Real-time neuromorphic vision vibration measurement method and system based on laser enhancement

Abstract

The present application relates to the technical field of machine structure component vibration measurement, and particularly relates to a real-time neuromorphic visual vibration measurement method and system based on laser enhancement, comprising the following steps: scanning the surface of a machine structure component to be measured by using a line laser, shooting the line laser scanning position of the machine to be measured by using an event camera, and obtaining event stream information; performing spatial hash indexing, filtering and polarity weighted coding on the event stream information, and obtaining discrete time signals of active pixels; performing batch parallel Fourier transform, peak frequency detection and histogram analysis, and obtaining candidate frequencies; performing spatial continuity verification, removing active pixels with isolated spatial positions, and obtaining effective vibration pixel regions; and performing aggregation and multi-layer pixel-level mapping on the discrete time signals of the active pixels, and obtaining vibration measurement results; the present application can improve the vibration measurement data density, and reduce the acquisition time and the calculation complexity.

Description

Technical Field

[0001] This invention relates to the field of vibration measurement technology for machine structural components, and specifically to a laser-enhanced real-time neuromorphic visual vibration measurement method and system. Background Technology

[0002] During the operation of mechanical equipment, the vibration state of structural components such as frames, shells, support beams, and connectors directly reflects the equipment's operational stability, assembly status, and potential failure risks. Therefore, vibration testing and monitoring of machine structural components are of great significance.

[0003] Existing vibration measurement technologies mainly fall into two categories: contact sensor measurement and non-contact measurement. Contact methods typically require fixing sensors such as accelerometers to the surface of the structure being measured. While this can acquire local vibration signals, sensor installation can affect the structural dynamics and makes it difficult to simultaneously measure large-area structural surfaces. Non-contact methods, while avoiding added mass and installation interference, often require point-by-point acquisition of vibration signals from the structural surface, making it difficult to quickly obtain the overall vibration distribution of machine structural components, resulting in limited measurement efficiency and real-time performance. On the other hand, existing auxiliary imaging or vision methods for structural vibration measurement, when inspecting complex mechanical components, typically only enhance local feature areas, failing to cover the entire structural surface. This leads to incomplete vibration information acquisition, affecting the reconstruction and analysis of the entire field vibration modes.

[0004] The aforementioned problems mean that existing technologies still suffer from limitations in full-field, high-precision, real-time vibration testing of machine structural components, including limited measurement range, low acquisition efficiency, and insufficient result reconstruction. Summary of the Invention

[0005] In view of the above problems, the present invention provides a laser-enhanced real-time neuromorphic visual vibration measurement method and system, which solves the technical problems of sparse vibration measurement data, long acquisition time and high computational complexity in the prior art.

[0006] On one hand, the present invention provides a real-time neuromorphic visual vibration measurement method based on laser enhancement, comprising the following steps: Step S1: Use a line laser to scan the surface of the structural component of the machine under test, and use an event camera to capture the line laser scan position of the machine under test to obtain event flow information; Step S2: Perform spatial hash indexing and filtering on the event stream information to obtain the active pixel event stream; perform polarity weighted encoding on the active pixel event stream based on the polarity of the event to obtain the discrete time signal of the active pixel; Step S3: Perform batch parallel Fourier transform and peak frequency detection on the discrete-time signals of active pixels, and perform histogram analysis on the peak frequencies of each active pixel to obtain multiple candidate frequencies. Step S4: Based on the pixel position and peak frequency of each active pixel, perform spatial coherence verification, remove active pixels with isolated spatial positions, and obtain the effective vibration pixel region; Step S5: Based on the multiple candidate frequencies and the effective vibration pixel region, aggregate and perform multi-layer pixel-level mapping on the discrete-time signals of active pixels to obtain vibration measurement results. The vibration measurement results include the vibration frequency and amplitude of all active pixels within the line laser scanning position.

[0007] Preferably, step S1 specifically includes: Step S1-1: Set the laser line to scan the surface of the machine structural component to be tested at a preset position at a preset speed; Step S1-2: Use an event camera to capture the line laser scanning position of the machine under test and obtain the asynchronous event stream: ;in, It is an asynchronous event stream. Indicates the first One event, Indicates the total number of events. Indicates the first The pixel coordinates of each event. For the first The timestamp of each event For the first The polarity of an event.

[0008] Preferably, step S2 specifically includes: Step S2-1: Using the spatial hash indexing method, a spatial hash table is constructed based on the pixel coordinates of each event in the asynchronous event stream. The spatial hash table includes the event set of each pixel in the image. Step S2-2: Filter the event set of each pixel in the grid coordinates of the spatial hash table by event trigger rate, including: calculating the average event trigger rate of all pixels in the image, removing pixels that exceed the average event trigger rate from the spatial hash table, and taking the remaining pixels as the active pixels; Steps 2-3: Construct discrete-time signals for each active pixel based on the sampling rate and the polarity of the event.

[0009] Preferably, in steps 2-3, the expression for the discrete-time signal of the active pixel is:

[0010] in, Indicates active pixels In the index brightness amplitude, Represents a discrete-time index. Indicates active pixels The corresponding event set, Indicates an indicator function, Indicates the start time. Indicates the sampling interval.

[0011] Preferably, step S3 specifically includes: Step S3-1: Determine the batch size, group all active pixels evenly according to the batch size to form multiple batches, perform Fast Fourier Transform on the discrete-time signal of each active pixel in each batch, perform peak detection on the obtained frequency domain signal, and obtain the peak frequency of each active pixel. Step S3-2: Determine the power spectral density of each active pixel, filter active pixels based on the determination results, and obtain effective active pixels; Step S3-3: Construct a frequency distribution histogram based on the peak frequency of the effective active pixels, perform numerical judgment on the frequency distribution histogram, and obtain multiple candidate frequencies.

[0012] Preferably, in step S3-1, the expression for the peak frequency of each active pixel is:

[0013] in, Indicates active pixels peak frequency, Represents frequency exist Find the maximum value within the frequency range. These are the minimum and maximum boundaries of the frequency range, respectively. Indicates active pixels The Fourier transform of a discrete-time signal at frequency The value at that location, Indicates the magnitude of a complex number; In step S3-2, the expression for determining the power spectral density of each active pixel is:

[0014] in, Represents pixels The power spectral density, Represents pixels The average power, where γ is the signal-to-noise ratio threshold; In step S3-3, the expression for the frequency distribution histogram is:

[0015] in, Represents frequency Histogram values ​​at the location, Expressed as frequency resolution; In step S3-3, the expression for numerically judging the frequency distribution histogram is:

[0016] in, This is the voting threshold coefficient. This represents the maximum value of the histogram. Indicates candidate frequency, This represents the histogram value at the candidate frequency.

[0017] Preferably, step S4 specifically includes: For each active pixel, the peak frequency of the active pixel itself and the peak frequencies of the eight spatially adjacent pixels, including those in the top, bottom, left, right and four diagonal directions, are obtained. The consistency of the peak frequency is evaluated by judging whether the peak frequency of the active pixel and its adjacent pixels meets the preset frequency tolerance. Pixels with the same peak frequency are classified into the same connected component. For all connected components, discard connected components with an area smaller than a preset threshold, and use the remaining connected components as valid vibration pixel regions. For the remaining connected components, a disjoint-set data structure is used to merge connected components with similar frequencies, ultimately obtaining the effective vibration pixel region and its corresponding reference frequency.

[0018] Preferably, step S5 specifically includes: Step S5-1: Aggregate the brightness amplitude of the effective vibration pixel region to obtain the aggregated region signal; Step S5-2: Perform local high-resolution spectrum analysis on the signal in the aggregation region based on the multiple candidate frequencies to obtain preliminary refined frequencies; Step S5-3: Based on the initial refined frequency, perform frequency extraction in three levels to gradually narrow the range, obtain the vibration frequency of all active pixels in the entire image, and obtain the amplitude of all active pixels.

[0019] Preferably, in step S5-1, the expression for the aggregation region signal is:

[0020] in, Indicates the effective vibration pixel area The aggregation region signal; In step S5-2, the expression for the preliminary refining frequency is:

[0021] in, Indicates the initial refining frequency. Indicates frequency exist Find the maximum value within the frequency range. For local range parameters, This represents the total number of discrete time points of the signal in the aggregation region. Represents the imaginary unit. Indicates the sampling frequency; In step S5-3, the step of performing a three-level progressive narrowing of the frequency range based on the initial refined frequency specifically includes: Use the initial refining frequency as the center frequency of the current level of refining, and repeat the frequency extraction process three times to narrow down the range. The expression for the frequency extraction of the narrowed range is:

[0022] in, This indicates the frequency after narrowing down the current level. Indicates the center frequency of the current level of refinement. This indicates the width of the local range for the current level of refinement; The expression for the amplitude of all active pixels is:

[0023] in, Represents pixels The amplitude of vibration.

[0024] On one hand, the present invention provides a laser-enhanced real-time neuromorphic visual vibration measurement system, characterized in that it comprises: The event stream acquisition module is used to scan the surface of the structural components of the machine under test using a line laser, and to capture the line laser scan position of the machine under test using an event camera to obtain event stream information. The discrete-time signal extraction module is used to perform spatial hash indexing and filtering on the event stream information to obtain the active pixel event stream; and to perform polarity-weighted encoding on the active pixel event stream based on the polarity of the event to obtain the discrete-time signal of the active pixel. The frequency domain analysis module is used to perform batch parallel Fourier transform and peak frequency detection on the discrete-time signals of active pixels, and to perform histogram analysis on the peak frequencies of each active pixel to obtain multiple candidate frequencies. The spatial coherence checking module is used to perform spatial coherence verification based on the pixel position and peak frequency of each active pixel, remove active pixels with isolated spatial positions, and obtain effective vibration pixel regions. A multi-level refining module is used to aggregate and perform multi-level pixel-level mapping on the discrete-time signals of active pixels based on the multiple candidate frequencies and the effective vibration pixel region to obtain vibration measurement results, which include the vibration frequency and amplitude of all active pixels within the line laser scanning position.

[0025] Compared with the prior art, the present invention has at least the following beneficial effects: (1) This invention uses a midline laser to scan the surface of the machine structural component under test and uses an event camera to simultaneously acquire event flow information at the line laser scanning position. This transforms the original event imaging method, which relied on natural edges or weak textures, into an event generation method actively enhanced by the laser, thereby significantly improving the event response density and spatial coverage of the vibration area. It breaks through the limitations of traditional non-enhanced neuromorphic methods, which suffer from sparse events and difficulty in covering the entire field, and achieves a more complete reconstruction of the vibration mode within the entire line laser scanning position.

[0026] (2) This invention fully utilizes the asynchronous and sparse characteristics of event streams in the processing link. Through steps such as spatial hash indexing and filtering, polarity weighted encoding, batch parallel Fourier transform and peak frequency detection, it achieves efficient processing of massive event data. At the same time, by combining histogram analysis and spatial coherence verification, the candidate frequencies and effective vibration pixel regions are refined layer by layer, achieving microsecond-level time resolution and high-precision frequency estimation while maintaining low computational complexity.

[0027] (3) Compared with the traditional point scanning LDV, the present invention does not require repeated measurement point by point and line by line, which significantly improves the detection efficiency. It can realize real-time full-field vibration measurement and the measurement speed can be increased by more than 50 times compared with the traditional point scanning method. It is especially suitable for micron-level vibration monitoring, equipment health status assessment and online detection scenarios of high-speed mechanical structures. Attached Figure Description

[0028] The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of the invention.

[0029] Figure 1 The flowchart shows the laser-enhanced real-time neuromorphic visual vibration measurement method provided by this invention.

[0030] Figure 2 The flowchart illustrates the module of the laser-enhanced real-time neuromorphic visual vibration measurement method provided by this invention.

[0031] Figure 3 A schematic diagram of the laser-enhanced real-time neuromorphic visual vibration measurement experimental device provided by the present invention.

[0032] Figure 4A physical image of the laser-enhanced real-time neuromorphic visual vibration measurement experimental device provided by the present invention.

[0033] Figure 5 This is a schematic diagram of the object to be tested A and the input signal provided by the present invention.

[0034] Figure 6 This is a schematic diagram of the vibration frequency reconstruction result of the test object A provided by the present invention.

[0035] Figure 7 This is a schematic diagram of the spectrum of the object A to be tested provided by the present invention.

[0036] Figure 8 This is a schematic diagram of the object combination B and the input signal provided by the present invention.

[0037] Figure 9 This is a schematic diagram of the vibration frequency reconstruction results of the test object combination B provided by the present invention.

[0038] Figure 10 This is a schematic diagram of the spectrum of the object combination B provided by the present invention. Detailed Implementation

[0039] To better understand the above-described objectives, features, and advantages of the present invention, the 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 of the present invention and the features thereof can be combined with each other. Furthermore, the present invention can be implemented in other ways different from those described herein; therefore, the scope of protection of the present invention is not limited to the specific embodiments disclosed below.

[0040] To illustrate the effectiveness of the method proposed in this invention, the following detailed description of the above technical solution is provided through a specific embodiment, such as... Figure 2 As shown, the laser scanning module first emits a laser to illuminate the object under test. The reflected light from the surface of the object enters the event camera. After the event camera acquires the event stream, it undergoes subsequent processing, including: first, spatial hash indexing and filtering are performed by the spatial hash indexing module; then, polarity weighted encoding is performed on the events by the polarity weighted encoding module; subsequently, frequency domain analysis is performed by the parallel FFT analysis module, and candidate frequencies are extracted by the spectral voting module; next, the spatial coherence verification module is used to remove active pixels with isolated spatial locations; finally, the hierarchical Zoom-DFT refining module further refines the data and outputs the final result.

[0041] like Figure 1 This invention discloses a real-time neuromorphic visual vibration measurement method based on laser enhancement, the specific implementation steps of which are as follows: Step S1: Use a line laser to scan the surface of the structural component of the machine under test, and use an event camera to capture the line laser scan position of the machine under test to obtain event flow information; The device for visual vibration measurement according to the present invention is as follows: Figure 3 , Figure 4 As shown, it includes a line laser, an event camera, and a computer. The computer is used to control the line laser and the event camera, and it is able to receive and process event stream data from the event camera.

[0042] In this step, the present invention uses a line laser to scan the surface of the object being tested. The laser emitted by the line laser forms a thin laser line on the surface of the object being tested, and the laser line scans the entire surface area to be tested along a predetermined trajectory.

[0043] In some embodiments, a laser line can be set to sweep across a preset position on the surface of the machine structural component under test at a preset rate.

[0044] When a minute vibration occurs on the surface of the object under test, the vibration-induced displacement modulates the intensity of the reflected light. The vibration displacement of the object's surface causes rapid changes in the incident angle and reflection intensity of the laser reflected light, thus artificially creating a rich array of light intensity variation signals on the object's surface. This laser enhancement mechanism breaks through the limitations of traditional visual methods that rely solely on the object's own texture or edge features, achieving full-field event coverage of textureless, uniform surfaces.

[0045] An event camera is used to capture the line laser scan position of the machine under test. The event camera captures the asynchronous event stream generated by vibration-induced light intensity changes with microsecond-level time resolution. Each pixel of the event camera operates independently; when a light intensity change exceeds a preset threshold, each pixel immediately outputs an event asynchronously. The asynchronous event stream output by the event camera is represented as follows:

[0046] in, It is an asynchronous event stream. Indicates the first One event, Indicates the total number of events. Indicates the first The pixel coordinates of each event represent the spatial location of the event in the image. , They represent the first The x and y coordinates of each event, For the first The timestamp of each event For the first The polarity of an event is 1 when the light intensity increases and -1 when the light intensity decreases.

[0047] The rate at which events are triggered corresponds directly to the vibration frequency of the object being tested; the higher the vibration frequency, the more events are generated per unit time.

[0048] Through the above steps, this invention uses line laser scanning to scan the surface of the structural components of the machine under test, and uses an event camera to capture the line laser scanning position of the machine under test to obtain event flow information for subsequent data analysis.

[0049] Step S2: Perform spatial hash indexing and filtering on the event stream information to obtain the active pixel event stream; perform polarity weighted encoding on the active pixel event stream based on the polarity of the event to obtain the discrete time signal of the active pixel; In this step, the present invention first performs spatial hash indexing on the event stream information. Specifically, the present invention can use existing spatial hash indexing methods to construct a spatial hash table in real time, mapping the massive asynchronous events in the event stream to the corresponding spatial grid according to pixel coordinates.

[0050] The mapping expression for the spatial hash table is:

[0051] in, Represents pixels in a spatial hash table Content, Represents pixels The set of all events for a given location.

[0052] By establishing the above spatial hash index, the system can quickly retrieve the event sequence of any pixel in O(1) time complexity, where O(1) represents constant time complexity.

[0053] Next, the event set of each pixel in the grid coordinates of the spatial hash table is filtered by event trigger rate. Specifically, the average event trigger rate of the global space is calculated. The event trigger rate of each pixel is calculated individually. If the event trigger rate of a pixel exceeds the average event trigger rate, the pixel is determined to be a hot pixel caused by a hardware defect and is removed from the spatial hash table. The remaining pixels are then considered as active pixels. After hot pixel filtering, the set of active pixels that truly reflect vibration information and their corresponding event stream are obtained.

[0054] Next, the present invention uses polarity-weighted encoding to convert asynchronous and irregular event streams into discrete-time signals with equal time intervals.

[0055] For active pixels Based on sampling rate The constructed discrete-time signal expression is as follows:

[0056] in, Indicates active pixels In the index brightness amplitude, Represents a discrete-time index. Indicates active pixels The corresponding event set, For the first The polarity of an event, This indicates an indicator function that takes the value 1 when the condition inside the parentheses is true, and 0 otherwise. Indicates the first The timestamp of each event Indicates the start time. Indicates the sampling interval. .

[0057] Using the above expression, for the nth discrete-time index, the polarities of all events falling within that time interval are weighted and accumulated. If the accumulated light intensity within that time slot is positive, the signal value is positive; if the accumulated intensity is negative, the signal value is negative.

[0058] Through the above polarity weighted encoding, the present invention successfully converts the originally chaotic asynchronous event stream into a discrete-time signal with uniform time intervals, suitable for digital signal processing.

[0059] Step S3: Perform batch parallel Fourier transform and peak frequency detection on the discrete-time signals of active pixels, and perform histogram analysis on the peak frequencies of each active pixel to obtain multiple candidate frequencies. In this step, the discrete-time signals of active pixels are first subjected to batch parallel Fourier transform.

[0060] In some embodiments, a batch size B can be set, and all active pixels can be evenly grouped according to the batch size B to form multiple batches. Multi-core parallel computing can be used for these multiple batches, performing a Fast Fourier Transform (FFT) on the discrete-time signal of each active pixel within each batch to obtain the frequency domain signal. The peak value within a preset frequency range in the frequency domain signal is detected, ultimately obtaining the peak frequency of each active pixel. The expression for the above steps is:

[0061] in, Indicates active pixels peak frequency, Represents frequency exist Find the maximum value within the frequency range. These represent the minimum and maximum boundaries of the frequency range, respectively. Indicates active pixels The Fourier transform of a discrete-time signal at frequency The value at that location, This indicates the magnitude of a complex number.

[0062] In some embodiments, the signal-to-noise ratio (SNR) filtering criterion of the present invention is as follows: For each active pixel, the corresponding power spectral density is calculated. and average power Active pixels are identified as valid active pixels only when the power spectral density is significantly higher than the average power.

[0063] Where γ is the signal-to-noise ratio (SNR) threshold. Pixels that do not meet the SNR filtering criterion are considered noise, and their peak frequencies are discarded. The above formula means calculating the average power of the spectrum. Next, determine whether the above inequality conditions are satisfied. If satisfied, retain the peak frequency of this pixel. The pixel is then processed further; if the conditions are not met, the pixel is considered invalid and will not be included in subsequent processing.

[0064] Subsequently, the present invention constructs a frequency distribution histogram for the peak frequencies of all effective active pixels. The expression for the histogram construction is as follows:

[0065] in, Represents frequency Histogram values ​​at the location, Indicates frequency resolution.

[0066] This histogram reflects the frequency distribution characteristics of all pixels.

[0067] The frequency distribution histogram is then numerically analyzed to obtain a candidate frequency set, which contains multiple candidate frequencies. Candidate frequency The following expression must be satisfied:

[0068] in, This is the voting threshold coefficient. This represents the maximum value of the histogram.

[0069] The candidate frequencies obtained through the steps of this invention represent the main vibration modes that may exist on the object under test. Histogram analysis effectively suppresses the influence of discrete noise pixels, retaining only frequency components that have sufficient spatial support.

[0070] Step S4: Based on the pixel position and peak frequency of each active pixel, perform spatial coherence verification, remove active pixels with isolated spatial positions, and obtain the effective vibration pixel region; In this step, the present invention is verified based on the spatial continuity of physical vibration. Real physical vibration sources typically form continuous vibration regions in space, while spurious events caused by random noise or hardware errors exhibit isolated distributions in space.

[0071] This invention employs an 8-connected component analysis algorithm to spatially group pixels with similar frequencies, as detailed below.

[0072] For each active pixel, the peak frequency of the active pixel itself and the peak frequencies of its eight spatially adjacent pixels, including those in the top, bottom, left, right and four diagonal directions, are obtained. The consistency of the peak frequencies is evaluated by judging whether the peak frequencies of the active pixel and its adjacent pixels meet the preset frequency tolerance. Pixels with the same peak frequencies are classified into the same connected component.

[0073] For all connected components, connected components with an area smaller than a preset threshold are identified as spatially isolated noise clusters and removed. The remaining connected components are considered as effective vibration pixel regions.

[0074] Furthermore, for the remaining connected components, a union-find data structure is used to merge connected components with similar frequencies, ultimately obtaining the effective vibration pixel region and its corresponding reference frequency.

[0075] After spatial coherence verification, the present invention obtained several effective vibration pixel regions and their corresponding reference frequencies, which represent the distribution of vibration modes that actually exist on the object under test.

[0076] Step S5: Based on the multiple candidate frequencies and the effective vibration pixel region, aggregate and perform multi-layer pixel-level mapping on the discrete-time signals of active pixels to obtain vibration measurement results. The vibration measurement results include the vibration frequency and amplitude of all active pixels within the line laser scanning position.

[0077] In this step, the present invention employs a multi-level frequency refinement strategy to achieve frequency estimation with sub-resolution accuracy.

[0078] First, the brightness amplitude of the effective vibration pixel region is aggregated to obtain the aggregated region signal, expressed as:

[0079] in, Indicates the effective vibration pixel area The aggregated region signal.

[0080] Subsequently, based on the multiple candidate frequencies, local high-resolution spectral analysis is performed on the signal in the aggregation region to obtain preliminary refined frequencies. The expression is:

[0081] in, Indicates the initial refining frequency. Indicates frequency exist Find the maximum value within the frequency range. For local range parameters, This represents the total number of discrete time points of the signal in the aggregation region. Represents the imaginary unit. Represents a discrete-time index. Indicates the sampling frequency.

[0082] To optimize computational efficiency while maintaining accuracy, the frequency was initially refined. Based on this, the present invention adopts a three-level progressively narrowing refinement strategy to obtain the vibration frequency of all active pixels in the entire image.

[0083] Specifically, the first level of refining is... Search within the range, where This represents the first-level local range parameter; the second-level refinement is centered on the first-level result, within a width of... Search within the specified range to obtain the second-level refining results, where This represents the second-level local range parameter; the third-level refinement is centered on the second-level result, within a width of... Search within the range, where This represents the second-level local range parameters. Each level of refinement uses denser frequency sampling points, gradually increasing the frequency resolution. The result after three levels of refinement is the vibration frequency of all active pixels in the entire image.

[0084] As mentioned above, the initial refining frequency is used as the center frequency of the current level of refining, and the frequency extraction is repeated three times to narrow the range. The expression for the frequency extraction of the narrowed range is:

[0085] in, This indicates the frequency after narrowing down the current level. Indicates the center frequency of the current level of refinement. This indicates the width of the local range for the current level of refinement.

[0086] By limiting the search range to a small interval around the global frequency, pixel-level refinement ensures computational efficiency while avoiding frequency estimation bias that may occur when the signal-to-noise ratio of a single pixel is low.

[0087] After frequency refinement, this invention further calculates the vibration amplitude distribution using the following method to achieve high-precision synchronous output of frequency and amplitude. The pixel-level vibration amplitude calculation expression is as follows:

[0088] In the above formula, Represents pixels The amplitude of vibration; The value equals the observation duration. With sampling rate The product is rounded down, i.e. ; The complex exponential term is used to calculate the discrete Fourier transform coefficients at precise frequency points; the coefficients... It is the standard single-sided spectral amplitude recovery coefficient, used to restore the DFT coefficients to the true amplitude of the corresponding sine wave.

[0089] The formula for calculating the regional (global) vibration amplitude is:

[0090] In the above formula, express The overall vibration amplitude of the region.

[0091] Using the two amplitude calculation formulas mentioned above, this invention can directly output a complete vibration mode distribution map after layered refinement, including the frequency-amplitude distribution of each pixel and the overall frequency-amplitude information of each effective vibration object.

[0092] The vibration amplitude reflects the vibration intensity at that pixel location, and combined with frequency information, it can fully describe the local vibration characteristics.

[0093] The final output includes complete vibration measurement results, including: (1) the spatial coordinates and precise frequency of all effective active pixels within the line laser scanning position. Vibration amplitude These data are presented as two-dimensional images with pixel-level resolution, forming a full-field vibration frequency distribution map and vibration amplitude distribution map, realizing comprehensive, high-precision, and real-time measurement of the vibration state of the machine structural components under test.

[0094] Preferably, the specific parameters in the above technical solutions of the present invention can be set as follows: , , , The values ​​in the third-level refinement are as follows: , , , , , Batch size .

[0095] Analysis shows that the computational complexity of the data processing procedure in this invention is... ,in The total number of events, Number of active pixels, This represents the total number of discrete time points of the signal in the aggregation region. For the number of pixels of the object, This represents the number of frequency samples.

[0096] like Figure 5 , Figure 6 , Figure 7 As shown, this diagram displays a single object A under test, the input frequency, the reconstructed vibration frequency, and the spectrum of object A.

[0097] like Figure 8 , Figure 9 , Figure 10 As shown, the spectrum of the test object combination B, the input frequency, the reconstructed vibration frequency, and the spectrum of the test object combination B are displayed.

[0098] This invention provides a laser-enhanced real-time neuromorphic visual vibration measurement system, comprising: The event stream acquisition module is used to scan the surface of the structural components of the machine under test using a line laser, and to capture the line laser scan position of the machine under test using an event camera to obtain event stream information. The discrete-time signal extraction module is used to perform spatial hash indexing and filtering on the event stream information to obtain the active pixel event stream; and to perform polarity-weighted encoding on the active pixel event stream based on the polarity of the event to obtain the discrete-time signal of the active pixel. The frequency domain analysis module is used to perform batch parallel Fourier transform and peak frequency detection on the discrete-time signals of active pixels, and to perform histogram analysis on the peak frequencies of each active pixel to obtain multiple candidate frequencies. The spatial coherence checking module is used to perform spatial coherence verification based on the pixel position and peak frequency of each active pixel, remove active pixels with isolated spatial positions, and obtain effective vibration pixel regions. A multi-level refining module is used to aggregate and perform multi-level pixel-level mapping on the discrete-time signals of active pixels based on the multiple candidate frequencies and the effective vibration pixel region to obtain vibration measurement results, which include the vibration frequency and amplitude of all active pixels within the line laser scanning position.

[0099] Compared to traditional point-scan laser Doppler vibration meters (LDV), this invention can simultaneously acquire vibration information across the entire field of view of a machine structural component's surface during a single scan. This avoids the efficiency limitations of point-by-point measurements of structures such as frames, shells, and plates, significantly improving the speed and coverage of structural vibration testing. Compared to traditional frame-based high-speed camera methods, this invention utilizes the microsecond-level temporal resolution and asynchronous sparse imaging characteristics of an event camera. While reducing data acquisition volume and subsequent computational complexity, it achieves highly sensitive detection and full-field reconstruction of minute vibrations in machine structural components, making it more suitable for real-time vibration testing and analysis of mechanical equipment in operation.

[0100] While the specific embodiments of the present invention depict actions or steps in a particular order, this should be understood as requiring such actions or steps to be performed in the shown specific order or sequential order, or requiring all illustrated actions or steps to be performed to achieve the desired result. In certain environments, multitasking and parallel processing may be advantageous. Similarly, although several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of the invention. Certain features described in the context of individual embodiments may also be implemented in combination in a single implementation. Conversely, various features described in the context of a single implementation may also be implemented individually or in any suitable sub-combination in multiple implementations. The above descriptions are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

[0101] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A real-time neuromorphic visual vibration measurement method based on laser enhancement, characterized in that, Includes the following steps: Step S1: Use a line laser to scan the surface of the structural component of the machine under test, and use an event camera to capture the line laser scan position of the machine under test to obtain event flow information; Step S2: Perform spatial hash indexing and filtering on the event stream information to obtain the active pixel event stream; Polarity-weighted encoding is performed on the active pixel event stream based on the polarity of the event to obtain the discrete-time signal of the active pixel; Step S3: Perform batch parallel Fourier transform and peak frequency detection on the discrete-time signals of active pixels, and perform histogram analysis on the peak frequencies of each active pixel to obtain multiple candidate frequencies. Step S4: Based on the pixel position and peak frequency of each active pixel, perform spatial coherence verification, remove active pixels with isolated spatial positions, and obtain the effective vibration pixel region; Step S5: Based on the multiple candidate frequencies and the effective vibration pixel region, aggregate and perform multi-layer pixel-level mapping on the discrete-time signals of active pixels to obtain vibration measurement results. The vibration measurement results include the vibration frequency and amplitude of all active pixels within the line laser scanning position.

2. The laser-enhanced real-time neuromorphic visual vibration measurement method according to claim 1, characterized in that, Step S1 specifically includes: Step S1-1: Set the laser line to scan the surface of the machine structural component to be tested at a preset position at a preset speed; Step S1-2: Use an event camera to capture the line laser scanning position of the machine under test and obtain the asynchronous event stream: ;in, It is an asynchronous event stream. Indicates the first One event, Indicates the total number of events. Indicates the first The pixel coordinates of each event. For the first The timestamp of each event For the first The polarity of an event.

3. The laser-enhanced real-time neuromorphic visual vibration measurement method according to claim 2, characterized in that, Step S2 specifically includes: Step S2-1: Using the spatial hash indexing method, a spatial hash table is constructed based on the pixel coordinates of each event in the asynchronous event stream. The spatial hash table includes the event set of each pixel in the image. Step S2-2: Filter the event set of each pixel in the grid coordinates of the spatial hash table by event trigger rate, including: calculating the average event trigger rate of all pixels in the image, removing pixels that exceed the average event trigger rate from the spatial hash table, and taking the remaining pixels as the active pixels; Steps 2-3: Construct discrete-time signals for each active pixel based on the sampling rate and the polarity of the event.

4. The laser-enhanced real-time neuromorphic visual vibration measurement method according to claim 3, characterized in that, In steps 2-3, the expression for the discrete-time signal of the active pixel is: in, Indicates active pixels In the index brightness amplitude, Represents a discrete-time index. Indicates active pixels The corresponding event set, Indicates an indicator function, Indicates the start time. Indicates the sampling interval.

5. The real-time neuromorphic visual vibration measurement method based on laser enhancement according to claim 4, characterized in that, Step S3 specifically includes: Step S3-1: Determine the batch size, group all active pixels evenly according to the batch size to form multiple batches, perform Fast Fourier Transform on the discrete-time signal of each active pixel in each batch, perform peak detection on the obtained frequency domain signal, and obtain the peak frequency of each active pixel. Step S3-2: Determine the power spectral density of each active pixel, filter active pixels based on the determination results, and obtain effective active pixels; Step S3-3: Construct a frequency distribution histogram based on the peak frequency of the effective active pixels, perform numerical judgment on the frequency distribution histogram, and obtain multiple candidate frequencies.

6. The laser-enhanced real-time neuromorphic visual vibration measurement method according to claim 5, characterized in that, In step S3-1, the expression for the peak frequency of each active pixel is: in, Indicates active pixels peak frequency, Represents frequency exist Find the maximum value within the frequency range. These represent the minimum and maximum boundaries of the frequency range, respectively. Indicates active pixels The Fourier transform of a discrete-time signal at frequency The value at that location, Indicates the magnitude of a complex number; In step S3-2, the expression for determining the power spectral density of each active pixel is: in, Represents pixels The power spectral density, Represents pixels The average power, where γ is the signal-to-noise ratio threshold; In step S3-3, the expression for the frequency distribution histogram is: in, Represents frequency Histogram values ​​at the location, Indicates frequency resolution; In step S3-3, the expression for numerically judging the frequency distribution histogram is: in, This is the voting threshold coefficient. This represents the maximum value of the histogram. Indicates candidate frequency, This represents the histogram value at the candidate frequency.

7. The laser-enhanced real-time neuromorphic visual vibration measurement method according to claim 6, characterized in that, Step S4 specifically includes: For each active pixel, the peak frequency of the active pixel itself and the peak frequencies of the eight spatially adjacent pixels, including those in the top, bottom, left, right and four diagonal directions, are obtained. The consistency of the peak frequency is evaluated by judging whether the peak frequency of the active pixel and its adjacent pixels meets the preset frequency tolerance. Pixels with the same peak frequency are classified into the same connected component. For all connected components, discard connected components with an area smaller than a preset threshold, and use the remaining connected components as valid vibration pixel regions. For the remaining connected components, a disjoint-set data structure is used to merge connected components with similar frequencies, ultimately obtaining the effective vibration pixel region and its corresponding reference frequency.

8. The real-time neuromorphic visual vibration measurement method based on laser enhancement according to claim 7, characterized in that, Step S5 specifically includes: Step S5-1: Aggregate the brightness amplitude of the effective vibration pixel region to obtain the aggregated region signal; Step S5-2: Perform local high-resolution spectrum analysis on the signal in the aggregation region based on the multiple candidate frequencies to obtain preliminary refined frequencies; Step S5-3: Based on the initial refined frequency, perform frequency extraction in three levels to gradually narrow the range, obtain the vibration frequency of all active pixels in the entire image, and obtain the amplitude of all active pixels.

9. The laser-enhanced real-time neuromorphic visual vibration measurement method according to claim 8, characterized in that, In step S5-1, the expression for the aggregation region signal is: in, Indicates the effective vibration pixel area The aggregation region signal; In step S5-2, the expression for the preliminary refining frequency is: in, Indicates the initial refining frequency. Indicates frequency exist Find the maximum value within the frequency range. For local range parameters, This represents the total number of discrete time points of the signal in the aggregation region. Represents the imaginary unit. Indicates the sampling frequency; In step S5-3, the step of performing a three-level progressive narrowing of the frequency range based on the initial refined frequency specifically includes: Use the initial refining frequency as the center frequency of the current level of refining, and repeat the frequency extraction process three times to narrow down the range. The expression for the frequency extraction of the narrowed range is: in, This indicates the frequency after narrowing down the current level. Indicates the center frequency of the current level of refinement. This indicates the width of the local range for the current level of refinement; The expression for the amplitude of all active pixels is: in, Represents pixels The amplitude of vibration.

10. A real-time neuromorphic visual vibration measurement system based on laser enhancement, characterized in that, include: The event stream acquisition module is used to scan the surface of the structural components of the machine under test using a line laser, and to capture the line laser scan position of the machine under test using an event camera to obtain event stream information. The discrete-time signal extraction module is used to perform spatial hash indexing and filtering on the event stream information to obtain the active pixel event stream; Polarity-weighted encoding is performed on the active pixel event stream based on the polarity of the event to obtain the discrete-time signal of the active pixel; The frequency domain analysis module is used to perform batch parallel Fourier transform and peak frequency detection on the discrete-time signals of active pixels, and to perform histogram analysis on the peak frequencies of each active pixel to obtain multiple candidate frequencies. The spatial coherence checking module is used to perform spatial coherence verification based on the pixel position and peak frequency of each active pixel, remove active pixels with isolated spatial positions, and obtain effective vibration pixel regions. A multi-level refining module is used to aggregate and perform multi-level pixel-level mapping of the pixel discrete-time signals of active images based on the multiple candidate frequencies and the effective vibration pixel regions to obtain vibration measurement results, which include the vibration frequency and amplitude of all active pixels within the line laser scanning position.

Images