Two-dimensional high-resolution single-pixel imaging system under low-resolution coding

By using an imaging system with low-resolution encoding, combined with digital micromirror devices and a ring-coded mask for two-stage modulation and fast Fourier reconstruction, the limitations of high-speed imaging and high-resolution reconstruction in single-pixel imaging technology are solved, achieving efficient high-resolution imaging.

CN118984419APending Publication Date: 2024-11-19DALIAN UNIV OF TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410982982.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-22
Publication Date
2024-11-19

AI Technical Summary

Technical Problem

Existing single-pixel imaging technologies have limitations in high-speed imaging and high-resolution reconstruction, especially due to the slow switching rate of digital micromirror devices and the increased volume of ring-coded masks.

Method used

An imaging system employing low-resolution encoding achieves high-resolution imaging through two modulations using a digital micromirror device and a ring-coded mask, combined with fast Fourier transform reconstruction.

Benefits of technology

It achieves high-speed, high-resolution single-pixel imaging, avoiding the problems of switching rate limitations of digital micromirror devices and increased volume of ring-coded masks, thus reducing system cost and power consumption.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118984419A_ABST
    Figure CN118984419A_ABST
Patent Text Reader

Abstract

The invention discloses a two-dimensional high-resolution single-pixel imaging system under low-resolution coding. The system mainly comprises three parts: an imaging coding module, a signal collection module and an algorithm reconstruction module. An imaging coding module of the system realizes image low-resolution coding by using an annular coding mask, and full-sampling high-resolution reconstruction of a two-dimensional image is realized by combining sub-pixel sampling in a rotation direction with modulation of a spatial light modulator in a direction orthogonal to the rotation direction. The imaging coding module designed by the invention is composed of the annular coding mask and the spatial light modulator, the coding advantages of the mask and the spatial light modulator are combined, and meanwhile, the limitation of the sharp increase of the radius of the mask and the switching speed of the spatial light modulator under high-resolution imaging is avoided; and rapid two-dimensional high-resolution single-pixel imaging under full sampling is realized.
Need to check novelty before this filing date? Find Prior Art

Description

Technical field:

[0001] The present invention belongs to the field of coded imaging, and in particular relates to a single-pixel high-resolution imaging method and system. Background technology:

[0002] Single pixel imaging (SPI) is a two-dimensional computational reconstruction imaging method under one-dimensional sampling by modulating and demodulating spatial light information (light source or image plane). Specifically, the two-dimensional light source plane or image plane is cut into multiple small units by the modulation module, each small unit represents a pixel, and each pixel is assigned a different weight according to the modulation information. Finally, a single-point detector measures the weighted sum of all pixels. Through multiple (not less than the image resolution) different modulation measurements, the linear equation of weighted pixels can be inversely solved according to the detector detection value and modulation information to complete image reconstruction under full sampling.

[0003] As an emerging imaging method, compared with traditional multi-pixel array imaging, single-pixel detection has higher energy collection efficiency and is suitable for extremely weak light and long-distance scenes; it has high detection sensitivity, obvious advantages in invisible light bands and unconventional imaging, and very low cost. In addition, its structure is relatively simple, and its power consumption is significantly lower than that of array cameras. Although SPI increases the sampling time, it can meet general frame rate requirements under certain high-speed sampling and compression algorithms.

[0004] At present, there are two main types of common spatial light encoding methods: spatial light modulator (SLM) and annular mask. Among them, SLM mainly uses digital micromirror device (DMD), which uses a high-speed digital light reflection switch array to control the micromirror flip angle and adjusts the grayscale value through the duty cycle of the reversal time within the cycle to achieve optical signal reflection control.

[0005] When making a ring-coded mask, a long strip (01) matrix is ​​first generated. The long strip matrix consists of multiple sub-coding areas of the same size as the imaging. According to the different mapping methods of the long strip rectangle to the ring area, the production of the ring mask can be divided into two types. One is to map the sub-coding area of ​​the long rectangle with the element product of each segment of the ring-shaped light-transmitting area, and each small unit is a rectangle or dot. Although this method is simple, there is information loss at the edge of the sub-coding area; the second method is to directly transform the long rectangle into a ring area, and each small unit is a fan-shaped. This method avoids the above problems, but because it is a fan-shaped sub-area, the reconstructed image will have a small amount of image distortion, but the distortion can be reduced through mask design.

[0006] In addition to the above two methods, frosted glass or integrating sphere is also used as a modulation module. However, due to the random unknown nature of the encoding, the above two methods often require a large amount of data training, so they are generally less used.

[0007] Although the use of digital micromirror devices as spatial optical signal modulation modules can complete modulation tasks at different positions and resolutions in experiments with great flexibility, the speed of completing an image modulation has a low upper limit without considering undersampling. Taking the reconstruction of an image with a resolution of 41*43 as an example, at least 1763 modulations are required to complete a full sampling. Since the current maximum refresh rate of digital micromirror devices is 32kHz (binary image), the fastest time to complete an image modulation is still 55ms, and this time increases linearly with the increase of resolution. Therefore, it is difficult to perform high-speed imaging due to the influence of the switching frame rate of the digital micromirror device.

[0008] The emergence of the ring coding mask solves the problem of slow switching rate of digital micromirror devices. However, since its low-resolution ring coding area is transformed from a long rectangular coding area, when the reconstruction resolution increases and the pixel coding matrix increases, the radius of the ring coding mask increases and the weight increases, which puts higher requirements on the subsequent drive motor load, increases the cost and is not conducive to the miniaturization and integration of the overall system. Summary of the invention:

[0009] In order to solve at least one technical problem in the mainstream encoding method in the above background technology, an improved single pixel imaging system is proposed

[0010] To this end, the present application proposes a two-dimensional high-resolution single-pixel imaging system under low-resolution coding, which includes an imaging coding module, a signal collection module and an algorithm reconstruction module; wherein the imaging and imaging coding modulation of the target space by the imaging coding module include: an imaging mirror group and a digital micromirror device arranged on the same first optical axis as the target space, wherein the imaging mirror group is configured to image the target space in a stripe coding area of ​​the digital micromirror device; the digital micromirror device is configured to perform a first imaging and a first modulation in a first direction on the light signal of the target space from the imaging mirror group through the stripe coding area to form a first modulated light signal; a reflector group is configured to reflect the first modulated light signal to a second optical axis parallel to the first optical axis; a reflector group is arranged on a first optical axis to reflect the first modulated light signal to a second optical axis parallel to the first optical axis; a reflector group is arranged on a second ... The zoom lens group and the annular coding mask on the second optical axis, wherein the zoom lens group is configured to zoom the first modulated light signal to image the low-resolution annular coding area of ​​the annular coding mask, and the annular coding mask is configured to use the low-resolution annular coding area to perform a second modulation on the first modulated light signal in a rotation direction to form a second modulated light signal, wherein the rotation direction is orthogonal to the first direction; and a rotation driving module for driving the annular coding mask; the signal collection module is configured to collect the second modulated light signal; an algorithm reconstruction module is configured to perform fast Fourier calculation reconstruction based on the set modulation characteristics and the corresponding detection values ​​to restore the image information of the target space to achieve two-dimensional high-resolution single-pixel imaging under low-resolution coding.

[0011] In some embodiments, the reflector group includes a first reflector and a second reflector; the first reflector and the second reflector are adjusted by placement positions and lens pitch angles so that the second optical axis is parallel to the first optical axis.

[0012] In some embodiments, the imaging lens group is composed of one or more lenses so that the imaging size of the target space on the digital micromirror device is the same as the size of the stripe coding area of ​​the digital micromirror array.

[0013] In some embodiments, the digital micromirror device is tilted relative to the first optical axis so that the direction of the stripe coding region is the same as the direction of the sub-coding region of the annular coding mask.

[0014] In some embodiments, the digital micromirror device realizes spatial light encoding in the first direction by changing the stripes in the first direction, and has no resolution in the rotation direction orthogonal to the first direction. The digital micromirror device includes multiple groups of switchable stripe codes, each group of stripe codes includes non-reflective stripes and reflective stripes alternately arranged in the first direction; the non-reflective stripes correspond to a first value signal, and the reflective stripes correspond to a second value signal; the size of the change direction of each group of stripe codes is the same as the size of the radial direction of a sub-coding area; the number of switching times of the stripe code required to complete a high-resolution reconstruction is linearly related to the resolution of the annular coding mask and the desired reconstruction resolution; the change direction of the stripe code is orthogonal to the rotation direction of the annular coding mask, and the switching frequency of the stripe code is the same as the coding period of the annular coding mask; the stripe coding matrix corresponding to the stripe code of one modulation period is full rank.

[0015] In some embodiments, the zoom lens group is composed of two or more lenses, and is configured to image the first modulated light signal for a second time in the sub-coding area of ​​the annular coding mask, and to make the size ratio of the first image and the second image equal to the area ratio of the first modulation area and the second modulation area.

[0016] In some embodiments, the annular coding mask is used as a second spatial light modulation unit for low-resolution two-dimensional coding; wherein the annular coding mask includes a low-resolution annular coding area including multiple sub-coding areas obtained by orthogonal transformation of a binary coding matrix of a long rectangular strip, and an opaque marking area arranged after the low-resolution annular binary; wherein a supplementary column is provided between the last sub-coding area of ​​the low-resolution annular binary coding area and the marking area, and the encoding of the supplementary column is the same as the encoding of the last column of the first sub-coding area of ​​the low-resolution annular binary coding area; one rotation of the annular coding mask is one coding cycle; each of the sub-coding areas has cyclicity and sparsity.

[0017] In some embodiments, the rotation drive module uses a micro motor in conjunction with a micro motor controller.

[0018] In some embodiments, the signal collection module includes a converging lens group and a single-point detector, and the converging lens group is configured so that the second modulated light signal is converged through the converging lens group and completely collected by the single-point detector, and finally measured and transmitted by the single-point detector; the converging lens group is composed of one or more lenses to ensure that the spot size of the second modulated light signal after the second modulation is not larger than the detection area of ​​the single-point detector after converging.

[0019] In some embodiments, high-resolution reconstruction of the circular coding mask in its rotation direction is achieved by performing fast Fourier calculation reconstruction through sub-pixel sampling of the circular coding mask, and high-resolution reconstruction in a first direction orthogonal to the rotation direction is achieved by switching the stripe coding of the digital micromirror device during the rotation cycle of the circular coding mask, thereby achieving two-dimensional high-resolution single-pixel imaging under low-resolution coding.

[0020] The algorithm reconstruction module of the high-resolution single-pixel imaging system under low-resolution coding described in the present invention completes the reception of the measurement signal and the algorithm reconstruction of the image. Fast Fourier calculation reconstruction is performed according to the cyclicity and sparsity of the two codes and their corresponding detection values ​​to realize the image information recovery of the target space. Furthermore, high-resolution reconstruction of the rotation direction of the annular mask is realized by fast Fourier calculation reconstruction through sub-pixel sampling of the annular coding mask. And high-resolution reconstruction in the other direction is completed through the stripe coding of the digital micromirror device in each rotation cycle, thereby realizing two-dimensional high-resolution single-pixel imaging under low-resolution coding. The high-resolution single-pixel imaging system under low-resolution coding provided by the present invention combines the advantages of two coding methods, realizes high-speed, high-resolution single-pixel imaging, breaks through the limitation of the switching rate of the digital micromirror device, and avoids the surge in the volume of the annular coding mask under high-resolution reconstruction. Description of the drawings:

[0021] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0022] Figure 1 1 is a schematic diagram of the optical path structure of a two-dimensional high-resolution single-pixel imaging system under low-resolution encoding according to an embodiment of the present invention;

[0023] Figure 2 is a schematic diagram of a circular coding mask in an embodiment of the present invention;

[0024] FIG3 is a schematic diagram of a long rectangular code for a ring-shaped coding mask;

[0025] Figures 4A to 4D They are schematic diagrams of four-frame stripe-type encoding of a digital micromirror device when m=4 used in the embodiments of the present application;

[0026] Figure 5A is a control flow chart d of a two-dimensional high-resolution single-pixel imaging system under low-resolution encoding according to an embodiment of the present invention;

[0027] Figure 5B is a two-dimensional image reconstruction flow chart of a two-dimensional high-resolution single-pixel imaging system under low-resolution encoding according to an embodiment of the present invention;

[0028] Figure 6 is a schematic diagram of the principle of low-resolution image reconstruction based on single-pixel imaging according to an embodiment of the present invention;

[0029] Figure 7 1 is a schematic diagram of the principle of unidirectional high-resolution reconstruction under the regulation of long rectangular low-resolution coding in an embodiment of the present invention;

[0030] Fig. 8A It is a simulated waveform diagram of the light intensity of a single-point detector with low resolution sampling under the long rectangular coding of an embodiment of the present invention;

[0031] Figure 8B It is a simulated waveform diagram of the light intensity sampled in a single direction with high resolution by a single-point detector under the long rectangular low-resolution coding control of an embodiment of the present invention;

[0032] Fig. 9A is a randomly generated image of size 44×52 in an embodiment of the present invention;

[0033] Fig. 9B is a reconstructed image of the 44×52 image under the encoding of only the low-resolution circular encoding mask, wherein the horizontal and vertical axes represent the position and the grayscale represents the relative amplitude;

[0034] Fig. 9C is the result of stitching four groups of sub-pixel images by fast Fourier inverse solution according to the sampling signal when n=4, that is, the high-resolution reconstruction image in the rotation direction;

[0035] Fig.9D It is a high-resolution two-dimensional analog simulation reconstruction image after two modulations based on a digital micromirror device and a rotating annular coding mask in an embodiment of the present invention;

[0036] Among them, (1) target space, (2) imaging lens group, (3) first reflector, (4) digital micromirror device, (5) second reflector, (6) zoom lens group, (7) annular coding mask, (8) rotation drive module, (9) convergence lens group, (10) single point detector, and (11) PC signal processing end. Specific implementation method:

[0037] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0038] This example provides a new single-pixel imaging solution, which avoids the infinite expansion of the ring-coded mask through two low-resolution modulations of the digital micromirror array and the ring-coded mask, and realizes fast high-resolution image reconstruction under low-speed switching.

[0039] Unless otherwise defined, the circular coding mask rotation period, low-resolution sampling period, sub-pixel sampling period, and coding period in this application all refer to the period of one rotation of the circular coding mask, among which the low-resolution sampling period and sub-pixel sampling period are used to distinguish the circular coding in the low-resolution algorithm and the high-resolution algorithm.

[0040] Full rank period, modulation period, and reconstruction period should all refer to the sampling period that completes a complete reconstruction process, that is, a DMD switching period (multiple ring coding mask sampling periods). Full rank period specifically describes the coding characteristics of a DMD period.

[0041] Reference Figure 1 In some embodiments of the present application, the two-dimensional high-resolution single-pixel imaging system structure under low-resolution encoding includes an imaging encoding module 100, a signal collection module 200, and an algorithm reconstruction module 300.

[0042] The imaging encoding module 100 includes an imaging lens group 2, a digital micromirror device 3, a first reflector 4, a second reflector 5, a zoom lens group 6, an annular encoding mask 7, and a rotation drive module 8. The target 1 is first clearly imaged on the digital micromirror device 3 through the imaging lens group 2 and the digital micromirror device 3 of the first optical axis set horizontally. According to the characteristics of the digital micromirror device 3, the digital micromirror device 3 can be placed tilted so that the direction of its stripe encoding area is consistent with the direction of the pixel encoding area of ​​the annular encoding mask 7. The digital micromirror device 3 first loads the first group of stripe codes, and performs the first modulation, i.e., the first spatial encoding, on the signal light carrying the imaging information of the target from the imaging lens group 2, to obtain a first modulated light signal; the first modulated light signal adjusts the optical axis through the first reflector 4 and the second reflector 5, so that the second optical axis of the outgoing light of the second reflector 5 is horizontal, so that the first modulated light signal is clearly imaged for the second time on the annular coding mask 7 through the zoom lens group 6 arranged on the second optical axis. The zoom lens group 6 can perform high-quality scaling on the image after the first modulation, so that the second imaging matches the size of a sub-coding area of ​​the annular coding mask 7, such as 701a, so that the annular coding mask 7 performs the second modulation, i.e., the second spatial encoding, on the imaging information, to obtain a second modulated light signal.

[0043] The signal collection module 200 includes a converging lens group 9 and a single point detector 10 arranged on the second optical axis. The converging lens group 9 and the single point detector 10 are configured so that the second modulated light signal after the second spatial encoding is converged through the converging lens group 9 so that the second modulated light signal carrying the twice modulated information of the target 1 can be completely collected by the single point detector 10.

[0044] The algorithm reconstruction module 300 may be a PC signal processing terminal 11, which is configured to receive data of at least one modulation cycle from the single point detector 10 and complete a two-dimensional high-resolution image reconstruction process. The two-dimensional high-resolution image reconstruction process of the PC signal processing terminal 11 divides the cycle by calculating the temporal correlation, and determines the starting position of each cycle by the characteristics of the marking area 702 of the annular coding mask 7. For example, the starting position is that the target image completely overlaps with the first sub-coding area 701a.

[0045] In the two-dimensional high-resolution image reconstruction process, it is assumed that the resolution of the low-resolution image reconstructed using the annular coding mask 7 is p×q (p is the longitudinal resolution, q is the lateral resolution). The reconstruction resolution of the high-resolution image reconstructed through sub-pixel sampling of the annular coding mask 7 and the first modulation of the digital micromirror device 3 is mp×nq.

[0046] The rotation driving module 8 is used as a switching driving module for the low-resolution annular binary coding area of ​​the annular coding mask 7. The rotation driving module 8 can use a micro motor in conjunction with a micro motor controller to achieve low-cost micro unit control.

[0047] The specific structure of the annular coding mask 7 in this embodiment, the specific structure of the stripe coding of the digital micromirror device 3, the low-resolution single-pixel computational imaging principle, the two-dimensional high-resolution single-pixel imaging principle under low-resolution coding, and a two-dimensional high-resolution single-pixel imaging system under low-resolution coding are described in detail below:

[0048] Fringe encoding of digital micromirror device 3:

[0049] Since the switching speed of binary images is much higher than the switching frequency of grayscale images, the fastest speed can reach 32kHz, therefore, the stripe encoding of the digital micromirror device 3 adopts only two states, namely reflection and non-reflection, corresponding to 1 and 0 of the digital signal respectively. Figures 4A to 4D In the figure, black lines are used to represent the stripes in the reflective state of the digital micromirror device 3 and blank areas are used to represent the stripes in the non-reflective state. The digital micromirror device 3 in this embodiment corresponds to the situation where imaging is performed at two positions in the longitudinal direction of the annular mask.

[0050] This stripe coding in the vertical direction makes the digital micromirror device 3 have no resolution in the rotation direction of the annular coding mask 7 of this embodiment, that is, in the horizontal direction, but has a resolution of m×p in the vertical direction.

[0051] Each frame of the stripe code of the digital micromirror device 3 is composed of p groups of repeated m×1-dimensional [0-1] matrices. Furthermore, each m frames of the stripe code of the digital micromirror device 3 is set as a complete two-dimensional reconstruction sampling period, and the switching frequency of each frame of the stripe code is the same as the rotation frequency of the annular coding mask 7. The m×m-dimensional [0-1] matrix composed of the m×1-dimensional vectors switched m times within the two-dimensional reconstruction sampling period is full rank. For example, when m=4, the encoding method of the four-frame stripe code of the digital micromirror device 3 within a complete two-dimensional reconstruction sampling period can be as follows: FIG. 4A to FIG. 4D It should be noted that 4A to 4D The four specific stripe codes in the first stripe code 31a, the second stripe code 31b, the third stripe code 31c, and the fourth stripe code 31d only represent a stripe coding method of the digital micromirror device 3 when m=4, and any loading order can be used, for example, according to FIG. 4A to FIG. 4D The loading order, or FIG. 4D to FIG. 4A The loading order, or Figure 4A , Figure 4C , Figure 4D , Figure 4B Other matrices that satisfy the above m×m dimensional 0-1 matrix full rank condition can all be the fringe encoding method of the digital micromirror device 3.

[0052] The stripe codes 31a, 31b, 31c, and 31d of the digital micromirror device 3 are respectively used as high-resolution reconstruction codes of the p×q low-resolution image in the longitudinal direction, which actually subdivides each longitudinal pixel into m stripes. Therefore, in this embodiment, the coordinate system of the stripe coding area of ​​the digital micromirror device 3 should be consistent with the imaging coding coordinate system of the annular coding mask 7, and the longitudinal size of each stripe in the m×1-dimensional stripe coding area is the same as the longitudinal size of a single pixel in the sub-coding area. For example, the lateral direction of the stripe extension of the stripe coding area of ​​the digital micromirror device 3 is consistent with the tangent direction of the rotation direction of the sub-coding area of ​​the fan, and the longitudinal direction of the stripe superposition of the stripe coding area of ​​the digital micromirror device 3 is consistent with the radial direction of the sub-coding area of ​​the fan.

[0053] In this embodiment, the stripe coding of the digital micromirror device 3 has no resolution in the horizontal direction, and the length of the horizontal stripe is not less than the imaging length of the target.

[0054] However, it should be understood that the stripe coding corresponding to different imaging positions, such as the two lateral positions of the annular coding mask, can also be in the horizontal direction, that is, the coding stripes are spaced apart on the left and right. In this way, it has no resolution in the longitudinal direction in the rotation direction of the annular coding mask 7 of other embodiments, but has a resolution of p×m in the lateral direction.

[0055] The code of the ring coding mask 7 is:

[0056] The structure of the ring coding mask 7 is as follows: Figure 2 As shown, it is composed of a mask plate, and the annular coding mask described in this embodiment has a low-resolution annular binary coding area 701 and a marking area 702 connected between the beginning and the end of the annular binary coding area 701. The annular binary coding area 701 includes a plurality of sub-coding areas, such as a first sub-coding area 701a, a second sub-coding area 701b, a last sub-coding area 701n, and a plurality of sub-coding areas represented by ellipsis in the middle, and the coding of each sub-coding area has two states of light-transmitting and light-impermeable, which correspond to 1 and 0 of the digital signal respectively.

[0057] The low-resolution annular binary coding region can be obtained by linear transformation of the long rectangular coding region 601. Figure 3A As shown, it can include a plurality of continuous rectangular sub-coding regions, such as 601a and 601b. In some embodiments, the rectangular sub-coding regions partially overlap, for example, the last few columns of the coding of the previous rectangular sub-coding region 601a are the same as the first few columns of the coding of the next rectangular sub-coding region 601b, forming a repeated region 601ab. Figure 3B As shown in the area within the dotted box in . This setting can improve the utilization rate of the low-resolution rectangular binary coding area, and at the same time can avoid the problem of increased area of ​​the entire annular coding mask caused by the length of the coding mask. For example, in some embodiments, a dual prime algorithm can be used to generate a basis function, and a pixel coding matrix is ​​generated by cyclic shift as the long rectangular coding. Taking the 17×19 low-resolution image reconstruction coding as an example, the size of each rectangular sub-coding area 601a is a 17×19 pixel coding matrix, and the last 18 columns of the coding of the previous rectangular sub-coding area 601a overlap with the first 18 columns of the coding of the next sub-coding area 601b, such as shown in the overlapping area 601ab. At the same time, a column of the last column of the first rectangular sub-coding area 601a is added to the end of the entire low-resolution rectangular binary coding area 601. The low-resolution long rectangular binary coding area finally generated is as shown in FIG. Figure 2 shown.

[0058] The low-resolution annular binary coding area 701 finally generated by the long rectangular coding area 601 is as follows Figure 2 As shown, its structure is the same as that of the low-resolution rectangular binary coding area, except that each rectangular sub-coding area is converted into a fan-shaped sub-coding area. The annular coding mask 7 is preferably designed to ensure that the position of the coding in the annular coding mask 7 is close to the edge of the mask plate, and the minimum area size should be much smaller than the radius of the mask plate, so as to reduce the image distortion caused by the difference in the inner and outer diameters of the fan-shaped sub-coding area of ​​the annular coding mask 7.

[0059] 2D high-resolution single-pixel imaging with low-resolution encoding:

[0060] The high-resolution single-pixel imaging system in the present application realizes high-resolution reconstruction of the annular coding mask 7 in its rotation direction through sub-pixel sampling. Every time the annular coding mask 7 rotates one circle, the digital micromirror device 3 performs full-rank stripe switching to perform high-resolution encoding in the direction orthogonal to the rotation direction. When the digital micromirror device 3 completes one full-rank cycle switching, the high-resolution reconstruction signal sampling and inverse decoding in this direction are realized, thereby completing the two-dimensional high-resolution reconstruction under low-resolution coding.

[0061] The above process is divided into two parts: sampling control process and two-dimensional image reconstruction process. The sampling control process is as follows: Figure 5A As shown; the process of two-dimensional image reconstruction is as follows Figure 5B shown.

[0062] like Figure 5A As shown, the sampling control process includes configuring the digital micromirror device to load the first stripe code 31a, step S101; driving the annular coding mask 7 to perform sub-pixel sampling for one cycle, step S102; configuring the digital micromirror device to load the stripe code 31b, step S103; determining whether the digital micromirror device has completed the switching of a reconstruction cycle, step S104; if not, continue to execute steps S102 and S103; if it has been completed, the image data collected in the whole process can be used as a data set for image reconstruction, step S105.

[0063] As shown in the figure, for the two-dimensional image reconstruction process, in general, Figure 5B As shown, it includes the following main steps: extracting a sampling data group in a sub-pixel sampling process of a ring-coded mask under a stripe code in the data set, such as a first sampling data group corresponding to a first stripe code, step S201; performing high-resolution sub-pixel reconstruction on the first sampling data group, step S202; switching to another sampling data group under another stripe code, such as a second sampling data group under a second stripe code, step S203; performing high-resolution sub-pixel reconstruction on the second sampling data group, step S204; judging whether the high-resolution sub-pixel reconstruction of all sampling data groups in the data set is completed, step S205; if not, returning to step S203, continuing to switch the sampling data group and performing high-resolution sub-pixel reconstruction; if it is completed, completing the two-dimensional high-resolution reconstruction, step S206.

[0064] It should be understood that the above control process and two-dimensional image reconstruction process can also be adjusted to be performed in real time, that is, after obtaining a sampling data group corresponding to a stripe code, it is reconstructed with high resolution sub-pixel without waiting until all the data sets are collected.

[0065] The two-dimensional high-resolution single-pixel imaging method under low-resolution coding proposed in the present application combines the digital micromirror device 3 with the annular coding mask 7, thereby avoiding the problems of increased switching times of the digital micromirror device 3, linear decrease in imaging speed, increased area of ​​the annular coding mask 7, increased cost, and increased motor load caused by increased reconstruction resolution.

[0066] The specific working principle of the system proposed in this application is as follows:

[0067] The pixel size reconstructed under low-resolution full sampling is the same as the minimum coding area size. The moving distance of each sampling is the same as the length of the minimum coding area. The sampling value of each time is the product of the image and the element of the current sub-coding area, as shown in the attached figure. Figure 7 As shown in the figure, the initial two adjacent sampling signal values ​​are [y1, y2]. In a low-resolution sampling period, that is, during the process of the annular coding mask 7 rotating one circle, the normalized analog sampling light intensity signal in its marking area 702 is as shown in the figure. Fig. 8A As shown, the horizontal axis represents the sampling time order.

[0068] However, since the rotation process of the annular coding mask 7 is continuous, the sliding window size of the sampling coding can be set to be much smaller than the minimum region length under low-resolution coding.

[0069] By utilizing the above characteristics, the two-dimensional high-resolution single-pixel imaging under the low-resolution encoding can adopt sub-pixel sampling. Figure 7 As shown in the figure, when the resolution in the rotation direction needs to be increased by 4 times, the moving distance of each sampling is 1 / 4 times the length of the minimum encoding area, and the signal [y1, y2] between two pixels at low resolution is split into In one sub-sampling period, that is, the annular coding mask rotates one circle, the normalized analog sampled light intensity signal in its mark area 702 is as follows: Figure 8B As shown, the horizontal axis represents the sampling time order.

[0070] At low resolution, the image P = [p1, p2, ..., p q ], pi is a column vector, and at high resolution p i was evenly cleaved. in

[0071] Therefore, the sub-pixel size is Among them, L p It represents the minimum region length under low-resolution encoding, and n represents the number of subdivisions in the rotation direction of the annular encoding mask.

[0072] In this embodiment, a randomly generated image of size 44×52 is taken as an example. The original image of the image can be found in Fig. 9A, the sub-pixel sampling value from one low-resolution sub-coding area to the next low-resolution sub-coding area during the rotation of the annular coding mask is (n+1) times: Among them, the subscript s represents the subcoding region sequence, Represents the i-th column of the first subcoding region.

[0073] Since the pixel encoding matrix has a cyclic characteristic, the above formula can be rewritten as:

[0074]

[0075] Subtracting adjacent sequences (2)-(1), ..., (n)-(n-1), (n+1)-(n), we get Right now:

[0076] Since s1 and s2 are both cyclic arrays, (s2-s1) still maintains the cyclic characteristics. Therefore, the sub-pixel solution can still be solved by fast Fourier transformation, that is, in This method can avoid the time increase problem caused by the increase in dimension. At the same time, the operations between each sub-pixel are independent. In this embodiment, parallel operations are applied. Reassemble and splice to get That is the original image P. When p=11, q=13, and n=4, the specific reconstruction effect can be found in Appendix 9C, where the horizontal and vertical coordinates represent positions, and the grayscale represents relative amplitude.

[0077] When the annular mask rotates once, the digital micromirror device 3 performs full-rank stripe switching to perform high-resolution encoding in the vertical direction. When the digital micromirror device 3 completes a full-rank cycle switching, the sub-pixel sampling in the rotation direction orthogonal to the vertical direction is completed. According to the full-rank characteristics of the stripe encoding of the digital micromirror device 3 within a sub-pixel sampling cycle, the high-resolution pixel value in the vertical direction is reversed to complete the two-dimensional high-resolution reconstruction. When p = 11, q = 13, m = 4, the specific two-dimensional high-resolution reconstruction effect is shown in the attached figure. Fig.9D In the figure, both the horizontal and vertical axes represent the position, and the grayscale represents the relative amplitude. Fig.9D and the original image, i.e. Fig. 9A It can be seen that the two are the same.

[0078] Comparative example: low-resolution single-pixel computational imaging.

[0079] The specific principle of low-resolution single-pixel computational imaging based on annular coded mask is as follows:

[0080] Single-pixel imaging is the process of using a single-point bucket detector to collect the weighted signal sum of each pixel in the image and the weighted value to invert the image information. Single-pixel imaging can be divided into the following categories: Figure 6 The two processes shown are: image modulation signal acquisition and image reconstruction.

[0081] As attached Figure 6 As shown, during the image acquisition process, the target object 81 generates a clear image P on the imaging plane of the imaging lens group through the imaging lens group 82, and a ring-shaped coding mask 87 is placed on the imaging plane. The image P is the same size as a sub-coding area of ​​the ring-shaped coding mask 87, as shown in the attached figure. Figure 6 As shown, a sub-coding area is composed of a plurality of continuous small "areas" with adjustable light transmittance. Each minimum area is a low-resolution pixel. Assuming that the size of a sub-coding area is p×q, each time the sub-coding area is switched, it moves a pixel length at a low resolution. After moving p×q times, the modulation of the object's transmitted light in a complete coding cycle is completed. The modulated light passes through the converging lens group 89 and is finally measured by a single-point detector 810. The modulated light signal y i (i=1, 2…p×q), complete the acquisition of image modulation signal. In order to simplify the coding, usually any t i Any "area" of the mask corresponding to the moment has only two states: transparent and opaque. In digital signals, it can be directly represented by 1 and 0. The signal collected by the single-point detector 810 at this moment can be expressed as in Indicates t i The two-dimensional sampling matrix corresponding to the time, the superscript i represents the order number of the sub-coding area, j and k represent the row and column of the i-th sub-coding area respectively, and y i The measurement value corresponding to a single-point detector is the sum of the weighted signals of all pixels. Therefore, the image reconstruction and demodulation process is essentially a process of solving a multivariate linear equation system.

[0082] In this comparative example, since the ring coding is cyclic, the above linear equation can be replaced by circular convolution: Where s1 represents the first sub-coding matrix. According to circular convolution theorem 2, in this embodiment, FFT is used to represent the sampling equation: F{y}=F{s1}·F{x}, where represents Fourier operation, then in this embodiment, the pixel matrix can be inversely solved as:

[0083] In the same Fig. 9A In the image shown, the imaging effect of only reconstructing the image by encoding the low-resolution ring-coded mask is poor. The results of reconstructing the image by encoding the low-resolution ring-coded mask are shown in the attached figure. Fig. 9B, and the two-dimensional high-resolution single-pixel imaging system under low-resolution coding described in this application can achieve higher-resolution image reconstruction by combining the low-resolution annular coding mask with a digital micromirror device, and the imaging effect is significantly improved. The results are shown in Fig.9D .

[0084] The above contents are further detailed descriptions of the present invention in combination with the preferred technical solutions, and the specific implementation of the present invention cannot be limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, simple deductions and substitutions can be made without departing from the concept of the present invention, which should be regarded as the protection scope of the present invention.

Claims

1. A two-dimensional high-resolution single-pixel imaging system under low-resolution coding, characterized in that: It includes imaging coding module, signal collection module and algorithm reconstruction module; The imaging and encoding modulation of the target space by the imaging and encoding module include: An imaging lens group (2) and a digital micromirror device (3) are arranged on the same first optical axis as the target space, wherein the imaging lens group is configured to image the target space in a stripe coding region of the digital micromirror device (3); the digital micromirror device (3) is configured to perform a first imaging and a first modulation in a first direction on a light signal of the target space from the imaging lens group (2) through the stripe coding region to form a first modulated light signal; a reflector assembly configured to reflect the first modulated optical signal to a second optical axis parallel to the first optical axis; a zoom lens group (6) and an annular coding mask (7) arranged on the second optical axis, wherein the zoom lens group (6) is configured to zoom the first modulated light signal to form an image on a low-resolution annular coding area of ​​the annular coding mask (7), and the annular coding mask (7) is configured to perform a second modulation on the first modulated light signal in a rotation direction using the low-resolution annular coding area to form a second modulated light signal, wherein the rotation direction is orthogonal to the first direction; and A rotation driving module (8) for driving the annular coding mask (7); The signal collection module is configured to collect the second modulated optical signal; The algorithm reconstruction module is configured to perform fast Fourier calculation reconstruction based on the set modulation characteristics and the corresponding detection values ​​to restore the image information of the target space so as to achieve two-dimensional high-resolution single-pixel imaging under low-resolution coding.

2. The two-dimensional high-resolution single-pixel imaging system under low-resolution coding according to claim 1, characterized in that: The reflector group comprises a first reflector (4) and a second reflector (5); the first reflector (4) and the second reflector (5) are arranged so that the second optical axis is parallel to the first optical axis by adjusting the placement position and the pitch angle of the lens.

3. The two-dimensional high-resolution single-pixel imaging system under low-resolution coding according to claim 1, characterized in that: The imaging mirror group is composed of one or more lenses, so that the imaging size of the target space on the digital micromirror device (3) is the same as the size of the stripe coding area of ​​the digital micromirror array.

4. The two-dimensional high-resolution single-pixel imaging system under low-resolution coding according to claim 1, characterized in that: The digital micromirror device (3) is arranged to be inclined with respect to the first optical axis so that the direction of the stripe coding area is the same as the direction of the sub-coding area of ​​the annular coding mask.

5. The two-dimensional high-resolution single-pixel imaging system under low-resolution coding according to claim 1, characterized in that: The digital micromirror device (3) realizes spatial light encoding in the first direction by changing the stripes in the first direction, and has no resolution in the rotation direction orthogonal to the first direction. The digital micromirror device (3) includes a plurality of groups of switchable stripe codes, each group of stripe codes includes non-reflective stripes and reflective stripes alternately arranged in the first direction; the non-reflective stripes correspond to a first value signal, and the reflective stripes correspond to a second value signal; The size of the changing direction of each group of stripe codes is the same as the size of a sub-coding area in the radial direction; The number of switching times of the stripe code required to complete a high-resolution reconstruction is linearly related to the resolution of the annular coding mask and the resolution of the desired reconstruction; The changing direction of the stripe code is orthogonal to the rotation direction of the annular code mask (7), and the switching frequency of the stripe code is the same as the coding period of the annular code mask; The fringe coding matrix corresponding to the fringe coding of one modulation period is full rank.

6. The two-dimensional high-resolution single-pixel imaging system under low-resolution coding according to claim 1, characterized in that: The zoom lens group is composed of two or more lenses, and is configured to image the first modulated light signal for a second time in the sub-coding area of ​​the annular coding mask, and to make the size ratio of the first image and the second image equal to the area ratio of the first modulation area and the second modulation area.

7. The two-dimensional high-resolution single-pixel imaging system under low-resolution coding according to claim 1, characterized in that: The annular coding mask (7) is used as a second spatial light modulation unit for low-resolution two-dimensional coding; wherein the annular coding mask (7) comprises a low-resolution annular coding area including a plurality of sub-coding areas obtained by orthogonal transformation of a long rectangular binary coding matrix, and an opaque marking area arranged after the low-resolution annular binary; wherein a supplementary column is provided between the last sub-coding area of ​​the low-resolution annular binary coding area and the marking area, and the code of the supplementary column is the same as the code of the last column of the first sub-coding area of ​​the low-resolution annular binary coding area; One rotation of the annular coding mask constitutes one coding cycle; Each of the sub-coding regions has cyclicity and sparseness.

8. The two-dimensional high-resolution single-pixel imaging system under low-resolution coding according to claim 1, characterized in that: The rotation drive module adopts a micro motor and a micro motor controller.

9. The two-dimensional high-resolution single-pixel imaging system under low-resolution coding according to claim 1, characterized in that: The signal collection module comprises a converging lens group (9) and a single-point detector (10); the converging lens group (9) is configured so that the second modulated light signal is converged by the converging lens group (9) and completely collected by the single-point detector (10), and finally measured and transmitted by the single-point detector (10); the converging lens group is composed of one or more lenses so as to achieve a light spot size after the second modulated light signal is converged and is not larger than the detection area of ​​the single-point detector (10).

10. The two-dimensional high-resolution single-pixel imaging system under low-resolution coding according to claim 1, characterized in that: The high-resolution reconstruction of the annular coding mask in its rotation direction is achieved by performing fast Fourier calculation reconstruction through sub-pixel sampling of the annular coding mask, and the high-resolution reconstruction in a first direction orthogonal to the rotation direction is achieved through the stripe coding switching of the digital micromirror device (3) during the rotation cycle of the annular coding mask, thereby achieving two-dimensional high-resolution single-pixel imaging under low-resolution coding.