A method for detecting leakage in the valve hall of a controllable commutator valve based on a moving laser.

Abstract

This invention discloses a method for detecting leakage in the valve hall of a controllable commutator valve based on a moving laser. The method utilizes a pan-tilt unit mounted on the inner wall of the valve hall, capable of moving along a preset track. A point laser and an image acquisition device are mounted on the pan-tilt unit to achieve dynamic inspection of the entire area. Background images are acquired when the laser is off. After the laser is turned on, multiple frames are continuously acquired within a preset time. Highlighted areas caused by laser irradiation are extracted through pixel-level interpolation processing, and laser trajectory interference is removed using contour recognition. Subsequently, connected component analysis is performed on the bright areas to extract candidate spot regions, and their roundness, grayscale profile fit goodness of fit, and spectral energy characteristics are calculated. The presence of droplet reflection is determined from multiple dimensions, including geometric morphology, grayscale distribution, and texture characteristics, thereby identifying leakage. This method effectively improves detection accuracy and coverage, and is suitable for high-reliability automated leakage monitoring in complex environments.

Description

Technical Field

[0001] This invention belongs to the field of power system leakage monitoring technology, specifically relating to a method for detecting leakage in the valve hall of a controllable commutator valve based on a moving laser. Background Technology

[0002] The controllable commutation converter valve hall is a critical site in the receiving-end controllable commutation converter station of HVDC transmission, providing a stable temperature and humidity operating environment for important equipment such as the controllable commutation converter valves. With increasing years of operation, the ceiling and walls of the valve hall often suffer from metal corrosion, organic material aging, and localized structural deformation due to long-term rain, snow, sun, and wind exposure, resulting in cracks in the ceiling. Once cracks appear in the ceiling, rain or snow can easily lead to water leakage in the valve hall, thereby affecting the safe operation of the converter valves and the HVDC transmission project within the hall. Water leakage failures have occurred multiple times in engineering practice, highlighting the urgent need for early and reliable identification and warning of leakage risks.

[0003] To prevent water leakage in the valve hall ceiling from affecting the operation of the controllable commutator valve and DC transmission line, it is necessary to adopt certain technical means to identify water leakage faults in the valve hall in advance. Existing technology CN202311048594.9 proposes a spatial liquid dripping detection system, method, and computer equipment, which proposes a method and equipment for making water leakage events explicit in space using a line laser and realizing water leakage fault monitoring through image analysis. Because a line laser is used for scanning, it has the ability to scan a spatial area, and the resolution is basically unaffected by distance. However, this method has two drawbacks: firstly, the line laser used is relatively expensive; secondly, a set of line lasers can only cover a certain area and cannot cover the entire commutator valve hall.

[0004] See also Chinese patent CN113793303A, which proposes using a fixed laser pointer to continuously illuminate a pipe in a leak scenario. A camera device captures video of the laser path, and the leak is identified by judging bright spots through keyframe extraction, grayscale conversion, and threshold segmentation. However, this method has the following limitations: 1. The laser pointer is fixed in position, which can only detect one laser path in the pipe and cannot cover a larger or more complex monitoring area; 2. The laser is always on, which can easily be confused with ambient light and other reflective objects, resulting in a high false alarm rate; 3. It is based only on a single frame of bright area judgment, without multi-frame redundancy or dynamic analysis, and is prone to missed / false alarms due to instantaneous changes in illumination or noise spots.

[0005] Therefore, existing technologies still have shortcomings in terms of coverage, cost, detection reliability, and anti-interference ability, and further technical means are needed to improve the leakage detection effect in large-area and complex environments of valve halls. Summary of the Invention

[0006] The purpose of this invention is to overcome the defects of the prior art and provide a method for detecting leakage in the valve chamber of a controllable commutator valve based on a moving laser.

[0007] The objective of this invention can be achieved through the following technical solutions:

[0008] This invention provides a method for detecting leakage in the valve chamber of a controllable commutator valve based on a moving laser, comprising the following steps:

[0009] A movable gimbal is installed along a preset track on the inner wall of the controllable commutation valve chamber. A point laser and an image acquisition device are respectively configured on the movable gimbal.

[0010] Control the movable gimbal to move to the preset detection position;

[0011] With the point laser off, the first image of that location is acquired by the image acquisition device;

[0012] Turn on the point laser and control the image acquisition device to acquire multiple second images at a preset frequency within a preset time.

[0013] Pixel-level difference processing is performed between the first image and multiple second images to obtain a difference image;

[0014] The difference image is subjected to a linear contour extraction algorithm for laser trajectory recognition and separation to obtain the bright area after removing the laser trajectory;

[0015] Based on the highlighted regions, connected component analysis is performed to extract candidate spot regions, and spatial water droplet features of the candidate spot regions are extracted.

[0016] Determine if there is a water leakage fault based on the characteristics of water droplets in the space;

[0017] Control the movable gimbal to move to the next preset detection position until all preset detection positions are completed.

[0018] Furthermore, the step of performing pixel-level interpolation processing on the first image and multiple second images to obtain a difference image specifically includes:

[0019] Convert the first image and multiple second images into grayscale images to obtain the first grayscale image. With multiple second grayscale images ,in, Indicates the first i The grayscale image of the second image;

[0020] Based on the grayscale values ​​of each pixel in each second grayscale image, calculate the average grayscale value of each second grayscale image, and select the second grayscale image with the highest average grayscale value as the third grayscale image. ;

[0021] Based on the first grayscale image With the third grayscale image Pixel-level interpolation is performed to obtain the interpolated image. .

[0022] Furthermore, the first grayscale image With the third grayscale image Pixel-level interpolation is performed to obtain the interpolated image. The formula is:

[0023]

[0024] in, This indicates the difference image at the pixel point. grayscale value at that location These represent the first grayscale image. With the third grayscale image At pixel The grayscale value at that location.

[0025] Furthermore, the step of using a linear contour extraction algorithm to identify and separate the laser trajectory in the difference image to obtain the highlighted area after removing the laser trajectory specifically includes:

[0026] For difference images Perform a light Gaussian smoothing process to obtain the smoothed difference image. ;

[0027] For the smoothed difference image The Canny edge detection algorithm is used to extract edge contours and obtain an edge map. ;

[0028] In the edge map The probabilistic Hough transform is used to identify the most salient line and obtain the polar coordinate parameters of the line. ;

[0029] Based on the polar coordinate parameters of the straight line obtained from the identification , compared to the typical half-width of the laser trajectory Constructing a laser trajectory mask The formula is:

[0030]

[0031] in, This is the typical half-width of the laser trajectory, which is half the width of the bright line formed by the laser in a grayscale image.

[0032] Based on the laser trajectory mask For difference images Perform a masking operation to obtain an image of the highlighted area after removing the laser trajectory:

[0033]

[0034] in, Highlighting the image at pixel points The grayscale value at that location.

[0035] Furthermore, the step of extracting candidate spot regions based on connected component analysis of the highlighted regions specifically includes:

[0036] For the highlighted area image Perform binarization to obtain a binary image. :

[0037]

[0038] in, Represents the binary image at pixel points The value at that location, A preset grayscale threshold is used to distinguish between bright areas that may be water droplets and the background;

[0039] For binary images Using the eight-neighbor connectivity analysis method, for each pixel with a value of 1 in the binary image, region growing is performed based on the connectivity relationships of its eight neighboring pixels to obtain the set of all connected regions. Each of them This represents a candidate spot region.

[0040] Furthermore, the extraction of spatial water droplet features from the candidate spot region specifically includes:

[0041] For each candidate spot region The roundness characteristics, intensity profile fitting characteristics, and local spectral analysis characteristics are calculated sequentially to form the spatial droplet characteristics of the candidate spot region.

[0042] Furthermore, the roundness feature is calculated as follows:

[0043] For candidate spot regions Extract its area With the perimeter of the area The roundness characteristic is calculated using the following formula:

[0044]

[0045] in, Indicates candidate spot regions pixel area Indicates candidate spot regions The perimeter, that is, the length of the line connecting the boundary pixels; Indicates candidate spot regions The roundness characteristic reflects how close the outline of the candidate spot region is to an ideal circle.

[0046] Furthermore, the calculation process for the intensity profile fitting features includes:

[0047] For each candidate spot region Extract its center pixel coordinates ;

[0048] Using center pixel coordinates Based on this, grayscale profile curves are extracted along the horizontal and vertical directions respectively to obtain the actual horizontal grayscale profile curve. Vertical grayscale profile curve ;

[0049] Extracted actual horizontal grayscale profile curve Vertical grayscale profile curve Gaussian function fitting was performed separately to obtain the horizontal Gaussian fitting curve and the vertical Gaussian fitting curve.

[0050] Calculate the goodness of fit between the Gaussian fitted curve and the actual grayscale profile curve, and define it as the coefficient of determination:

[0051]

[0052] in, Indicates the actual profile curve at position grayscale value, The corresponding Gaussian fitted value is... This represents the average grayscale value of the actual profile curve. The goodness-of-fit coefficient;

[0053] The average of the goodness-of-fit coefficients in the horizontal and vertical directions is used as the candidate spot region. Strength profile fitting characteristics .

[0054] Furthermore, the calculation process for the local spectral analysis features includes:

[0055] For each candidate spot region In the difference image Candidate spot regions between China and Israel Extract a fixed-size local image patch centered on the target area. ;

[0056] For local image patches Perform a two-dimensional fast Fourier transform to obtain local image patches. Spectrum The calculation formula is as follows:

[0057]

[0058] in, For local image patches at pixels grayscale value at that location The imaginary unit, Frequency coordinates in a spectrum diagram The complex spectrum value; N is the local image patch. The side length;

[0059] For the spectrum Take the amplitude spectrum to obtain the spectral energy map:

[0060]

[0061] in, Represents frequency coordinates Spectral amplitude at that location Frequency coordinates in a spectrum diagram Complex spectrum values The real part, Frequency coordinates in a spectrum diagram Complex spectrum values The imaginary part;

[0062] Based on the spectral energy map, the concentration of spectral energy is calculated as a characteristic of local spectral analysis. Defined as the proportion of energy in the low-frequency region to the total energy:

[0063]

[0064] in, Candidate spot regions Local spectral analysis characteristics, The set of frequency coordinates in the low-frequency region satisfies the following conditions:

[0065]

[0066] in, The low-frequency radius threshold is set to a value of [value to be filled in]. , indicating the boundary of the low-frequency range.

[0067] Furthermore, the step of determining whether a water leakage fault exists based on the characteristics of spatial water droplets specifically includes:

[0068] For each candidate spot region Based on candidate spot regions Spatial water droplet characteristics Each is compared with the preset roundness threshold. Intensity profile fitting threshold With local spectral analysis threshold The comparison is performed, and when all three features are greater than the threshold, the candidate spot region is determined to be a real water droplet spot;

[0069] If at least one candidate spot area is a real water droplet spot, then it is determined that there is a water leakage fault at the current detection location.

[0070] Compared with the prior art, the present invention has the following advantages:

[0071] (1) This invention uses a movable gimbal installed along a preset track on the inner wall of the valve hall. The point laser can move or rotate along the track to scan the entire valve hall area sequentially. Compared with existing technologies where laser pointers or line lasers can only cover a single path or a limited fixed area, the movable scanning method of this invention not only achieves full coverage and blind spot monitoring of the valve hall space, significantly improving the comprehensiveness and reliability of leakage detection, but also greatly reduces the number of lasers and related hardware, thus lowering the overall system cost, because only one movable device is needed to complete the detection of a large area. At the same time, the flexibility of the movable gimbal in translation and rotation also enables this invention to adapt to the complex structural layout within the valve hall, providing higher deployment efficiency and economy for on-site implementation.

[0072] (2) This invention acquires multiple second image sequences at a preset frequency after the laser is turned on, and then selects the frame with the largest average gray value as the analysis object after gray-scale conversion. This solves the problem that the timing of single-frame acquisition is difficult to synchronize with the appearance of water droplets. The moment when water droplets and laser beams meet is often extremely short. By acquiring multiple redundant frames and selecting the frame with the strongest gray value, it can not only ensure that the bright spots generated by the scattering of liquid droplets are captured, but also eliminate low-contrast interference caused by ambient light or noise, thereby significantly improving the hit rate and robustness of water leakage detection.

[0073] (3) Existing technologies typically perform threshold segmentation on single-frame laser images to identify bright areas, making it difficult to distinguish between bright spots caused by droplet scattering and false spots generated by changes in ambient light, reflections from equipment surfaces, or fixed laser bands. Static or semi-static optical interferences such as background textures, fluctuations in sunlight illuminance, and wall reflections can generate or amplify non-leaking bright spots in a single frame, leading to false alarms. This invention acquires images in both laser-off and laser-on states sequentially and performs pixel-level differences between them, retaining only the grayscale increments caused by laser irradiation and droplet scattering in the on state, fundamentally eliminating the influence of static backgrounds and fixed spots. Subsequently, Gaussian smoothing, Canny edge detection, and Hough linear transformation are applied to the difference image to accurately identify and shield the main laser trajectory, ultimately obtaining a bright remaining image containing only droplet scattering spots. This solves the technical problems of false alarms caused by static or semi-static light source interference and the inability to eliminate laser band artifacts in existing methods, significantly improving the contrast between signal and noise, reducing the false alarm rate, and ensuring efficient and reliable detection of truly leaking droplet bright spots.

[0074] (4) Existing technologies often rely solely on threshold segmentation after obtaining bright pixels, treating all bright spots as leak points and triggering alarms. As a result, non-droplet spots such as wall reflections, stains, and rust are also identified as leaks, leading to a high false alarm rate. This invention first performs eight-neighbor connected component analysis on the binarized bright areas, aggregating those discrete and regular bright pixel blocks into several candidate spot regions. Subsequently, it calculates three types of spatial water droplet features—roundness, grayscale profile fitting goodness, and low-spectral energy proportion—on each connected region, and uses the threshold values ​​of these three features to jointly determine whether it is a real droplet spot. Connected component analysis ensures that the detection object is a complete bright spot region, while multi-dimensional feature extraction utilizes the three physical properties of water droplets—geometric, light intensity, and texture—to distinguish real water droplet reflections from other interferences. The aforementioned distinguishing technical features utilize the typical structural, optical, and textural patterns of droplet reflection, effectively eliminating bright interference spots caused by non-droplet factors. This solves the problem of lack of substantial discrimination in existing technologies, which leads to a high false alarm rate. It achieves high accuracy and robustness in leak point identification in complex valve hall environments. Attached Figure Description

[0075] Figure 1 This is a flowchart of the water leakage detection method according to an embodiment of the present invention;

[0076] Figure 2 This is a schematic diagram of the arrangement of laser light sources on a linear track according to an embodiment of the present invention;

[0077] Figure 3 This is a schematic diagram of the laser source movement method of the serpentine track according to an embodiment of the present invention;

[0078] Figure 4This is a schematic diagram of a laser light source horizontal rotation arrangement method according to an embodiment of the present invention;

[0079] Figure 5 This is a schematic diagram of the arrangement method for horizontal rotation and translation of the laser light source according to an embodiment of the present invention. Detailed Implementation

[0080] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0081] Example 1:

[0082] This embodiment provides a method and device for detecting water leakage in the controllable commutation valve hall of a converter station based on moving laser scanning. The basic principle is to use a point laser to illuminate the area to be detected, with the light trajectory being a straight line. When a water leakage fault occurs in the ceiling of the converter valve hall, bright spots appear in the space when the leaking water droplets or columns intersect with the laser trajectory. By identifying these bright spots through image acquisition and processing, the leaking water droplets or columns in the space can be detected. By translating and rotating the point laser in the direction intersecting the laser trajectory, the water leakage detection area can be expanded.

[0083] like Figure 1 As shown, the specific implementation steps of the above-mentioned leak detection method include the following steps:

[0084] Step S1: Install a movable gimbal along a preset track on the inner wall of the controllable commutation valve chamber. A point laser and an image acquisition device are respectively configured on the movable gimbal.

[0085] The image acquisition device is located next to the point laser and faces the same direction. The image acquisition device is a regular camera. The movable gimbal is a rotatable gimbal that can control the orientation angle of the point laser and the image acquisition device.

[0086] This embodiment provides a preset track, such as Figure 2 As shown, the laser moves along a straight track on the wall of the valve hall. The direction of movement of the point laser is parallel to the track direction, and the irradiation direction of the point laser is a horizontal direction perpendicular to the track.

[0087] This embodiment also provides a preset track, such as Figure 3As shown, the laser moves along a curved track on the valve hall wall. This curved track allows it to bypass obstacles in both the vertical and horizontal directions, adapting to more complex application environments. It not only enables the monitoring of leakage faults in the valve hall but also allows for the monitoring of leakage faults in the lower-level converter valve layer through the movement of the laser. The direction of movement of the point laser is parallel to the local tangent direction of the track, and the laser's illumination direction is also perpendicular to the track and horizontal.

[0088] In both of the above preset tracks, the movable gimbal does not rotate; it only extends the scanning range by translation.

[0089] This embodiment also provides a movable gimbal movement method, such as... Figure 4 As shown, in this movement method, the movable pan-tilt unit does not translate, but only rotates. The movable pan-tilt unit expands the scanning range by rotating, achieving extended leakage monitoring within a certain angular range without translation.

[0090] This embodiment also provides a movable gimbal movement method, such as... Figure 5 As shown, in this method, the movable pan-tilt unit not only translates but also rotates. The movable pan-tilt unit expands the leakage monitoring range through both translation and rotation.

[0091] Step S2: Control the movable gimbal to move to the preset detection position;

[0092] Step S3: With the point laser off, acquire the first image of the location using the image acquisition device;

[0093] Step S4: Turn on the point laser and control the image acquisition device to acquire multiple second images at a preset frequency within a preset time.

[0094] The step of acquiring an image of the area to be inspected using a point laser involves activating the point laser to emit a cylindrical laser beam with a linear trajectory of a certain diameter, which then illuminates and passes through the area to be inspected. An image acquisition device then captures an image of this area.

[0095] Step S5: Perform pixel-level difference processing on the first image and multiple second images to obtain a difference image, specifically including:

[0096] Convert the first image and multiple second images into grayscale images to obtain the first grayscale image. With multiple second grayscale images ,in, Indicates the first i The grayscale image of the second image;

[0097] Based on the grayscale values ​​of each pixel in each second grayscale image, calculate the average grayscale value of each second grayscale image, and select the second grayscale image with the highest average grayscale value as the third grayscale image. ;

[0098] Based on the first grayscale image With the third grayscale image Pixel-level interpolation is performed to obtain the interpolated image. The formula is:

[0099]

[0100] in, This indicates the difference image at the pixel point. grayscale value at that location These represent the first grayscale image. With the third grayscale image At pixel The grayscale value at that location.

[0101] In existing technologies, the falling of droplets is highly random, and their distribution over time is difficult to predict. This makes it easy to miss critical moments of droplet falling or reflection when only single-frame or short-time images are captured. Especially in practical engineering environments such as converter valve halls, where the number of droplets is small, the intervals are long, and the falling speed is fast, traditional methods often fail to capture the complete reflection process or only capture unrepresentative, blurry reflections, leading to missed or false detections.

[0102] To overcome this problem, this embodiment controls the image acquisition device to continuously acquire multiple second images at a preset frequency within a preset time window, forming an image sequence. This approach significantly improves the probability of capturing water droplet reflections, ensuring that the laser can illuminate the water droplet and form a bright spot with typical characteristics in at least one frame. Simultaneously, the time sequence of images can be used to select the frame with the best quality for subsequent processing, further avoiding the impact of blurriness, underexposure, or background interference in individual frames on the overall detection accuracy. Furthermore, this embodiment uses the image with the highest average grayscale value as the third image. This is because when the laser illuminates the water droplet, the droplet's surface has excellent specular reflection characteristics, forming a bright spot in the image with significantly higher brightness than the background, thus increasing the average grayscale value of the entire frame. Therefore, selecting the image with the highest average grayscale value among multiple images often means that the frame is most likely to have captured the strong reflection signal of the droplet. Compared to randomly selecting images or directly processing all images, this selection strategy improves the focus of the analysis and reduces the computational burden of subsequent image processing.

[0103] This embodiment uses a difference formula to retain only the brightness-enhanced areas, that is, the highlights actually added due to laser irradiation. This method is more stable and robust than simple subtraction, and can significantly suppress errors introduced by background interference and image jitter.

[0104] Step S6: Apply a linear contour extraction algorithm to the difference image to identify and separate the laser trajectory, obtaining the highlighted area after removing the laser trajectory. Specifically, this includes:

[0105] Specifically, it includes:

[0106] For difference images Perform a light Gaussian smoothing process to obtain the smoothed difference image. ;

[0107] For the smoothed difference image The Canny edge detection algorithm is used to extract edge contours and obtain an edge map. ;

[0108] In the edge map The probabilistic Hough transform is used to identify the most salient line and obtain the polar coordinate parameters of the line. ;

[0109] Based on the polar coordinate parameters of the straight line obtained from the identification , compared to the typical half-width of the laser trajectory Constructing a laser trajectory mask The formula is:

[0110]

[0111] in, This is the typical half-width of the laser trajectory, which is half the width of the bright line formed by the laser in a grayscale image.

[0112] Based on laser trajectory masking For difference images Perform a masking operation to obtain an image of the highlighted area after removing the laser trajectory:

[0113]

[0114] in, Highlighting the image at pixel points The grayscale value at that location.

[0115] Step S6 employs a linear contour extraction algorithm to identify and separate laser trajectories in the difference image. This is primarily to address the interference of long, bright lines formed by laser excitation on subsequent leak spot detection. In existing technologies, the trajectory lines generated by laser irradiation exhibit obvious linear high-brightness characteristics in the image. If these laser trajectories are not removed, the leak detection algorithm may mistakenly identify them as leak spots, severely impacting the accuracy and reliability of detection. Step S6 resolves the interference of laser trajectories on leak detection, enabling accurate identification and extraction of bright leak spots, significantly improving the system's accuracy and robustness in complex environments. In this way, the leak monitoring system can more effectively distinguish between laser trajectories and actual leak spots, ensuring the reliability and real-time nature of leak alarms.

[0116] Step S7: Based on the highlighted areas, perform connected component analysis to extract candidate spot regions, and extract the spatial droplet features of the candidate spot regions, specifically including:

[0117] For the highlighted area image Perform binarization to obtain a binary image. :

[0118]

[0119] in, Represents the binary image at pixel points The value at that location, A preset grayscale threshold is used to distinguish between bright areas that may be water droplets and the background;

[0120] For binary images Using the eight-neighbor connectivity analysis method, for each pixel with a value of 1 in the binary image, region growing is performed based on the connectivity relationships of its eight neighboring pixels to obtain the set of all connected regions. Each of them This represents a candidate spot region.

[0121] For each candidate spot region The roundness characteristics, intensity profile fitting characteristics, and local spectral analysis characteristics are calculated sequentially to form the spatial droplet characteristics of the candidate spot region.

[0122] The roundness feature calculation process includes:

[0123] For candidate spot regions Extract its area With the perimeter of the area The roundness characteristic is calculated using the following formula:

[0124]

[0125] in, Indicates candidate spot regions pixel area Indicates candidate spot regions The perimeter, that is, the length of the line connecting the boundary pixels; Indicates candidate spot regions The roundness characteristic reflects how close the outline of the candidate spot region is to an ideal circle.

[0126] The strength profile fitting characteristics and calculation process include:

[0127] For each candidate spot region Extract its center pixel coordinates ;

[0128] Using center pixel coordinates Based on this, grayscale profile curves are extracted along the horizontal and vertical directions respectively to obtain the actual horizontal grayscale profile curve. Vertical grayscale profile curve ;

[0129] Extracted actual horizontal grayscale profile curve Vertical grayscale profile curve Gaussian function fitting was performed separately to obtain the horizontal Gaussian fitting curve and the vertical Gaussian fitting curve.

[0130] Calculate the goodness of fit between the Gaussian fitted curve and the actual grayscale profile curve, and define it as the coefficient of determination:

[0131]

[0132] in, Indicates the actual profile curve at position grayscale value, The corresponding Gaussian fitted value is... This represents the average grayscale value of the actual profile curve. The goodness-of-fit coefficient;

[0133] The average of the goodness-of-fit coefficients in the horizontal and vertical directions is used as the candidate spot region. Strength profile fitting characteristics .

[0134] Local spectral analysis features, the calculation process includes:

[0135] For each candidate spot region In the difference image Candidate spot regions between China and Israel Extract a fixed-size local image patch centered on the target area. ;

[0136] For local image patches Perform a two-dimensional fast Fourier transform to obtain local image patches. Spectrum The calculation formula is as follows:

[0137]

[0138] in, For local image patches at pixels grayscale value at that location The imaginary unit, Frequency coordinates in a spectrum diagram The complex spectrum value; N is the local image patch. The side length;

[0139] For the spectrum Take the amplitude spectrum to obtain the spectral energy map:

[0140]

[0141] in, Represents frequency coordinates Spectral amplitude at that location Frequency coordinates in a spectrum diagram Complex spectrum values The real part, Frequency coordinates in a spectrum diagram Complex spectrum values The imaginary part;

[0142] Based on the spectral energy map, the concentration of spectral energy is calculated as a characteristic of local spectral analysis. Defined as the proportion of energy in the low-frequency region to the total energy:

[0143]

[0144] in, Candidate spot regions Local spectral analysis characteristics, The set of frequency coordinates in the low-frequency region satisfies the following conditions:

[0145]

[0146] in, The low-frequency radius threshold is set to a value of [value to be filled in]. , indicating the boundary of the low-frequency range.

[0147] When a laser beam strikes a liquid droplet, laser scattering occurs. This means the laser light is reflected and refracted in all directions by the droplet. By capturing the area to be monitored with a camera, an image containing the laser spot can be obtained. Due to surface tension, water droplets are nearly spherical, resulting in a relatively rounded and regular outline of the laser spot. This naturally formed near-circular structure makes roundness a direct and effective feature for determining whether a candidate spot is a water droplet. Non-water droplet interference, such as rust, stains, and cracks on walls, are often irregular in shape, with jagged edges and lower roundness, making them easily distinguishable from water droplet spots. Furthermore, the reflection and scattering of laser light by the water droplet generally results in a Gaussian distribution of the spot's grayscale, bright at the center and gradually weakening outwards. This is because the laser light is refracted and scattered on the surface of the water droplet, concentrating the intensity in the central region and attenuating symmetrically along both the horizontal and vertical directions. By performing Gaussian fitting on the grayscale profile of the light spot, this distribution pattern can be quantified. A high goodness of fit indicates that the light spot conforms to the typical light intensity distribution of a water droplet, while the opposite may indicate other interfering light spots.

[0148] Local spectral analysis utilizes frequency domain characteristics to reflect the texture and smoothness of light spots. Water droplet spots, due to their simple shape and smooth surface, have most of their spectral energy concentrated in the low-frequency range, manifesting as energy accumulation in the spectrum. In contrast, bright areas caused by wall reflections, rust, or cracks typically have complex structures and rich textures, resulting in more dispersed spectral energy and a lower proportion of low-frequency energy. By calculating the proportion of low-frequency energy, it is possible to further distinguish between real water droplets and background interference in the frequency domain.

[0149] This embodiment uses these three features to comprehensively and multidimensionally reflect the physical nature and imaging rules of the light spots formed by water droplets after being irradiated by laser from three levels: geometric shape, grayscale distribution, and texture frequency. This ensures that water leakage detection can accurately distinguish between real water droplets and various environmental interferences, achieving a high-precision and low-false-alarm technical effect.

[0150] Step S8: Determine if there is a water leakage fault based on the characteristics of the water droplets in the space, specifically including:

[0151] For each candidate spot region Based on its spatial water droplet characteristics Each is compared with the preset roundness threshold. Intensity profile fitting threshold With local spectral analysis threshold The comparison is performed, and when all three features are greater than the threshold, the candidate spot region is determined to be a real water droplet spot;

[0152] If at least one candidate spot area is a real water droplet spot, then it is determined that there is a water leakage fault at the current detection location.

[0153] Step S9: Control the movable gimbal to move to the next preset detection position until all preset detection positions have been detected.

[0154] Example 2:

[0155] This embodiment provides a mobile laser scanning valve hall leakage detection device, which consists of a track, a mobile pan-tilt unit, a point laser, a control unit, a computing unit, and a human-computer interaction unit. These components are connected into a whole through a certain medium.

[0156] The control unit, computing unit, and human-computer interaction unit are implemented by a microcomputer. Through corresponding boards, interfaces, cables, displays, speakers, processors, and other core components and computer programs, the internal units of the device are connected, controlled, calculated, and interact with each other.

[0157] The track gimbal is an electrically controlled gimbal with remote wireless control for movement and rotation, and multi-directional movement and rotation capabilities.

[0158] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0159] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A mobile laser-based controllable commutation converter valve hall water leakage detection method, characterized in that, The method comprises the following steps: A movable holder is installed on the inner wall of the valve hall of the controllable commutation converter valve along a preset track, and a point laser and an image acquisition device are respectively arranged on the movable holder; The movable holder is controlled to move to a preset detection position; In a closed state of the point laser, a first image of the position is acquired by the image acquisition device; The point laser is turned on, and the image acquisition device is controlled to acquire a plurality of second images within a preset time at a preset frequency; Pixel-level difference processing is performed on the first image and the plurality of second images to obtain a difference image; Linear contour extraction algorithm is used to perform laser trajectory recognition and separation on the difference image to obtain a highlight region after removing the laser trajectory; Based on the highlight region, connected domain analysis is performed to extract a candidate spot region, and spatial water droplet features of the candidate spot region are extracted; It is judged whether there is a water leakage fault according to the spatial water droplet features; The movable holder is controlled to move to a next preset detection position until the detection of all preset detection positions is completed.

2. The mobile laser-based controllable commutation converter valve hall water leak detection method of claim 1, wherein, The pixel-level difference processing on the first image and the plurality of second images to obtain the difference image specifically comprises: convert the first image and the plurality of second images into gray scale images to obtain a first gray scale image and a plurality of second gray scale images wherein, a gray scale image of the i plurality of second images According to the gray value of each pixel point of each second gray image, the average gray value of each second gray image is calculated, and the second gray image with the maximum average gray value is taken as a third gray image ; based on the first grayscale image with the third grayscale image , pixel-level difference processing is performed to obtain a difference image .

3. The mobile laser-based controllable commutation converter valve hall water leak detection method of claim 2, wherein, The first gray scale image is based on The third gray scale image is based on The pixel-level difference processing is performed to obtain a difference image The formula is: wherein, denotes the gray value of the difference image at the pixel point , denote the gray values of the first gray image and the third gray image at the pixel point , respectively.

4. The mobile laser-based controllable commutation converter valve hall water leak detection method of claim 1, wherein, The linear contour extraction algorithm is used to perform laser trajectory recognition and separation on the difference image to obtain the highlight region after removing the laser trajectory, specifically comprising: to the difference image performing light Gaussian smoothing to obtain a smoothed difference image ; the smoothed difference image edge profiles are extracted using a Canny edge detection algorithm, resulting in an edge map ; In the edge map The most significant straight line is identified using a probabilistic Hough transform, obtaining the polar coordinates parameters of the straight line ; According to the polar coordinate parameters of the identified straight line , with a typical half-width of the laser track , constructing a laser track mask , the formula is: wherein, is the typical half-width of the laser track, i.e. half the width of the bright line formed by the laser in the grey scale image; based on the laser trajectory mask to the difference image performing a masking operation to obtain a highlight region image after removing the laser trajectory wherein, is the gray value of the highlight region image at pixel point .

5. The mobile laser-based controllable commutation converter valve hall water leak detection method of claim 1, wherein, The connected domain analysis is performed based on the highlight region to extract the candidate spot region, specifically comprising: Image for a highlight region : binaryzation processing is performed to obtain a binary image : wherein, represents a value of the binary image at a pixel point , is a preset gray threshold value for distinguishing the highlight region possibly being the water droplet and the background; Binary image Using eight-neighbor connected component analysis method, for each pixel point with pixel value 1 in the binary image, based on the connected relationship of its adjacent 8 direction pixels, region growing is carried out to obtain all connected region sets , wherein each represents a candidate spot region.

6. The mobile laser-based controllable commutation valve hall water leak detection method of claim 1, wherein, The spatial water droplet features of the candidate spot region are extracted, specifically comprising: for each candidate spot region The circularity feature, the intensity profile fitting feature and the local spectrum analysis feature are calculated in sequence to form the spatial water droplet feature of the candidate spot region.

7. The mobile laser-based controllable commutation converter valve hall water leak detection method of claim 6, wherein, The roundness feature, the calculation process comprising: for candidate blob regions extracting its area and the region perimeter the circularity feature is calculated according to the following formula: wherein, represents the pixel area of the candidate blob region represents the perimeter of the candidate blob region , i.e. the length of the connecting line of the border pixels; represents the circularity feature of the candidate blob region , reflecting the degree to which the contour of the candidate blob region approximates an ideal circle.​ 8. The mobile laser-based controllable commutation converter valve hall water leak detection method of claim 6, wherein, The intensity profile fitting feature, the calculation process comprising: for each candidate blob region , extract the center pixel coordinate ; with the center pixel coordinate The actual horizontal gray profile curve is obtained by extracting the gray profile curve along the horizontal direction and the vertical direction respectively and the actual vertical gray profile curve wherein, is the difference image; actual horizontal gray profile curve actual vertical gray profile curve Gaussian fitting is performed respectively to obtain horizontal Gaussian fitting curve and vertical Gaussian fitting curve; The goodness of fit between the Gaussian fitting curve and the actual gray profile curve is calculated, defined as a goodness of fit coefficient: wherein, represents the gray value of the actual profile curve at the position , is the Gaussian fitting value of the corresponding position, is the gray mean value of the actual profile curve, is the goodness-of-fit coefficient; The average of the goodness of fit coefficients in the horizontal and vertical directions is taken as the candidate spot region The intensity profile of the fitted feature .

9. The mobile laser-based controllable commutation converter valve hall water leak detection method of claim 6, wherein, The local spectral analysis feature, the calculation process comprising: for each candidate blob region in the difference image a fixed size local image patch is extracted centered on the candidate blob region ; for the local image block performing a two-dimensional fast Fourier transform to obtain a spectral map for the local image block , the calculation formula being as follows: in, For local image patches at pixels grayscale value at that location The imaginary unit, Frequency coordinates in a spectrum diagram The complex spectrum value; N is the local image patch. The side length; to the spectrogram Taking the magnitude spectrum, the spectrogram energy map is obtained: wherein denotes the spectral amplitude at the frequency coordinate , denotes the real part of the complex spectral value at the frequency coordinate in the spectral diagram , denotes the imaginary part of the complex spectral value at the frequency coordinate in the spectral diagram . According to the spectral energy map, a concentration degree of spectral energy is calculated as a local spectral analysis feature defined as the proportion of low-frequency regional energy to total energy: wherein is a local spectral analysis feature of the candidate spot region is a set of frequency coordinates of the low frequency region, satisfying the following conditions: is a set of frequency coordinates of the low frequency region, satisfying the following conditions: wherein is a low frequency radius threshold, taking values in the range , representing the boundary of the low frequency range.

10. The mobile laser-based controllable commutation converter valve hall water leak detection method of claim 1, wherein, The spatial water droplet features are used to judge whether there is a water leakage fault, specifically comprising: for each candidate blob region , a spatial droplet feature of the candidate blob region is compared with a preset roundness threshold , an intensity profile fitting threshold and a local spectrum analysis threshold respectively, and when all the three features are greater than the thresholds, the candidate blob region is determined as a real water droplet blob, wherein denotes a roundness feature of the candidate blob region denotes an intensity profile fitting feature of the candidate blob region denotes a local spectrum analysis feature of the candidate blob region ​​​​ If there is at least one candidate spot region that is a real water droplet spot, it is judged that there is a water leakage fault at the current detection position.

Images