Destructive discharge spark detection method and system based on visual technology

Abstract

The present invention discloses a destructive discharge spark detection method and system based on visual technology. The destructive discharge spark detection method of the present invention comprises: using an industrial camera device to capture the spark image of the discharge area in real time at an adaptive frame rate, and using an adaptive time-space transformation algorithm to convert the captured spark image into a spark image after time-space transformation; performing feature extraction on the spark image after time-space transformation to obtain a spark intensity characteristic value, a spark shape change speed characteristic value, and an edge stability characteristic value; based on the spark intensity characteristic value, the spark shape change speed characteristic value, and the edge stability characteristic value, using a spark feature measurement algorithm to calculate a comprehensive measurement value of whether the spark is destructive, the comprehensive measurement value is compared with a preset threshold value, and the current spark is accurately judged whether it is destructive, providing the ability to respond and take measures in real time, avoiding potential safety risks, and ensuring safety and stability in industrial applications.

Description

Technical Field

[0001] The invention belongs to the technical field of visual detection, and in particular to a destructive discharge spark detection method and system based on visual technology. Background Art

[0002] Destructive discharge sparks are caused by the voltage exceeding the dielectric strength of the material, resulting in the discharge phenomenon of current passing through the air gap or the surface of the material. They exist in high-voltage equipment, electrical devices and industrial environments with special working conditions. The transient and high-temperature characteristics of discharge sparks can cause the combustion of surrounding materials or damage to materials, thereby causing unforeseen economic losses and safety risks. Due to the randomness and instantaneousness of discharge sparks, the detection of destructive discharge sparks becomes particularly difficult. Timely and accurate detection of destructive discharge sparks is crucial and of great significance to ensuring industrial safety and stable operation of equipment.

[0003] Traditional destructive discharge spark detection methods rely on the detection of abnormal changes in current and voltage or acoustic characteristics, and have the limitations of slow response speed, low detection accuracy, and great interference from environmental noise. With the rapid development of computer vision technology and image processing technology, destructive discharge spark detection methods based on vision technology provide a new way to solve the shortcomings of traditional destructive discharge spark detection methods, further improve the accuracy, real-time and stability of spark detection, and have broad application prospects in the field of industrial safety monitoring.

[0004] However, the above technology has at least the following technical problems: the generation and change of discharge sparks is a high-speed and dynamic process, and a single static detection method is difficult to accurately capture the occurrence and destructive characteristics of sparks; the existing spark detection technology only considers information in a single dimension of space or time, and has limited effects on spark detection and identification; the existing feature extraction methods lack robustness and stability in complex scenarios of spark features; there is a lack of effective comprehensive measurement methods to evaluate the destructiveness of sparks. Summary of the invention

[0005] The technical problem to be solved by the present invention is to overcome the defects of the above-mentioned prior art and provide a destructive discharge spark detection method and system based on visual technology to accurately judge whether the current spark is destructive, provide the ability to respond and take measures in real time, avoid potential safety risks, and ensure safety and stability in industrial applications.

[0006] To this end, the present invention adopts the following technical solution.

[0007] In a first aspect, the present invention provides a destructive discharge spark detection method based on visual technology, which comprises the steps of:

[0008] S1, uses industrial camera equipment to capture spark images of the discharge area in real time at an adaptive frame rate, and uses an adaptive spatiotemporal transformation algorithm to convert the captured spark images into spatiotemporal transformed spark images;

[0009] S2, feature extraction is performed on the spark image after time-space transformation to obtain the spark intensity characteristic value, the spark shape change speed characteristic value and the edge stability characteristic value;

[0010] S3, based on the spark intensity characteristic value, spark shape change speed characteristic value and edge stability characteristic value, use the spark characteristic measurement algorithm to calculate the comprehensive measurement value of whether the spark is destructive. The comprehensive measurement value is compared with the preset threshold to accurately judge whether the current spark is destructive.

[0011] Furthermore, in S1, the adaptive spatiotemporal transformation algorithm uses spatial cosine transformation and Gaussian attenuation function to transform and process the captured spark image in spatial and temporal dimensions.

[0012] Furthermore, in S1, the spatial cosine transform is used to transform the spatial features of the spark image from the spatial domain to the frequency domain, including calculating the inner product of the spatial coordinates and the spatial coordinates after the space-time transformation, and performing a cosine transform on the inner product.

[0013] Furthermore, in S1, the Gaussian attenuation function is used to perform weighted and smoothing processing on the spark image in the time dimension, focusing on the spark image in the current time window and ignoring historical data far away from the current time point, which helps to improve real-time performance and accuracy.

[0014] Furthermore, the S2 specifically includes:

[0015] Based on the spark image after time-space transformation, the spark intensity feature extraction algorithm is used to calculate the spark intensity feature value; the spark intensity feature extraction algorithm aims to accurately identify the spark intensity change by performing feature analysis on the spark image after time-space transformation, introduce an attenuation factor to characterize the local attenuation of the spark intensity, suppress the area with high spark brightness through the exponential attenuation function, enhance the stability of the overall spark feature, further adjust the nonlinear response of the spark intensity feature through power transformation, for sparks with different brightness distributions, by selecting appropriate nonlinear response adjustment parameters, it is possible to sensitively detect subtle brightness changes, and apply inverse power transformation to normalize the spark intensity feature value.

[0016] Furthermore, the S2 further specifically includes:

[0017] Based on the spark image after time-space transformation, the spark shape change speed characteristic value is calculated using the spark shape change speed detection algorithm; the spark shape change speed detection algorithm reflects the intensity of spark brightness change by calculating the gradient amplitude of the spark image after time-space transformation, and calculates the ratio of the change rate in different directions through the inverse tangent function, thereby determining the change direction of the spark shape, and comprehensively evaluating the complexity and speed of the spark shape change.

[0018] Furthermore, the S2 further specifically includes:

[0019] Based on the spark image after time-space transformation, the edge stability characteristic value is calculated using an edge stability detection algorithm; the edge stability detection algorithm obtains the change amount by comparing the difference in pixel values ​​between the current time point and the previous time point, introduces a nonlinear enhancement parameter to enhance the responsiveness to subtle changes, and further processes the nonlinearly enhanced change amount through a hyperbolic cosine function, providing a smooth and sensitive detection capability for edge changes, calculates the product of the nonlinear enhancement term and the hyperbolic cosine enhancement term of all pixel points and takes the average, and uses a logarithmic function to smooth the cumulative average change to reduce the impact of extreme changes on edge stability detection.

[0020] Furthermore, in S3, the spark feature measurement algorithm reflects the synergy between the spark intensity feature, the shape change rate feature and the edge stability feature by multiplying the spark intensity feature value, the spark shape change rate feature and the edge stability feature, thereby balancing the influence of each feature value and effectively reflecting the overall state of the spark.

[0021] Furthermore, in S3, in order to further balance the impact of various features on the comprehensive metric value, a hyperbolic tangent function is used to smooth the ratio between spark intensity and shape change speed and edge stability, thereby avoiding the impact of extreme situations on the comprehensive metric value and enhancing the overall judgment ability of spark behavior.

[0022] In a second aspect, the present invention provides a destructive discharge spark detection system based on visual technology, which is used to implement the above-mentioned destructive discharge spark detection method, and comprises:

[0023] Time-space transformation unit: uses industrial camera equipment to capture spark images of the discharge area in real time at an adaptive frame rate, and uses an adaptive time-space transformation algorithm to convert the captured spark images into spark images after time-space transformation;

[0024] Feature extraction unit: performs feature extraction on the spark image after time-space transformation to obtain the spark intensity characteristic value, the spark shape change speed characteristic value and the edge stability characteristic value;

[0025] Spark destructiveness judgment unit: Based on the spark intensity characteristic value, spark shape change speed characteristic value and edge stability characteristic value, the spark characteristic measurement algorithm is used to calculate the comprehensive measurement value of whether the spark is destructive. The comprehensive measurement value is compared with the preset threshold to accurately judge whether the current spark is destructive.

[0026] The beneficial effects of the present invention are as follows: the present invention improves the real-time and accuracy of spark image processing through an adaptive space-time transformation algorithm, reduces computational complexity, and enhances the sensitive response capability to spark state changes; the present invention uses a spark feature measurement algorithm to calculate a comprehensive measurement value of whether a spark is destructive, and the comprehensive measurement value is compared with a preset threshold to accurately determine whether the current spark is destructive, providing the ability to respond and take measures in real time, avoiding potential safety risks, and ensuring safety and stability in industrial applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is a flow chart of a destructive discharge spark detection method based on visual technology of the present invention;

[0028] Figure 2 This is an architectural diagram of a destructive discharge spark detection system based on visual technology of the present invention. DETAILED DESCRIPTION

[0029] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0030] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0031] Refer to the attached Figure 1 , which shows a flow chart of a destructive discharge spark detection method based on visual technology provided by an embodiment of the present invention, the method comprising the following steps:

[0032] S1. Use industrial camera equipment to capture spark images of the discharge area in real time at an adaptive frame rate, and use an adaptive spatiotemporal transformation algorithm to convert the captured spark images into spatiotemporal transformed spark images.

[0033] The adaptive spatiotemporal transformation algorithm adopts spatial cosine transformation and Gaussian decay function to transform and process the spark image in spatial and temporal dimensions.

[0034] The spatial cosine transform is used to transform the spatial features of the spark image from the spatial domain to the frequency domain. Specifically, the inner product of the spatial coordinates and the spatial coordinates after the space-time transform is calculated, and the inner product is cosine transformed.

[0035] The Gaussian decay function is used to perform weighted and smoothing processing on the spark image in the time dimension, focusing on the spark image in the current time window and ignoring the historical data far away from the current time point, which helps to improve real-time performance and accuracy.

[0036] The calculation formula of the eigenvalue in the spark image after time-space transformation is:

[0037]

[0038] in, Represents the spatial coordinates in the spark image after time-space transformation and time coordinates The eigenvalues ​​of Indicates at time Time and space location The pixel value on ; and are the width and height of the spark image, in pixels, respectively; represents spatial cosine transform, which is used to transform the spatial features of spark images from spatial domain to frequency domain; Represents a constant used to convert angles to radians; It represents the inner product of the spatial coordinates and the spatial coordinates after space-time transformation; Indicates the width of the spark image, used as a normalization factor to ensure that the frequency calculation is consistent with the image size; Represents the Gaussian decay function part, which is used to perform weighted smoothing on the time dimension; represents the natural exponential function, used for exponential decay; represents the square of the time difference, representing the spark image capture time The time point after time and space transformation The distance can achieve smooth weighted processing in the time dimension; Represents the denominator of the Gaussian function, which is used to control the width of time smoothing; is the standard deviation of the Gaussian decay function, which controls the degree of smoothing of the time weighting and is set according to the time scale of spark variation.

[0039] S2. Perform feature extraction on the spark image after time-space transformation to obtain the spark intensity characteristic value, the spark shape change speed characteristic value and the edge stability characteristic value.

[0040] Based on the spark image after time-space transformation, the spark intensity feature value is calculated using the spark intensity feature extraction algorithm.

[0041] The spark intensity feature extraction algorithm aims to accurately identify the change of spark intensity by performing feature analysis on the spark image after time-space transformation. An attenuation factor is introduced to characterize the local attenuation of the spark intensity. The area with high spark brightness is suppressed by the exponential attenuation function to enhance the stability of the overall spark characteristics. The nonlinear response of the spark intensity characteristics is further adjusted by power transformation. For sparks with different brightness distributions, subtle brightness changes can be sensitively detected by selecting appropriate nonlinear response adjustment parameters, and the spark intensity characteristic values ​​are normalized by applying inverse power transformation.

[0042] The calculation formula of the spark intensity characteristic value is:

[0043]

[0044] in, Indicates at time Spark intensity characteristic value at the moment; represents a double summation of all spatial coordinates; Represents the spatial coordinates in the spark image after time-space transformation and time coordinates The eigenvalues ​​of Represents the attenuation factor, which is used to characterize the local attenuation of the spark intensity. The exponential attenuation function is used to suppress the area with high spark brightness and enhance the stability of the overall characteristics of the spark. represents the attenuation control parameter, which is used to control the nonlinear response of the spark intensity. ; It represents the nonlinear response adjustment parameter, which is used to adjust the nonlinear response degree of the spark intensity; Used to normalize the spark intensity feature value.

[0045] Based on the spark image after time-space transformation, the spark shape change speed characteristic value is calculated using the spark shape change speed detection algorithm.

[0046] The spark shape change speed detection algorithm reflects the intensity of spark brightness change by calculating the gradient amplitude of the spark image after time-space transformation, and calculates the ratio of the change rate in different directions through the inverse tangent function to determine the change direction of the spark shape, and comprehensively evaluates the complexity and speed of the spark shape change.

[0047] The calculation formula for the spark shape change speed is:

[0048]

[0049] in, Indicates at a point in time The characteristic value of the spark shape change speed on represents a double summation of all spatial coordinates; Represents the spatial coordinates in the spark image after time-space transformation and time coordinates The eigenvalues ​​of and Respectively expressed in space coordinates and The pixel value gradient in the direction, that is, the rate of change; It represents the gradient amplitude, reflecting the speed of change of the spark shape; The absolute value of the arc tangent value representing the change in the gradient direction at the pixel point is used to capture the direction change information of the spark edge. The direction of change is determined by calculating the inclination angle of the spark edge. The absolute value sign ensures that the positive or negative direction change does not affect the calculation of the spark shape change speed.

[0050] Based on the spark image after time-space transformation, the edge stability feature value is calculated using the edge stability detection algorithm.

[0051] The edge stability detection algorithm obtains the change amount by comparing the difference in pixel values ​​between the current time point and the previous time point, introduces a nonlinear enhancement parameter to enhance the responsiveness to subtle changes, and further processes the nonlinearly enhanced change amount through a hyperbolic cosine function, providing a smooth and sensitive detection capability for edge changes. The product of the nonlinear enhancement term and the hyperbolic cosine enhancement term of all pixel points is calculated and averaged, and the cumulative average change is smoothed using a logarithmic function to reduce the impact of extreme changes on edge stability detection.

[0052] The calculation formula of the edge stability eigenvalue is:

[0053]

[0054] in, Indicates at time The marginal stability eigenvalue at the moment; represents the natural logarithm function, which is used for smoothing to reduce the impact of extreme values ​​on edge stability assessment; represents the normalization coefficient, where and Respectively represent the number of feature points in the horizontal and vertical directions; Indicates time Moment feature points The change in the value of with respect to the previous time point captures the change in the spark edge between two consecutive time frames; represents the nonlinear enhancement parameter, which is used to enhance the responsiveness to subtle changes; represents the hyperbolic cosine function, introducing a smooth nonlinear response; Represents the parameter that adjusts the response of the hyperbolic cosine function, which is used to control the sensitivity of the hyperbolic cosine function to changes.

[0055] S3. Based on the spark intensity characteristic value, the spark shape change speed characteristic value, and the edge stability characteristic value, a spark characteristic measurement algorithm is used to calculate a comprehensive measurement value of whether the spark is destructive. The comprehensive measurement value is compared with the preset threshold to accurately determine whether the current spark is destructive.

[0056] The spark feature measurement algorithm reflects the synergy between the spark intensity feature, the shape change speed feature and the edge stability feature by multiplying the spark intensity feature value, the spark shape change speed feature and the edge stability feature. It can balance the influence of each feature and effectively reflect the overall state of the spark. In order to further balance the influence of each feature on the measurement value, the hyperbolic tangent function is used to smooth the ratio between the spark intensity and the shape change speed and the edge stability to avoid the influence of extreme situations on the comprehensive measurement value and enhance the overall judgment ability of the spark behavior.

[0057] The formula for calculating the composite measure of whether a spark is destructive is:

[0058]

[0059] in, Indicates at time A comprehensive measure of whether the moment is destructive to the spark; Indicates at time Spark intensity characteristic value at the moment; Indicates at a point in time The characteristic value of the spark shape change speed on Indicates at time The marginal stability eigenvalue at the moment; It represents the square root of the product of spark intensity, shape change speed and edge stability. Through nonlinear combination, it comprehensively reflects the overall characteristics of the spark. The square root operation is used for smoothing. The hyperbolic tangent function representing the ratio of the spark intensity to the sum of the shape change rate and the edge stability is used to smooth the ratio of the spark intensity to the shape change rate and the edge stability.

[0060] By calculating the comprehensive measurement value of whether the spark is destructive in real time and comparing it with the preset threshold, it is accurately determined whether the current spark is destructive. The process is as follows:

[0061]

[0062] in, It represents a preset threshold value, which can be set according to the specific implementation scenario and is not limited here.

[0063] By comparing with the preset threshold, it can determine in real time whether the current spark state is destructive, ensuring that in actual applications, it can respond quickly and take corresponding measures to avoid potential safety risks.

[0064] Refer to the attached Figure 2 , which shows an architecture diagram of a destructive discharge spark detection system based on visual technology provided by another embodiment of the present invention, which is used to implement the above-mentioned destructive discharge spark detection method, and includes:

[0065] Time-space transformation unit: uses industrial camera equipment to capture spark images of the discharge area in real time at an adaptive frame rate, and uses an adaptive time-space transformation algorithm to convert the captured spark images into spark images after time-space transformation;

[0066] Feature extraction unit: performs feature extraction on the spark image after time-space transformation to obtain the spark intensity characteristic value, the spark shape change speed characteristic value and the edge stability characteristic value;

[0067] Spark destructiveness judgment unit: Based on the spark intensity characteristic value, spark shape change speed characteristic value and edge stability characteristic value, the spark characteristic measurement algorithm is used to calculate the comprehensive measurement value of whether the spark is destructive. The comprehensive measurement value is compared with the preset threshold to accurately judge whether the current spark is destructive.

[0068] It should be noted that each unit in the above-mentioned destructive discharge spark detection system based on visual technology can be implemented in whole or in part by software, hardware and a combination thereof. The above-mentioned units can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above-mentioned modules. For the specific definition of a destructive discharge spark detection system based on visual technology, please refer to the definition of a destructive discharge spark detection method based on visual technology above. The two have the same functions and effects, which will not be repeated here.

[0069] The beneficial effects of the technical solution of the present invention are:

[0070] 1. The real-time and accuracy of spark image processing are improved through the adaptive space-time transformation algorithm, the computational complexity is reduced, and the sensitive response to spark state changes is enhanced.

[0071] 2. The detection accuracy of spark intensity changes is improved by using the spark intensity feature extraction algorithm, especially under spark conditions with different brightness distributions, which can accurately identify local changes in spark intensity and enhance the comprehensive judgment ability of spark characteristics.

[0072] 3. The spark shape change speed detection algorithm is used to effectively capture the change speed and direction of the spark shape, which improves the detection accuracy of the dynamic changes of the spark shape, especially the ability to identify complex shape change behaviors.

[0073] 4. The edge stability detection algorithm is used to improve the detection accuracy of the spark edge stability feature, especially showing good sensitivity and stability when dealing with subtle edge changes, which helps to accurately evaluate the overall stability state of the spark.

[0074] 5. Perform a square root operation on the product of the spark intensity, shape change speed, and edge stability eigenvalues ​​to comprehensively reflect the overall characteristics of the spark. Further use the hyperbolic tangent function for smoothing to balance the impact of each feature on the metric value and avoid extreme situations affecting the results. By integrating the synergistic effect of multiple features, the overall state of the spark can be effectively reflected to ensure the stability and reliability of the spark destructiveness judgment.

[0075] 6. By comparing with the preset threshold, it can accurately determine whether the current spark is destructive, providing the ability to respond and take measures in real time, avoiding potential safety risks, and ensuring safety and stability in industrial applications.

[0076] The sequence of the embodiments of the present invention is for description only and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0077] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A destructive discharge spark detection method based on visual technology, characterized in that: Includes steps: S1, uses industrial camera equipment to capture spark images of the discharge area in real time at an adaptive frame rate, and uses an adaptive spatiotemporal transformation algorithm to convert the captured spark images into spatiotemporal transformed spark images; S2, feature extraction is performed on the spark image after time-space transformation to obtain the spark intensity characteristic value, the spark shape change speed characteristic value and the edge stability characteristic value; S3, based on the spark intensity characteristic value, the spark shape change speed characteristic value and the edge stability characteristic value, a spark characteristic measurement algorithm is used to calculate a comprehensive measurement value of whether the spark is destructive, and the comprehensive measurement value is compared with a preset threshold to accurately determine whether the current spark is destructive; In S2, the calculation formula of the spark intensity characteristic value is: in, Indicates at time Spark intensity characteristic value at the moment; represents a double summation of all spatial coordinates; Represents the spatial coordinates in the spark image after time-space transformation and time coordinates The eigenvalues ​​of represents the attenuation factor; represents the attenuation control parameter, ; represents the nonlinear response adjustment parameter; Used to normalize the spark intensity characteristic value; The calculation formula for the characteristic value of the spark shape change speed is: in, Indicates that at a point in time The characteristic value of the spark shape change speed on represents a double summation of all spatial coordinates; Represents the spatial coordinates in the spark image after time-space transformation and time coordinates The eigenvalues ​​of and Respectively expressed in space coordinates and The pixel value gradient in the direction; represents the gradient amplitude; Indicates the absolute value of the arc tangent of the gradient direction change at the pixel point; The calculation formula of the edge stability eigenvalue is: in, Indicates at time The marginal stability eigenvalue at the moment; represents the natural logarithm function; represents the normalization coefficient, where and Respectively represent the number of feature points in the horizontal and vertical directions; Indicates time Moment feature points The change of the value from the previous time point; represents the nonlinear enhancement parameter; represents the hyperbolic cosine function; Represents the parameters that adjust the response of the hyperbolic cosine function; In S3, the calculation formula for the comprehensive measurement value of whether the spark is destructive is: in, Indicates at time A comprehensive measure of whether the moment is destructive to the spark; Indicates at time Spark intensity characteristic value at the moment; Indicates that at a point in time The characteristic value of the spark shape change speed on Indicates at time The marginal stability eigenvalue at the moment; It represents the square root of the product of spark intensity, shape change speed and edge stability; Hyperbolic tangent function representing the ratio of spark intensity to the sum of shape change rate and edge stability.

2. The destructive discharge spark detection method according to claim 1, characterized in that: In S1, the adaptive spatiotemporal transformation algorithm uses spatial cosine transformation and Gaussian attenuation function to transform and process the captured spark image in spatial and temporal dimensions.

3. The destructive discharge spark detection method according to claim 2, characterized in that: In S1, the spatial cosine transform is used to transform the spatial features of the spark image from the spatial domain to the frequency domain, including calculating the inner product of the spatial coordinates and the spatial coordinates after the space-time transform, and performing a cosine transform on the inner product.

4. The destructive discharge spark detection method according to claim 2, characterized in that: In S1, the Gaussian decay function is used to perform weighted and smoothing processing on the spark image in the time dimension, focusing on the spark image in the current time window and ignoring historical data far away from the current time point.

5. The destructive discharge spark detection method according to claim 1, characterized in that: The S2 specifically includes: Based on the spark image after time-space transformation, the spark intensity feature extraction algorithm is used to calculate the spark intensity feature value; the spark intensity feature extraction algorithm aims to introduce an attenuation factor by performing feature analysis on the spark image after time-space transformation, suppress the area with high spark brightness through an exponential attenuation function, enhance the stability of the overall spark characteristics, further adjust the nonlinear response of the spark intensity feature through power transformation, for sparks with different brightness distributions, detect subtle brightness changes by selecting appropriate nonlinear response adjustment parameters, and apply inverse power transformation to normalize the spark intensity feature value.

6. The destructive discharge spark detection method according to claim 1, characterized in that: The S2 specifically includes: Based on the spark image after time-space transformation, the spark shape change speed characteristic value is calculated using the spark shape change speed detection algorithm; the spark shape change speed detection algorithm reflects the intensity of the spark brightness change by calculating the gradient amplitude of the spark image after time-space transformation, and calculates the ratio of the change rate in different directions through the inverse tangent function, thereby determining the change direction of the spark shape.

7. The destructive discharge spark detection method according to claim 1, characterized in that: The S2 specifically includes: Based on the spark image after time-space transformation, the edge stability characteristic value is calculated using an edge stability detection algorithm; the edge stability detection algorithm obtains the change amount by comparing the difference in pixel values ​​between the current time point and the previous time point, introduces a nonlinear enhancement parameter to enhance the responsiveness to subtle changes, further processes the change amount after nonlinear enhancement through a hyperbolic cosine function, calculates the product of the nonlinear enhancement term and the hyperbolic cosine enhancement term of all pixel points and takes the average, and uses a logarithmic function to smooth the cumulative average change to reduce the impact of extreme changes on edge stability detection.

8. The destructive discharge spark detection method according to claim 1, characterized in that: In S3, the spark feature measurement algorithm reflects the synergy between the spark intensity feature, the shape change speed feature and the edge stability feature by multiplying the spark intensity feature value, the spark shape change speed feature and the edge stability feature value.

9. The destructive discharge spark detection method according to claim 8, characterized in that: In S3, in order to further balance the influence of each feature on the comprehensive metric value, a hyperbolic tangent function is used to smooth the ratio between the spark intensity and the shape change speed and edge stability.

10. A destructive discharge spark detection system based on visual technology, used to implement the destructive discharge spark detection method according to any one of claims 1 to 9, characterized in that: include: Time-space transformation unit: uses industrial camera equipment to capture spark images of the discharge area in real time at an adaptive frame rate, and uses an adaptive time-space transformation algorithm to convert the captured spark images into spark images after time-space transformation; Feature extraction unit: performs feature extraction on the spark image after time-space transformation to obtain the spark intensity characteristic value, the spark shape change speed characteristic value and the edge stability characteristic value; Spark destructiveness judgment unit: Based on the spark intensity characteristic value, spark shape change speed characteristic value and edge stability characteristic value, the spark characteristic measurement algorithm is used to calculate the comprehensive measurement value of whether the spark is destructive. The comprehensive measurement value is compared with the preset threshold to accurately judge whether the current spark is destructive.

Images