An encoding-based frequency up-conversion imaging system and method

Abstract

The application discloses a kind of based on encoding frequency up-conversion imaging system and method, the system is mainly by light field modulation module, pump light beam module, frequency up-conversion module, filter system and signal acquisition module etc. Part composition.This method utilizes the characteristics that wavelet decomposition of image can separate high-frequency component in image, by wavelet decomposition to the low-resolution image of target scene, and according to the high-frequency component in the image obtained by decomposition estimates the region where the details in target scene, finally measures the wavelet coefficient of these regions under high resolution and reconstructs image, realizes the purpose of reducing measurement times and improving imaging speed under the premise of retaining image detail information.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of infrared imaging technology, and specifically to a frequency up-conversion imaging system and method based on coding. BACKGROUND

[0002] In recent years, high-sensitivity infrared imaging and infrared spectroscopy have attracted more and more attention because many molecules and compounds have strong absorption transitions in the infrared band. This makes weak signal imaging and fast infrared spectroscopy helpful for identifying complex molecules without staining and sample preparation. With the emergence of supercontinuum (SC) light sources and quantum cascade lasers (QCLs), replacing traditional infrared optical parametric oscillators (OPOs) or OPAs, the latest progress in infrared light sources has been greatly changed. However, the technology of weak infrared detection is still far behind. Currently, infrared detectors exist in different forms, but they are all limited by the inevitable dark noise caused by the thermal radiation of the detector itself at a limited temperature. Therefore, high-sensitivity infrared detectors need to be cooled to reduce noise. This is in sharp contrast to visible (VIS) or near-infrared (NIR) detectors, as their thermal radiation (dark noise) is not as noticeable at room temperature.

[0003] The upconversion detection overcomes the common difficulties of traditional mid-infrared detectors, and can work at room temperature. The essence of the upconversion detection is a sum-frequency generation process, that is, the weak MIR signal and the strong pump laser interact with each other in a nonlinear medium, and the MIR information is transmitted to the near-infrared or visible wavelength range, and then a high-efficiency, sensitive and reasonably-priced silicon-based photonic detector or a CCD camera working in the near-infrared region is used to detect the upconversion signal. Although the frequency upconversion imaging technology has been greatly improved in conversion efficiency and imaging speed, there are still some contradictions in the analysis of the point spread function of the frequency upconversion imaging system: when the pump light power is constant, the smaller the beam waist diameter of the pump light is, the higher the conversion efficiency is, and at the same time, the point spread function of the system is also closer to the two-dimensional Gaussian function, and the image is more blurred. Christian Pedersen and others pointed out this contradiction when they proposed the mathematical model of frequency upconversion imaging. They believed that although the use of pump light with a larger beam diameter or strong convergence of signal light is beneficial to improve the image resolution, the former will cause the conversion efficiency to decrease, and the latter is limited by the size of the nonlinear crystal, which leads to the reduction of the field of view of the system. They proposed that when constructing the frequency upconversion imaging system, only one of the image resolution and the conversion efficiency can be optimized, and the other unoptimized performance is sacrificed. In addition, when the spectral range of the signal light is narrow, if the temperature scanning method is used to obtain the infrared image of the entire field of view, the imaging speed will be greatly reduced; if a wide-spectrum light source is used to generate pump light, the signal light in the entire field of view can be phase-matched at the same time, which is easy to cause the phenomenon of target blur in the experimental results of Romain Demur and others. SUMMARY

[0004] The present application aims to provide an encoding-based frequency upconversion imaging system to reduce the measurement times of the encoding frequency upconversion imaging method and improve the imaging speed.

[0005] The technical scheme for achieving the object of the present application is as follows: a frequency up-conversion imaging system based on coding, comprising: a light field modulation module, a pump light beam module, a frequency up-conversion module, a filtering system and a signal acquisition module, the light field modulation module is used for adjusting signal light to a specified polarization state and measuring the light intensity of multiple positions of a target scene, realizing the coding function of the target scene, the pump light beam module is used for providing pump light of a specified polarization state, the frequency up-conversion module comprises a dichroic mirror, a PPLN crystal, a temperature control device and a first 4f system, the dichroic mirror is used for combining the signal light and the pump light provided by the pump light beam module, so that the two beams of light are collinearly propagated to the incident surface of the PPLN crystal, the PPLN crystal is arranged in the temperature control device, and the temperature control device is used for providing a constant temperature environment; the first 4f system is used for correcting the chromatic aberration of the sum frequency light of the PLN crystal exit surface and scaling the image ratio; the filtering system is used for filtering out the noise mixed in the sum frequency light; and the detection module is used for acquiring the sum frequency light filtered by the filtering system.

[0006] Preferably, the light field modulation module comprises a first optical fiber collimator, a first half-wave plate, a resolution plate, a second 4f system, a digital micro mirror and a third 4f system, after the signal light is collimated by the first optical fiber collimator, the signal light passes through the first half-wave plate to irradiate the resolution plate, and the image passing through the resolution plate is scaled by the 4f system at a proper ratio, so that the digital micro mirror array working area of the digital micro mirror can accommodate the target image; the digital micro mirror is used for modulating the target according to the loaded measurement matrix data, and the modulated light pulse reflected by the digital micro mirror is collimated and propagated, and enters the frequency up-conversion module through the third 4f system.

[0007] Preferably, the period determination formula of the PPLN crystal is as follows:

[0008]

[0009] Wherein, Λ is the period of the PPLN crystal, n i is the refractive index corresponding to different wavelengths, i=1, 2, 3, λ i are the wavelengths of the light to be measured, the pump light and the up-converted light respectively.

[0010] Preferably, the calculation formula of the refractive index is as follows:

[0011]

[0012] Wherein, ni is the refractive index corresponding to different wavelengths λi, f=(T-T0)(T+T0+2×273.16), T0 is a reference temperature, and T is the temperature of the PPLN crystal.

[0013]

[0014] Preferably, the length of the PPLN crystal along the light propagation direction is greater than a threshold l, and the threshold l is specifically:

[0015] l = l 25 [a(T-25)+b(T-25) 2 ]

[0016] Where l is the length of the PPLN crystal at temperature T, l 25 The length of the PPLN crystal at 25℃ is given by the empirical coefficient a = 1.54 × 10⁻⁶. -5 b = 5.3 × 10 -9 .

[0017] Preferably, the pump beam module includes a fiber laser, a fiber collimator and a half-wave plate arranged in sequence, and the fiber laser, the fiber collimator and the half-wave plate are located in the same optical path.

[0018] Preferably, the signal / pump light incident surface of the PPLN crystal is coated with an antireflection film.

[0019] Preferably, the sum-frequency light emitting surface of the PPLN crystal is coated with an anti-reflection film in the sum-frequency light band.

[0020] This invention also proposes a frequency-to-image conversion method based on coding, the specific steps of which are as follows:

[0021] Step 1: Acquire low-resolution image: at resolution M J ×M J The sum-frequency light intensity at various locations in the scene was measured, specifically by loading digital micromirrors. The positive and negative parts of a Hadamard matrix correspond to the matrix Φ. HP and Φ HN The target scene is modulated using the measurement matrix, and the results of each measurement are recorded sequentially. M is then reconstructed based on the measurement results and the measurement matrix. J ×M J Frequency upconversion of resolution image;

[0022] Step 2: Perform a two-dimensional discrete wavelet transform on the frequency upconversion image, and calculate the threshold for filtering wavelet coefficients that need to be measured again at high resolution based on the results;

[0023] Step 3: Generate the label matrix: Based on the threshold of the wavelet coefficients determined in Step 2, filter out the wavelet coefficients that need to be measured again at scale j-1, specifically:

[0024] Let the high-frequency component corresponding to scale j be LH j HL j HH j , LHj , HL j , HH j , respectively, and the wavelet coefficients not lower than the threshold value are screened out, and the pixel positions corresponding to the wavelet coefficients are marked, to generate the marked matrix M (j-J) C LH , 2 (j-J) C HL , 2 (j-J) C HH Comparison is made, and the wavelet coefficients not lower than the threshold value are screened out, and the pixel positions corresponding to the wavelet coefficients are marked, to generate the marked matrix M j , MHL j , MHH j , wherein the marked positions correspond to elements with a value of 1 in the matrix, and the unmarked positions correspond to elements with a value of 0 in the matrix, and M j-1 ×M j-1 is generated according to the marking result;

[0025] Step 4: For the four pixels at the j-1 scale, the four pixels are combined two by two to obtain the wavelet coefficients corresponding to the four pixels;

[0026] The digital micro-mirror loads the measurement matrix, and the photodetector measures the sum frequency light intensity of the marked positions, and the output voltage enters the acquisition card and is recorded and saved as the measurement result at a resolution of M j-1 ×M j-1 , and the corresponding wavelet coefficients are calculated according to the measurement matrix and the measurement value, to obtain LH j-1 , HL j-1 , HH j-1 , and the current scale is adjusted from j to j-1;

[0027] Step 5: It is judged whether the current scale is 1, if yes, the initial resolution image is taken as the low-frequency component, and a series of wavelet coefficients obtained by measurement are taken as the high-frequency component to reconstruct a high-resolution frequency up-conversion image, otherwise, the step 2 is returned.

[0028] Preferably, the threshold value of the wavelet coefficient that needs to be measured again is determined by the following method:

[0029] The image at a resolution of M J ×M J is subjected to two-dimensional discrete wavelet decomposition to obtain a low-frequency component matrix LL and three high-frequency component matrices LH, HL and HH, wherein LL, LH, HL and HH each contain M J / 2×M J / 2 elements, wherein the value of each element in the high-frequency component matrices LH, HL and HH represents the change amplitude of the four elements at the corresponding positions in the horizontal, vertical and diagonal directions;

[0030] Measure the positions corresponding to larger wavelet coefficients in LH, HL, and HH. Assume the wavelet coefficient sampling rate is SR, where 0 ≤ SR ≤ 1. Arrange the absolute values ​​of the wavelet coefficients in LH, HL, and HH in descending order, and take the nth... The absolute value C of each wavelet coefficient LH C HL and C HH As a threshold used to determine whether the position corresponding to the wavelet coefficient is significant in the horizontal, vertical, and diagonal directions, C LH C is used to filter wavelet coefficients for high-frequency components in the horizontal direction. HL C is used to filter wavelet coefficients for high-frequency components in the vertical direction. HH Wavelet coefficients used to filter high-frequency components in the diagonal direction.

[0031] Compared with the prior art, the present invention has the following significant advantages: (1) The upconversion process in the present invention is a second-order nonlinear effect, which does not involve absorption in the crystal. Unnecessary radiation from the crystal's surroundings will also be strongly filtered in space and frequency. Therefore, the upconversion in the present invention has room-temperature single-photon detection capability; (2) The crystal in the present invention has high nonlinearity, wide MIR transparency, low absorption loss and high damage threshold. Using 1064nm as a hybrid laser, phase matching can be performed in the entire mid-infrared spectrum range, providing upconversion signals in the 785-877nm range, which is suitable for silicon-based detectors. The present invention can be used for any range of light source power, only the exposure time needs to be matched with the spectral radiation level being analyzed; (3) During the conversion process, due to the use of a high-power pump laser, even in the mid-infrared band, 10 -17 WHz -1 / 2 The noise equivalent power (NEP) is at a level far exceeding that of cryogenically cooled mercury cadmium telluride (HgCdTe) detectors (~10). -12 WHz -1 / 2 This also means that the detection sensitivity of the present invention is five orders of magnitude higher than that of the current mainstream mid-infrared detectors; (4) The present invention makes full use of the advantage that the image resolution in the image is not affected by the size of the pump beam in the frequency upconversion based on coding, and improves the conversion efficiency through optical methods, thereby reducing the power requirements of the pump light source, and can obtain higher conversion efficiency under the same pump light power. Attached Figure Description

[0032] Figure 1 It is based on the frequency conversion of the encoding into the image system schematic.

[0033] Figure 2 This is a infographic related to DMD.

[0034] Figure 3is a contrast diagram of a coded frequency up-conversion imaging system and a traditional conversion imaging effect. DETAILED DESCRIPTION

[0035] The features of the present application are described in further detail below with reference to the accompanying drawings, so as to be understood by those skilled in the art.

[0036] A coded frequency up-conversion imaging system, comprising: a light field modulation module, a pump beam module, a frequency up-conversion module, a filtering system and a signal acquisition module, the light field modulation module is used to adjust the signal light to a specified polarization state and measure the light intensity of multiple positions of the target scene, realizing the coding function of the target scene, the pump beam module is used to provide pump light of a specified polarization state, the frequency up-conversion module comprises a dichroic mirror, a PPLN crystal, a temperature control device and a first 4f system, the dichroic mirror is used to combine the signal light and the pump light provided by the pump beam module, so that the two beams of light are collinearly propagated to the incident surface of the PPLN crystal, the PPLN crystal is arranged in the temperature control device, and the temperature control device is used to provide a constant temperature environment; the first 4f system is used to correct the chromatic aberration of the sum frequency light of the PLN crystal exit surface and scale the image ratio; the filtering system is used to filter out the noise mixed in the sum frequency light; and the detection module is used to acquire the sum frequency light filtered by the filtering system. In some embodiments, the dichroic mirror is used to realize the combination of the 1550nm pulse and the 1950nm pump light transmitted through the target, so that the two beams of light are collinearly propagated.

[0037] The light field modulation module comprises a first fiber collimator, a first half-wave plate, a resolution plate, a second 4f system, a digital micro-mirror and a third 4f system, after the signal light is collimated by the first fiber collimator, it is irradiated to the resolution plate through the first half-wave plate, the image transmitted through the resolution plate is scaled by the 4f system to an appropriate scale, so that the digital micro-mirror array working area of the digital micro-mirror can accommodate the target image; the digital micro-mirror is used to modulate the target according to the loaded measurement matrix data, the modulated light pulse reflected by the digital micro-mirror is collimated and propagated, and enters the frequency up-conversion module through the third 4f system.

[0038] In further embodiments, the pump beam module comprises a fiber laser, a fiber collimator and a half-wave plate arranged in sequence, and the fiber laser, the fiber collimator and the half-wave plate are located in the same optical path. The main function of this module is to provide pump light beams for the frequency up-conversion process and ensure that the pump light beams are in a specified polarization state.

[0039] In further embodiments, the signal acquisition module adopts one or more combinations of a single photon detector, a spectrometer and an EMC CD camera.

[0040] In further embodiments, the period determination formula of the PPLN crystal is:

[0041]

[0042] wherein, Λ is the period of the PPLN crystal, n i is the refractive index corresponding to different wavelengths, i = 1, 2, 3, λ i are the wavelengths of the light to be measured, the pump light and the up-converted light respectively.

[0043] Preferably, the calculation formula of the refractive index is:

[0044]

[0045] wherein, n i is the refractive index corresponding to different wavelengths λ i, f = (T - T 0 ) (T + T 0 + 2 × 273.16), T 0 is the reference temperature, and T is the temperature of the PPLN crystal;

[0046]

[0047] Preferably, the length of the PPLN crystal along the light propagation direction is greater than a threshold value l, and the threshold value l is specifically:

[0048] l = l 25 [a (T - 25) + b (T - 25) 2 ]

[0049] wherein, l is the length of the PPLN crystal when the temperature is T, l 25 is the length of the PPLN crystal when the temperature is 25°C.

[0050] In further embodiments, the filter system comprises a combination of FES0900 short-pass, FEL0800 long-pass and FB860-10 band-pass filter.

[0051] In further embodiments, the signal acquisition module adopts one or more combinations of a single-photon detector, a spectrometer and an EMC CD camera.

[0052] The working process of the application is as follows: the collimated output signal light irradiates the USAF1951 standard resolution plate as a target after passing through a half-wave plate, and a 4f system composed of lenses L1 and L2 scales the image passing through the resolution plate at a proper ratio so that the digital micro-mirror array working area of the DMD can accommodate the target image. Then the DMD is used to modulate the target according to the loaded measurement matrix data. In order to keep the collimated propagation of the modulated light pulse reflected by the DMD, the reflected light pulse enters a frequency up-conversion module through a 4f system composed of lenses L3 and L4. Meanwhile, the collimated propagation of the pump light beam in free space enters the PPLN crystal to generate the sum-frequency light together with the signal light after adjusting the polarization state through a half-wave plate, wherein L5 is an achromatic lens that can correct the chromatic aberration of light of different wavebands, and L6 is another 4f system composed of lenses that can properly scale the up-converted image.

[0053] For the pump light with a spot radius of r2, which can be approximated as a collimated light beam, after being focused by the lens L5 with a focal length of f, the spot radius at the center of the PPLN is:

[0054]

[0055] Generally, the spot section of the pump light beam satisfies the Gaussian distribution. Since the process of converging the collimated light beam by the lens can be regarded as the Fourier transform of the light field, for the pump light focused by the lens, the Gaussian distribution expression after the Fourier transform is substituted, and the sum-frequency light intensity of the mth measurement is obtained:

[0056]

[0057] wherein A Gauss represents the amplitude field of the Gaussian distribution of the pump light, and represents the corresponding Fourier transform form. When the spot radius of the pump light beam in the PPLN changes from r2 to r 2f , the theoretical conversion efficiency will be increased to times of the original. Since the spot radius r2 of the pump light used in the frequency up-conversion is in the sub-millimeter to millimeter level, the wavelength λ2 of the pump light is in the hundreds of nanometers to several micrometers level, the focal length f of the lens is in the tens of millimeters to hundreds of millimeters level, and the refractive index n2 of the PPLN to the pump light is slightly less than π, therefore can be made to be much greater than 1 by reasonably designing the pump light beam and the lens parameters, so that the conversion efficiency is observably improved.

[0058] The spatial resolution of the application can be calculated as:

[0059]

[0060] wherein d resis the resolution of the system, A is the ratio of the size of the target scene after being projected onto the DMD and the original size, all the digital micromirrors in the DMD are square with side length S DMD According to the Shannon-Nyquist sampling theorem, when two adjacent targets are imaged on the DMD, if the distance between the two targets is greater than two pixels, the two targets can be distinguished. It can be seen from the formula that there is no related term of the pump light spot radius r2, that is, the image resolution of the frequency up-conversion imaging system based on the coding frequency will not deteriorate due to the too small pump light beam. The imaging resolution in the traditional method is affected by the size of the pump light. Therefore, this method also effectively solves the contradiction that the frequency up-conversion imaging system can only optimize one of the image resolution and the conversion efficiency.

[0061] The DMD is composed of a digital micromirror array, and the size of each digital micromirror can reach microns, which is comparable to the size of a single pixel of a camera. The DMD can be individually controlled by a control circuit to make any one digital micromirror in one of the two flip states of +12° or -12°. In a single-pixel imaging system, according to the measurement mode and installation method of the DMD, the digital micromirror in one of the +12° or -12° flip states can be used to reflect the target scene light, and a focusing lens can be used to collect the reflected target scene light onto the target surface of the detector. Since the DMD loads the flip state of each digital micromirror first and then flips all the digital micromirrors at the same time when it works, the DMD can reflect the target scene light at a specific position to the photodetector each time to measure the light intensity at these positions. The output voltage of the photodetector represents the light intensity entering the detector, and the measurement value obtained after the analog-to-digital conversion of the voltage also reflects the light intensity of the target scene reflected by the digital micromirror. Assuming that the DMD in the single-pixel imaging system modulates the target scene using N×N digital micromirrors, an image containing N×N pixels can be reconstructed after N×N measurements of the scene by the system. The result of the mthmeasurement can be represented by the following formula:

[0062]

[0063] A row vector f containing N 2 elements is used to represent the target scene image, and f n in formula (3.1) represents the nth element in the vector f corresponding to the pixel located in the nth r row and the nth l column of the target scene image. Wherein, n r is the integer part of n divided by N, and n l is the remainder part of n divided by N. D m is a column vector representing the flip state of each digital micromirror of the DMD, also containing N 2An element of D. mn is a column vector D m is the n-th element of D r , representing the digital micro-mirror at the n-th l row and n-th column position on the DMD, whose flipping state D mn takes value 0 or 1. Considering that the DMD can only load non-negative measurement patterns, it is not harmful to define that when D mn = 0, the light reflected by the digital micro-mirror does not enter the single-pixel detector; when D mn = 1, the light reflected by the digital micro-mirror enters the single-pixel detector and is detected by the detector response. dc represents the direct current bias, i.e. the detector output voltage when the light in the scene is reflected by the DMD and does not enter the single-pixel detector.

[0064] An encoding-based frequency up-conversion imaging method, the specific steps are:

[0065] Step 1: Obtain a low-resolution image: measure the sum frequency light intensity at each position in the scene under the resolution M J × M J , specifically: use the digital micro-mirror to load the positive and negative parts of the Hadamard matrix of order M × M HP and Φ HN as the measurement matrix to modulate the target scene, and record the results of each measurement in order, and reconstruct the frequency up-conversion image of the resolution M J × M J according to the measurement results and the measurement matrix;

[0066] Step 2: Perform two-dimensional discrete wavelet transform on the frequency up-conversion image, and calculate the threshold value of the wavelet coefficient that needs to be measured again under high resolution according to the result;

[0067] The determination method of the threshold value of the wavelet coefficient that needs to be measured again is:

[0068] Perform two-dimensional discrete wavelet decomposition on the image of the resolution M J × M J to obtain a low-frequency component matrix LL and three high-frequency component matrices LH, HL and HH, LL, LH, HL and HH each contain M J / 2× M J / 2 elements, wherein the value of each element in the high-frequency component matrices LH, HL and HH represents the change amplitude of the corresponding four elements in the horizontal, vertical and diagonal directions;

[0069] Measure the corresponding positions of the wavelet coefficients with larger values in LH, HL and HH, set the wavelet coefficient sampling rate as SR, 0≤SR≤1, arrange the absolute values of the wavelet coefficients in LH, HL and HH in descending order respectively, and take the first the absolute value of the wavelet coefficient C LH , C HL , and C HH as a threshold value for determining whether the position corresponding to the wavelet coefficient is significant in the horizontal, vertical, and diagonal directions, wherein C LH the wavelet coefficient for screening the high-frequency component in the horizontal direction, C HL the wavelet coefficient for screening the high-frequency component in the vertical direction, C HH the wavelet coefficient for screening the high-frequency component in the diagonal direction.

[0070] Step 3: Generate a marked matrix: according to the threshold screening of the wavelet coefficient determined in step 2, the wavelet coefficient that needs to be measured again at scale j-1 is determined, specifically:

[0071] Let the high-frequency component corresponding to scale j be LH j , HL j , and HH j . The wavelet coefficients in LH j , HL j , and HH j are compared with the threshold values 2 (j-J) C LH , 2 (j-J) C HL , and 2 (j-J) C HH respectively, and the wavelet coefficients not lower than the threshold values are screened out and marked with the pixel positions corresponding to these wavelet coefficients. The marked matrices MLH j , MHL j , and MHH j in the horizontal, vertical, and diagonal directions are generated respectively. The marked positions correspond to elements with a value of 1 in the matrix, and the unmarked positions correspond to elements with a value of 0 in the matrix. According to the marked results, M j-1 ×M j-1 resolution measurement matrix is generated;

[0072] Step 4: For the four pixels at scale j-1, the four pixels are combined two by two to obtain the wavelet coefficients corresponding to the four pixels;

[0073] The digital micro-mirror loads the measurement matrix, and the photodetector measures the sum frequency light intensity of the marked positions. The output voltage enters the acquisition card after analog-digital conversion and is recorded and saved as the measurement result at M j-1 ×M j-1 resolution. According to the measurement matrix and the measurement value, the corresponding wavelet coefficients are calculated, and LH j-1 , HL j-1 , and HH j-1 are obtained. The current scale is adjusted from j to j-1;

[0074] Step 5: Determine whether the current scale is 1, if it is 1, take the initial resolution image as the low frequency component, measure the obtained series of wavelet coefficients as the high frequency component to reconstruct the high resolution frequency up-converted image, otherwise return to step 2.

[0075] Embodiment

[0076] An encoding-based frequency up-conversion imaging system, as shown in Figure 1 The above modules are introduced one by one as follows:

[0077] 1) Light field modulation module

[0078] The light field modulation module is composed of a signal light source, a DMD, a half-wave plate, and lenses L1-L4. The main functions of this module include: generating light pulses by the signal light source and adjusting the polarization state of the signal light using the half-wave plate HWP1. The light pulses pass through the half-wave plate and illuminate the USAF1951 standard resolution plate as the target, and then the 4f system composed of lenses L1 and L2 scales the image transmitted through the resolution plate by an appropriate ratio, so that the digital micromirror array working area of the DMD can accommodate the target image. Then the DMD modulates the target according to the loaded measurement matrix data. In order to keep the reflected modulated light pulses collimated, the reflected light pulses pass through the 4f system composed of lenses L3 and L4 into the frequency up-conversion module.

[0079] The system uses a 1550nm-CoLID-I fiber laser as a signal light source for illuminating the resolution plate as the target. The light pulses generated by the 1550nm-CoLID-I laser pass through the FC / APC interface connecting the ends of the polarization maintaining fiber into free space. In the experiment, the fiber collimator of the FC / APC interface is connected to the output end of the 1550nm-CoLID-I fiber laser to ensure that the light pulses propagate collimated after entering the free space. When the fiber laser is working, the laser repetition rate is set to 10kHz and the pulse width is 10ns. According to the measurement results, the peak power of the light pulses illuminating the standard resolution plate is about 200mW. Considering that the 1550nm pulse is output through the polarization maintaining fiber connected to the 1550nm-CoLID-I laser, the signal light used by the system is linearly polarized light.

[0080] Considering the existing resources in the laboratory and the need to ensure that the signal light power is much lower than the pump light power in the experiment, the DMD used in this paper is a V-7000 series ultra-high-speed spatial light modulator from Germany ViALUX company, and the detailed parameters are shown in Table 1.

[0081] Although the wavelength window of DMD in the experiment only includes visible light and ultraviolet light, the reflectivity in the short-wave infrared region is not less than 80%, and the signal light needs to be much lower than the pump light in the experiment, so this type of DMD can meet the experimental requirements. In addition, the maximum micro-mirror flipping rate of DMD is much larger than the pulse repetition frequency of 1550nm, in order to ensure the stable operation of the system and avoid frame errors and other problems, the micro-mirror flipping rate of DMD should not be greater than the pulse repetition frequency of the signal light, so the micro-mirror flipping rate of DMD in the experiment is set to 100Hz.

[0082] Table 1 DMD technical parameters

[0083]

[0084]

[0085] 2) Pump light beam module

[0086] According to the existing resources in the laboratory and the requirements of the polarization state of the pump light, the Powerwave 2000 type continuous wave laser is used as the pump light source in this system. Therefore, the pump light beam module is composed of this fiber laser, a fiber collimator and a half-wave plate. The main function of this module is to provide a pump light beam for the frequency up-conversion process and to ensure that the pump light beam is in the specified polarization state.

[0087] The Powerwave 2000 type laser uses a polarization maintaining fiber output, and the maximum output power is 10W. When the laser is working, only the control knob on the front panel needs to be adjusted to change the control current inside the laser to change the laser output power. Before the system works, the Powerwave 2000 type laser needs to be preheated for about 10 minutes after being turned on and the output power is set by the front panel knob. After the output power is stable, the system can work. After that, no additional operation is required for the laser.

[0088] The laser beam output by the Powerwave 2000 type laser is continuous light, and as a pump light source, it does not need to consider the synchronization of the signal light and the pump light in the time domain, and the walk-off effect can be ignored. Only the power of the pump light beam needs to be greater than the peak power of the signal light pulse and not damage the PPLN crystal. The pump light generated by the Powerwave 2000 type laser is also output to free space by the polarization maintaining fiber connected to the SMA905 interface at the end. In order to make the pump light propagate in free space in a collimated manner, a fiber collimator with an SMA905 interface is connected to the fiber output port of the Powerwave 2000 type laser. Finally, the pump light beam propagating in free space collimated passes through the half-wave plate HWP2 to adjust the polarization state before entering the frequency up-conversion module.

[0089] 3) Frequency up-conversion module

[0090] The frequency up-conversion module is composed of a dichroic mirror, a PPLN crystal, a temperature control device, and an achromatic lens L5 and a lens L6. The main function of the module is to use the dichroic mirror to realize the combination of the 1550 nm pulse and the 1950 nm pump light, so that the two beams of light are co-linearly propagated. The reflectivity of the dichroic mirror at 1550 nm is 98.92%, and the transmittance at 1950 nm is 98.15%. Then the mixed light beam is focused to the center of the PPLN crystal through the achromatic lens L5 to generate the 863 nm sum frequency light. Considering that the signal light and the pump light have different wavelengths, the lens L5 should use an achromatic lens to make the focal planes of the two beams of light overlap as much as possible. During the operation of the system, the temperature control device is needed to ensure that the temperature of the PPLN crystal is maintained constant.

[0091] The core device of the frequency up-conversion module is the PPLN crystal, and the design of the PPLN crystal directly determines the result of the frequency up-conversion. In the system, the 1550 nm signal light and the 1950 nm pump light propagate in the PPLN in the form of an extraordinary light, that is, the linear polarization directions of the two beams of light are both vertical polarization. The period of the PPLN crystal can be calculated according to the quasi-phase matching condition (QPM), the function relationship between the refractive index of the PPLN crystal and the temperature, and the function relationship between the length of the PPLN crystal and the temperature when the PPLN crystal is heated and expanded. The function relationship between the refractive index of the PPLN crystal and the temperature is as follows:

[0092]

[0093]

[0094] wherein, n e is the refractive index of the PPLN crystal to be solved, T0 is the reference temperature, and T is the temperature of the PPLN crystal.

[0095] The period determination formula of the PPLN crystal is:

[0096]

[0097] wherein, Λ is the period of the PPLN crystal, n i is the refractive index corresponding to different wavelengths λ i , i=1, 2, 3, respectively corresponding to the wavelength of the signal light source, the pump light source and the up-converted light.

[0098] The coefficients in formula (1) are shown in Table 2. The length of the PPLN crystal in the horizontal direction is calculated as:

[0099] l = l 25 [a(T-25) + b(T-25) 2 ] (4)

[0100] where, / is the length of PPLN crystal at temperature T, / 25 is the length of PPLN crystal at temperature 25°C. 25 The empirical coefficient of PPLN crystal used in the system is a = 1.54 x 10 -5 , b = 5.3 x 10 -9 .

[0101] When the optical power of the incident PPLN crystal is too large, it will cause irreversible damage to the crystal, which is called photorefractive effect. In order to increase the optical damage threshold of PPLN crystal, the PPLN crystal doped with 5% magnesium oxide produced by Nanjing Kuike Optoelectronics Co., Ltd. is selected, and the optical damage threshold of the crystal can reach 500 MW / cm 2 , and the effective nonlinear coefficient is d eff = 14 pm / V.

[0102] Table 2 Thermal expansion equation parameters

[0103]

[0104] Considering the cost of the crystal and the processing technology, the polarization period of the final PPLN crystal is designed to be 24.1 μm, and the working temperature is about 58.4°C under this condition. The parameters of the PPLN crystal used in the system are shown in Table 3.

[0105] Table 3 PPLN crystal parameters

[0106]

[0107] In order to reduce the energy loss of signal light and pump light in the propagation process, the signal light / pump light incident surface of the PPLN crystal is coated with an antireflection film to ensure that the reflectivity of the signal light and pump light incident crystal is less than 0.2%. The 863 nm film is coated on the sum frequency light exit surface to make the reflectivity of the 863 nm exit crystal less than 0.2%.

[0108] The PPLN crystal is installed in a temperature control furnace. The PPLN crystal in the temperature control furnace can be heated from room temperature to 200°C, and when in a constant temperature state, the temperature in the temperature control furnace fluctuates no more than ±0.1°C. Since the temperature of the PPLN needs to be kept constant during the experiment, the temperature control device needs to be turned on before the frequency up-conversion imaging system based on coding works and wait for the temperature shown on the temperature control furnace panel to be stable.

[0109] 4) Filtering system

[0110] According to the spectral characteristics of the sum frequency light, the filter system selects FES0900 type short pass, FEL0800 long pass and FB860-10 type band pass filter, the combined transmittance of which at 863 nm is 34%. After the sum frequency light is denoised by the filter system, it is converged to the signal acquisition module by a lens with a focal length of 50 mm.

[0111] 5) Signal acquisition module

[0112] The signal acquisition module includes one or more of a single photon detector, a spectrometer and an EMCCD camera. The main functions of the module include: filtering out noise mixed in the sum frequency light, measuring the pulse energy of the sum frequency light at 863 nm, converting the voltage output by the single photon detector into a digital signal and recording and processing it by the computer.

[0113] In order to improve the signal-to-noise ratio, the system uses Hamamatsu H7422-20 type photomultiplier tube (PMT) as a single pixel detector.

[0114] According to the focal length of the lens for converging the sum frequency light is 50 mm and the target surface diameter of H7422-20 is 7 mm, the field of view angle of the detector can be calculated as 7.97°. The response sensitivity of the PMT at 863 nm is 30 mA / W, and the transmittance of the filter FB860-10 at 863 nm is 34.68%. In order to reduce the situation that the sum frequency light power corresponding to the scene modulated by the DMD is too low to be responded by the photomultiplier tube or interfered by noise, the gain voltage of the photomultiplier tube is set to 0.75 V, at which the corresponding gain is 10 5 Finally, the system uses the PXIE-5122 type acquisition card of NI (National Instruments) company to convert the voltage output by the PMT from an analog signal to a digital signal and send it to the computer for recording by the control software Labview of the system.

[0115] As shown in Figure 2 , the digital micro-mirror device (DMD) is the core device of the modulation module. When the DMD works, the entire measurement matrix is first loaded into the memory, and then the measurement matrix is read out row by row and converted into a matrix. The converted matrix contains the flipping state of each digital micro-mirror, and each converted matrix corresponds to a measurement mode. Then the flipping state of each micro-mirror is independently controlled by the CMOS storage circuit located below the micro-mirror. The CMOS storage circuit controls the electrode voltage of the corresponding digital micro-mirror to make the hinge as the axis to rotate the micro-mirror, and adjusts the digital micro-mirror to the two flipping states of on and off to realize the function of reflecting the incident light at +12° or -12°. Figure 2(b) is a scanning electron microscope (SEM) image of a 3x3 size digital micromirror array, where each micromirror has a size of 13.68 μm x 13.68 μm and the pixel pitch is about 13.7 μm. Figure 2 (c) is a real photo of the DMD, where the encapsulated DMD micromirror array is the effective part of the DMD for optical intensity modulation. In order to protect the DMD micromirror array, the incident direction of the scene light is isolated from the outside by an optical glass. In order to make the DMD suitable for a variety of different occasions, the transmittance of different wavelengths of light in the range of 300 nm to 2700 nm through the optical glass can be increased by applying different coatings on the optical glass, thereby improving the reflection ability of the digital micromirror array.

[0116] As shown in Figure 3 , a comparison diagram of the coded frequency up-conversion imaging system and the traditional up-conversion imaging effect is shown. In order to verify that the coded frequency up-conversion imaging method can more effectively utilize the convergent pump beam to achieve imaging while obtaining higher conversion efficiency, and eliminate the contradiction between resolution and conversion efficiency, a group of comparison experiments will be carried out. When the pump beam is converged by a lens with a shorter focal length, the pump spot radius r2 on the focal plane is also smaller, and the light intensity of the pump light in the PPLN will be improved, and at this time the conversion efficiency will also be improved. Figure 3 The imaging results of the coded frequency up-conversion imaging method and the traditional frequency up-conversion imaging method when the pump light is converged by lenses with different focal lengths are given. Since the conversion efficiency is related to the focal length of the lens, that is, the shorter the focal length of the lens, the greater the conversion efficiency, and considering that the simulation result is intended to analyze the image resolution, the pixel values in Figure 3 are normalized. For the same target shown in Figure 3 (a), Figure 3 (b-d) gives the results of coded frequency up-conversion imaging. Although the lens focal length is reduced, the target profile is still clear. As a comparison, Figure 3 (e-g) gives the results of traditional frequency up-conversion imaging. By comparing Figure 3 (b-e), it can be found that the profile of the target image appears serious distortion as the focal length of the lens becomes shorter, and the convolution effect of the target amplitude field and the pump light amplitude field is also more obvious. As can be seen from Table 4, the image based on the coded frequency up-conversion imaging can better preserve the information of the target and reduce the distortion in the frequency up-conversion process compared with the traditional frequency up-conversion imaging. At the same time, by comparing horizontally, it can be obtained that as the spot radius r2 of the pump light on the focal plane decreases, the SSIM decay rate of the image based on the coded frequency up-conversion imaging is much lower than that of the traditional frequency up-conversion imaging, that is, when the pump beam is optimized by focusing, the coded frequency up-conversion imaging method has better robustness.

[0117] Table 4 Comparison of simulated SSIM

[0118]

[0119]

[0120] An encoding-based frequency up-conversion imaging method, the specific steps are:

[0121] Step 1: Obtain a low-resolution image. In order to estimate the detailed information in the scene, it is necessary to first measure the scene at a lower resolution M J ×M J Measure the sum frequency light intensity at each position in the scene at this time, and the corresponding initial scale is J (J ∈ Z + ). Use the DMD to load the positive and negative parts of the Hadamard matrix of order J and Φ HP and Φ HN as the measurement matrix modulates the target scene, and records the results of each measurement in order. After the measurement is completed, the M J ×M J resolution frequency up-converted image can be obtained according to the measurement results and the measurement matrix.

[0122] Step 2: Select a threshold. After obtaining the M J ×M J resolution frequency up-converted image, a two-dimensional discrete wavelet transform is performed on the image, and a threshold value for screening wavelet coefficients that need to be measured again at high resolution is calculated according to the results.

[0123] First, perform two-dimensional discrete wavelet decomposition on the M J ×M J resolution image to obtain a low-frequency component matrix LL and three high-frequency component matrices LH, HL and HH. LL, LH, HL and HH each contain M J / 2×M J / 2 elements. Among them, the value of each element in the high-frequency component matrices LH, HL and HH represents the change amplitude of the four elements in the corresponding position in the horizontal, vertical and diagonal directions. In order to improve the imaging speed as much as possible while preserving details, the corresponding positions with larger wavelet coefficients in LH, HL and HH should be measured first. Let the wavelet coefficient sampling rate be SR (0 ≤ SR ≤ 1), arrange the absolute values of the wavelet coefficients in LH, HL and HH in descending order respectively, and take the absolute value C of the wavelet coefficient of the LH th HL th HHThis serves as a threshold for determining whether the position corresponding to the wavelet coefficient is significant in the horizontal, vertical, and diagonal directions. Wherein, C... LH C is used to filter wavelet coefficients for high-frequency components in the horizontal direction. HL C is used to filter wavelet coefficients for high-frequency components in the vertical direction. HH Wavelet coefficients used to filter high-frequency components in the diagonal direction.

[0124] Step 3: Generate the label matrix. According to the wavelet tree model, the coefficients exhibit persistence across different scales. When the current scale j > 1, a threshold C needs to be applied. LH C HL and C HH Select the wavelet coefficients that need to be measured again at scale j-1. Let the high-frequency component corresponding to scale j be LH. j HL j HH j Especially when j = J, where J is the initial scale, the high-frequency component LH j HL j HH j These correspond to LH, HL, and HH, respectively. (The LH...) j HL j HH j The wavelet coefficients in the data are respectively related to the threshold 2. (j-J) C LH 2 (j-J) C HL 2 (j-J) C HH The wavelet coefficients that are not lower than the threshold are compared and selected, and the pixel positions corresponding to these wavelet coefficients are marked. Marking matrices MLH are generated in the horizontal, vertical and diagonal directions respectively. j MHL j MHH j The marked positions correspond to elements with a value of 1 in the matrix, and the sum-frequency light intensity at these positions will be at a resolution M corresponding to the scale factor j-1. j-1 ×M j-1 The sum-frequency light intensity at these locations will be measured again at higher resolution; the corresponding element in the matrix for unmarked locations will be 0, and the sum-frequency light intensity at these locations will not be measured at higher resolution.

[0125] Step 4: High-resolution sampling. After generating the label matrix at scale j, M needs to be generated based on the labeling results. j-1 ×M j-1 The resolution measurement matrix. For four pixels at scale j-1, if it is necessary to collect a certain wavelet coefficient of these four pixels, it is not necessary to measure the sum-frequency light intensity of each pixel. It is only necessary to combine these four pixels in pairs, that is, first measure the sum-frequency light intensity of two specified pixels, and then combine these four pixels at scale M. j ×Mj The sum frequency light intensity of the corresponding pixels at the resolution is subtracted from the measured value, and finally the wavelet coefficients corresponding to the four pixels can be obtained. The calculation process is shown in equations (1), (2) and (3).

[0126]

[0127]

[0128]

[0129] Then, the DMD loads the measurement matrix, while the photodetector measures the sum frequency light intensity of the mark position, and the output voltage enters the acquisition card after analog-digital conversion and is recorded and saved as M j-1 × M j-1 The measurement result at the resolution. Finally, the corresponding wavelet coefficients are calculated according to the measurement matrix and the measured value, and the current scale is adjusted from j to j-1. j-1 , HL j-1 , HH j-1 , and the current scale is adjusted from j to j-1.

[0130] Step 5: High-resolution image reconstruction. When j = 1, it can be considered that the resolution corresponding to the current scale is the final resolution. At this time, the initial resolution image is taken as the low-frequency component, and a series of wavelet coefficients obtained by measurement are taken as the high-frequency component to reconstruct the high-resolution frequency up-conversion image.

Claims

1. A frequency up-conversion imaging method, characterized in that, The specific steps are as follows: Step 1: Obtain low-resolution image: In resolution M J × M J Measure the sum frequency light intensity at each position in the scene, specifically: use digital micro-mirrors to load The positive and negative parts of the Hadamard matrix correspond to the matrices Φ HP and Φ HN As the measurement matrix modulates the target scene, and records the results of each measurement in order, according to the measurement results and the measurement matrix, the frequency up-conversion image of M J × M J resolution is obtained; Step 2: performing two-dimensional discrete wavelet transform on the frequency up-converted image, and calculating a threshold value for screening wavelet coefficients that need to be measured again at a high resolution according to the result; Step 3: generating a marking matrix: screening wavelet coefficients that need to be measured again at scale j-1 according to the threshold value of the wavelet coefficients determined in step 2, specifically: Let the high frequency component corresponding to the scale j be LH j , HL j , HH j , respectively, and the wavelet coefficients in LH j , HL j , HH j be respectively compared with the threshold 2 (j-J) C LH , 2 (j-J) C HL , 2 (j-J) C HH , respectively, and the wavelet coefficients not lower than the threshold are screened out, and the pixel positions corresponding to these wavelet coefficients are marked, respectively, to generate the marked matrices MLH j , MHL j , MHH j in the horizontal, vertical, and diagonal directions, respectively, and the marked positions correspond to the elements with a value of 1 in the matrix, and the unmarked positions correspond to the elements with a value of 0 in the matrix, and M j-1 ×M j-1 is generated according to the marking result to be a measurement matrix with a resolution of Step 4: for the four pixels at scale j-1, combining the four pixels two by two to obtain wavelet coefficients corresponding to the four pixels; Digital micro-mirror loads the measurement matrix, and the photodetector measures the sum frequency light intensity of the mark position, and the output voltage enters the acquisition card after analog-digital conversion and is recorded and saved as M j-1 ×M j-1 The measurement result under the resolution, according to the measurement matrix and the measurement value, the corresponding wavelet coefficient is calculated, and the LH j-1 , HL j-1 , HH j-1 , and adjust the current scale from j to j-1; Step 5: determining whether the current scale is 1, if yes, taking the initial resolution image as a low-frequency component, and taking a series of wavelet coefficients obtained by measurement as a high-frequency component to reconstruct a high-resolution frequency up-converted image, otherwise returning to step 2.

2. The coded-based frequency upconversion imaging method of claim 1, wherein, The threshold value of the wavelet coefficients that need to be measured again is determined by the following method: M J x M J A two-dimensional discrete wavelet decomposition is performed on the image with resolution of M J / 2 x M J / 2 elements, wherein the value of each element in the high-frequency component matrix LH, HL and HH represents the change amplitude of the four elements at the corresponding position in the horizontal, vertical and diagonal directions, respectively. The positions corresponding to the wavelet coefficients with larger values in LH, HL and HH are measured, and the wavelet coefficient sampling rate is SR, 0≤SR≤1. The absolute values of the wavelet coefficients in LH, HL and HH are arranged in descending order, and the absolute values C , LH , HL and C HH of the first wavelet coefficients are taken as threshold values for judging whether the positions corresponding to the wavelet coefficients are significant in the horizontal, vertical and diagonal directions, wherein C LH is used for screening the wavelet coefficients of the high-frequency components in the horizontal direction, C HL is used for screening the wavelet coefficients of the high-frequency components in the vertical direction, and C HH is used for screening the wavelet coefficients of the high-frequency components in the diagonal direction.

Images