Rapid single-pixel imaging method based on galvanometer scanning

Abstract

The invention discloses a quick single-pixel imaging method based on galvanometer scanning. The quick single-pixel imaging method comprises the following steps: slicing and reconstructing a Hadamard matrix to obtain an innovative scanning base graph; a galvanometer is used for continuously and rapidly scanning and traversing the modulation process of a target image, and the number of mask patterns needing to be modulated is reduced; obtaining a Hadamard coefficient of each scanning base image based on a differential measurement mode, including obtaining a light intensity value after differential modulation by using a positive Hadamard scanning base image and a negative Hadamard scanning base image, and calculating to obtain a Hadamard coefficient of the target image; and after the Hadamard coefficient of the target image is obtained, forming a Hadamard frequency spectrum, and performing inverse transformation on the Hadamard frequency spectrum to obtain a reconstructed target image. According to the invention, the problems of a large number of modulation masks and large memory consumption in the traditional single-pixel imaging technology are solved, the generation duration and imaging time of the mask pattern are shortened, and the application prospect of the single-pixel imaging technology is expanded, especially in the fields of industrial part detection, medical imaging and the like.

Description

Technical Field

[0001] The invention relates to the technical field of single pixel imaging, and in particular to a fast single pixel imaging method based on galvanometer scanning. Background Art

[0002] Traditional passive Hadamard single-pixel imaging technology uses a light source to illuminate the object to be imaged. The light reflected by the object is focused by a lens or a lens group onto the surface of a spatial light modulator with a preset structured illumination pattern. Different areas will produce different light intensity changes depending on their characteristics. This series of light intensity values ​​is reflected or transmitted to the photosensitive surface of a single-point detector. The optical signal is converted into an electrical signal and stored through a data acquisition device. This method obtains the Hadamard spectrum of the target area through differential measurement, and then uses a reconstruction algorithm to reconstruct the object image.

[0003] The shortcomings of the existing technology are that the existing traditional single-pixel imaging technology has problems such as a large number of modulation patterns, long generation time, large memory resource usage, long preloading time, long modulation time, and low DMD utilization. Summary of the invention

[0004] The purpose of the present invention is to overcome the shortcomings of the prior art. To achieve the above purpose, a fast single-pixel imaging method based on galvanometer scanning is adopted to solve the problems raised in the above background technology.

[0005] A fast single-pixel imaging method based on galvanometer scanning comprises the following steps:

[0006] Step S1, slicing and reconstructing the Hadamard base image according to the required resolution of the target to generate a scanning base image;

[0007] Step S2, utilizing the periodic and continuous scanning characteristics of the galvanometer to accurately traverse the modulation process of the target area;

[0008] Step S3, capturing the light intensity value modulated by the positive Hadamard scanning base image and the negative Hadamard scanning base image based on a differential measurement method;

[0009] Step S4, calculating and obtaining the Hadamard coefficients of the corresponding frequency points to form a Hadamard spectrum;

[0010] Step S5: perform inverse transformation on the Hadamard spectrum to obtain a reconstructed target image.

[0011] As a further technical solution of the present invention, the specific steps in step S1 include:

[0012] Through the study and analysis of Hadamard matrix, the rows or columns of Hadamard matrix are sliced ​​to obtain several first-order row or column vectors;

[0013] The slices are reconstructed by copying to obtain an innovative scanning base map, which is denoted as D M ; Among them, the scanning base graph is composed of M column vectors, each column vector is denoted by d M :

[0014] D M =[d1,d2,···,d M ].

[0015] As a further technical solution of the present invention, the specific steps in step S2 include:

[0016] Using the scanning characteristics of the galvanometer to traverse multiple first-order modulation trains in the target area;

[0017] Among them, the lateral resolution of the scanned base image depends on the image resolution requirement.

[0018] As a further technical solution of the present invention, the specific steps in step S3 and step S4 include:

[0019] Using differential measurement and adapting to the structural characteristics of the digital micromirror array (DMD), the mirror of the digital micromirror array (DMD) is tilted at different angles during operation to reflect or retain light information in the light trap, thereby obtaining a specific coded illumination pattern.

[0020] The Hadamard coefficient Q(u,v) of each scanning base graph is obtained based on the differential measurement method:

[0021] Q(u,v)=Y + -Y -

[0022] Among them, Y + and Y - are the light intensity values ​​corresponding to the target image after being differentially modulated by the positive Hadamard scanning base image and the negative Hadamard scanning base image, and the expression for obtaining the light intensity value is:

[0023]

[0024] In the above formula, Y(N) is the light intensity value, i·K is the number of measurements corresponding to each coding pattern, K is the number of repeated acquisitions of the light intensity value during each modulation process, N is the modulation corresponding to the Nth coding pattern and the target information, l is the number of light intensity values ​​that are discarded and belong to the transition state of the digital micromirror array DMD, and y(x) is the original light intensity value received by the photodetector after being modulated by each mask pattern;

[0025] After calculating the Hadamard coefficient corresponding to the target image by differential method, the Hadamard spectrum is obtained:

[0026] Q=[Q1,Q2,Q3,···,QM ].

[0027] As a further technical solution of the present invention, the specific steps in step S5 include:

[0028] By inverse transforming the Hadamard spectrum, we can get the reconstructed target image:

[0029] P(x,y)=H -1 {Q(u,v)}

[0030] The formula for the reconstruction process is:

[0031] P = H·Q.

[0032] Compared with the prior art, the present invention has the following technical effects:

[0033] By adopting the above-mentioned technical solution, an innovative mask pattern is realized by slicing reconstruction based on the Hadamard matrix, and the column modulation technical solution reduces the number of masks required for imaging to 1 / N. Since the Hadamard matrix is ​​a square matrix, only the N*N target area can be reconstructed, but the modulation mode of this technology breaks the modulation limitation. Column scanning can realize image modulation of non-square targets, and then the modulation of the target area is traversed by the periodic galvanometer scanning light path to complete the image reconstruction. The extremely small number of masks greatly reduces the memory consumption and generation and burning time of the pre-loaded mode. At the same time, the reduction in the number of masks brings about a fast modulation process, which reduces the imaging time by T / N relative to the traditional HIS. Unlike the traditional HIS, due to the 22K memory upper limit of the DMD, the reconstruction resolution of the traditional HIS is limited to less than 256*256. The advantages of the present invention in modulation technology greatly improve the utilization rate of the DMD, and can realize one-time higher resolution image reconstruction, which significantly improves the imaging efficiency of single-pixel imaging. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings:

[0035] Figure 1 A schematic diagram of the steps of the fast single-pixel imaging method according to the embodiment disclosed in the present application;

[0036] Figure 2 A schematic diagram of the Hadamard coefficient acquisition principle of the disclosed embodiment of the present application;

[0037] Figure 3 Generate a schematic diagram for scanning a base image of an embodiment disclosed in the present application;

[0038] Figure 4 An experimental device diagram of an embodiment disclosed in this application;

[0039] Figure 5 This is a numerical simulation reconstruction effect diagram at different sampling rates of the embodiment disclosed in this application;

[0040] Figure 6 This is a reconstruction result diagram based on rapid scanning according to an embodiment disclosed in the present application. DETAILED DESCRIPTION

[0041] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 creative work are within the scope of protection of the present invention.

[0042] Please refer to Figure 1 , a fast single-pixel imaging method based on galvanometer scanning, uses a straight laser equipped with a galvanometer to achieve a precise scanning optical path, and cooperates with innovative slice reconstruction scanning base image to traverse the partial modulation of single-pixel imaging, thereby reducing the modulation pattern by 1 / N. It includes the following steps:

[0043] In this embodiment, Figure 2 As shown, the fast scanning modulation scheme proposed in this implementation is based on Hadamard single-pixel imaging. This technology obtains the Hadamard spectrum of the target image and then reconstructs the target image using the inverse Hadamard transform. Each row of the target is coded and modulated by the corresponding shadow part of each scanning base image. The lateral resolution of the scanning base image depends on the image resolution requirement. After completing multiple cycles of scanning, a series of Hadamard coefficients can be obtained to form the Hadamard spectrum. Figure 2 The diagram is the principle diagram for obtaining the Hadamard coefficient, and the formula for obtaining it is:

[0044] Q i =P / i·D M

[0045] Among them, Q i is the Hadamard coefficient, P is the target image, D M For the innovative scanning base map, the scanning base map in this implementation is obtained by slicing and copying the Hadamard matrix. First of all, orthogonal transformation is an important mathematical tool in image processing, which can simplify the calculation, ensure that the signals of different bases are still independent of each other after transformation, and understand and analyze the signals more completely and accurately. Orthogonality is expressed as:

[0046]

[0047] In the above formula, H NThe transposed matrix of E is an orthogonal matrix. Any two rows (or columns) of the Hadamard matrix are orthogonal. This orthogonal matrix consists of two elements, +1 and -1, which ensures that the signals of different bases are still independent of each other after transformation. The Hadamard matrix H N It is expressed as:

[0048]

[0049] Among them, h N is the Nth row vector of the Hadamard matrix, is the transpose of the row vector. This implementation designs a Hadamard basis graph that conforms to the fast scanning mode based on the transformation of the Hadamard matrix. By slicing the original N-order Hadamard basis graph by row (or column), N different first-order row (or column) vectors can be obtained, and their expressions are:

[0050]

[0051] Step S1: Slice and reconstruct the Hadamard base image according to the target required resolution to generate a scan base image. The scan base image is denoted as D M ; Among them, the scanning base graph is composed of M column vectors, each column vector is denoted by d M :

[0052] D M =[d1,d2,···,d M ]

[0053] In this embodiment, a specific implementation step is to use Hadamard single-pixel imaging technology to quickly scan and collect light intensity through the scanning base image reconstructed by DMD projection slices, and obtain a 256x256 resolution target image of the target through inverse transformation.

[0054] Step S2, using the periodic and continuous scanning characteristics of the galvanometer to accurately traverse the modulation process of the target area. Specifically, using the scanning characteristics of the galvanometer to traverse multiple first-order modulation sequences of the target area; wherein the lateral resolution of the scanning base image depends on the image resolution requirement.

[0055] In this embodiment, a specific implementation step is:

[0056] Step S3: Based on the differential measurement method, the positive Hadamard scanning base image and the negative Hadamard scanning base image are captured to obtain the modulated light intensity value. The specific steps include:

[0057] Using differential measurement and adapting to the structural characteristics of the digital micromirror array (DMD), the mirror of the digital micromirror array (DMD) is tilted at different angles during operation to reflect or retain light information in the light trap, thereby obtaining a specific coded illumination pattern.

[0058] Step S4, calculating and obtaining the Hadamard coefficients of the corresponding frequency points to form a Hadamard spectrum; wherein the specific steps in step S3 and step S4 include:

[0059] The Hadamard coefficients Q(u, v) of each scanning base graph are obtained based on differential measurement:

[0060] Q(u,v)=Y + -Y -

[0061] Among them, Y + and Y - are the light intensity values ​​corresponding to the target image after being differentially modulated by the positive Hadamard scanning base image and the negative Hadamard scanning base image, and the expression for obtaining the light intensity value is:

[0062]

[0063] In the above formula, Y(N) is the light intensity value, i·K is the number of measurements corresponding to each coding pattern, K is the number of repeated acquisitions of the light intensity value during each modulation process, N is the modulation corresponding to the Nth coding pattern and target information, l is the number of light intensity values ​​that are discarded in the transition state of the digital micromirror array DMD, and y(x) is the original light intensity value received by the photodetector after being modulated by each mask pattern.

[0064] In this embodiment, a specific implementation step is as follows: Figure 3 As shown, the figure is a schematic diagram of the scanning base map generation principle; it is an example of generating a scanning base map by slicing reconstruction of an 8-order Hadamard matrix, and a scanning base map is composed of M column vectors. The scanning base map designed in this implementation is denoted as D M The modulation of the target image area is completed by scanning with a precision optical system, cooperating with the multi-frame projection of the DMD, and synchronizing the drive device and the acquisition device. In addition to the projection speed of the DMD, a faster scanning device means a higher imaging rate. The scanning base image is composed of multiple column vectors, each of which is denoted by d M :

[0065] D M =[d1,d2,···,d M ]

[0066] In order to enhance the anti-interference ability of Hadamard single-pixel imaging, differential measurement is usually used. At the same time, the differential mode is well adapted to the structural characteristics of DMD. When DMD is working, the mirror is tilted at different angles to achieve the function of reflecting or retaining light information in the light trap, and then complete the realization of a specific coded illumination pattern. Differential HSI is a manifestation of HSI, which allows the various Hadamard coefficients Q(u, v) to be obtained by differential measurement:

[0067] Q(u,v)=Y + -Y1

[0068] Among them, Y + and Y - are the light intensity values ​​corresponding to the target image after being differentially modulated by the positive Hadamard scanning base image and the negative Hadamard scanning base image, and the expression for obtaining the light intensity value is:

[0069]

[0070] In the above formula, Y(N) is a set of light intensity values, i·K is the number of measurements corresponding to each coding pattern, K is the number of repeated acquisitions of the light intensity value during each modulation process, N is the modulation corresponding to the Nth coding pattern and the target information, l is the number of light intensity values ​​that are discarded in the transition state of the digital micromirror array DMD, and y(x) is the original light intensity value received by the photodetector after being modulated by each mask pattern.

[0071] After calculating the Hadamard coefficient corresponding to the target image by differential method, the Hadamard spectrum is obtained:

[0072] Q=[Q1,Q2,Q3,···,Q M ].

[0073] Step S5: perform inverse transformation on the Hadamard spectrum to obtain a reconstructed target image.

[0074] In this embodiment, the Hadamard spectrum is inversely transformed to obtain the reconstructed target image:

[0075] P(x,y)=H -1 {Q(u,v)}

[0076] The formula for the reconstruction process is:

[0077] P = H·Q.

[0078] In this example, the galvanometer just traverses all areas (N*M) of the image during the rapid scanning process, and achieves 1 / 2N measurements in each frame. Therefore, only 2N coded patterns are needed when using the scanning base map proposed in this implementation for Hadamard single-pixel imaging, which reduces 1 / N pre-loaded modes compared to the original traditional Hadamard imaging technology, greatly reducing the loading time and the modulation time required for imaging. The imaging feature of row (column) modulation means that the imaging method proposed in this implementation is no longer limited to a square target area in terms of resolution, and no longer requires partition measurement and splicing to be compatible with image reconstruction of different resolutions.

[0079] like Figure 4As shown, the diagram is a diagram of the experimental device, which shows the fast column scanning single pixel imaging experimental architecture in this embodiment. The light source is an external focusing line laser with a wavelength of 650nm and a power of 50mw. It is fixed using a multi-axis translation stage with a micrometer head. The laser beam is irradiated onto a galvanometer mirror driven by a stable continuous triangular wave signal. The voltage value and frequency value of the driving signal form a certain mathematical logical relationship with the size of the scanning area and the scanning speed, generating a continuous and periodic precision optical scanning area. That is, through the synchronous cooperation between the hardware devices, a fast scanning optical path is realized to illuminate the image of the target area. The illumination light signal carrying the target information is synchronously reflected to the DMD micromirror with a preset coding pattern, and then the fixed photodetector captures the change in light intensity information. The optical signal is converted into an electrical signal and then synchronously collected by an oscilloscope.

[0080] In order to verify the effect of reconstructed images, two images of "Bird" and "flower" with different features and complexity with a resolution of 512*768 are used in numerical simulation to simulate the proposed method, and the structural similarity (SSIM), one of the more recognized indicators in the field of digital images, is used to quantitatively analyze the reconstruction performance of the method in principle. The value of SSIM is between [-1, 1]. The closer this value is to 1, the closer the image reconstruction effect is to the original image. Its expression is:

[0081]

[0082] Among them, u i and u j are the average values ​​of the original image and the reconstructed image, respectively. and σj 2 are the variances of i and j, σ ij is the covariance of i and j.

[0083] like Figure 5 As shown in the figure, the reconstruction effect diagram and the corresponding structural similarity index under different sampling rates (sap) in the numerical simulation are displayed. As the sampling rate decreases, the image quality begins to deteriorate. When the sampling ratio is too low, mosaic artifacts will be introduced.

[0084] Table 1 Comparison of various parameters between traditional HIS and this technology

[0085]

[0086] As shown in Table 1, the numerical comparison shows the performance of various parameters of the two technologies in image reconstruction (the graphics card of the computing device used in the data in this table can be NVIDIA GeForce RTX4080). The comparison shows that this technology has obvious advantages, which can greatly reduce the number of modulation patterns projected, thereby reducing the memory requirements and burn-in time of the preload mode, and has obvious advantages in reducing the image reconstruction time. This implementation has made great innovations in improving single-pixel imaging efficiency and reducing memory requirements.

[0087] like Figure 6 As shown in the figure, the reconstruction result based on fast scanning is shown, which shows the image reconstruction effect of 256*256 resolution under different complexity features. This implementation only uses 512 mask patterns to achieve modulation reconstruction of the target area. The traditional Hadamard single-pixel imaging technology requires 131072 mask patterns, and even under the condition of 1% compressed sampling, 1250 modulation patterns are required. The modulation scheme used in this implementation is even less than the number of masks for the traditional HIS undersampling of 1%, demonstrating the great potential of reducing memory resource consumption and improving imaging efficiency in the field of single-pixel imaging.

[0088] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents, and all should be included within the scope of protection of the present invention.

Claims

1. A fast single-pixel imaging method based on galvanometer scanning, characterized in that: The following steps are involved: Step S1, slicing and reconstructing the Hadamard base image according to the required resolution of the target to generate a scanning base image; Step S2, utilizing the periodic and continuous scanning characteristics of the galvanometer to accurately traverse the modulation process of the target area; Step S3, capturing the light intensity value modulated by the positive Hadamard scanning base image and the negative Hadamard scanning base image based on a differential measurement method; Step S4, calculating and obtaining the Hadamard coefficients of the corresponding frequency points to form a Hadamard spectrum; Step S5: perform inverse transformation on the Hadamard spectrum to obtain a reconstructed target image.

2. The fast single-pixel imaging method based on galvanometer scanning according to claim 1, characterized in that: The specific steps in step S1 include: Through the study and analysis of Hadamard matrix, the rows or columns of Hadamard matrix are sliced ​​to obtain several first-order row or column vectors; The slices are reconstructed by copying to obtain an innovative scanning base map, which is denoted as D M ; Among them, the scanning base graph is composed of M column vectors, each column vector is denoted by d M : D M =[d1,d2,···,d M ]。 3. The fast single-pixel imaging method based on galvanometer scanning according to claim 1, characterized in that: The specific steps in step S2 include: Using the scanning characteristics of the galvanometer to traverse multiple first-order modulation trains in the target area; Among them, the lateral resolution of the scanned base image depends on the image resolution requirement.

4. The fast single-pixel imaging method based on galvanometer scanning according to claim 1, characterized in that: The specific steps in step S3 and step S4 include: Using differential measurement and adapting to the structural characteristics of the digital micromirror array (DMD), the mirror of the digital micromirror array (DMD) is tilted at different angles during operation to reflect or retain light information in the light trap, thereby obtaining a specific coded illumination pattern. The Hadamard coefficient Q(u,v) of each scanning base graph is obtained based on the differential measurement method: Q(u,v)=Y + -AND - Among them, Y + and Y - are the light intensity values ​​corresponding to the target image after being differentially modulated by the positive Hadamard scanning base image and the negative Hadamard scanning base image, and the expression for obtaining the light intensity value is: In the above formula, Y(N) is the light intensity value, i·K is the number of measurements corresponding to each coding pattern, K is the number of repeated acquisitions of the light intensity value during each modulation process, N is the modulation corresponding to the Nth coding pattern and the target information, l is the number of light intensity values ​​that are discarded and belong to the transition state of the digital micromirror array DMD, and y(x) is the original light intensity value received by the photodetector after being modulated by each mask pattern; After calculating the Hadamard coefficient corresponding to the target image by differential method, the Hadamard spectrum is obtained: Q=[Q1,Q2,Q3,···,Q M ]。 5. The fast single-pixel imaging method based on galvanometer scanning according to claim 4, characterized in that: The specific steps in step S5 include: By inverse transforming the Hadamard spectrum, we can get the reconstructed target image: P(x,y)=H -1 {Q(u,v)} The formula for the reconstruction process is: P = H·Q.

Images