High-speed single-photon detection imaging method based on time encoding

Abstract

The present invention discloses a high-speed single-photon detection imaging method based on time coding. The steps of the method are: first, calibrate the relationship curve between the photon detection efficiency and the bias voltage of a single single-photon avalanche diode SPAD; then, emit a pulsed laser to the scene, use an array composed of multiple SPADs to collect the light returned from the scene, and modulate the bias voltage on the SPAD to achieve modulation of the photon detection efficiency, that is, introduce time coding for SPAD imaging detection; by performing different time coding on the SPAD, different total detection photon numbers are collected; finally, based on the least squares reconstruction of the gradient descent algorithm, the calculation and reconstruction of the returned light intensity at different times. The present invention is an indirect photon time domain information acquisition method, which avoids the light self-stacking phenomenon due to time-correlated single-photon counting imaging, and can achieve single-photon high-speed imaging that is robust to environmental noise and scattering.

Description

Technical Field

[0001] The invention relates to a high-speed single-photon detection imaging method based on time coding. Background Art

[0002] Single-photon avalanche diode (SPAD) arrays detect photons propagating in a scene with single-photon sensitivity, and have important applications in long-range radar detection. Existing detection devices that can achieve single-photon sensitivity have different forms, including photomultiplier tubes (PMT Photomultiplier), EMCCD (Electron multiply charge coupled device), ICCD (Intensified charged coupled device), SPAD, etc. Among them, the imaging array that may achieve low cost and high sensitivity is the integrated CMOS (Complementary metal oxide semiconductor) SPAD. Since the SPAD integrated on silicon was formally proposed in 2003, the array scale and detection efficiency have been continuously improved and developed. Imaging technology based on CMOS SPAD can achieve single-photon sensitivity and array-level imaging detection of scenes, and has been widely valued and applied in biology, astronomy, military, etc., such as biological fluorescence molecular lifetime imaging, Raman scattering imaging, single molecule localization imaging, quantum super-resolution imaging, non-time domain imaging, non-time domain tracking, and three-dimensional depth imaging, etc.

[0003] The three-dimensional imaging technology based on SPAD is the time-correlated single-photon imaging technology TCSPC (Time correlated single photon counting). By directly collecting the propagation time of multiple photons in the scene, the propagation time of the photons in the scene is counted to obtain the time distribution function of the light reflected from the scene. However, this imaging technology only detects the propagation time of the first photon in the scene under a single laser pulse, which will cause the detection rate of the photons that arrive first to be higher than the detection efficiency of the photons that arrive later. The probability distribution of the detected photon arrival time has the effect that the probability increases when the propagation time is relatively short and decreases when the propagation time is relatively long. The statistical photon time probability distribution does not conform to the real distribution, which is referred to as the photon stacking effect. This effect leads to large errors in the measurement of the photon propagation time function when the ambient light is strong and the backscattering in the scene is severe. At the same time, the shortcomings of TCSPC also restrict the robust and accurate detection of three-dimensional information of close-range scenes by strong laser pulses.

[0004] In the field of computational imaging, high-speed imaging technology based on time coding has been successfully applied. By adding different time modulation functions at different times and then solving them through reverse calculation, scene images collected at different times can be reconstructed to achieve high-speed imaging. However, these imaging technologies are still limited to high-speed imaging at the level of thousands of frames per second, and the speed of time coding is limited by optical modulation devices. Summary of the invention

[0005] In view of the above defects in the prior art, the present invention proposes a single-photon high-speed detection imaging method based on time coding.

[0006] The technical solution adopted by the present invention is as follows:

[0007] A high-speed single-photon detection imaging method based on time encoding, the steps of the method are as follows:

[0008] (1) First, the relationship curve between the photon detection efficiency and bias voltage of a single single-photon avalanche diode (SPAD) is calibrated: under the same incident light intensity, different bias voltages are applied to the SPAD, and the corresponding total number of detected photons is collected to obtain the relationship curve ρ(x) between the SPAD bias voltage x and the detection efficiency; based on the calibrated curve ρ(x) and the intensity of the detection efficiency corresponding to the time code at different times, the bias voltage value that should be added at different times to achieve a certain time code can be obtained;

[0009] (2) emitting laser pulses to the scene, using an array composed of multiple SPADs to collect the light returned from the scene, and changing the detection efficiency of the SPAD by adjusting the bias voltage on the SPAD during the exposure time, thereby realizing the temporal encoding of the SPAD detection; specifically: within the exposure time T, with Δt as the minimum time unit, adjusting the bias voltage on the SPAD so that the detection efficiency of the SPAD is 0 or 1 or any decimal between 0 and 1, so that the corresponding time code is a random time code sequence;

[0010] (3) For each exposure, a time code sequence is randomly generated, and the total number of photons detected by the SPAD under this time code is the measurement value corresponding to the time code;

[0011] (4) repeating the random time coding and acquisition process of steps (2) and (3) multiple times, generating a different random time coding sequence each time, and correspondingly obtaining a different measurement value;

[0012] (5) Based on different time coding sequences and their corresponding measurement values, an optimization objective function is established, and computational reconstruction based on gradient descent is performed to obtain the light intensity values ​​returned by the scene at different times under the pulsed laser, thereby realizing high-speed single-photon detection imaging.

[0013] Furthermore, in the step (2), the bias voltage on the SPAD is adjusted to a range of 0 to 3V.

[0014] Furthermore, in step (2), the array unit includes a single SPAD, a quenching circuit, a shaping circuit and a counter circuit. The SPAD converts incident photons into electrical pulse signals, outputs standard pulse signals after passing through the shaping circuit, and the counter circuit records the number of photons.

[0015] The present invention proposes a single-photon high-speed detection imaging method based on time coding, which is an indirect photon time-domain information acquisition method. It does not require direct measurement of the photon arrival time, avoids the light self-stacking phenomenon existing in time-correlated single-photon counting imaging, and can achieve single-photon high-speed imaging that is robust to environmental noise and scattering. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic flow diagram of the method of the present invention.

[0017] Figure 2 Schematic diagram of a device for realizing detection in an embodiment of the present invention.

[0018] Figure 3 This is a graph showing the relationship between the bias voltage on the SPAD and the detection efficiency.

[0019] Figure 4 This is a schematic diagram of the SPAD acquisition process.

[0020] Figure 5 This is a schematic diagram of the principle of single-photon time-coded high-speed acquisition. DETAILED DESCRIPTION

[0021] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0022] The existing SPAD array combined with time-coupled single-photon counting (TCSPC) technology can directly measure the time it takes for light to propagate in a scene, thereby collecting three-dimensional information about the scene. However, this method is affected by the photon pile-up effect, has limited collection accuracy, and is easily affected by ambient light and scattering in the scene. Therefore, this embodiment provides a single-photon high-speed detection imaging method based on time coding, including SPAD bias voltage regulation to implement time coding technology, cumulative counting, and computational reconstruction. The main process is as follows: Figure 1 As shown:

[0023] (1) First, it is necessary to calibrate the relationship curve between the photon detection efficiency and bias voltage of a single single-photon avalanche diode (SPAD): under the same incident light intensity, add different bias voltages to the SPAD, collect the corresponding total number of detected photons, and obtain the relationship curve ρ(x) between the SPAD bias voltage x and the detection efficiency. Based on the calibrated ρ(x) and the intensity of the detection efficiency corresponding to the time code at different times, the bias voltage that should be added at different times to achieve a certain time code can be obtained. For different random coding sequences, the code value is between 0-1, that is, the detection efficiency of the code is between 0-1. According to the calibrated bias voltage and detection efficiency, the corresponding bias voltage loaded on the SPAD can be found to make the SPAD detection efficiency reach the target code value. The random distribution for generating random codes between 0-1 can include a random {0, 1} binary distribution, a uniform random distribution of 0-1, or a truncated Gaussian distribution with random values ​​between 0-1 (mean is 0.5, variance is 0.5), etc.

[0024] (2) Emitting laser pulses to the scene, using the SPAD array to collect the light returned from the scene, by adjusting the bias voltage on the SPAD during the exposure time, the detection efficiency of the SPAD at different times during the exposure time can be changed, and the time coding of the SPAD detection can be realized. Specifically, the time coding used in this embodiment is a {0, 1} coding sequence: defining the bias overvoltage less than or equal to 0V as coding '0', at which time the detection efficiency of the SPAD device is 0, and it has no detection capability for the light signal; defining a certain detection efficiency bias overvoltage (usually around 3V) as coding '1', the detection efficiency under this bias voltage is a constant greater than 0, and it has normal light detection capability. Within the exposure time T, with Δt as the minimum time unit, changing the bias voltage on the SPAD to 0 or 3V, the corresponding SPAD detection efficiency is 0 or 1, so that the corresponding time coding is a {0, 1} sequence, and the {0, 1} sequence in each exposure time is randomly generated. For example, the exposure time T is set to 10us and the minimum time unit Δt is 50ns. Then the time code length corresponding to the exposure time range of 10us is 200, that is, the random time code can be represented as a vector with a dimension of 200*1, and each element of the vector is 0 or 1.

[0025] (3) For each exposure, the total number of photons detected by the SPAD under a random time code is the measurement value corresponding to that time code.

[0026] (4) Repeat the above random time encoding and acquisition process multiple times, generating a different random {0, 1} sequence each time, and correspondingly obtaining a different measurement value.

[0027] (5) Based on different time codes and their corresponding measurement values, an optimization objective function is established, and computational reconstruction based on gradient descent is performed to obtain the light intensity values ​​returned by the scene at different times under the pulsed laser, thereby realizing high-speed imaging based on SPAD on-chip time coding.

[0028] like Figure 2 As shown in the figure, the unit of the SPAD array includes a single-photon avalanche diode (SPAD), a quenching circuit, a shaping circuit and a counter circuit. When a photon is incident, the SPAD will convert the photon into an electrical pulse signal, which will be output as a standard pulse signal after passing through the shaping circuit, and the number of photons will be recorded by the counter circuit. The single-photon avalanche diode works in Geiger mode, and the bias voltage applied to it directly determines the efficiency of the SPAD in detecting photons. Figure 3 As shown in the figure, the magnitude of the bias voltage applied to the SPAD device has a very significant effect on the detection efficiency of photons. Therefore, by changing the bias voltage of the single-photon avalanche diode, the detection efficiency of the single-photon avalanche diode for photons can be changed, and the time encoding of the SPAD photon detection can be realized. Specifically, the number of photons detected on the SPAD depends not only on the total number of photons reaching the SPAD during the exposure time, but also on the bias voltage applied thereto. Different bias voltages correspond to different SPAD detection efficiencies. Specifically, when the number of photons reaching the SPAD is N, and the photon detection efficiency of the SPAD is ρ(x) under the bias voltage x, the total number of photons detected by the SPAD is Nρ(x). Through multiple measurements under different random coding bases, the projection measurement values ​​of the photon intensity at different times within the exposure time T under the random 0-1 coding measurement can be obtained, and the intensity of the photons arriving at different times within the time T can be obtained by solving the augmented Lagrangian algorithm, thereby realizing high-speed single-photon detection and imaging.

[0029] like Figure 4 As shown, the process of SPAD acquisition is that the laser light source emits a pulse laser to the scene, and the photons reflected from the scene are detected by the detection array SPAD and accumulated and counted during the exposure time. The different modulation signals received by the laser when emitting the pulse laser and the detection array SPAD when receiving the reflected light are used to reconstruct the scene in three dimensions. Affected by the geometry and material properties of the scene, the number of photons reflected from the scene changes with time, that is, it is a function of time, which is briefly described as the pulse response P(t) of the scene. In order to detect the pulse response P(t), the present invention proposes to introduce time coding during the exposure time to code and modulate the photon detection efficiency at different times, and accumulate and count the total number of photons arriving during the exposure time under the coding modulation. By repeating this process multiple times and using different time coding modulation in each process, the cumulative number of photons detected by the SPAD under different time coding of P(t) can be collected. The pulse response P(t) of the scene can be reconstructed by least squares reconstruction based on gradient descent.

[0030] like Figure 5 As shown in the figure, different time coding modulations are performed on the SPAD photon detection bias voltage, and the total number of detected photons under different codings can be collected by the counter in the SPAD pixel. The time coding here uses random 0-1 coding, that is, the efficiency ρ(x) of SPAD for photon detection is randomly 0 or 1 for different collections. Figure 5 Given a simple example, in w1 = [1, 0, 0, 1, 0, ..., 1, 0, 0] T Under the time coding, the SPAD accumulates the photons m1 that arrive intermittently during the exposure time, corresponding to w2 = [1, 0, 1, 1, ..., 0, 0, 0] T The cumulative number of photons under time coding is m2, corresponding to w3 = [1, 1, 1, 0, 1, ..., 0, 0, 1] T The cumulative number of photons under time coding is m3, and so on. Through multiple measurements, each measurement uses a different random 0-1 code, and we can get the random coding modulation W = [w1, w2, w3, ...., w N ] T The measured value under is m=[m1,m2,m3,....,m N ] T . By modeling the problem of reconstructing P(t) as a least squares problem:

[0031]

[0032] Here p is the vectorized form of P(t), H t is the gradient operator along the time dimension, where the first term is the data term, which constrains the projection value of the reconstructed time function p under the time encoding matrix W to be as close as possible to the measured value, and the second term is the sparse constraint term, which constrains the time function to be sparse. Based on the gradient descent algorithm, p can be reconstructed, that is, the light intensity reflected from the scene at different times under the pulsed laser. A single-photon detection array is used here. By performing the same time encoding on all pixel units of the entire array, the time response function of each point in the scene to the pulsed laser can be reconstructed, realizing high-speed imaging of the single-photon sensitivity of the scene.

Claims

1. A high-speed single-photon detection imaging method based on time coding, characterized in that: The steps of this method are as follows: (1) First, the relationship curve between the photon detection efficiency and bias voltage of a single single-photon avalanche diode (SPAD) is calibrated: under the same incident light intensity, different bias voltages are applied to the SPAD, and the corresponding total number of detected photons is collected to obtain the relationship curve ρ(x) between the SPAD bias voltage x and the detection efficiency; based on the calibrated curve ρ(x) and the intensity of the detection efficiency corresponding to the time code at different times, the bias voltage value that should be added at different times to achieve a certain time code can be obtained; (2) Emitting laser pulses to the scene, using an array of multiple SPADs to collect the light returning from the scene, and changing the detection efficiency of the SPAD by adjusting the bias voltage on the SPAD during the exposure time to achieve temporal encoding of the SPAD detection; Specifically, within the exposure time T, with Δt as the minimum time unit, the bias voltage on the SPAD is adjusted so that the detection efficiency of the SPAD is 0 or 1 or any decimal between 0 and 1, so that the corresponding time code is a random time code sequence; (3) For each exposure, a time code sequence is randomly generated, and the total number of photons detected by the SPAD under this time code is the measurement value corresponding to the time code; (4) repeating the random time coding and acquisition process of steps (2) and (3) multiple times, generating a different random time coding sequence each time, and correspondingly obtaining a different measurement value; (5) Based on different time coding sequences and their corresponding measurement values, an optimization objective function is established, and computational reconstruction based on gradient descent is performed to obtain the light intensity values ​​returned by the scene at different times under the pulsed laser, thereby realizing high-speed single-photon detection imaging.

2. The high-speed single-photon detection imaging method based on time coding according to claim 1 is characterized in that: In the step (2), the bias voltage on the SPAD is adjusted to a range of 0 to 3V.

3. The high-speed single-photon detection imaging method based on time coding according to claim 1 is characterized in that: In the step (2), the array unit includes a single SPAD, a quenching circuit, a shaping circuit and a counter circuit. The SPAD converts incident photons into electrical pulse signals, outputs standard pulse signals after passing through the shaping circuit, and the counter circuit records the number of photons.

Images