A laser seeker adaptive laser pulse detection method

Abstract

The application discloses a kind of laser seeker adaptive laser pulse detection methods, the module including ADC sampling module, noise modeling module, matched filter module, data preprocessing module, dynamic threshold detection module and pulse positioning module;The application acquires noise sample under the condition of no laser irradiation, calculates noise mean and variance, establishes environmental noise adaptive model;Then the baseline correction is carried out to input signal, and the signal envelope is extracted by matched filter based on raised cosine function construction;Based on noise model, set dynamic detection threshold, only search effective pulse peak value in the set sampling point.The application has low computational complexity, does not need multiple frame average, and can be run in real time on embedded platform such as Zynq.Compared with traditional FFT method, the application improves the success rate under low SNR condition, reduces processing time, and is suitable for laser ranging, laser guidance and other high real-time requirement application scenarios.

Description

Technical Field

[0001] This invention belongs to the field of signal processing technology, and specifically relates to an adaptive laser pulse detection method for laser seekers suitable for embedded systems. Background Technology

[0002] In laser-guided applications, it is typically necessary to detect nanosecond-level laser pulse signals. Existing detection technologies mainly face the following challenges:

[0003] First, the traditional FFT frequency domain denoising method has high computational complexity and requires a large number of floating-point operations, making it difficult to achieve real-time processing in resource-constrained embedded systems. Moreover, when the signal energy is dispersed across multiple frequency bands, frequency domain filtering is prone to introducing spectral leakage, leading to pulse waveform distortion.

[0004] Second, a general-purpose matched filter is adopted. Since traditional matched filters usually use the system's transmitted pulse as a template, in actual system applications, the received pulse waveform is often distorted due to factors such as optical path transmission and detector response, resulting in a decrease in matching performance. In addition, improper boundary processing can also seriously affect the detection accuracy of the first pulse (usually the nearest target echo).

[0005] Third, when using a fixed threshold detection method, the fixed threshold cannot be adaptively adjusted in scenarios with changing environmental noise (such as temperature drift or power fluctuation), resulting in a high false detection rate or decreased detection sensitivity.

[0006] Fourth, the use of a general-purpose IIR / FIR filter without parameter optimization for specific pulse widths and sampling rates makes it impossible to maximize noise suppression while preserving pulse characteristics.

[0007] Fifth, the multi-frame averaging method relies on multiple samplings, which cannot meet the requirements of real-time detection of a single frame and is not suitable for high-speed dynamic scenes.

[0008] Therefore, there is an urgent need for a single-frame laser pulse detection scheme that is computationally efficient, highly adaptable, and can be optimized for specific pulse widths. Summary of the Invention

[0009] To address the aforementioned problems, the purpose of this invention is to provide an adaptive laser pulse detection method for laser seekers that integrates adaptive noise estimation and optimized raised cosine matched filtering, thereby solving the problems of high computational complexity, poor adaptability, significant boundary effects, and DC offset interference in existing detection methods.

[0010] To achieve the above objectives, the present invention adopts the following technical solution: an adaptive laser pulse detection method for a laser seeker, comprising modules including an ADC sampling module, a noise modeling module, a matched filtering module, a data preprocessing module, a dynamic threshold detection module, and a pulse positioning module. The specific implementation steps of the detection method are as follows:

[0011] S1. Noise samples are collected by the ADC sampling module under conditions without laser irradiation, and then the noise mean and variance are calculated by the noise modeling module to establish an adaptive environmental noise model.

[0012] S2, the ADC sampling module synchronously triggers laser emission and data acquisition, the data preprocessing module uses the modeled noise mean to perform baseline correction on the input signal, and the matched filter module based on the raised cosine function extracts the signal envelope;

[0013] S3, the dynamic threshold detection module is based on an adaptive environmental noise model. It sets a dynamic detection threshold (3σ principle) and searches for valid pulse peaks only within the set sampling points. The specific operation is as follows:

[0014] Set the detection threshold to ,in Let k be the noise standard deviation, and k be the sensitivity coefficient (usually taken as 3.0, corresponding to 3). in principle).

[0015] S4. The pulse positioning module searches for peak values ​​within a physically reasonable window of the set sampling points. That is, on the envelope signal output by the matched filter, it searches for peak values ​​exceeding the threshold within the set sampling points based on the characteristics of the laser pulse width, and returns the precise position.

[0016] Furthermore, the establishment of the environmental noise adaptive model in step S1 involves acquiring a frame of noise data of length N through a noise modeling module under conditions of no laser irradiation. Through the calculation formula:

[0017]

[0018] Calculate the noise mean and variance An adaptive model for environmental noise was established.

[0019] Furthermore, the implementation steps for baseline correction and signal envelope extraction in step S2 are as follows:

[0020] S2.1, Prepare the filter template

[0021] S2.1.1 Set the basic parameters of the template, where the template length sampling point is set to M; the center position is set to... That is, the center point of symmetry of the template; the time constant is set to The sampling frequency determines the width of the raised cosine function;

[0022] S2.1.2 Calculate the template coefficients. For each index k (0 ≤ k ≤ M), perform the following calculation steps:

[0023] Calculate the time offset: Map index k to a time coordinate system with the center as the origin, so that t is in the interval Internal changes;

[0024] Conditional judgments and function calculations:

[0025]

[0026] Because the template length just covers All points satisfy The condition is such that each point is calculated using the raised cosine function; this function is... It reaches a maximum value of 1.0 at that point. When the value is 0, a perfect bell-shaped curve is formed;

[0027] The template coefficients are normalized as follows:

[0028]

[0029] Calculate the sum of all M coefficients Then divide each coefficient by S;

[0030] Normalization ensures that the sum of the template coefficients is 1 ( To maintain signal energy conservation, avoid amplitude distortion after filtering, and ensure that the amplitude of the matched filter output has physical meaning (comparable to the amplitude of the input signal).

[0031] S2.2, Obtaining the normalized template Then, the baseline-corrected signal The steps for performing matched filtering are as follows:

[0032] Perform the sliding dot product calculation. For each output point n (0 ≤ n < M), calculate:

[0033]

[0034] For a fixed n, let the template index m slide from 0 to M with the following index: ,make sure:

[0035] when At that time, the corresponding signal point n (the center of the template is aligned with the current output point);

[0036] when When accessing the right side of signal point n ;

[0037] when When accessing the left side of signal point n ;

[0038] The above indexing arrangement makes the template center ( It is always aligned with the current calculation point n.

[0039] Compared with existing technologies, the core of this invention lies in establishing an adaptive noise model using noise samples in a laser-free state, and achieving high-precision laser pulse detection under single-frame data conditions. This invention has low computational complexity, requires no multi-frame averaging, and can run in real-time on embedded platforms such as Zynq. Compared with traditional FFT methods, this invention improves detection success rate and reduces processing time under low signal-to-noise ratio conditions, making it suitable for applications with high real-time requirements such as laser ranging and laser guidance. Attached Figure Description

[0040] Figure 1 This is a system architecture diagram of the method of the present invention;

[0041] Figure 2 This is the ideal template diagram for the matched filter;

[0042] Figure 3 This is a comparison chart of the actual detection effects of FFT filtering and the matched filtering of this invention at a signal-to-noise ratio of -5dB.

[0043] Figure 4 This is a comparison chart of the actual detection effects of FFT filtering and the matched filtering of this invention at a signal-to-noise ratio of 0dB.

[0044] Figure 5 This is a comparison chart of the actual detection effects of FFT filtering and the matched filtering of this invention at a signal-to-noise ratio of 5dB.

[0045] Figure 6 This is a comparison chart of the actual detection effects of FFT filtering and the matched filtering of this invention at a signal-to-noise ratio of 10dB. Detailed Implementation

[0046] The technical solution of the present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0047] like Figure 1 As shown in this embodiment, an adaptive laser pulse detection method for a laser seeker employs modules including an ADC sampling module, a noise modeling module, a matched filtering module, a data preprocessing module, a dynamic threshold detection module, a pulse positioning module, and a result output module.

[0048] Among them, the noise modeling module collects a frame of noise data under the condition of no laser signal, calculates the noise mean and variance, and establishes an environmental noise model.

[0049] The matched filtering module performs a sliding dot product calculation on the acquired laser signal according to the filtering template, thereby filtering the signal waveform.

[0050] The dynamic threshold detection module is based on an environmental noise model and sets the detection threshold to [value missing]. ,in Let k be the noise standard deviation, and k be the sensitivity coefficient (usually taken as 3.0, corresponding to 3). in principle);

[0051] The pulse positioning module searches for peak values ​​exceeding a threshold within the sampling points set according to the laser pulse width characteristics on the envelope signal output by the matched filter, and returns the precise position.

[0052] The hardware platform deployment in this embodiment can be implemented on the Zynq-7000 series embedded platform, and the specific hardware configuration is as follows:

[0053] The main control chip is a Xilinx Zynq-7020 (dual-core Cortex-A9 @ 667MHz).

[0054] ADC sampling module: 14-bit precision, 250MSPS sampling rate, supports hardware-triggered synchronization.

[0055] Memory configuration: 512MB DDR3, with 2KB allocated for core algorithms.

[0056] The system in this embodiment adopts a bare-metal real-time architecture to avoid operating system scheduling delays. The software structure is as follows:

[0057] 1. Pulse detector configuration structure

[0058] This structure is the core data container of the entire laser pulse detection system, used to store all key parameters and status information, as detailed below:

[0059] Noise mean: The average value of the ambient noise, calculated after collecting pure noise samples during the noise calibration phase, and used for baseline correction of subsequent signals;

[0060] Noise variance: The variance of the noise in the storage environment. This parameter is crucial to the system's adaptability and directly determines the setting level of the dynamic threshold.

[0061] Template coefficients: Store the coefficients of the matched filter, which have been fixed-point processed;

[0062] Template length: The actual length of the filter template used is dynamically adjusted according to different laser pulse widths and sampling rates;

[0063] Dynamic threshold: Stores a detection threshold calculated based on the 3σ principle. This threshold is automatically adjusted according to changes in environmental noise to ensure stable detection performance under different working conditions.

[0064] 2. System state machine enumeration, which defines four key states of system operation:

[0065] Idle state: The standby state of the system in its initial state or after a task is completed. In this state, the system maintains low power consumption and waits for external trigger commands or timer events to wake it up.

[0066] Noise calibration status: The state the system enters upon initial deployment or restart; in this state, the laser source is disabled, the system collects pure ambient noise samples, calculates and updates the noise mean and noise variance parameters, and establishes an initial noise model;

[0067] Real-time detection status: The system is in normal working condition. In this state, the system synchronously triggers laser emission and signal acquisition, executes core algorithm processes such as baseline correction, matched filtering, dynamic threshold detection and pulse positioning, and outputs detection results.

[0068] The entire system in this embodiment is uniformly scheduled through the control and coordination module, and the modules adopt a zero-copy memory sharing mechanism to avoid redundant data transmission; the system supports a periodic noise recalibration mechanism to automatically adapt to temperature changes and device aging.

[0069] The implementation steps of the adaptive laser pulse detection method in this embodiment are as follows:

[0070] S1. Noise samples are collected by the ADC sampling module under conditions without laser irradiation, and then the noise mean and variance are calculated by the noise modeling module to establish an adaptive environmental noise model.

[0071] Specifically, noise data acquisition involves continuously acquiring noise samples by triggering the ADC sampling module through hardware, while disabling the laser emitter and ensuring no external light interference. The data is then directly stored in a pre-allocated buffer using a DMA transfer mechanism.

[0072] Noise statistical modeling is used to construct statistical models of environmental noise under conditions without laser irradiation; the function receives raw noise sample data and calculates it using the following formula:

[0073]

[0074] Calculate and update the noise mean and variance parameters in the global configuration.

[0075] S2, the ADC sampling module synchronously triggers laser emission and data acquisition, the data preprocessing module uses the modeled noise mean to perform baseline correction on the input signal, and the matched filter module based on the raised cosine function extracts the signal envelope;

[0076] Specifically, the preparation process of the filter template is as follows:

[0077] (1) Determine the template length point based on the laser pulse width and sampling rate;

[0078] (2) For each index position k ( Perform time coordinate mapping; calculate the time offset relative to the template center. This operation converts the index space into a time coordinate system with the template center as the origin; for example, when template_len=64, k=32 corresponds to t=0 (center point).

[0079] The raised cosine function is calculated as follows:

[0080] Verify that the time offset is within the valid range:

[0081] Calculate the raised cosine value:

[0082] (3) Template normalization and fixed-point stage

[0083] Set the fixed-point scaling factor and perform the following for each template coefficient:

[0084] Normalization process: This ensures that the sum of all template coefficients is 1.0, maintains signal energy conservation, and avoids amplitude distortion after filtering.

[0085] Specifically, the implementation steps of the real-time detection phase of the data preprocessing module are as follows:

[0086] (1) Data preprocessing

[0087] As the first step in the real-time detection process, this function performs baseline correction on the raw acquired signal, providing high-quality input data for subsequent matched filtering;

[0088] Preprocessing parameters: raw_signal: contains the raw signal data acquired by the ADC, including laser pulses, ambient noise, and system DC offset; centered_signal: used to store the signal result after baseline correction; length: indicates the number of sampling points of the signal to be processed;

[0089] The preprocessing process is as follows:

[0090] The function performs a very concise but crucial operation:

[0091] Point-by-point correction: Traverse each sampling point of the input signal (0 ≤ i < length).

[0092] Baseline elimination: Perform a subtraction operation on each sample point:

[0093]

[0094] The config.noise_mean value is the average environmental noise value pre-calculated and stored during the noise calibration phase. This value represents the inherent offset of the system in the absence of a laser signal and the DC component of the environmental noise.

[0095] (2) Implementation of efficient matched filtering

[0096] A lookup table method is used to accelerate cosine calculation, a circular buffer is used to avoid data copying, and SIMD instructions are used to process the four data points in parallel.

[0097] Core processing flow

[0098] 1. Pointwise convolution calculation

[0099] Outer loop: Iterate through each position n of the output signal ( );

[0100] 2. Template sliding and alignment

[0101] Inner loop: Iterate through each coefficient position m of the matched filter template ( )

[0102] Intelligent index calculation:

[0103]

[0104] This calculation ensures the center point of the template ( Always precisely aligned with the current output point n;

[0105] when At that time, visit the right side (future point) of signal point n.

[0106] when At that time, visit the left side of signal point n (historical point).

[0107] This type of index protects the integrity of the first pulse segment (the nearest target echo);

[0108] 3. Border Protection Mechanism

[0109] Safety boundary check: Ensures system stability by automatically adjusting the effective calculation range at the start and end of the signal region to avoid distortion and spurious pulses introduced by traditional filters at the boundary.

[0110] 4. Multiplication, accumulation, and addition operations

[0111] Energy convergence: The accumulator gathers the similarity measure between the template and the signal at the current position. When the signal waveform is highly matched with the template, the accumulated value reaches the peak value. Noise components are effectively suppressed due to randomness, thereby improving the signal-to-noise ratio.

[0112] S3, the dynamic threshold detection module is based on an adaptive environmental noise model. It sets a dynamic detection threshold (3σ principle) and searches for valid pulse peaks only within the set sampling points. The specific operation is as follows:

[0113] Set the detection threshold to ,in Let k be the noise standard deviation, and k be the sensitivity coefficient (usually taken as 3.0, corresponding to 3). in principle);

[0114] The core processing flow is as follows:

[0115] (1) Dynamic threshold adaptive setting

[0116] Achieving the 3σ principle: The detection threshold is dynamically calculated based on a pre-established noise model. The 3σ principle statistically guarantees that approximately 99.7% of pure noise samples will not exceed this threshold. This mechanism enables the system to automatically adapt to changes in environmental noise (such as temperature drift and power fluctuations).

[0117] (2) Peak search within a restricted window

[0118] Search scope limitation: The search is strictly limited to a physically reasonable window, significantly reducing computational load. This window is pre-calculated based on laser emission time, light speed, and the system's maximum detection distance. For example, for a system with a maximum detection distance of 100m, the search window only covers peak sampling within approximately 667ns.

[0119] (3) Secondary verification of pulse morphology

[0120] Boundary safety check: Ensures that candidate peaks have complete left and right neighborhoods to avoid boundary effects;

[0121] Local morphological verification:

[0122] Verify whether the peak point is significantly higher than the adjacent points (1.2 times relationship). This condition ensures that the detected signal is a sharp pulse rather than a smooth noise fluctuation. The 1.2 times factor has been optimized through a large number of experiments to balance detection sensitivity and anti-false alarm capability, and effectively filter out false signals caused by electromagnetic interference, ADC glitches, etc.

[0123] S4. The pulse positioning module searches for peak values ​​within a physically reasonable window of the set sampling points. That is, on the envelope signal output by the matched filter, it searches for peak values ​​exceeding the threshold within the set sampling points based on the characteristics of the laser pulse width, and returns the precise position.

[0124] A valid pulse is one that passes all verification conditions and returns the precise peak position, peak_pos.

[0125] Invalid result: If any verification fails, -1 is returned to indicate that no valid pulse was detected;

[0126] Result determinism: No probability output, provides a clear binary decision result, which facilitates decision-making by the upper-level system;

[0127] For a 300ns laser pulse and a 250MHz sampling rate scenario: template length M = 40 points (corresponding to a 160ns pulse width);

[0128] Search window: starting 500ns after laser emission and lasting for 10μs, with sensitivity coefficient k = 3.0 (3σ principle);

[0129] Actual performance measured on the Zynq-7020 platform: Single frame processing time: 2μs (far lower than the 10ms frame interval);

[0130] Resource usage: RAM 1.8KB, BRAM 2KB;

[0131] The pulse was successfully detected under low signal-to-noise ratio (-5dB) conditions, such as Figure 3 As shown.

[0132] Based on the above detection method, the detection pulse under low signal-to-noise ratio (0dB) conditions can be obtained, such as... Figure 4 As shown; the detection pulse under low signal-to-noise ratio (5dB) conditions, such as Figure 5 As shown; the detection pulse under low signal-to-noise ratio (10dB) conditions, such as Figure 6 As shown.

[0133] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for adaptive laser pulse detection of a laser seeker, characterized in that, The modules adopted include an ADC sampling module, a noise modeling module, a matched filtering module, a data preprocessing module, a dynamic threshold detection module, and a pulse positioning module. The specific implementation steps of the detection method are as follows: S1. Collect noise samples through the ADC sampling module under the condition of no laser irradiation, and then calculate the noise mean and variance through the noise modeling module to establish an ambient noise adaptive model; S2. The ADC sampling module synchronously triggers laser emission and data acquisition. The data preprocessing module uses the modeled noise mean to perform baseline correction on the input signal, and the matched filtering module constructed based on the cosine roll-off function extracts the signal envelope; S3. The dynamic threshold detection module sets a dynamic detection threshold based on the ambient noise adaptive model and adopts the 3σ principle to search for valid pulse peaks only within the set sampling points. The specific operations are as follows: The detection threshold is set as wherein is the noise standard deviation, k is the sensitivity coefficient; S4. The pulse positioning module searches for peaks within the physically reasonable window of the set sampling points, that is, on the envelope signal output by the matched filtering, and searches for peaks exceeding the threshold within the sampling points according to the laser pulse width characteristics to return the accurate position.

2. The method according to claim 1, wherein, The establishment of the environmental noise adaptive model in step S1 involves acquiring a frame of noise data of length N using a noise modeling module under conditions of no laser irradiation. Through the calculation formula: Calculate the noise mean and variance An adaptive model for environmental noise was established.

3. The method according to claim 2, wherein, The implementation steps of the baseline correction and signal envelope extraction described in step S2 are as follows: S2.

1. Prepare a filter template S2.1.1 Set the basic parameters of the template, where the template length sampling point is set to M; the center position is set to... That is, the center point of symmetry of the template; the time constant is set to The sampling frequency determines the width of the raised cosine function; S2.1.

2. Calculate the template coefficients. For each index k, where the value range of k is 0 ≤ k ≤ M, perform the following calculation steps: Computing the time offset: The index k is mapped to a time coordinate system with the center as the origin, so that t varies in the interval . Condition judgment and function calculation: The template length just covers All points satisfy The condition is that each point is calculated using the raised cosine function; this function is... It reaches a maximum value of 1.0 at that point. When the value is 0, a perfect bell-shaped curve is formed; Normalize the template coefficients as follows: Sum all M coefficients and then divide each coefficient by S; S2.2, obtaining a normalization template After the baseline correction of the signal The matched filtering is performed, implementing the following steps: Perform sliding dot product calculation. For each output point n, where the value range of n is 0 ≤ n < M, calculate: For a fixed n, let the template index m slide from 0 to M over the signal index as , ensuring that: When the corresponding signal point n is the template center aligning the current output point; when When accessing the right side of signal point n ; When the access signal point n left side ; The above indexing arrangement keeps the template center always aligned with the current computation point n.

Images