed sensing imaging system, adopt an advanced measurement matrix, and adopt a compressed sensing scheme based on spatial light modulation to image the sample under low sampling rate conditions and in a very short time, thereby greatly improving the imaging speed of the terahertz continuous wave imaging system and improving the image reconstruction quality of the terahertz compressed sensing imaging system.
[0051] like Figure 1 As shown, the terahertz continuous wave compressed sensing imaging method provided in the embodiment includes the following steps:
[0052] S1. Design the Hadamard matrix according to the imaging requirements, and perform rearrangement optimization and underdetermined sampling processing on the Hadamard matrix based on connectivity to obtain the measurement matrix.
[0053] In the embodiment, when designing the Hadamard matrix according to the imaging requirements, the Hadamard matrix is designed according to the resolution of the spatial light modulator used in the terahertz continuous wave compressed sensing imaging system, the imaging accuracy requirements, and the data acquisition requirements. , wherein the spatial light modulator includes a light-controlled semiconductor, a rotating metal mask or a terahertz metasurface, etc.
[0054] Specifically, the number of image pixels is determined according to the spatial light modulator resolution, imaging accuracy requirements, and data acquisition requirements. , according to the number of image pixels Determine the Hadamard matrix Dimensions ,get dimensional Hadamard matrix .
[0055] In the embodiment, the Hadamard matrix is further subjected to rearrangement optimization based on connectivity and underdetermined sampling processing to obtain a measurement matrix , including:
[0056] Connectivity-based rearrangement optimization of the Hadamard matrix: No. The connectivity of a row is defined as The rows are split and concatenated into Dimensional matrix ,matrix Each local maximum rectangular block composed of the same elements contributes a connectivity value of 1; the Hadamard matrix is sorted in order of connectivity from small to large Rearrange each row of to obtain the Cake-cutting Hadamard matrix The mask matrix with small connectivity can provide more image information, so placing it in the front position can achieve better imaging quality at low sampling rate.
[0057] Cake-cutting Hadamard Matrix Perform underdetermined sampling processing: Determine the measurement matrix based on the sampling rate The row dimension , and extract the Cake-cutting Hadamard matrix Before Row Formation Dimensional measurement matrix .
[0058] S2. Use the measurement matrix to perform spatial domain amplitude encoding on the terahertz continuous wave signal, and measure the transmission signal formed after the encoded terahertz continuous wave signal is transmitted through the sample to obtain a measurement signal.
[0059] In the embodiment, the measurement signal ,in, It is a terahertz continuous wave signal. Specifically, it is impossible to directly realize the measurement matrix due to the spatial light modulation device based on optically controlled semiconductors, metal rotating masks or terahertz metasurfaces. The modulation effect of the -1 element in the equation is thus obtained, so a positive measurement matrix for differential measurement is constructed based on the measurement matrix: and the negative measurement matrix ,in, Middle position Positive measurement at , Middle position Negative measurements at Calculated as:
[0060] ;
[0061] ;
[0062] Align the measurement matrix separately Matrix each row in to get individual dimensional mask matrix ; For the negative measurement matrix Matrix each row in to get indivual dimensional negative mask matrix , is the mask matrix index;
[0063] Using spatial light modulators The mask matrix and Negative mask matrix Play them in sequence to realize spatial domain amplitude encoding of the terahertz continuous wave signal. The encoded terahertz continuous wave signal passes through the sample to form a transmission signal. The transmission signal is measured to obtain a measurement signal. , which can be expressed as:
[0064] ;
[0065] in, is the signal to be reconstructed.
[0066] S3. Select a sparse representation basis, and obtain a sparse expression of the signal according to the measurement matrix, the sparse representation basis, and the measurement signal.
[0067] In the embodiment, a one-dimensional discrete cosine transform basis is selected as the sparse representation basis of the signal , which can be expressed as:
[0068] ;
[0069] in, Indicates the number of signal points to be reconstructed;
[0070] Then based on the measurement matrix and sparse representation basis Get the matrix ; Based on the sparse representation basis and the signal to be reconstructed Get sparse signal ; The signal is sparsely expressed as: the measured signal .
[0071] S4. Based on the sparse expression of the signal, a sparse reconstructed signal is obtained by solving a convex optimization problem.
[0072] In the embodiment, first, a sparse expression based on the signal is constructed. Norm minimization problem:
[0073] ;
[0074] in, For sparsely reconstructed signals, represents the variable value that minimizes the objective function, express of norm;
[0075] Then based on The convex relaxation of the norm minimization problem is equivalent to Norm minimization problem:
[0076] ;
[0077] in, express of norm;
[0078] Finally, the basis pursuit method based on linear programming is used to solve The norm minimization problem is to obtain the sparse reconstructed signal .
[0079] S5. Perform two-dimensional graphics processing based on the sparsely reconstructed signal to obtain a compressed sensing imaging result.
[0080] In the embodiment, the sparse reconstruction signal With sparse representation basis Multiply and perform sparse inverse transform to obtain the reconstructed signal ; Then reconstruct the signal The matrix is dimensional matrix, the The dimensional matrix is the compressed sensing imaging result.
[0081] An implementation example is given below to illustrate the performance of the above imaging method and compare it with a technical method based on a general Hadamard matrix.
[0082] like Figure 2 As shown, the terahertz continuous wave compressed sensing imaging experimental system implemented by the present invention includes a transmitting link, a spatial light modulation link, a receiving link and a sample, and the sample is placed after the transmitting link and the spatial light modulation link and before the receiving link. Among them, the transmitting link is used to generate a terahertz continuous wave signal and transmit it to free space. The terahertz continuous wave signal is transmitted through a modulated optically controlled semiconductor in free space to form a spatial domain amplitude-coded terahertz continuous wave signal. The spatial domain amplitude-coded terahertz continuous wave signal is then transmitted through the sample to form a transmission signal. The spatial light modulation link is used to convert the amplitude of the terahertz continuous wave signal emitted by the transmitting link in the spatial domain according to a positive mask matrix and the negative mask matrix The receiving link is used to receive the transmission signal formed by the spatial amplitude coded terahertz signal passing through the sample, and reconstruct the collected terahertz signal amplitude information using the compressed sensing reconstruction algorithm to obtain the reconstructed signal, and perform two-dimensional image processing to obtain the sample imaging result.
[0083] like Figure 2 As shown, the transmission link includes a terahertz continuous wave transmitter, which uses an electronics-based terahertz radiation source or an optics-based terahertz radiation source to generate a terahertz signal, wherein the electronics-based terahertz radiation source includes a terahertz radiation source based on vacuum electronics technology, a terahertz quantum cascade laser or a terahertz free electron laser; the optics-based terahertz radiation source includes a terahertz radiation source based on ultrafast laser technology or a terahertz radiation source based on optical beat frequency technology.
[0084] like Figure 2 As shown, the spatial light modulation link includes an infrared laser, a digital micromirror, and a light-controlled semiconductor. The digital micromirror is used to play the measurement matrix and the positive mask matrix. and the negative mask matrix The infrared laser is reflected by the digital micromirror and projects a positive mask matrix on the optical control semiconductor. and the negative mask matrix The mask pattern is used to modulate the terahertz signal based on the carrier characteristics of the light-controlled semiconductor.
[0085] like Figure 2 As shown, the receiving link includes a terahertz receiver and a signal processing module; wherein the terahertz receiver is used to receive the transmission signal formed by the spatial amplitude-encoded terahertz signal through the sample transmission. Specifically, the reflected signal of the terahertz signal is received by down-conversion, including a subharmonic mixer, a photoconductive antenna, a Schottky barrier diode or a quantum well photodetector. In order to better receive the reflected signal, the system also includes a terahertz prism, which is arranged at the input end of the receiving link, that is, at the input end of the terahertz receiver, and is used to collimate and focus the reflected signal and then input it to the terahertz receiver to reduce the free space propagation loss. The signal processing module is used to collect the measurement signal and post-process the measurement signal to obtain a reconstructed image. Specifically, a high-speed data acquisition card is used to collect the amplitude of the received signal, and a method based on convex optimization is used to implement compressed sensing reconstruction using a computer to obtain a reconstructed image.
[0086] In the implementation example, the transmission link operates at a frequency of 300 GHz, and the sample is placed about 10 cm away from the transmission link and close to the optical control semiconductor. In order to obtain high-precision imaging, the number of imaging pixels is set to 64×64. If the traditional mechanical scanning method is used, it will take a lot of time. In this implementation example, the image can be reconstructed with high quality when the sampling rate is 50%, and the measurement matrix The dimension is 4096×4096, and a total of 4096 mask patterns need to be played, which are 2048 by The 64×64 positive mask matrix formed and 2048 by The 64×64 negative mask matrix formed , setting the playback rate of the digital micromirror to 1 kHz, all data acquisition can be completed in about 4.1 s. Figure 3 The image reconstruction results of the technical method with a sampling rate of 20% and a general Hadamard matrix as the measurement matrix are given, where X represents the pixel position of the image horizontal axis, Y represents the pixel position of the image numerical axis, Figure 4 The image reconstruction results of the method proposed in the present invention when the sampling rate is 20% are given. It can be observed that under the condition of a lower sampling rate such as 20%, the method proposed in the present invention can reconstruct the original image, and the imaging quality is higher than the technical method using the general Hadamard matrix as the measurement matrix. Figure 5-Figure 7 The image reconstruction results of the method proposed in the present invention are given when the sampling rate is 30%, 40%, and 50%. It can be observed that under the sampling rate conditions of 30%, 40%, and 50%, the method proposed in the present invention can reconstruct the original image, and the image reconstruction quality is improved when the sampling rate is increased.
[0087] The above method is based on compressed sensing technology and provides an effective solution for terahertz continuous wave imaging systems under high speed, high resolution and low sampling rate conditions.
[0088] The specific implementation methods described above provide a detailed description of the technical solutions and beneficial effects of the present invention. It should be understood that the above is only the most preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, supplements and equivalent substitutions made within the scope of the principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A terahertz continuous wave compressed sensing imaging method, characterized in that: The following steps are involved: Design the Hadamard matrix according to the imaging requirements, and perform rearrangement optimization and underdetermined sampling processing on the Hadamard matrix based on connectivity to obtain the measurement matrix, including: Rearrangement optimization of the Hadamard matrix based on connectivity: No. The connectivity of a row is defined as The rows are split and concatenated into Dimensional matrix ,matrix Each local maximum rectangular block composed of the same elements contributes a connectivity value of 1; the Hadamard matrix is sorted in order of connectivity from small to large Rearrange each row of to obtain the Cake-cutting Hadamard matrix ; Cake-cutting Hadamard Matrix Perform underdetermined sampling processing: Determine the measurement matrix based on the sampling rate The row dimension , and extract the Cake-cutting Hadamard matrix Before Row Formation Dimensional measurement matrix ; Using a measurement matrix to perform spatial domain amplitude encoding on a terahertz continuous wave signal, including: constructing a positive measurement matrix for differential measurement based on the measurement matrix and the negative measurement matrix ,in, Middle position Positive measurement at , Middle position Negative measurements at Calculated as: ; ; Align the measurement matrix separately Matrix each row in to get individual dimensional mask matrix ; For the negative measurement matrix Matrix each row in to get indivual dimensional negative mask matrix , is the mask matrix index; using the spatial light modulator The mask matrix and Negative mask matrix Play them sequentially to achieve spatial domain amplitude coding of terahertz continuous wave signals; measuring a transmission signal formed after the encoded terahertz continuous wave signal is transmitted through the sample to obtain a measurement signal; Selecting a sparse representation basis, and obtaining a sparse representation of the signal according to the measurement matrix, the sparse representation basis, and the measurement signal; According to the sparse expression of the signal, a sparse reconstructed signal is obtained by solving a convex optimization problem; The compressed sensing imaging result is obtained by performing two-dimensional graphics processing based on the sparsely reconstructed signal.
2. The terahertz continuous wave compressed sensing imaging method according to claim 1, characterized in that: Design the Hadamard matrix based on imaging requirements, including: Determine the number of image pixels based on the spatial light modulator resolution, imaging accuracy requirements, and data acquisition requirements , according to the number of image pixels Determine the Hadamard matrix Dimensions ,get dimensional Hadamard matrix .
3. The terahertz continuous wave compressed sensing imaging method according to claim 1, characterized in that: The transmission signal formed after the encoded terahertz continuous wave signal is transmitted through the sample is measured to obtain a measurement signal, including: ; in, To measure the signal, is the signal to be reconstructed.
4. The terahertz continuous wave compressed sensing imaging method according to claim 1, characterized in that: Selecting a one-dimensional discrete cosine transform basis as the sparse representation basis of the signal , expressed as: ; in, Indicates the number of signal points to be reconstructed.
5. The terahertz continuous wave compressed sensing imaging method according to claim 1, characterized in that: The signal sparse expression is obtained according to the measurement matrix, the sparse representation basis, and the measurement signal, including: Based on the measurement matrix and sparse representation basis Get the matrix ; Based on the sparse representation and the signal to be reconstructed Get sparse signal ; The signal is sparsely expressed as: the measured signal .
6. The terahertz continuous wave compressed sensing imaging method according to claim 5, characterized in that: According to the sparse expression of the signal, a sparse reconstructed signal is obtained by solving a convex optimization problem, including: Based on the sparse expression of signals Norm minimization problem: ,in, For sparse reconstruction signal, represents the variable value that minimizes the objective function, express of norm; Will be based on The convex relaxation of the norm minimization problem is equivalent to Norm minimization problem: , express of norm; Solve using the basis pursuit method based on linear programming The norm minimization problem is to obtain the sparse reconstructed signal .
7. The terahertz continuous wave compressed sensing imaging method according to claim 1, characterized in that: The compressed sensing imaging results are obtained by performing two-dimensional graphics processing based on the sparse reconstruction signal, including: The sparsely reconstructed signal With sparse representation basis Multiply and perform sparse inverse transform to obtain the reconstructed signal ; The reconstructed signal The matrix is dimensional matrix, the The dimensional matrix is the compressed sensing imaging result.
