Portable device and method for identifying and rapidly detecting particles in insulating oil

Abstract

The invention relates to a portable device and a method for identifying and rapidly detecting particles in insulating oil. The portable device and the method are combined with a micro-flow imaging technology and a Raman spectrum analysis technology. The device can detect the particles in the insulating oil with high precision and high speed and obtain the physical characteristics and chemical components of the particles. Physical characteristics of particles are detected through a microflow imaging technology, and meanwhile chemical component information of the particles is analyzed through a surface enhanced Raman scattering technology. The device has a real-time online monitoring function, data transmission and remote control can be realized through the wireless communication module, and the detection efficiency and the intelligent level are remarkably improved. And the high-resolution CMOS camera and the SERS module are adopted, so that the detection sensitivity and precision are remarkably improved. In combination with Gaussian blur and histogram equalization technologies, noise can be effectively removed, the contrast of the particle image can be enhanced, and the particles can be accurately identified even under the condition of weak light or low contrast.

Description

Technical Field

[0001] The present invention belongs to the field of transformers, and particularly relates to a portable device and method for particle identification and rapid detection in insulating oil. Background Art

[0002] During the long-term operation of electrical equipment such as power transformers, insulating oil, as a key insulating medium, the stability of its performance directly affects the reliability and safety of the equipment. As the operation time of the equipment increases, tiny particles will gradually form in the insulating oil due to reasons such as thermal aging, electrical stress, mechanical wear, and intrusion of external impurities. Under the action of the electric field, the particles aggregate as the insulating oil flows, thus triggering electrical faults such as partial discharge, further leading to the deterioration of insulating materials, and even possibly causing breakdown accidents, seriously threatening the normal operation of power equipment. Therefore, the accurate detection and rapid identification of particles in insulating oil are of great significance for the fault prevention and insulation status monitoring of power equipment.

[0003] Currently, there are many limitations in traditional particle detection methods. First of all, common detection techniques mainly rely on the measurement of particle physical properties, such as particle size and quantity, and it is difficult to identify the types of particles in the oil, which limits the comprehensive analysis of the potential hazards of particles. Secondly, the detection sensitivity of traditional methods is relatively low, especially in the case of dealing with tiny particles, it is difficult to achieve accurate detection. In addition, these techniques usually take a long time and cannot meet the requirements of on-line monitoring and rapid response of insulating oil.

[0004] In recent years, as an emerging particle detection means, microfluidic imaging technology has received extensive attention because it can capture images of tiny particles in liquid flow with high precision. However, a single physical imaging ability is not sufficient to meet the analysis requirements of particle types. At the same time, Raman spectroscopy technology, as a spectroscopic analysis method based on molecular vibration information, can provide the chemical composition information of particles, and then deduce the types of particles. Therefore, combining microfluidic imaging technology with Raman spectroscopy technology can achieve multi-parameter detection of particles in insulating oil, filling the gap in the comprehensive detection of particle physical and chemical information in the existing technology. Summary of the Invention

[0005] To address the deficiencies in the existing technology, the present invention proposes a portable device and method for particle identification and rapid detection in insulating oil. Based on microfluidic imaging and Raman spectroscopy analysis, it aims to achieve high-precision and rapid detection of particles in insulating oil. This device can not only simultaneously obtain the physical characteristics (such as size and shape) and chemical composition information (such as type) of particles, but also has a real-time online monitoring function, making up for the deficiencies of the existing technology in detection accuracy, detection range, and multi-parameter comprehensive analysis. The innovative application of this technology is of great significance for improving the operation safety of power equipment, extending the service life of equipment, and promoting the academic research and industrial application of insulating oil particle detection technology.

[0006] The present invention adopts the following technical solutions: A portable device for particle identification and rapid detection in insulating oil, comprising:

[0007] A microfluidic imaging module, including an upper flow channel for focusing particles in insulating oil, two lower flow channels parallel to the upper flow channel, and a vertical connection channel for connecting the upper flow channel and the lower flow channels to ensure stable transmission of particles to the lower flow channels. The two lower flow channels include a No. 1 lower flow channel for optical imaging of particles and a No. 2 lower flow channel for spectral analysis of particles.

[0008] A flow control system for regulating the flow rate and pressure of insulating oil in the upper flow channel and the lower flow channels.

[0009] A high-resolution imaging device for real-time imaging of particles in the No. 1 lower flow channel and transmitting image data to the data processing unit;

[0010] A surface-enhanced Raman scattering (SERS) module for exciting and collecting signals from particles in the No. 2 lower flow channel to obtain the chemical composition information of particles; The Raman spectrometer is equipped with an excitation light source with adjustable wavelength and power, and is combined with the SERS module to significantly enhance the Raman scattering signal.

[0011] A Raman spectroscopy analyzer is used to analyze the molecular vibration information of particles in the spectral analysis area of the No. 2 lower flow channel; a data processing unit is used to process the image data transmitted by a high-resolution imaging device to obtain a final image with clear particle contours, and obtain the physical characteristics of the particles from the final image; it is also used to comprehensively analyze the physical characteristics and chemical composition information of the particles to achieve particle classification [here, the comprehensive analysis refers to correlating and comparing the physical characteristics of the particles (such as the size, shape, surface characteristics, etc. of the particles) and the chemical composition information (such as the chemical composition, molecular structure, etc. of the particles), and then realizing the identification and classification of the particles], and generate a detection report. The following contents are presented in the detection report: 1) Physical characteristics of the particles: including the concentration, shape, and particle size distribution of the particles; 2) Chemical composition information of the particles: the composition substances and molecular structure of the particles; 3) Classification of the particles. According to the physical properties and chemical composition, the types of particles in the oil can be output. The device uses Gaussian blur and histogram equalization techniques to process the image data to reduce noise interference and improve the contrast of the particle images. The device has a modular design and can be flexibly adjusted according to the requirements of different application scenarios, and is applicable to the detection of particles in other liquid media.

[0012] A wireless communication module is used to transmit the physical property information, chemical composition information, image data, molecular vibration information, and detection report of the particles to a remote device in real time.

[0013] Further, the upper flow channel is a spiral microchannel, and the secondary flow and inertial focusing effect are used to concentrate the particles to the center of the upper flow channel.

[0014] Further, the high-resolution imaging device uses a CMOS camera with a resolution of 1920×1200 pixels and can capture the image data of the particles in the insulating oil at a speed of 155 frames per second.

[0015] Further, the wireless communication module uses a Wi-Fi module and performs data transmission through the MQTT protocol, supporting remote real-time monitoring and control.

[0016] A method for identifying and rapidly detecting particles in insulating oil uses the above-mentioned portable device for identifying and rapidly detecting particles in insulating oil, and includes the following steps:

[0017] Step 1: Pretreat the insulating oil sample to be tested to remove air bubbles and large particle impurities;

[0018] Step 2: Inject the pretreated insulating oil sample into the upper flow channel and use the spiral microchannel to focus the particles;

[0019] Step 3: Transmit the focused particles to the lower flow channel through a vertical connection channel;

[0020] Step 4: Use a high-resolution imaging device to perform real-time imaging on the particles in the No. 1 lower channel, and transmit the image data to the data processing unit. The data processing unit processes the image data using Gaussian blur and histogram equalization to obtain the final image;

[0021] Step 5: Perform spectral analysis on the particles in the No. 2 lower channel through the surface-enhanced Raman scattering (SERS) module to obtain chemical composition information;

[0022] Step 6: The data processing unit comprehensively analyzes the collected physical properties and chemical composition information of the particles, generates a test report, and performs data transmission through the wireless communication module.

[0023] Further, the process of the data processing unit processing the image data using Gaussian blur and histogram equalization to obtain the final image is as follows:

[0024] 1) Convert the image data into a grayscale image, and the formula is as follows:

[0025] Gray = 0.299×R + 0.587×G + 0.114×B

[0026] Where: R represents the red component of the pixel, with a value range of [0, 255], G represents the green component of the pixel, with a value range of [0, 255], B represents the blue component of the pixel, with a value range of [0, 255], and Gray is the converted grayscale value, with a value range of [0, 255].

[0027] 2) Use the Gaussian blur method to eliminate the noise in the grayscale image obtained in 1) to obtain a Gaussian blurred image:

[0028] Gaussian blur is the process of convolving the grayscale image with a Gaussian kernel function. It smooths the image through the weighted average of neighboring pixels to reduce noise interference. The Gaussian kernel function formula is as follows:

[0029]

[0030] Where: G(x, y) represents the Gaussian kernel function, indicating the weight at the position (x, y), σ: the standard deviation of the Gaussian function, and x and y are the horizontal and vertical coordinates of the pixel points in the image.

[0031] Use the weights generated by the Gaussian function to convolve the grayscale image. The convolution formula is:

[0032]

[0033] Where: I ′(x, y) is the pixel value of the blurred image; I(x + i, y + j) is the pixel value at the position (x + i, y + j) in the grayscale image; G(i, j) represents the weight at the position (i, j), and k is the size of the convolution kernel.

[0034] 3) Calculate the gradient of the Gaussian blurred image to detect edges

[0035] The gradient of the Gaussian blurred image is calculated by the Sobel operator, and the gradients in the horizontal and vertical directions are calculated respectively:

[0036]

[0037] Where: G x and G y are the gradients in the horizontal and vertical directions respectively, and I is the Gaussian blurred image.

[0038] The gradient magnitude G of the Gaussian blurred image is calculated by taking the square root of the sum of the squares of the horizontal and vertical direction gradients:

[0039]

[0040] The gradient direction θ of the Gaussian blurred image is calculated as:

[0041]

[0042] 4) Use the non-maximum suppression method and the double-threshold method to process the Gaussian blurred image noise to obtain a clear particle contour image:

[0043] Non-maximum suppression method: Check adjacent pixels along the gradient direction θ. If the gradient value of the current pixel point is not a local maximum, set the current pixel to 0.

[0044] Double-threshold method: Set two thresholds, which are the high threshold T high and the low threshold T low :

[0045] If the gradient value of the current pixel point is greater than the high threshold T high , it is considered a strong edge.

[0046] If the gradient value of the current pixel point is less than the low threshold T low , it is considered noise and discarded.

[0047] If the gradient value of the current pixel point is between the two thresholds, the current pixel point is retained as a weak edge, but it is only retained if it is connected to a strong edge, otherwise it is also removed.

[0048] 5) Use the histogram equalization method to enhance the clear particle contour image obtained in 4) to obtain the final image;

[0049] 5-1) Calculate the number of pixel points with each gray value in the clear particle contour image to generate a gray histogram H(r k ), where r k represents the k-th gray level, with a value range of [0, L-1], where L is the total number of gray levels. Usually, L = 256 corresponds to an 8-bit image.

[0050] The normalized histogram p(r k ) represents the probability of each gray value occurring:

[0051]

[0052] where N is the total number of pixels in the image, and n k represents the number of pixels with gray level r k .

[0053] 5-2) Calculate the cumulative distribution function CDF(r k ) for each gray level r k . CDF(r k ) represents the cumulative probability from the lowest gray value to the gray value r k . The formula is as follows:

[0054]

[0055] where: p(r j ) is the normalized probability of the gray value r j .

[0056] 5-3) Map the gray value r k in the clear particle contour image to the equalized gray value s k . The mapping formula is as follows:

[0057] s k = (L-1) × CDF(r k )

[0058] where: s k is the equalized gray value, and L-1 is the maximum value of the gray levels.

[0059] Compared with the prior art, the present invention has at least the following beneficial effects:

[0060] 1. The device of the present invention can simultaneously obtain the physical information (such as shape, size, distribution) and chemical composition (such as type, molecular vibration information) of particles, greatly improving the accuracy and comprehensiveness of detection. By combining microfluidic imaging and SERS technology, it overcomes the limitation that traditional methods can only detect certain aspects of particle information, meeting the requirements of multi-parameter detection of particles in insulating oil.

[0061] 2. The present invention adopts a high-resolution CMOS camera and an SERS module, greatly improving the detection accuracy and sensitivity of particles. Through the electromagnetic field effect enhanced by nanostructures, the SERS technology significantly enhances the Raman scattering signal, enabling the detection of the chemical composition of particles to have higher sensitivity and realizing the identification and analysis of trace particles and low-concentration particles.

[0062] 3. The detection device of the present invention has a real-time online monitoring function and can realize the instant transmission and remote monitoring of data through a wireless communication module. The device integrates a multi-dimensional data processing unit, which can comprehensively analyze the collected particle images and Raman spectra and generate a detailed detection report, greatly improving the detection efficiency and the intelligent level of data processing.

[0063] 4. The method of the present invention uses Gaussian blur and histogram equalization techniques, enabling the device of the invention to effectively reduce noise interference and improve image contrast, so that clear particle contours can still be obtained in a complex imaging environment (such as weak light or low-contrast conditions), ensuring the accuracy of edge detection and thus improving the overall detection reliability.

[0064] 5. The device of the present invention is compact and portable, suitable for rapid detection on-site. The device is easy to operate and has a high integration level, and can easily realize the capture, positioning, imaging and analysis of particles. By optimizing the flow channel design and precise flow control, the stable flow and accurate detection of particles in the detection area are ensured.

[0065] 6. The present invention is not only applicable to the detection of insulating oil particles, but can also be extended to the particle analysis in other liquid media according to needs. The modular design of the device facilitates adjustment according to the requirements of different application scenarios and has good scalability and application prospects.

[0066] 7. The device of the present invention can simultaneously obtain the physical information of particles (such as morphology, size, distribution) and chemical composition (such as type, molecular vibration information), greatly improving the detection accuracy and comprehensiveness. By combining microfluidic imaging and SERS technology, the limitation that traditional methods can only detect certain aspects of particle information is overcome, meeting the requirements of multi-parameter detection of particles in insulating oil.

[0067] 8. The present invention adopts a high-resolution CMOS camera and an SERS module, greatly improving the detection accuracy and sensitivity of particles. Through the electromagnetic field effect enhanced by nanostructures, the SERS technology significantly enhances the Raman scattering signal, enabling the detection of the chemical composition of particles to have higher sensitivity and realizing the identification and analysis of trace particles and low-concentration particles.

[0068] 9. The detection device of the present invention has a real-time online monitoring function, and can achieve instant data transmission and remote monitoring through a wireless communication module. The device integrates a multi-dimensional data processing unit, which can comprehensively analyze the collected particle images and Raman spectra, and generate a detailed detection report, greatly improving the detection efficiency and the intelligent level of data processing.

[0069] 10. The method of the present invention uses Gaussian blur and histogram equalization techniques, enabling the device of the invention to effectively reduce noise interference and improve image contrast, so that clear particle contours can still be obtained in a complex imaging environment (such as under conditions of weak light or low contrast), ensuring the accuracy of edge detection, and thus improving the overall detection reliability.

[0070] 11. The device of the present invention is designed to be compact and portable, suitable for rapid on-site detection. The device is easy to operate and has a high degree of integration, and can easily achieve particle capture, positioning, imaging and analysis. By optimizing the flow channel design and precise flow control, the stable flow and accurate detection of particles in the detection area are ensured.

[0071] 12. The present invention is not only applicable to the detection of insulating oil particles, but can also be extended to the particle analysis in other liquid media according to needs. The modular design of the device facilitates adjustment according to the requirements of different application scenarios, and has good scalability and application prospects. Brief Description of the Drawings

[0072] Figure 1 is the flow chart of the oil particle detection method of the detection method provided by the present invention.

[0073] Figure 2 is the schematic diagram of the principle of the portable device for particle identification and rapid detection in insulating oil provided by the present invention.

[0074] Figure 3 is the structural schematic diagram of the portable device for particle identification and rapid detection in insulating oil provided by the present invention. Detailed Embodiments

[0075] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. The embodiments described in this application are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the spirit of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0076] The present invention proposes a method for particle identification and rapid detection in insulating oil, which combines microfluidic imaging technology and Raman spectroscopy analysis technology to achieve comprehensive detection of the physical properties and chemical compositions of particles. It mainly includes the following five steps.

[0077] Sample pretreatment, sample injection into the upper flow channel, particle flow focusing, vertical transmission of particles to the lower flow channel, real-time imaging of particles, Raman spectroscopy analysis of particles, and data processing and comprehensive analysis.

[0078] Example 1: A portable device for precise identification and rapid detection of particles in insulating oil, mainly composed of the following components and structures: an upper flow channel (focusing layer), a lower flow channel (detection layer), a vertical connection channel, a particle capture and positioning module, a high-resolution imaging device, a Raman spectrometer, a flow control system, a data processing unit, and a remote device (or called a display and control interface).

[0079] The upper flow channel (focusing layer) is designed as a spiral microchannel with a specific geometric structure. This flow channel design utilizes secondary flow (Dean vortices) and inertial focusing effects, causing the particles to be subjected to a centripetal force in the flow channel and gradually gather towards the center of the flow channel, thereby achieving effective focusing of the particles.

[0080] The lower flow channel (detection layer) is parallel to the upper flow channel and is located at the lower part of the device. It is designed as a flow channel area made of an optically transparent material, facilitating imaging and spectral analysis. It is used for high-resolution imaging and Raman spectroscopy analysis of particles to detect the physical properties and chemical compositions of the particles in the flow channel.

[0081] The vertical connection channel connects the upper flow channel and the lower flow channel together. It is designed as a straight-through structure, using the action of gravity to ensure stable transmission of particles, reducing the possibility of deviation or retention. The vertical channel is divided into two paths. One path enters the No. 1 lower flow channel for particle imaging, and the other path enters the No. 2 lower flow channel for spectral analysis, ensuring stable transmission of particles from the focusing layer to the detection layer, avoiding confusion or loss during the transmission process, thereby improving the detection accuracy and efficiency.

[0082] The high-resolution imaging device is integrated in the imaging area of the lower flow channel. It uses a high-resolution CMOS camera with the model Basler acA1920-155um, having a resolution of 1920×1200 pixels and a capture ability of up to 155 frames per second. When the particles flow through the lower flow channel, this CMOS camera can capture particle images in real time and transmit the image data to the data processing unit for image processing.

[0083] The Raman spectrometer is installed in the spectral analysis area of the second lower flow channel, equipped with an excitation light source with adjustable wavelength and power, as well as a surface-enhanced Raman scattering (SERS) technology module. The laser emission source selects the Cobolt 08-01 series, which has an adjustable wavelength range such as 532 nm and 785 nm. The specific models include Cobolt 08-01 532 nm or Cobolt 08-01 785 nm, with high wavelength stability and adjustable power output, suitable for a variety of spectral application scenarios. The SERS technology module uses the Horiba ST Japan G-SERS substrate, which is based on a metal nanoparticle structure and can locally generate an enhanced electromagnetic field on the metal surface, significantly amplifying the Raman scattering signal. It is particularly suitable for the detection of trace particles and low-concentration molecules. Through the combination of the above devices, the molecular vibration information of the particles in the flow channel can be collected and analyzed, and then the chemical composition information of the particles can be accurately extracted.

[0084] The flow control system consists of a precise pump and valve system, which is used to regulate the flow rate and pressure of the fluid in each laminar flow channel. The pump selects the Harvard Apparatus PHD Ultra model, which is commonly used in microfluidic systems and can accurately control the flow rate of microfluids. It supports precise flow rate adjustment from nanoliters to milliliters per minute to ensure the stable transmission of the fluid. Its adjustable flow rate and pressure functions meet the precise control requirements under a variety of experimental conditions and can handle fluids with a variety of different viscosities, suitable for diverse application scenarios. The valve selects the The Lee Company LHL series of miniature inert valves, which is a solenoid valve commonly used in microfluidic systems and is designed specifically for precise control of small-flow liquids. This valve can achieve high-precision fluid control within a small flow rate range to ensure that the particles flow stably in the flow channel according to the preset path. In addition, its low energy consumption characteristics are suitable for long-term operation to ensure the high efficiency and stability of the entire device during operation. The combined use of the pump and the valve ensures the stable flow of the fluid in each flow channel layer and provides reliable flow control. By precisely adjusting the flow rate and pressure of the sample, the flow path of the particles in the flow channel can be effectively regulated to achieve high-precision detection.

[0085] The data processing unit integrates a high-performance computing chip and an algorithm module, which is used to process the data acquired by the imaging device and the spectral analyzer. The computing chip is selected as NVIDIA Jetson Xavier NX, which has excellent image and data processing capabilities and can efficiently handle complex computing tasks. The algorithm module uses OpenCV for image processing and feature extraction, and uses the MATLAB Signal Processing Toolbox for spectral data analysis and report generation. Through this processing unit, the image and spectral data of the particles can be comprehensively processed, the physical properties and chemical composition information of the particles can be accurately extracted, and a detailed detection report can be generated. The wireless communication module, namely the communication module, selects the Wi-Fi module ESP8266 to communicate with the data processing unit through the wireless network, supports remote data transmission, and enables users to monitor and control the device in real time through the touch screen or computer terminal. The wireless communication module supports high-bandwidth data transmission, is suitable for transmitting large-scale data such as images and detection reports, and has the characteristics of high integration, low power consumption, and small volume, which is very suitable for embedded system applications. The communication protocol selects the MQTT protocol, which is lightweight and efficient, suitable for devices with limited resources, and realizes data transmission under low-bandwidth conditions. Its publish-subscribe mode supports multiple terminals to monitor the detection status or receive detection reports simultaneously. The MQTT protocol can effectively reduce communication energy consumption and is suitable for systems with high power consumption requirements.

[0086] The display and control interface is equipped with a touch screen or a computer terminal, which is used to operate the device and monitor the detection process in real time. Users can set detection parameters, monitor the detection status, and view or export the detection report through the interface.

[0087] This portable device for precise identification and rapid detection of particles in insulating oil realizes multi-parameter detection of particles in insulating oil through an optimized multi-layer flow channel design and advanced imaging and spectral analysis technologies. Through accurate particle capture, positioning, and detection processes, the device can provide high-precision physical and chemical information, providing important data support for the condition monitoring and fault diagnosis of power equipment. The device is flexibly designed and can be adjusted according to actual application requirements, and is suitable for particle detection tasks in various industrial and scientific research environments.

[0088] Embodiment 2: A method for identifying and rapidly detecting particles in insulating oil uses the portable device for precise identification and rapid detection of particles in insulating oil defined in Embodiment 1, and includes the following steps:

[0089] Step1: Pretreatment of the insulating oil sample to be measured

[0090] Pretreat the insulating oil sample to be tested to remove large particle impurities and bubbles, ensure the uniformity of the fluid in the microchannel, and reduce interference with the test results. A micro vacuum degassing module is used to remove bubbles. The sample is locally decompressed by a micro vacuum pump to reduce the solubility of gas in the liquid and promote the escape of bubbles. The escaped bubbles are discharged along a specific channel to ensure that the bubbles in the sample are completely removed. For large particle impurities, a disposable microporous filter membrane is used. The insulating oil sample is filtered through the microporous filter membrane, which can effectively intercept particle impurities larger than a specific size, ensure that large particles in the sample are removed, and ensure the uniformity and smooth flow of the fluid in the microchannel.

[0091] Step2: Inject the pretreated insulating oil sample to focus the particles in the insulating oil

[0092] Inject the pretreated insulating oil sample into the upper flow channel (focusing layer) of the microfluidic device. The upper flow channel adjusts the fluid through the spiral microchannel. This structure adjusts the fluid by inducing secondary flow effects and inertial flow, and uses the centrifugal force and inertial migration effects of the fluid to cause the particles in the insulating oil to converge in the flow channel. Larger particles gradually gather towards the center of the flow channel due to the greater centrifugal force and inertia, forming a stable particle flow path, providing a more accurate particle distribution for subsequent detection.

[0093] Step3: Shunt and transport the insulating oil sample for physical and chemical property detection

[0094] Through the vertical connection channel, accurately transfer the focused particles from the upper flow channel to the lower flow channel (detection layer). Use the gravity to ensure the stable transfer of the particles and reduce the possibility of deviation or retention. The vertical channel is divided into two paths. One path enters the No. 1 lower flow channel for particle optical imaging, and the other path enters the No. 2 lower flow channel for spectral analysis, ensuring the stable transfer of the particles from the focusing layer to the detection layer, avoiding confusion or loss during the transfer process, and thus improving the detection accuracy and efficiency.

[0095] Step4: Image the particles in the insulating oil

[0096] In the imaging area of the No. 1 lower flow channel, a high-resolution CMOS camera is used to perform real-time imaging on the particles and transmit the image data to the data processing unit. The data processing unit processes the acquired image data using an edge detection algorithm. First, the image data is converted into a grayscale image to simplify the subsequent processing process. Then, Gaussian blur technology is applied to eliminate noise and avoid noise interfering with the accuracy of edge detection. The edges in the image are identified by calculating the gradient values of the grayscale image, and then non-maximum suppression is performed to retain the edges of local maxima while removing weaker edges. By setting high and low thresholds, strong edges and weak edges are further distinguished, effectively eliminating the influence of false edges and noise to obtain a clear image of the particle contour, thereby accurately extracting the contour of the particle and measuring physical characteristics such as the size and shape of the particle, meeting the detection requirements with a relatively high clarity of the particle boundary.

[0097] To improve the contrast of the image and the accuracy of particle feature extraction, histogram equalization technology is used to enhance the clear image of the particle contour. First, the number of pixel points at each gray level in the clear image of the particle contour is counted to generate a gray histogram. Subsequently, the cumulative distribution function of each gray value is calculated based on the gray histogram, and based on this, the image pixel values are redistributed. By adjusting the gray range, the gray values in the darker and brighter regions of the image are stretched, thereby enhancing the overall contrast of the image. Finally, the equalized gray values are applied to the image to generate an image with enhanced contrast, making the contour of the particle clearer, especially suitable for imaging environments with weak light or low contrast, and further improving the accuracy of edge detection and shape analysis.

[0098] Specifically, 1) An edge detection algorithm is used to process the image data collected by the CMOS camera and convert it into a grayscale image. The formula is as follows:

[0099] Gray = 0.299×R + 0.587×G + 0.114×B

[0100] Where: R represents the red component of the pixel, with a value range of [0, 255], G represents the green component of the pixel, with a value range of [0, 255], B represents the blue component of the pixel, with a value range of [0, 255], and Gray is the converted grayscale value, with a value range of [0, 255].

[0101] 2) The Gaussian blur technology reduces the influence of noise on gradient calculation by smoothing the grayscale image. After calculating the gradient values, non-maximum suppression and double-threshold processing are used to effectively eliminate noise interference and accurately extract the real edges in the image.

[0102] The Gaussian blur is a process of convolving the image with a Gaussian kernel function, which smooths the image by weighted averaging of neighboring pixels and reduces noise interference. The Gaussian kernel function formula is as follows:

[0103]

[0104] Among them: G(x, y) represents the Gaussian kernel function, indicating the weight at the position (x, y). σ represents the standard deviation of the Gaussian function, which determines the degree of blurring; the larger σ is, the higher the degree of blurring. x and y are the pixel positions in the image.

[0105] The weight kernel generated using the Gaussian function is convolved with the image, and the convolution formula is:

[0106]

[0107] Among them: I ′ (x, y) is the pixel value of the blurred image, I(x + i, y + j) is the pixel value at the position (x + i, y + j) in the original image, G(i, j) represents the weight at the position (i, j), and k is the size of the convolution kernel, which is determined by the size of the Gaussian kernel.

[0108] 3) After Gaussian blurring, calculate the gradient of the Gaussian blurred image to detect edges. The gradient of the Gaussian blurred image is calculated through the Sobel operator, and the gradients in the horizontal and vertical directions are calculated respectively:

[0109]

[0110] Among them: G x and G y are the gradients in the horizontal and vertical directions respectively, and I is the image after Gaussian blurring.

[0111] The gradient magnitude G of the image is calculated by taking the square root of the sum of the squares of the gradients in the horizontal and vertical directions:

[0112]

[0113] The Gaussian blurring technique eliminates noise and avoids noise interfering with the accuracy of edge detection. The formula is as follows:

[0114] Among them: G is the gradient magnitude of the image, indicating the intensity of pixel brightness change.

[0115] The gradient direction θ of the image can also be calculated as:

[0116]

[0117] 4) After the gradient calculation is completed, non-maximum suppression and double-threshold techniques are usually used to accurately retain important edges and remove noise. The non-maximum suppression analyzes the gradient direction θ, retains the edges of local maxima, and suppresses the edges of non-maxima to reduce noise. The specific steps are as follows: Check adjacent pixels along the gradient direction. If the gradient value of the current pixel is not a local maximum, set the current pixel to 0 to remove the weaker edges. For the double-threshold processing, two thresholds are set, namely the high threshold T high and the low threshold T low to distinguish strong edges and weak edges. The specific process is as follows:

[0118] If the gradient value of a pixel is greater than the high threshold T high , it is considered a strong edge.

[0119] If the gradient value of a pixel is less than the low threshold T low , it is considered noise and discarded.

[0120] If the gradient value of a pixel is between the two thresholds, it is retained as a weak edge, but it is only retained if it is connected to a strong edge; otherwise, it is also removed.

[0121] 5) The histogram equalization technique enhances the contrast of the image and improves the clarity of the particle contour, especially playing an important role in an imaging environment with weak light or low contrast. Through the equalization process, the brightness distribution of the image becomes more uniform, thereby enhancing the details and edge features in the image.

[0122] First, the first step of histogram equalization is to calculate the number of pixel points for each gray value in the image to generate a gray histogram H(r k )

[0123] r k : the k-th gray level, with a value range of [0, L - 1], where L is the total number of gray levels. Usually, L = 256 for an 8-bit image. n k : represents the number of pixels with gray level r k

[0124] From this, the normalized histogram p(r k ) can be obtained, which represents the probability of each gray value occurring:

[0125]

[0126] where N is the total number of pixels in the image.

[0127] To achieve equalization, the cumulative distribution function CDF(r k ) of each gray level r k ​), which represents the cumulative probability from the lowest gray value to the current gray value. The formula is as follows:

[0128]

[0129] Where: p(r j ) is the normalized probability of the gray value r j . CDF(r k ) is the cumulative probability value of the gray level r k and all gray levels below it.

[0130] Through the cumulative distribution function CDF(r k ), the gray value r k in the original image is mapped to the equalized gray value s k . The mapping formula is as follows:

[0131] s k =(L - 1)×CDF(r k )

[0132] Where: s k is the equalized gray value. L - 1 is the maximum value of the gray level (for 8 - bit images, usually L - 1 = 255).

[0133] By redistributing the gray values, histogram equalization stretches the pixels originally concentrated in a relatively small gray range to the entire gray range. This adjustment process effectively improves the contrast of the image, especially enhancing the details in darker or brighter regions. Finally, the equalized gray value s k is applied to the original image to generate an image with enhanced contrast. This process is especially suitable for imaging environments with weak light or low contrast, such as the imaging environment described in the paragraph, making the contours of the particles clearer and further improving the accuracy of edge detection and shape analysis.

[0134] Step5: Spectral analysis of particles in insulating oil

[0135] When the particles are in the spectral analysis area of the No. 2 lower laminar flow channel, an excitation light source with adjustable wavelength and power, as well as a surface - enhanced Raman scattering (SERS) technology module, are used for excitation and signal acquisition, and the particles are analyzed by a Raman spectrometer to extract the chemical composition information of the particles.

[0136] Step6: Detection and analysis of the physical and chemical properties of particles in insulating oil

[0137] Wireless communication is achieved through the Wi-Fi module. The acquired image data and Raman spectroscopy data are sent to the computing chip NVIDIA Jetson Xavier NX for comprehensive processing and analysis, extracting and integrating the geometric features and chemical components of the particles, generating multi-dimensional detection reports by combining offline and online data, providing detailed analysis results on the particles in the insulating oil, and fulfilling the requirements of remote real-time monitoring and control for users based on the wireless network module.

[0138] This method effectively combines the physical and chemical detection of particles through an optimized multi-layer flow channel structure, providing an efficient and accurate comprehensive detection means for insulating oil particles.

Claims

1. A portable device for identifying and quickly detecting particles in insulating oil, characterized in that: include: A microfluidic imaging module, comprising an upper flow channel for focusing particles in insulating oil, two lower flow channels parallel to the upper flow channel, and a vertical connection channel for connecting the upper flow channel and the lower flow channel to ensure stable transmission of particles to the lower flow channels, wherein the two lower flow channels include a No. 1 lower flow channel for performing particle optical imaging and a No. 2 lower flow channel for performing particle spectral analysis; A flow control system for adjusting the flow rate and pressure of the insulating oil in the upper flow channel and the lower flow channel; A high-resolution imaging device for real-time imaging of particles in the No. 1 lower flow channel and transmitting image data to a data processing unit; Surface Enhanced Raman Scattering (SERS) module, used to excite and collect signals from particles in the No. 2 lower flow channel to obtain information on the chemical composition of the particles; A Raman spectrometer is used to analyze the molecular vibration information of particles in the spectral analysis area of ​​the No. 2 lower flow channel; a data processing unit is used to process the image data transmitted by the high-resolution imaging device to obtain a final image with clear particle outlines, and obtain the physical characteristics of the particles from the final image; It is also used to comprehensively analyze the physical characteristics and chemical composition information of particles to classify the particles and generate a test report, which shows the physical characteristics of the particles, the chemical composition information of the particles and the classification of the particles; The wireless communication module is used to transmit the physical property information, chemical composition information, image data, molecular vibration information and detection reports of particles to the remote device in real time.

2. The portable device for identifying and quickly detecting particles in insulating oil according to claim 1, characterized in that: The upper flow channel is a spiral micro-channel, and the particles are concentrated to the center of the upper flow channel by using secondary flow and inertial focusing effect.

3. The portable device for identifying and quickly detecting particles in insulating oil according to claim 2, characterized in that: The high-resolution imaging device uses a CMOS camera with a resolution of 1920×1200 pixels and is capable of capturing image data of particles in insulating oil at a speed of 155 frames per second.

4. The portable device for identifying and quickly detecting particles in insulating oil according to claim 1, characterized in that: The wireless communication module adopts a Wi-Fi module and performs data transmission via the MQTT protocol.

5. A method for identifying and quickly detecting particles in insulating oil, characterized in that: The portable device for identifying and quickly detecting particles in insulating oil according to any one of claims 1 to 4 comprises the following steps: Step 1: Pre-treat the insulating oil sample to remove bubbles and larger particles; Step 2: Inject the pretreated insulating oil sample into the upper flow channel and use the spiral microchannel to focus the particles; Step 3: Transfer the focused particles to the lower flow channel through the vertical connecting channel; Step 4: Use a high-resolution imaging device to perform real-time imaging of the particles in the No. 1 lower flow channel, and transmit the image data to a data processing unit. The data processing unit uses Gaussian blur and histogram equalization to process the image data to obtain a final image; Step 5: Perform spectral analysis on the particles in the No. 2 lower flow channel through a surface enhanced Raman scattering (SERS) module to obtain chemical composition information; Step 6: The data processing unit conducts a comprehensive analysis of the collected particle physical properties and chemical composition information, generates a test report, and transmits data through the wireless communication module.

6. The method for identifying and quickly detecting particles in insulating oil according to claim 5, characterized in that: The process of the data processing unit using Gaussian blur and histogram equalization to process the image data to obtain the final image is as follows: 1) Convert the image data into a grayscale image using the following formula: Gray=0.299×R+0.587×G+0.114×B Where: R represents the red component of the pixel, the value range is [0,255], G represents the green component of the pixel, the value range is [0,255], B represents the blue component of the pixel, the value range is [0,255], Gray is the converted gray value, the value range is [0,255]; 2) Use Gaussian blur method to eliminate the noise in the grayscale image obtained in 1) to obtain a Gaussian blurred image: Gaussian blur is the process of convolving a grayscale image with a Gaussian kernel function. It smoothes the image by weighted averaging of neighborhood pixels to reduce noise interference. The formula of the Gaussian kernel function is as follows: Where: G(x,y) represents the Gaussian kernel function, which represents the weight at the position (x,y), σ: the standard deviation of the Gaussian function, and x and y are the horizontal and vertical coordinates of the pixel in the image; The weights generated by the Gaussian function are used to convolve the grayscale image. The convolution formula is: Where: I ′ (xy) is the pixel value of the blurred image; I(x+i,y+j) is the pixel value at position (x+i,y+j) in the grayscale image; G(i,j) represents the weight at position (i,j), and k is the size of the convolution kernel; 3) Calculate the gradient of the Gaussian blur image to detect edges: The gradient of the Gaussian blur image is calculated by the Sobel operator, which calculates the gradient in the horizontal and vertical directions respectively: Where: G x and G y are the gradients in the horizontal and vertical directions respectively, and I is the Gaussian blur image; The gradient amplitude G of the Gaussian blur image is calculated by taking the square root of the sum of the squares of the horizontal and vertical gradients: The gradient direction θ of the Gaussian blur image is calculated as: 4) Use the non-maximum suppression method and the double threshold method to process the Gaussian blur image noise to obtain a clear particle outline image: Non-maximum suppression method: check the adjacent pixels along the gradient direction θ. If the gradient value of the current pixel is not the local maximum, set the current pixel to 0. Double threshold method: set two thresholds, namely high threshold T high and low threshold T low : If the gradient value of the current pixel is greater than the high threshold T high , it is considered a strong edge; If the gradient value of the current pixel is less than the low threshold T low , it is considered as noise and discarded; If the gradient value of the current pixel is between the two thresholds, the current pixel is retained as a weak edge, but it is only retained if it is connected to a strong edge, otherwise it is removed; 5) Using histogram equalization method to enhance the particle outline clear image obtained in 4) to obtain the final image; 5-1) Calculate the number of pixels of each gray value in the particle outline clear image and generate the gray histogram H(r k ), r k represents the kth gray level, and its value range is [0, L-1], where L is the total number of gray levels; Normalized histogram p(r k ) represents the probability of each gray value appearing: Where N is the total number of pixels in the image, n k Represents gray level r k The number of pixels; 5-2) Calculate each gray level r k The cumulative distribution function CDF(r k ), CDF(r k ) represents the gray value from the lowest to r k The cumulative probability of the gray value is as follows: Where: p(r j ) is the gray value r j The normalized probability of 5-3) The gray value r in the particle outline clear image k Mapped to the balanced gray value s k , the mapping formula is as follows: s k =(L-1)×CDF(r k ) Where: s k is the grayscale value after equalization, and L-1 is the maximum grayscale level.

Images