High-speed imaging method, system and device based on spad array and time counting circuit

Abstract

The application discloses a high-speed imaging method, system and device based on a SPAD array and a time counting circuit, and the method comprises the following steps: collecting photon event original data to obtain an original photon counting matrix of each frame; performing statistical modeling based on the original photon counting matrix to obtain a dynamic threshold matrix; performing threshold filtering based on the dynamic threshold matrix and the original photon counting matrix to obtain a sparse matrix; and performing accumulation and image reconstruction based on multiple frames of original photon counting matrices to output a final image. The application solves the noise problem of high-frame-rate imaging under extremely low illumination by combining the high time resolution of a SPAD array and TCSPC with a BNQSA algorithm, and has the characteristics of high sensitivity, real-time performance and low power consumption, thereby providing a breakthrough solution for a weak light scene.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and more specifically, to a high-speed imaging method, system, and apparatus based on a SPAD array and a time counting circuit. Background Technology

[0002] With the development of imaging technology, traditional CMOS image sensors and CCD image sensors have been widely used in various vision systems. However, when the ambient illuminance is as low as 0.01 lux, the signal-to-noise ratio of these sensors drops sharply, making it impossible to accurately acquire effective image information and limiting their application capabilities in fields such as night surveillance, deep space exploration, and quantum communication.

[0003] In recent years, single-photon detection technology (such as SPAD, Single-Photon Avalanche Diode 2D Array) has gradually become an important means to overcome the bottleneck of low-light imaging. Among them, SPAD can detect a single photon, but because it operates on a sub-nanosecond timescale, it is easily affected by thermal noise, dark counting, crosstalk, background photons, etc. At the same time, how to effectively count, suppress noise, and output clear images at high speed remains a technical challenge.

[0004] Currently, existing technologies mainly focus on SPAD unit optimization, TDC precision control, or back-end image fusion strategies. An integrated solution for high frame rate real-time imaging systems under extremely low illumination conditions has not yet been formed, and there is a lack of quantum active suppression methods for background noise. Summary of the Invention

[0005] The purpose of this invention is to provide a high-speed imaging method, system, and device based on SPAD array and time-correlation counting circuit, especially a high-speed imaging system combining single-photon avalanche diode array and time-correlation counting circuit. It also proposes a quantum suppression algorithm for background noise to enhance image quality and signal-to-noise ratio, which is widely applicable to fields with extremely high requirements for imaging in low light conditions, such as night security, astronomical observation, and biological imaging.

[0006] The first aspect of this invention provides a high-speed imaging method based on a SPAD array and a time-counting circuit, comprising the following steps:

[0007] The raw photon count matrix for each frame is obtained by collecting raw photon event data;

[0008] Statistical modeling is performed based on the original photon counting matrix to obtain a dynamic threshold matrix;

[0009] A sparse matrix is ​​obtained by performing threshold filtering based on the dynamic threshold matrix and the original photon counting matrix.

[0010] Accumulation and image reconstruction are performed based on the original photon counting matrix of multiple frames to output the final image.

[0011] In this scheme, the acquisition of raw photon event data to obtain the raw photon counting matrix for each frame specifically includes:

[0012] Acquire the photon trigger pulse signal output by the SPAD array, specifically including the timestamp sequence;

[0013] The photon arrival time of each pixel is recorded based on the photon trigger pulse signal;

[0014] The original photon count matrix for each frame is obtained by counting the photons of each pixel according to a preset time window.

[0015] In this scheme, the step of obtaining a dynamic threshold matrix by performing statistical modeling based on the original photon counting matrix specifically includes:

[0016] Simultaneous modeling is performed while acquiring the original photon counting matrix, with the background model being updated in the background when a new frame is acquired.

[0017] Calculate the background mean and standard deviation for each pixel in each frame;

[0018] The dynamic threshold matrix is ​​calculated based on the background mean, the standard deviation, and the preset weights.

[0019] In this scheme, the step of performing threshold filtering based on the dynamic threshold matrix combined with the original photon counting matrix to obtain a sparse matrix specifically includes:

[0020] Compare the photon counts in the original photon counting matrix with the corresponding thresholds pixel by pixel;

[0021] During the comparison process, the background noise is reduced to zero to retain the effective signal, and then threshold filtering is performed to obtain the sparse matrix.

[0022] In this scheme, the step of accumulating and reconstructing the image based on multiple frames of original photon counting matrices to output the final image specifically includes:

[0023] The final image is obtained by performing temporal accumulation, normalization, and dynamic range stretching on the timestamp sequence.

[0024] The final image is used to generate a visualization image through linear stretching and pseudo-color mapping, while the fusion window is automatically adjusted according to the frame rate and exposure control.

[0025] A second aspect of the present invention also provides a high-speed imaging system based on a SPAD array and a time-counting circuit, comprising a memory and a processor. The memory includes a high-speed imaging method program based on a SPAD array and a time-counting circuit. When the high-speed imaging method program based on a SPAD array and a time-counting circuit is executed by the processor, it performs the following steps:

[0026] The raw photon count matrix for each frame is obtained by collecting raw photon event data;

[0027] Statistical modeling is performed based on the original photon counting matrix to obtain a dynamic threshold matrix;

[0028] A sparse matrix is ​​obtained by performing threshold filtering based on the dynamic threshold matrix and the original photon counting matrix.

[0029] Accumulation and image reconstruction are performed based on the original photon counting matrix of multiple frames to output the final image.

[0030] In this scheme, the acquisition of raw photon event data to obtain the raw photon counting matrix for each frame specifically includes:

[0031] Acquire the photon trigger pulse signal output by the SPAD array, specifically including the timestamp sequence;

[0032] The photon arrival time of each pixel is recorded based on the photon trigger pulse signal;

[0033] The original photon count matrix for each frame is obtained by counting the photons of each pixel according to a preset time window.

[0034] In this scheme, the step of obtaining a dynamic threshold matrix by performing statistical modeling based on the original photon counting matrix specifically includes:

[0035] Simultaneous modeling is performed while acquiring the original photon counting matrix, with the background model being updated in the background when a new frame is acquired.

[0036] Calculate the background mean and standard deviation for each pixel in each frame;

[0037] The dynamic threshold matrix is ​​calculated based on the background mean, the standard deviation, and the preset weights.

[0038] In this scheme, the step of performing threshold filtering based on the dynamic threshold matrix combined with the original photon counting matrix to obtain a sparse matrix specifically includes:

[0039] Compare the photon counts in the original photon counting matrix with the corresponding thresholds pixel by pixel;

[0040] During the comparison process, the background noise is reduced to zero to retain the effective signal, and then threshold filtering is performed to obtain the sparse matrix.

[0041] In this scheme, the step of accumulating and reconstructing the image based on multiple frames of original photon counting matrices to output the final image specifically includes:

[0042] The final image is obtained by performing temporal accumulation, normalization, and dynamic range stretching on the timestamp sequence.

[0043] The final image is used to generate a visualization image through linear stretching and pseudo-color mapping, while the fusion window is automatically adjusted according to the frame rate and exposure control.

[0044] A third aspect of the present invention provides a computer-readable storage medium comprising a machine program for a high-speed imaging method based on a SPAD array and a time-counting circuit, wherein when the high-speed imaging method program based on a SPAD array and a time-counting circuit is executed by a processor, it implements the steps of a high-speed imaging method based on a SPAD array and a time-counting circuit as described in any of the preceding claims.

[0045] A fourth aspect of the present invention provides a high-speed imaging device based on a SPAD array and a time-counting circuit, the device comprising:

[0046] The optical collection module specifically includes a high-transmittance lens that collects the maximum number of incident photons, and is equipped with an anti-reflective coating.

[0047] The SPAD two-dimensional detector array consists of several single-photon avalanche detectors. Each SPAD unit consists of an n+ buried layer, a p substrate, a ring anode, and an overvoltage control MOS.

[0048] The time-correlated single-photon counting module specifically includes acquiring the photon trigger time of each pixel in the SPAD unit, and using a dead-zone time management mechanism after triggering to record the count value of each pixel within a unit exposure window;

[0049] A multi-channel time-to-digital converter, wherein each row or column is provided with an independent time-to-digital converter channel, and the channel adopts a differential ring oscillator and a time interval latch structure;

[0050] The image processing module implements any one of the high-speed imaging methods based on a SPAD array and a time counting circuit through an FPGA or DSP.

[0051] The data interface module uses temporary buffering and high-speed data links to transmit image data streams.

[0052] This invention discloses a high-speed imaging method, system, and device based on a SPAD array and a time-counting circuit. By combining the high temporal resolution of a SPAD array with a Time-Correlated Single-Photon Counting (TCSPC) module and a Background Noise Quantum Suppression Algorithm (BNQSA), it solves the noise problem in high-frame-rate imaging under extremely low light conditions. It combines high sensitivity, real-time performance, and low power consumption, providing a breakthrough solution for low-light scenarios. Specific benefits include: achieving 1 million frames per second single-photon imaging in a 0.001 lux environment; real-time background noise suppression with a signal-to-noise ratio improvement of up to 23 dB; support for multi-channel parallel counting and parallel image reconstruction, possessing potential for miniaturization and on-chip integration; and broad application prospects in fields such as biofluorescence imaging, quantum optics, and nighttime drone navigation. Attached Figure Description

[0053] Figure 1 A flowchart of a high-speed imaging method based on a SPAD array and a time counting circuit according to the present invention is shown;

[0054] Figure 2 A schematic diagram of background noise spatial distribution function modeling is shown for a high-speed imaging method based on SPAD array and time counting circuit according to the present invention.

[0055] Figure 3 A block diagram of a high-speed imaging system based on a SPAD array and a time counting circuit according to the present invention is shown.

[0056] Figure 4 The SPAD array pixel structure and its working timing diagram are shown in the high-speed imaging method based on SPAD array and time counting circuit of the present invention.

[0057] Figure 5 The diagram shows the relationship between frame rate and illumination intensity in a high-speed imaging method based on a SPAD array and a time counting circuit according to the present invention.

[0058] Figure 6 The diagram shows a comparative imaging experiment of a high-speed imaging method based on a SPAD array and a time counting circuit according to the present invention. Detailed Implementation

[0059] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0060] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0061] Figure 1 A flowchart of a high-speed imaging method based on a SPAD array and a time-counting circuit is shown in this application.

[0062] like Figure 1 As shown, this application discloses a high-speed imaging method based on a SPAD array and a time-counting circuit, including the following steps:

[0063] S102, Collect raw photon event data to obtain the raw photon counting matrix for each frame;

[0064] S104, Statistical modeling is performed based on the original photon counting matrix to obtain a dynamic threshold matrix;

[0065] S106, based on the dynamic threshold matrix and the original photon counting matrix, threshold filtering is performed to obtain a sparse matrix;

[0066] S108 performs accumulation and image reconstruction based on the multi-frame original photon counting matrix to output the final image.

[0067] It should be noted that in this embodiment, under extremely low illumination, there are a large number of non-imaging photon events (such as dark counting, background scattering, etc.) in the acquired data. Therefore, a BNQSA algorithm based on a photon statistical distribution model is proposed. This includes acquiring raw photon event data to obtain the raw photon counting matrix for each frame, and performing statistical modeling based on the raw photon counting matrix to obtain a dynamic threshold matrix. Specifically, Poisson statistics are used to model the probability distribution of incident photon arrival, with a time window defined as... The number of incident photons is a random variable. Its probability distribution is: ,in This represents the average number of photons within a unit time window. For actual counts, , This refers to the detector's quantum efficiency (typ. 0.35–0.55). It is the incident photon flux per unit time per unit area (photon / s / mm²). It is the exposure time window (e.g.) The expectation and variance of the Poisson distribution are both 1. Therefore, in low light intensity ( Under these conditions, the fluctuations are very obvious, and the image noise mainly comes from statistical jitter rather than electronic interference.

[0068] Furthermore, in this embodiment, a sparse matrix is ​​obtained by combining the dynamic threshold matrix with the original photon counting matrix through threshold filtering. This is achieved by acquiring multiple frames of sample data under dark field conditions and extracting the background noise spatial distribution function. The process involves constructing a quantization threshold matrix and performing threshold filtering to obtain a sparse matrix. Finally, the process is accumulated and image reconstructed based on the original photon count matrix of multiple frames to output the final image. The specific process will be explained in detail in the subsequent manual.

[0069] According to an embodiment of the present invention, the acquisition of raw photon event data to obtain the raw photon counting matrix for each frame specifically includes:

[0070] Acquire the photon trigger pulse signal output by the SPAD array, specifically including the timestamp sequence;

[0071] The photon arrival time of each pixel is recorded based on the photon trigger pulse signal;

[0072] The original photon count matrix for each frame is obtained by counting the photons of each pixel according to a preset time window.

[0073] It should be noted that, in this embodiment, under extremely low illumination (0.001 lux) conditions, the effective photon triggering frequency of a single SPAD pixel is 10~100 photon / s. In order to obtain sufficient image information, this embodiment uses a 1,000,000 fps sampling time window (1µs) to perform triggering statistics on all pixels.

[0074] ;

[0075] in, For pixels In the The number of photons per frame, It is a unit impulse function. Let k be the timestamp of the k-th photon event. This is the total timestamp.

[0076] Furthermore, in this embodiment, when the SPAD is photon-triggered at a certain moment, its response signal can be expressed as a unit impulse function. ;in, For the first The timestamp of each photon event actually records the photon count value within each exposure window. , For the first The timestamp of a photon event.

[0077] According to an embodiment of the present invention, the step of obtaining a dynamic threshold matrix by performing statistical modeling based on the original photon counting matrix specifically includes:

[0078] Simultaneous modeling is performed while acquiring the original photon counting matrix, with the background model being updated in the background when a new frame is acquired.

[0079] Calculate the background mean and standard deviation for each pixel in each frame;

[0080] The dynamic threshold matrix is ​​calculated based on the background mean, the standard deviation, and the preset weights.

[0081] It should be noted that, in this embodiment, the background model is updated in the background when a new frame is acquired, and the average background value of each pixel in each frame is calculated. with standard deviation In this embodiment, the calculation process is not described in detail here, but specifically, the background expectation value is extracted based on multi-frame low-light sampling. and calculate the standard deviation. Multiple frames of sample data were acquired under dark conditions, and the spatial distribution function of background noise was extracted. And construct the quantization threshold matrix:

[0082] ;

[0083] in, , These are the mean and standard deviation of the background region, respectively. For adjustment coefficients, It can be adjusted according to actual application.

[0084] Furthermore, in this embodiment, modeling is based on a time-to-digital converter (TDC), where the TDC is used to map continuous-time events to discrete-time indices, and its theoretical accuracy is determined by the following: ,in, It is the frequency of the ring oscillator (typically 1 GHz). The time interpolation subdivision bits (usually 128 or 256) limit the frame rate upper limit and time jitter error in this embodiment.

[0085] According to an embodiment of the present invention, the step of performing threshold filtering based on the dynamic threshold matrix combined with the original photon counting matrix to obtain a sparse matrix specifically includes:

[0086] Compare the photon counts in the original photon counting matrix with the corresponding thresholds pixel by pixel;

[0087] During the comparison process, the background noise is reduced to zero to retain the effective signal, and then threshold filtering is performed to obtain the sparse matrix.

[0088] It should be noted that, in this embodiment, the background noise mainly comes from systematic deviations such as the inherent thermal noise of the dark count rate (DCR)-SPAD, external background light, electromagnetic crosstalk between neighboring pixels, and power supply noise.

[0089] Furthermore, in this embodiment, the photon counts in the original photon counting matrix are compared pixel by pixel with the corresponding thresholds. During the comparison process, background noise is reduced to zero to retain the effective signal, and threshold filtering is performed to obtain the sparse matrix. The original photon statistical map is then used to... Threshold processing is performed as follows: Thus, the sparse matrix is ​​obtained, where, as Figure 2 As shown, this is a schematic diagram illustrating the spatial distribution function modeling of background noise, where, Figure 2 In the diagram, A corresponds to the average background photon count. Figure 2 B in the diagram corresponds to the standard deviation distribution. Figure 2 C in the diagram corresponds to the pixel-level threshold map.

[0090] Specifically, based on extensive sampling and verification, assuming the background noise follows a Gaussian distribution, its probability density function is... The corresponding cumulative distribution function obtained by combining the set threshold is: ,in, It is the standard normal distribution function, therefore, At that time, approximately 97.7% of the background noise was effectively filtered, assuming the background intensity was low ( The target signal strength is Let the probability of misjudgment be... Its theoretical value is: When applying it, choose the appropriate Can make .

[0091] According to an embodiment of the present invention, the step of accumulating and reconstructing the image based on the multi-frame original photon counting matrix to output the final image specifically includes:

[0092] The final image is obtained by performing temporal accumulation, normalization, and dynamic range stretching on the timestamp sequence.

[0093] The final image is used to generate a visualization image through linear stretching and pseudo-color mapping, while the fusion window is automatically adjusted according to the frame rate and exposure control.

[0094] It should be noted that, in this embodiment, the final image is obtained through multi-frame accumulation processing after background filtering. The final image is generated into a visual image through linear stretching and pseudo-color mapping. At the same time, the fusion window is automatically adjusted according to the frame rate and exposure control to ensure that no ghosting or image retention occurs in dynamic scenes.

[0095] Figure 3 A block diagram of a high-speed imaging system based on a SPAD array and a time counting circuit according to the present invention is shown.

[0096] like Figure 3 As shown, this invention discloses a high-speed imaging system based on a SPAD array and a time-counting circuit, including a memory and a processor. The memory includes a high-speed imaging method program based on a SPAD array and a time-counting circuit. When the processor executes the high-speed imaging method program based on a SPAD array and a time-counting circuit, it implements the following steps:

[0097] The raw photon count matrix for each frame is obtained by collecting raw photon event data;

[0098] Statistical modeling is performed based on the original photon counting matrix to obtain a dynamic threshold matrix;

[0099] A sparse matrix is ​​obtained by performing threshold filtering based on the dynamic threshold matrix and the original photon counting matrix.

[0100] Accumulation and image reconstruction are performed based on the original photon counting matrix of multiple frames to output the final image.

[0101] It should be noted that in this embodiment, under extremely low illumination, there are a large number of non-imaging photon events (such as dark counting, background scattering, etc.) in the acquired data. Therefore, a BNQSA algorithm based on a photon statistical distribution model is proposed. This includes acquiring raw photon event data to obtain the raw photon counting matrix for each frame, and performing statistical modeling based on the raw photon counting matrix to obtain a dynamic threshold matrix. Specifically, Poisson statistics are used to model the probability distribution of incident photon arrival, with a time window defined as... The number of incident photons is a random variable. Its probability distribution is: ,in This represents the average number of photons within a unit time window, where k is the actual count. , This refers to the detector's quantum efficiency (typ. 0.35–0.55). It is the incident photon flux per unit time per unit area (photon / s / mm²). It is the exposure time window (e.g.) The expectation and variance of the Poisson distribution are both 1. Therefore, in low light intensity ( Under these conditions, the fluctuations are very obvious, and the image noise mainly comes from statistical jitter rather than electronic interference.

[0102] Furthermore, in this embodiment, a sparse matrix is ​​obtained by combining the dynamic threshold matrix with the original photon counting matrix through threshold filtering. This is achieved by acquiring multiple frames of sample data under dark field conditions and extracting the background noise spatial distribution function. The process involves constructing a quantization threshold matrix and performing threshold filtering to obtain a sparse matrix. Finally, the process is accumulated and image reconstructed based on the original photon count matrix of multiple frames to output the final image. The specific process will be explained in detail in the subsequent manual.

[0103] According to an embodiment of the present invention, the acquisition of raw photon event data to obtain the raw photon counting matrix for each frame specifically includes:

[0104] Acquire the photon trigger pulse signal output by the SPAD array, specifically including the timestamp sequence;

[0105] The photon arrival time of each pixel is recorded based on the photon trigger pulse signal;

[0106] The original photon count matrix for each frame is obtained by counting the photons of each pixel according to a preset time window.

[0107] It should be noted that, in this embodiment, under extremely low illumination (0.001 lux) conditions, the effective photon triggering frequency of a single SPAD pixel is 10~100 photon / s. In order to obtain sufficient image information, this embodiment uses a 1,000,000 fps sampling time window (1µs) to perform triggering statistics on all pixels.

[0108] ;

[0109] in, For pixels The number of photons in the t-th frame, It is a unit impulse function. For the first A timestamp of a photon event. This is the total timestamp.

[0110] Furthermore, in this embodiment, when the SPAD is photon-triggered at a certain moment, its response signal can be expressed as a unit impulse function. ;in, For the first The timestamp of each photon event actually records the photon count value within each exposure window. , For the first The timestamp of a photon event.

[0111] According to an embodiment of the present invention, the step of obtaining a dynamic threshold matrix by performing statistical modeling based on the original photon counting matrix specifically includes:

[0112] Simultaneous modeling is performed while acquiring the original photon counting matrix, with the background model being updated in the background when a new frame is acquired.

[0113] Calculate the background mean and standard deviation for each pixel in each frame;

[0114] The dynamic threshold matrix is ​​calculated based on the background mean, the standard deviation, and the preset weights.

[0115] It should be noted that, in this embodiment, the background model is updated in the background when a new frame is acquired, and the average background value of each pixel in each frame is calculated. with standard deviation In this embodiment, the calculation process is not described in detail here, but specifically, the background expectation value is extracted based on multi-frame low-light sampling. and calculate the standard deviation. Multiple frames of sample data were acquired under dark conditions, and the spatial distribution function of background noise was extracted. And construct the quantization threshold matrix:

[0116] ;

[0117] in, , These are the mean and standard deviation of the background region, respectively. For adjustment coefficients, It can be adjusted according to actual application.

[0118] Furthermore, in this embodiment, modeling is based on a time-to-digital converter (TDC), where the TDC is used to map continuous-time events to discrete-time indices, and its theoretical accuracy is determined by the following: ,in, It is the frequency of the ring oscillator (typically 1 GHz). The time interpolation subdivision bits (usually 128 or 256) limit the frame rate upper limit and time jitter error in this embodiment.

[0119] According to an embodiment of the present invention, the step of performing threshold filtering based on the dynamic threshold matrix combined with the original photon counting matrix to obtain a sparse matrix specifically includes:

[0120] Compare the photon counts in the original photon counting matrix with the corresponding thresholds pixel by pixel;

[0121] During the comparison process, the background noise is reduced to zero to retain the effective signal, and then threshold filtering is performed to obtain the sparse matrix.

[0122] It should be noted that, in this embodiment, the background noise mainly comes from systematic deviations such as the inherent thermal noise of the dark count rate (DCR)-SPAD, external background light, electromagnetic crosstalk between neighboring pixels, and power supply noise.

[0123] Furthermore, in this embodiment, the photon counts in the original photon counting matrix are compared pixel by pixel with the corresponding thresholds. During the comparison process, background noise is reduced to zero to retain the effective signal, and threshold filtering is performed to obtain the sparse matrix. The original photon statistical map is then used to... Threshold processing is performed as follows: Thus, the sparse matrix is ​​obtained, where, as Figure 2 As shown, this is a schematic diagram illustrating the spatial distribution function modeling of background noise, where, Figure 2 In the diagram, A corresponds to the average background photon count. Figure 2 B in the diagram corresponds to the standard deviation distribution. Figure 2 C in the diagram corresponds to the pixel-level threshold map.

[0124] Specifically, based on extensive sampling and verification, assuming the background noise follows a Gaussian distribution, its probability density function is... The corresponding cumulative distribution function obtained by combining the set threshold is: ,in, It is the standard normal distribution function, therefore, At that time, approximately 97.7% of the background noise was effectively filtered, assuming the background intensity was low ( The target signal strength is Let the probability of misjudgment be... Its theoretical value is: When applying it, choose the appropriate Can make .

[0125] According to an embodiment of the present invention, the step of accumulating and reconstructing the image based on the multi-frame original photon counting matrix to output the final image specifically includes:

[0126] The final image is obtained by performing temporal accumulation, normalization, and dynamic range stretching on the timestamp sequence.

[0127] The final image is used to generate a visualization image through linear stretching and pseudo-color mapping, while the fusion window is automatically adjusted according to the frame rate and exposure control.

[0128] It should be noted that, in this embodiment, the final image is obtained through multi-frame accumulation processing after background filtering. The final image is generated into a visual image through linear stretching and pseudo-color mapping. At the same time, the fusion window is automatically adjusted according to the frame rate and exposure control to ensure that no ghosting or image retention occurs in dynamic scenes.

[0129] A third aspect of the present invention provides a computer-readable storage medium comprising a high-speed imaging method program based on a SPAD array and a time-counting circuit. When executed by a processor, the high-speed imaging method program based on a SPAD array and a time-counting circuit implements the steps of the high-speed imaging method based on a SPAD array and a time-counting circuit as described in any of the preceding claims.

[0130] A fourth aspect of the present invention provides a high-speed imaging device based on a SPAD array and a time-counting circuit, the device comprising:

[0131] The optical collection module specifically includes a high-transmittance lens that collects the maximum number of incident photons, and is equipped with an anti-reflective coating.

[0132] The SPAD two-dimensional detector array consists of several single-photon avalanche detectors. Each SPAD unit consists of an n+ buried layer, a p substrate, a ring anode, and an overvoltage control MOS.

[0133] The time-correlated single-photon counting module specifically includes acquiring the photon trigger time of each pixel in the SPAD unit, and using a dead-zone time management mechanism after triggering to record the count value of each pixel within a unit exposure window;

[0134] A multi-channel time-to-digital converter, wherein each row or column is provided with an independent time-to-digital converter channel, and the channel adopts a differential ring oscillator and a time interval latch structure;

[0135] The image processing module implements any one of the high-speed imaging methods based on a SPAD array and a time counting circuit through an FPGA or DSP.

[0136] The data interface module uses temporary buffering and high-speed data links to transmit image data streams.

[0137] It should be noted that, in this embodiment, the purpose of the present invention is to provide a high frame rate (up to 1,000,000 fps) imaging system suitable for extremely low illumination conditions (such as 0.001 lux). This system combines a SPAD two-dimensional array with a multi-channel time-correlated single-photon counting (TCSPC) circuit, supplemented by the background noise quantum suppression algorithm (BNQSA) proposed in this invention, thereby significantly improving image clarity and signal-to-noise ratio, and achieving photon-level imaging. In this system, the SPAD is excited into an avalanche state when photons are incident, achieving single-photon-level detection. Its output signal is amplified at the front end and then input into the TDC module. The TCSPC module records the timestamp of each photon event. Multiple channels work in parallel to support high frame rate imaging.

[0138] Specifically, in this embodiment, as Figure 4 As shown, the diagram illustrates the pixel structure of a SPAD array and its timing sequence. Figure 4 A in the diagram is a schematic diagram of the SPAD array pixel structure, which is composed of N×N (e.g., 128×128) single-photon avalanche detectors. Each SPAD unit consists of an n+ buried layer, a p substrate, a ring anode, and an overvoltage control MOS, and operates in Geiger mode with a breakdown voltage of approximately 20–30V. Figure 4 B in the diagram represents the SPAD timing sequence, illustrating the key timing relationships between photon triggering, dead-time control, and TDC (Time-to-Digital Converter) timestamp recording. The optical collection module employs a high-throughput lens (F / 1.2, band-optimized) to collect the maximum number of incident photons, and incorporates an anti-reflective coating to reduce reflection loss. Specifically... Figure 5 As shown, this is a schematic diagram illustrating the relationship between frame rate and light intensity.

[0139] Furthermore, in this embodiment, the Time-Correlated Single-Photon Counting (TCSPC) module collects the photon trigger time of each pixel of the SPAD and adopts a dead-time management mechanism after triggering to record the count value of each pixel within a unit exposure window (e.g., 1µs). The Time-to-Digital Converter Channel (TDC) sets up an independent TDC channel for each row or column to avoid crosstalk interference and improve the frame rate. The TDC adopts a differential ring oscillator and a time interval latch structure, and the time resolution can reach 10–20ps.

[0140] Furthermore, in this embodiment, the image processing module implements a high-speed imaging method based on a SPAD array and a time counting circuit as described above through an FPGA or DSP, specifically realizing real-time background modeling, photon denoising, pseudo-color imaging, and multi-frame fusion; and the data interface module specifically uses DDR4 temporary cache and PCIe / USB3.0 high-speed data link to achieve image data stream transmission of up to several GB per second.

[0141] Furthermore, in this embodiment, in specific applications, such as Figure 6 As shown, this is a schematic diagram comparing imaging experiments, in which... Figure 6 In the image, A corresponds to the original image, showing the raw image without photon counting processing. This image is characterized by dense noise and a blurred target. Figure 6 In this example, B corresponds to the algorithm application diagram, where noise is significantly reduced and the target is clearly identifiable. Figure 6 The 'C' in the diagram represents the effect of traditional median filtering, which is effective in noise reduction but suffers from severe loss of detail and blurred edges. The overall performance of each module can be modeled as follows: maximum frame rate. ,in For the exposure time, For system bandwidth, For single-frame data volume, the relationship between image resolution and SNR is as follows: Where B is the total number of false triggers in the background, and DCR is the dark count rate per pixel.

[0142] This invention discloses a high-speed imaging method, system, and device based on a SPAD array and a time counting circuit. By combining the high temporal resolution of the SPAD array with the TCSPC and the BNQSA algorithm, it solves the noise problem of high frame rate imaging under extremely low light conditions, and has the characteristics of high sensitivity, real-time performance, and low power consumption, providing a breakthrough solution for low-light scenes.

[0143] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0144] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0145] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0146] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0147] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

Claims

1. A high-speed imaging method based on SPAD array and time counting circuit, characterized in that, Includes the following steps: The process of acquiring raw photon event data to obtain the raw photon counting matrix for each frame includes: acquiring the photon trigger pulse signal output by the SPAD array, specifically including a timestamp sequence; recording the photon arrival time of each pixel based on the photon trigger pulse signal; and calculating the photon count of each pixel according to a preset time window to obtain the raw photon counting matrix for each frame. Statistical modeling is performed based on the original photon counting matrix to obtain a dynamic threshold matrix. Poisson statistics are used to model the probability distribution of incident photon arrival. Specifically, this includes: synchronous modeling when acquiring the original photon counting matrix, wherein the background model is updated in the background when acquiring a new frame; calculating the background mean and standard deviation of each pixel in each frame; and calculating the dynamic threshold matrix based on the background mean, the standard deviation, and preset weights. The sparse matrix is ​​obtained by threshold filtering based on the dynamic threshold matrix and the original photon counting matrix. Specifically, this includes: comparing the photon counts in the original photon counting matrix with the corresponding thresholds pixel by pixel; and reducing the background noise to zero during the comparison process to retain the effective signal and then performing threshold filtering to obtain the sparse matrix. After filtering out background noise, the image is accumulated and reconstructed based on the original photon counting matrix of multiple frames to output the final image.

2. The high-speed imaging method based on SPAD array and time counting circuit according to claim 1, characterized in that, The process of accumulating and reconstructing images based on multi-frame raw photon counting matrices to output the final image specifically includes: The final image is obtained by performing temporal accumulation, normalization, and dynamic range stretching on the timestamp sequence. The final image is linearly stretched and pseudo-color mapped to generate a visualization image, while the fusion window is automatically adjusted according to the frame rate and exposure control.

3. A high-speed imaging system based on a SPAD array and a time-counting circuit, characterized in that, The system includes a memory and a processor. The memory contains a high-speed imaging method program based on a SPAD array and a time-counting circuit. When executed by the processor, the high-speed imaging method program based on a SPAD array and a time-counting circuit implements the following steps: The process of acquiring raw photon event data to obtain the raw photon counting matrix for each frame includes: acquiring the photon trigger pulse signal output by the SPAD array, specifically including a timestamp sequence; recording the photon arrival time of each pixel based on the photon trigger pulse signal; and calculating the photon count of each pixel according to a preset time window to obtain the raw photon counting matrix for each frame. Statistical modeling is performed based on the original photon counting matrix to obtain a dynamic threshold matrix. Poisson statistics are used to model the probability distribution of incident photon arrival. Specifically, this includes: synchronous modeling when acquiring the original photon counting matrix, wherein the background model is updated in the background when acquiring a new frame; calculating the background mean and standard deviation of each pixel in each frame; and calculating the dynamic threshold matrix based on the background mean, the standard deviation, and preset weights. The sparse matrix is ​​obtained by threshold filtering based on the dynamic threshold matrix and the original photon counting matrix. Specifically, this includes: comparing the photon counts in the original photon counting matrix with the corresponding thresholds pixel by pixel; and reducing the background noise to zero during the comparison process to retain the effective signal and then performing threshold filtering to obtain the sparse matrix. After filtering out background noise, the image is accumulated and reconstructed based on the original photon counting matrix of multiple frames to output the final image.

4. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a high-speed imaging method program based on a SPAD array and a time-counting circuit. When the high-speed imaging method program based on a SPAD array and a time-counting circuit is executed by a processor, it implements the steps of a high-speed imaging method based on a SPAD array and a time-counting circuit as described in any one of claims 1 to 2.

5. A high-speed imaging device based on a SPAD array and a time-counting circuit, characterized in that, The device includes: The optical collection module specifically includes a high-transmittance lens that collects the maximum number of incident photons, and is equipped with an anti-reflective coating. The SPAD two-dimensional detector array consists of several single-photon avalanche detectors. Each SPAD unit consists of an n+ buried layer, a p substrate, a ring anode, and an overvoltage control MOS. The time-correlated single-photon counting module specifically includes acquiring the photon trigger time of each pixel in the SPAD unit, and using a dead-zone time management mechanism after triggering to record the count value of each pixel within the unit exposure window; A multi-channel time-to-digital converter, wherein each row or column is provided with an independent time-to-digital converter channel, and the channel adopts a differential ring oscillator and a time interval latch structure; The image processing module implements the high-speed imaging method based on SPAD array and time counting circuit as described in any one of claims 1 to 2 through FPGA or DSP; The data interface module uses temporary buffering and high-speed data links to transmit image data streams.

Images