A Visual Enhancement-Based Real-Time Monitoring Method and System for Equipment Micro-Vibration

Abstract

This invention discloses a real-time monitoring method and system for equipment micro-vibration based on visual enhancement, belonging to the field of equipment micro-vibration monitoring. The method includes the following steps: Step S1, acquiring high frame rate video and recording acquisition parameters in real time; Step S2, constructing a Gaussian pyramid and then constructing a Laplace pyramid using the image difference between adjacent layers of the Gaussian pyramid to decompose the acquired video into different spatial frequency bands; Step S3, designing a bandpass filter to extract the specific frequency band of the equipment vibration signal and enhancing and reconstructing the micro-vibration signal; Step S4, selecting a region from the enhanced image frame, calculating the optical flow within the selected region using the optical flow method, and extracting the characteristic frequency of the vibration; Step S5, data visualization, generating a vibration signal image in real time to display the motion characteristics of the target region. This invention can solve the problems of insufficient sensitivity and poor real-time performance in traditional vibration monitoring methods for detecting minor equipment vibrations.

Description

Technical Field

[0001] This invention relates to the field of equipment micro-vibration monitoring technology, and in particular to a method and system for real-time monitoring of equipment micro-vibration based on visual enhancement. Background Technology

[0002] Equipment micro-vibration monitoring is a technology used to detect and analyze the subtle vibrations generated by mechanical equipment during operation. This technology is crucial in industry because it helps identify equipment health conditions, predict potential failures, and thus improve equipment reliability and extend its service life. Existing equipment micro-vibration monitoring methods, including accelerometer measurement, laser displacement sensor measurement, and eddy current sensor measurement, all have limitations and shortcomings.

[0003] 1) Accelerometer measurement method

[0004] Accelerometers operate based on the piezoelectric effect or capacitance principle. When equipment vibrates, the sensitive element inside the sensor experiences changes in charge (piezoelectric) or capacitance (capacitive) due to inertial forces. These changes are proportional to the acceleration of the vibration. By measuring these changes in physical quantities, the acceleration information of the equipment vibration can be obtained, thereby detecting whether the vibration frequency of the equipment is normal.

[0005] Disadvantages: Installing the sensor may require drilling, gluing, or using clamps to secure it to the device surface. These operations alter the local mass distribution and structural stiffness of the device, potentially introducing new vibration modes or changing existing vibration characteristics. Due to their piezoelectric or capacitive operating principles, accelerometers have specific environmental requirements. When the application scenario involves numerous electromagnetic devices, these devices generate complex electromagnetic fields. The accelerometer output signal may be affected by electromagnetic interference, which, when superimposed on the vibration signal, ultimately distorts the measurement signal. At high temperatures, the performance of piezoelectric materials may drift, leading to measurement deviations. High humidity environments may cause moisture buildup in the sensor's internal circuitry, affecting its electrical performance. Particularly for accelerometers containing electronic components, humidity may cause short circuits, leakage, or component aging.

[0006] 2) Laser displacement sensor measurement method

[0007] A laser displacement sensor works by emitting a laser beam onto the surface of the device being measured and then receiving the reflected light. The distance between the device surface and the sensor is calculated based on the laser's propagation time or the phase change of the reflected light. This distance changes when the device vibrates, and by continuously measuring these changes, vibration information can be obtained.

[0008] Disadvantages: It places excessive demands on the surface of the object being measured. Different materials exhibit significantly different reflection and absorption characteristics of laser light. Materials with strong light absorption will absorb most of the laser beam energy, resulting in weak reflected light signals and making accurate measurement difficult. Conversely, objects with strong reflection or specular reflection may cause light to return perpendicularly without diffuse reflection, leading to unstable or non-received signals. If the object being measured has deep grooves, complex curved surfaces, or obstructions, the light path will be blocked, preventing some light from being properly reflected to the receiver, thus making measurement impossible or resulting in inaccurate results. Temperature changes can alter the laser's wavelength, output power, and beam quality, thereby affecting measurement accuracy.

[0009] 3) Eddy current sensor measurement method

[0010] Eddy current sensors utilize the principle of electromagnetic induction. When a sensor coil is brought close to a metallic object, it generates an alternating magnetic field. This induces eddy currents on the surface of the object, which in turn generate a magnetic field opposite to the original magnetic field. This alters the electrical parameters of the sensor coil, such as inductance and impedance. These changes are related to the distance between the object and the sensor. Vibration of the equipment causes changes in this distance, allowing the vibration of the equipment to be detected by measuring these electrical parameters.

[0011] Disadvantages: Sensitive to the material properties of the measured object. Eddy current sensors operate based on electromagnetic induction, and the conductivity and permeability of the measured object significantly affect the sensor's sensitivity and measurement accuracy. When the measured object is a magnetic material, eddy current and magnetic effects coexist. The magnetic effect counteracts the eddy current effect, weakening it and thus reducing the sensor's sensitivity. They are also sensitive to temperature changes; the conductivity and permeability of the measured object change with temperature, further affecting the eddy current effect and the sensor's measurement accuracy. Although eddy current sensors have some resistance to electromagnetic interference, in strong electromagnetic field environments, such as near large motors or transformers, the magnetic field may alter the magnetic field distribution generated by the sensor probe, thus affecting the generation and detection of eddy currents. Summary of the Invention

[0012] The purpose of this invention is to provide a method and system for real-time monitoring of equipment micro-vibrations based on visual enhancement, which solves the problems of insufficient sensitivity and poor real-time performance in traditional vibration monitoring methods for detecting small vibrations of equipment.

[0013] To achieve the above objectives, the present invention provides a method for real-time monitoring of device micro-vibrations based on visual enhancement, comprising the following steps:

[0014] Step S1: Acquire high frame rate video, and select key parts of the target device as acquisition areas through manual calibration, and record acquisition parameters in real time, including frame rate, resolution and number of frames.

[0015] Step S2: Construct a Gaussian pyramid to perform layer-by-layer sampling and blurring of the image. Construct a Laplacian pyramid by the difference between adjacent layers of the Gaussian pyramid to decompose the acquired video into different spatial frequency bands.

[0016] Step S3: Design a bandpass filter to extract the specific frequency band of the equipment vibration signal, perform time-domain bandpass filtering on the time-varying signal of each pixel of the Laplace pyramid, and enhance and reconstruct the micro-vibration signal.

[0017] Step S4: Select a region from the enhanced image frame, calculate the optical flow within the selected region using the Farneback optical flow method in OpenCV, and extract the characteristic frequencies of vibration.

[0018] Step S5 involves data visualization, generating real-time vibration signal images to display the motion characteristics of the target area, and monitoring and recording the dynamic changes in vibration signal amplitude and characteristic frequency.

[0019] Preferably, in step S1, the minute vibrations of the target device are captured by a high frame rate industrial camera. The video frame rate acquired by the high frame rate industrial camera is not less than 120fps, and the video resolution is 720p or 1080p.

[0020] Preferably, step S2 includes steps S21 and S22.

[0021] Step S21, construct the Gaussian pyramid:

[0022] Step S211, Gaussian blur:

[0023] Let the input image be I0(x, y), and its resolution be W×H;

[0024] For image I i (x, y) is blurred using a Gaussian kernel to filter out high-frequency information:

[0025] I' i (x, y) = I i (x, y)*G(x, y);

[0026] Where G(x, y) is the standard Gaussian kernel,

[0027]

[0028] x and y represent the positions in the two-dimensional image coordinate system, and σ represents the standard deviation of the Gaussian distribution;

[0029] Step S212, image downsampling:

[0030] The blurred image I′i (x, y) is downsampled by a ratio of 1 / 2 to obtain the next layer image I. i+1 (x, y),

[0031] I i+1 (x, y) = I′ i (2x, 2y);

[0032] During downsampling, the resolution is halved until the predetermined minimum size is reached;

[0033] Repeat the above steps to construct multiple layers of Gaussian pyramids until the target size is reached. The result is:

[0034] Gaussian Pyramid={I0,I1,I2,...,I n};

[0035] I n It is the lowest resolution image.

[0036] Preferably, in step S22, the Laplace pyramid is constructed:

[0037] Step S221, Laplace layer calculation:

[0038] For each Gaussian image I i (x, y) and its lower-level image I i+1 The upsampling result I′ of (x, y) i+1 Find the difference between (x, y) to obtain the Laplace layer L. i (x, y):

[0039] L i (x, y) = I i (x, y) - I′ i+1 (x, y);

[0040] Among them, I′ i+1 (x, y) is I i+1 Upsampled image of (x, y);

[0041] Step S222, preserve the underlying Gaussian image:

[0042] Lowest resolution image I n (x, y) is used as the base of the Laplace pyramid. By adding details, we obtain the Laplace pyramid:

[0043] Laplacian Pyramid={L0, L1, L2,...,I n}

[0044] Preferably, step S3 includes steps S31 and S32.

[0045] Step S31, Bandpass filter design:

[0046] Step S311, Design a dual IIR low-pass filter:

[0047] Set the vibration frequency range [f] low f high The time-domain bandpass filter is applied to the time-varying signal of each pixel in the Laplacian pyramid, using two IIR low-pass filters (LPFs) with different cutoff frequencies. low and LPF high For each frame of video image, first use LPF low Filter out high-frequency components, retain low-frequency signals, and then use LPF. high Filtering out results in a wideband signal;

[0048] Step S312, Filter Differential:

[0049] via LPF high Filter result minus LPF low Filtering results:

[0050] BandPass = LPF high (Frame)-LPF low (Frame);

[0051] A frame represents a single static image in a video stream.

[0052] Preferably, step S32, vibration signal enhancement and reconstruction:

[0053] Step S321, enhance the signal:

[0054] For the separated vibration signal L k Amplification is achieved by applying a gain factor α:

[0055] L′ k =L k ×α;

[0056] Step S322, reconstruct the enhanced signal:

[0057] By reconstructing the Laplace pyramid layer by layer, video frames with enhanced micro-vibration signals were obtained:

[0058] I' i (x, y) = L′ i (x, y) + Upsample(I′) i+1 (x, y));

[0059] The enhanced video frames are upsampled and image fused using OpenCV functions to achieve layer-by-layer reconstruction. The reconstructed image I′0(x,y) is the enhanced result frame.

[0060] Preferably, step S4 includes steps S41 and S42.

[0061] Step S41, Region selection and optical flow calculation:

[0062] Step S411, ROI region selection:

[0063] From the enhanced image frame, manually or automatically select the region of interest (ROI);

[0064] Step S412, Optical flow calculation:

[0065] The Farneback optical flow method in OpenCV is used to calculate the optical flow within the ROI region for motion analysis. Multi-resolution block matching is used to optimize the optical flow vector of each pixel, and the output optical flow field flow(x,y)=(u,v) is generated, where u and v are the displacement components of the target device in the horizontal and vertical directions.

[0066] To calculate the motion amplitude, the optical flow field is converted to polar coordinates:

[0067]

[0068] The average value of the motion amplitude is calculated for the selected area, and the vibration signal amplitude and pixel motion direction of each frame are extracted.

[0069] Preferably, in step S42, vibration frequency extraction:

[0070] Step S421, Vibration signal construction:

[0071] By concatenating the average amplitude of optical flow motion in each frame according to the frame sequence, a temporal vibration signal sequence is obtained:

[0072]

[0073] Where N is the number of valid pixels in the target area;

[0074] Step S422, Fourier transform analysis:

[0075] Perform a Fast Fourier Transform (FFT) on the vibration signal sequence to convert the time-domain signal into a frequency-domain signal:

[0076]

[0077] Where t represents the time of each moment or frame, j represents the index of the spatial pixel, f represents the frequency describing the device's micro-vibrations, and T represents the total duration of the video;

[0078] Step S423, Feature Frequency Extraction:

[0079] Find the frequency component f with the largest amplitude in the frequency domain signal. max The characteristic vibration frequency of the equipment:

[0080] f max =arg max|F(f)|.

[0081] Preferably, in step S5, based on the dynamic changes in vibration signal amplitude and characteristic frequency, a threshold range is set, vibration frequency deviating from the normal range is detected, an alarm is triggered, and video and analysis data of the abnormal period are recorded;

[0082] Real-time generation of vibration signal time-domain waveform diagrams, frequency-domain vibration spectrum diagrams, and optical flow field animations displays the motion characteristics of the target area.

[0083] This invention also provides a system for real-time monitoring of device micro-vibrations based on visual enhancement, including...

[0084] The video acquisition module is used to acquire high frame rate video of the micro-vibrations of the target device and record the acquisition parameters in real time.

[0085] The signal enhancement module is used to input the original video frame, separate the vibration signal and background information through the Laplacian pyramid, enhance the signal after filtering and differential, and reconstruct the enhanced signal.

[0086] The region selection and optical flow calculation module is used for selecting the Region of Interest (ROI) in the enhanced image frame and calculating the optical flow within the ROI.

[0087] The vibration frequency extraction module is used for vibration signal construction, transformation, and characteristic frequency extraction.

[0088] The real-time monitoring and data output module is used for real-time monitoring of vibration signals, assessment of process status, early warning, and visualization of vibration signal data to display the motion characteristics of the target area.

[0089] Therefore, the present invention employs the above-mentioned method and system for real-time monitoring of device micro-vibrations based on visual enhancement, which has the following beneficial effects:

[0090] This invention employs non-contact optical imaging and visual enhancement methods, avoiding the problems of altered mass distribution and poor environmental adaptability caused by the physical contact required by accelerometers. It is also unaffected by electromagnetic interference or high humidity. In signal extraction, this invention utilizes bandpass filters and Laplace pyramid decomposition to enhance micro-vibration signals in specific frequency bands, eliminating background noise and overcoming the high requirements of laser displacement sensors on target surface materials and reflection characteristics, thus adapting to complex surfaces and environmental changes. The motion analysis based on optical flow avoids the dependence of eddy current sensors on the electromagnetic properties of the measured material, improving measurement adaptability and accuracy. Furthermore, this invention extracts vibration frequencies in real time through Fourier transform, providing an efficient solution for equipment condition monitoring and fault early warning, possessing significant technical advantages such as strong environmental adaptability, high measurement accuracy, and superior anti-interference capabilities.

[0091] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0092] Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention;

[0093] Figure 2 This is a schematic diagram of a Gaussian pyramid according to an embodiment of the present invention. Detailed Implementation

[0094] To make the objectives, technical solutions, and advantages disclosed in the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely illustrative of the embodiments of the present invention and are not intended to limit the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments in this application without creative effort are within the scope of protection of this application. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout.

[0095] It should be noted that the terms “comprising” and “having”, and any variations thereof, are intended to cover non-exclusive inclusion, such that a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to such processes, methods, products, or devices.

[0096] Similar labels and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0097] In the description of this invention, it should be noted that the terms "upper," "lower," "inner," "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship in which the product of this invention is usually placed when in use. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention.

[0098] In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," and "connect" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0099] Example

[0100] like Figure 1 As shown, the present invention provides a method for real-time monitoring of device micro-vibrations based on visual enhancement, comprising the following steps:

[0101] Step S1: Acquire high frame rate video, and select key parts of the target device as acquisition areas through manual calibration, and record acquisition parameters in real time, including frame rate, resolution and number of frames.

[0102] ① High frame rate video capture

[0103] High-frame-rate industrial cameras (at least 120fps) are used to ensure thorough capture of even the smallest vibrations of the target equipment. Video resolution of 720p or 1080p is used to meet real-time spatial resolution requirements.

[0104] ②Setting the data collection target

[0105] The key parts of the target device are selected as the acquisition area by manual calibration, and the acquisition parameters, including frame rate, resolution and number of frames, are recorded in real time.

[0106] Step S2: Construct a Gaussian pyramid to perform layer-by-layer sampling and blurring of the image. Construct a Laplacian pyramid by the difference between adjacent layers of the Gaussian pyramid to decompose the acquired video into different spatial frequency bands.

[0107] After inputting the original video frames, the core of Laplacian pyramid decomposition lies in separating vibration signals and background information through multi-scale image representation to highlight the micro-vibration characteristics of the target region. The specific decomposition process is as follows:

[0108] Step S21, construct the Gaussian pyramid:

[0109] The construction of the Gaussian pyramid is the foundation of the Laplacian pyramid decomposition, which is achieved by downsampling and blurring the image layer by layer, as follows:

[0110] Let the input image be I0(x, y) with a resolution of W×H.

[0111] Step S211, Gaussian blur:

[0112] For image I i (x, y) is blurred using a Gaussian kernel to filter out high-frequency information:

[0113] I' i (x, y) = I i (x, y)*G(x, y);

[0114] Where G(x, y) is the standard Gaussian kernel,

[0115]

[0116] x and y represent the positions in the two-dimensional image coordinate system, and σ represents the standard deviation of the Gaussian distribution; this operation is achieved using the cv2.GaussianBlur function in OpenCV.

[0117] Step S212, image downsampling:

[0118] The blurred image I′ i (x, y) is downsampled by a ratio of 1 / 2 to obtain the next layer image I. i+1 (x, y),

[0119] I i+1 (x, y) = I′ i (2x, 2y)

[0120] This process is implemented using the OpenCV function cv2.pyrDown. During downsampling, the resolution is halved until a predetermined minimum size is reached;

[0121] Repeat the above steps to construct multiple layers of Gaussian pyramids until the target size is reached. The result is:

[0122] Gaussian Pyramid={I0,I1,I2,...,I n};

[0123] I n It is the lowest resolution image, such as Figure 2 It is a multi-layered Gaussian pyramid.

[0124] Step S22, construct the Laplace Pyramid:

[0125] The Laplacian pyramid is constructed using the image difference between adjacent layers of the Gaussian pyramid, preserving detailed information in each layer. The specific process is as follows:

[0126] Step S221, Laplace layer calculation:

[0127] For each Gaussian image I i (x, y) and its lower-level image I i+1 The upsampling result I′ of (x, y) i+1 Find the difference between (x, y) to obtain the Laplace layer L. i (x, y):

[0128] L i (x, y) = I i (x, y) - I′ i+1 (x, y);

[0129] Among them, I′ i+1 (x, y) is I i+1 The upsampled image of (x, y) is obtained using the OpenCV function cv2.pyrUp.

[0130] Step S222, preserve the underlying Gaussian image:

[0131] Lowest resolution image I n (x, y) serves as the base of the Laplace pyramid. By adding detailed information, the final Laplace pyramid is obtained:

[0132] Laplacian Pyramid={L0, L1, L2,...,I n}

[0133] Step S3: Design a bandpass filter to extract the specific frequency band of the equipment vibration signal, perform time-domain bandpass filtering on the time-varying signal of each pixel of the Laplace pyramid, and enhance and reconstruct the micro-vibration signal.

[0134] Step S31, Bandpass filter design:

[0135] Based on the vibration characteristics of the target equipment, a bandpass filter is designed to extract the specific frequency band of the equipment's vibration signal, thereby obtaining a clearer input signal. The vibration frequency range [f] is then set. low f high The time-domain bandpass filtering of the time-varying signal of each pixel in the Laplacian pyramid is performed as follows:

[0136] Step S311, Design a dual IIR low-pass filter:

[0137] Choose two IIR low-pass filters LPF with different cutoff frequencies. low and LPF high For each frame of video image, first use LPF low Filter out high-frequency components, retain lower-frequency signals, and then use LPF. high Higher frequencies are filtered out to obtain a wider frequency band signal.

[0138] Step S312, Filter Differential:

[0139] via LPF high Filter result minus LPF low Filtering results:

[0140] BandPass = LPF high (Frame)-LPF low (Frame);

[0141] A frame represents a single static image in a video stream. After final filtering, the retained pixel vibration signals represent motion changes within the corresponding frequency range.

[0142] Step S32, Vibration signal enhancement and reconstruction:

[0143] Step S321, enhance the signal:

[0144] For the specific frequency component associated with the micro-vibration signal, i.e., the separated vibration signal L k Amplification is achieved by applying a gain factor α:

[0145] L′ k =L k ×α;

[0146] Gain-processed Laplace layer L′ k The amplified vibration signal characteristics were preserved, while the other layers remained unchanged.

[0147] Step S322, reconstruct the enhanced signal:

[0148] By reconstructing the Laplace pyramid layer by layer, video frames with enhanced micro-vibration signals were obtained:

[0149] I' i (x, y) = L′ i (x, y) + Upsample(I′) i+1 (x, y));

[0150] The OpenCV functions cv2.add and cv2.pyrUp are used to upsample and fuse the enhanced video frames, achieving layer-by-layer reconstruction. Finally, the reconstructed image I′0(x, y) is the enhanced result frame.

[0151] Step S4: Select a region from the enhanced image frame, calculate the optical flow within the selected region using the Farneback optical flow method in OpenCV, and extract the characteristic frequencies of vibration.

[0152] Step S41, Region selection and optical flow calculation:

[0153] Step S411, ROI region selection:

[0154] From the enhanced image frames, select the Region of Interest (ROI) manually or automatically to ensure that the motion area of ​​the target device is focused on, while avoiding interference from non-target areas.

[0155] Step S412, Optical flow calculation:

[0156] The Farneback optical flow method (cv2.calcOpticalFlowFarneback) in OpenCV is used to calculate the optical flow within the ROI region for motion analysis.

[0157] The optical flow vector of each pixel is optimized by multi-resolution block matching, and the output optical flow field flow(x, y) = (u, v) is generated, where u and v are the displacement components of the target device in the horizontal and vertical directions.

[0158] To calculate the motion amplitude, the optical flow field is converted to polar coordinates:

[0159]

[0160] The average value of the motion amplitude is calculated for the selected area, and the vibration signal amplitude and pixel motion direction of each frame are extracted.

[0161] Step S42, vibration frequency extraction:

[0162] Step S421, Vibration signal construction:

[0163] By concatenating the average amplitude of optical flow motion in each frame according to the frame sequence, a temporal vibration signal sequence is obtained:

[0164]

[0165] Where N is the number of valid pixels in the target area;

[0166] Step S422, Fourier transform analysis:

[0167] Performing a Fast Fourier Transform (FFT) on the vibration signal sequence converts the time-domain signal into a frequency-domain signal.

[0168]

[0169] Where t represents the time of each moment or frame, j represents the index of the spatial pixel, f represents the frequency describing the device's micro-vibrations, and T represents the total duration of the video.

[0170] Step S423, Feature Frequency Extraction:

[0171] Find the frequency component f with the largest amplitude in the frequency domain signal. max As the main vibration frequency of the equipment:

[0172] f max =arg max|F(f)|;

[0173] Characteristic frequencies correspond to the physical characteristics of equipment vibration and can be used for monitoring and fault diagnosis.

[0174] Step S5 involves data visualization, generating real-time vibration signal images to display the motion characteristics of the target area, and monitoring and recording the dynamic changes in vibration signal amplitude and characteristic frequency.

[0175] ① Status assessment and early warning

[0176] Based on the dynamic changes in vibration signal amplitude and characteristic frequency, a threshold range is set. Once the detected vibration frequency deviates from the normal range, the system triggers an alarm and records video and analysis data of the abnormal period.

[0177] ② Data visualization

[0178] Real-time generation of vibration signal time-domain waveform diagrams, frequency-domain vibration spectrum diagrams, and optical flow field animations displays the motion characteristics of the target area.

[0179] The system of the real-time monitoring method for device micro-vibration based on vision enhancement described in this invention includes:

[0180] The video acquisition module is used to acquire high frame rate video of the micro-vibrations of the target device and record the acquisition parameters in real time.

[0181] The signal enhancement module is used to input the original video frame, separate the vibration signal and background information through the Laplacian pyramid, enhance the signal after filtering and differential, and reconstruct the enhanced signal;

[0182] The region selection and optical flow calculation module is used for selecting the Region of Interest (ROI) in the enhanced image frame and calculating the optical flow within the ROI.

[0183] The vibration frequency extraction module is used for vibration signal construction, transformation, and characteristic frequency extraction.

[0184] The real-time monitoring and data output module is used for real-time monitoring of vibration signals, process status assessment and early warning, and visualization of vibration signal data to display the motion characteristics of the target area.

[0185] The system of this invention first decomposes the input video sequence into different spatial frequency bands by constructing a Laplacian pyramid, and applies the same time bandpass filter to all frequency bands. Then, the filtered spatial frequency bands are amplified by a given factor to obtain enhanced image frames, which are then added to the original signal and folded to generate a vibration-amplified output video. The amplitude of the vibration signal in each frame is enhanced by the above method, realizing vibration visualization. By performing optical flow calculation on the enhanced image frames, the equipment vibration spectrum can be obtained, thereby enabling equipment fault analysis based on the spectrum.

[0186] Through systematic optimization and methodological improvements, the following advantages and effects are mainly demonstrated:

[0187] ① High-sensitivity extraction of micro-vibration signals

[0188] Bandpass filter design:

[0189] Using dual IIR filters to perform time-domain bandpass filtering on video frame signals effectively extracts the frequency components of micro-vibrations, eliminates low-frequency interference (such as changes in device background light) and high-frequency noise, and ensures high-quality input signals.

[0190] Signal enhancement module:

[0191] The signal extracted by the Laplace pyramid decomposition and filtering is amplified to bring the subtle vibration signal into the observable range, while suppressing interference from background-irrelevant regions.

[0192] ② Efficient motion characteristic analysis

[0193] Optical flow combined with visual enhancement:

[0194] Optical flow is used to analyze vibration amplitude and direction on the enhanced image, avoiding the problem of insufficient sensitivity of traditional optical flow in detecting subtle motions, thus accurately capturing the dynamic changes of the target area.

[0195] ③ Accurate frequency response extraction

[0196] Frequency domain analysis:

[0197] By extracting the spectral characteristics of vibration signals through Fourier transform, the vibration frequency of the target equipment can be accurately calculated, providing core indicators of the equipment's status.

[0198] Combining the time domain and frequency domain:

[0199] By employing both time-domain filtering and frequency-domain analysis, the signal's distinguishability is enhanced, ensuring the accuracy and reliability of vibration characteristic extraction.

[0200] ④ Equipment safety and fault early warning

[0201] Based on the detection results of real-time vibration characteristics, this system can quickly determine the operating status of the equipment and issue early warning signals, thereby greatly improving the safety of equipment operation and maintenance efficiency.

[0202] The present invention employs the above-mentioned vision-enhanced real-time monitoring method and system for equipment micro-vibration, which can solve the problems of insufficient sensitivity and poor real-time performance in traditional vibration monitoring methods for detecting small vibrations of equipment.

[0203] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for real-time monitoring of device micro-vibrations based on visual enhancement, characterized in that: Includes the following steps: Step S1: Acquire high frame rate video, and select key parts of the target device as acquisition areas through manual calibration, and record acquisition parameters in real time, including frame rate, resolution and number of frames. Step S2: Construct a Gaussian pyramid to perform layer-by-layer sampling and blurring of the image. Construct a Laplacian pyramid by the difference between adjacent layers of the Gaussian pyramid to decompose the acquired video into different spatial frequency bands. Step S3: Design a bandpass filter to extract the specific frequency band of the equipment vibration signal, perform time-domain bandpass filtering on the time-varying signal of each pixel of the Laplace pyramid, and enhance and reconstruct the micro-vibration signal. Step S4: Select a region from the enhanced image frame, calculate the optical flow within the selected region using the Farneback optical flow method in OpenCV, and extract the characteristic frequencies of vibration. Step S5: Perform data visualization, generate vibration signal images in real time to display the motion characteristics of the target area, and monitor and record the dynamic changes in vibration signal amplitude and characteristic frequency; Step S4 includes steps S41 and S42. Step S41, Region selection and optical flow calculation: Step S411, ROI region selection: From the enhanced image frame, manually or automatically select the region of interest (ROI); Step S412, Optical flow calculation: The Farneback optical flow method in OpenCV is used to calculate the optical flow within the ROI region for motion analysis. Multi-resolution block matching is used to optimize the optical flow vector of each pixel, and the optical flow field is output. ,in These are the displacement components of the target device in the horizontal and vertical directions. To calculate the motion amplitude, the optical flow field is converted to polar coordinates: ； Calculate the average value of motion amplitude for the selected area, and extract the vibration signal amplitude and pixel motion direction for each frame; Step S42, vibration frequency extraction: Step S421, Vibration signal construction: By concatenating the average amplitude of optical flow motion in each frame according to the frame sequence, a temporal vibration signal sequence is obtained: ； in N The number of valid pixels within the target area; Step S422, Fourier transform analysis: Perform a Fast Fourier Transform (FFT) on the vibration signal sequence to convert the time-domain signal into a frequency-domain signal: ； in This represents the time of each moment or frame. Indicates the index of a spatial pixel. This indicates the frequency describing the micro-vibration of the equipment. Indicates the total duration of the video; Step S423, Feature Frequency Extraction: Find the frequency component with the largest amplitude in the frequency domain signal. The characteristic vibration frequency of the equipment: 。 2. The method for real-time monitoring of device micro-vibrations based on visual enhancement according to claim 1, characterized in that: In step S1, a high frame rate industrial camera is used to capture minute vibrations of the target device. The high frame rate industrial camera captures video with a frame rate of no less than 120fps and a video resolution of 720p or 1080p.

3. The method for real-time monitoring of device micro-vibrations based on visual enhancement according to claim 1, characterized in that: Step S2 includes steps S21 and S22. Step S21, construct the Gaussian pyramid: Step S211, Gaussian blur: Let the input image be Its resolution is ; For images Applying a Gaussian kernel for blurring to filter out high-frequency information: ； in, For standard Gaussian kernels, ； , Represents the position in a two-dimensional image coordinate system. This represents the standard deviation of a Gaussian distribution. Step S212, image downsampling: Blur the image The next layer image is obtained by downsampling by 1 / 2 ratio. , ； During downsampling, the resolution is halved until the predetermined minimum size is reached; Repeat the above steps to construct multiple layers of Gaussian pyramids until the target size is reached. The result is: ； It is the lowest resolution image.

4. The method for real-time monitoring of device micro-vibrations based on visual enhancement according to claim 3, characterized in that: Step S22, construct the Laplace Pyramid: Step S221, Laplace layer calculation: For each Gaussian image layer Its underlying image upsampling results Calculate the difference to obtain the Laplace layer. : ； in, yes Upsampled image; Step S222, preserve the underlying Gaussian image: Lowest resolution image As the base of the Pyramid of Laplace, by adding details, we obtain the Pyramid of Laplace: 。 5. The method for real-time monitoring of device micro-vibration based on visual enhancement according to claim 1, characterized in that: Step S3 includes steps S31 and S32. Step S31, Bandpass filter design: Step S311, Design a dual IIR low-pass filter: Set vibration frequency range Time-domain bandpass filtering is performed on the time-varying signal of each pixel in the Laplacian pyramid, using two IIR low-pass filters with different cutoff frequencies. and For each frame of video image, first use Filter out high-frequency components, retain low-frequency signals, and then use them. Filtering out results in a wideband signal; Step S312, Filter Differential: pass Filter result minus Filtering results: ； This represents a single static image in a video stream.

6. The method for real-time monitoring of device micro-vibration based on visual enhancement according to claim 5, characterized in that: Step S32, Vibration signal enhancement and reconstruction: Step S321, enhance the signal: For the separated vibration signal Application gain coefficient Enlarge: ； Step S322, reconstruct the enhanced signal: By reconstructing the Laplace pyramid layer by layer, video frames with enhanced micro-vibration signals were obtained: ； OpenCV functions are used to upsample and fuse enhanced video frames to achieve layer-by-layer reconstruction, resulting in reconstructed images. This is the enhanced result frame.

7. The method for real-time monitoring of device micro-vibrations based on visual enhancement according to claim 1, characterized in that: In step S5, based on the dynamic changes in vibration signal amplitude and characteristic frequency, a threshold range is set, vibration frequency deviations from the normal range are detected, an alarm is triggered, and video and analysis data of the abnormal period are recorded. Real-time generation of vibration signal time-domain waveform diagrams, frequency-domain vibration spectrum diagrams, and optical flow field animations displays the motion characteristics of the target area.

8. A system for real-time monitoring of device micro-vibrations based on visual enhancement according to any one of claims 1-7, characterized in that: include The video acquisition module is used to acquire high frame rate video of the micro-vibrations of the target device and record the acquisition parameters in real time. The signal enhancement module is used to input the original video frame, separate the vibration signal and background information through the Laplacian pyramid, enhance the signal after filtering and differential, and reconstruct the enhanced signal; The region selection and optical flow calculation module is used for selecting the Region of Interest (ROI) in the enhanced image frame and calculating the optical flow within the ROI. The vibration frequency extraction module is used for vibration signal construction, transformation, and characteristic frequency extraction. The real-time monitoring and data output module is used for real-time monitoring of vibration signals, assessment of process status, early warning, and visualization of vibration signal data to display the motion characteristics of the target area.

Images