Visual alignment guiding method and system for ash tank opening of thermal power plant

By installing cameras at the ash car tank openings of thermal power plants, dust-resistant image enhancement and target detection are performed to generate alignment guidance information, solving the problems of low efficiency and safety hazards associated with manual alignment at ash car tank openings, and achieving accurate and safe positioning and alignment of ash car tank openings.

CN123118745APending Publication Date: 2026-11-17HUANENG TAICANG POWER GENERATION CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611287850.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-24
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Manual alignment of ash truck openings in thermal power plants is inefficient and poses safety hazards in high-dust environments. The lack of quantitative measurement methods means that alignment accuracy depends on personal experience, affecting ash loading efficiency and operational safety.

Method used

The system uses a camera to capture images of the ash truck's tank opening. Through anti-dust image enhancement preprocessing and target detection models, it calculates the physical deviation between the ash truck's tank opening and the ash discharge port, generates alignment guidance information, and outputs precise alignment instructions to the operators.

Benefits of technology

It has achieved automated and objective positioning of the gray tank opening, improved alignment accuracy and operational safety, shortened the alignment time of a single vehicle, eliminated safety hazards for on-site personnel, and improved vehicle turnover efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123118745A_ABST
    Figure CN123118745A_ABST
Patent Text Reader

Abstract

This invention provides a visual alignment guidance method and system for ash truck openings in thermal power plants. The method includes: acquiring an image of the work area, including the ash truck opening, captured by a camera positioned above the ash discharge opening; performing anti-dust image enhancement preprocessing on the work area image, inputting it into a trained target detection model, and outputting the center pixel coordinates of the ash truck opening; calculating the pixel deviation of the center pixel coordinates of the ash truck opening relative to the reference point, using the position of the ash discharge opening center in the image as a reference point; converting the pixel deviation into a physical deviation of the ash truck opening relative to the ash discharge opening; generating alignment guidance information based on the physical deviation, and outputting the alignment guidance information to the ash truck operator. Using this invention, operators can independently complete precise alignment without relying on on-site manual command, eliminating safety hazards for on-site personnel working in high-dust areas, and improving alignment accuracy and operational safety.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of power plant automation technology, specifically to a visual alignment guidance method and system for ash truck openings in thermal power plants. Background Technology

[0002] Fly ash is a major industrial byproduct generated during coal-fired power generation in thermal power plants. Its loading and transportation typically involves tank trucks (i.e., ash trucks) driving into the ash discharge station, aligning the tank opening with the discharge port, and then opening the discharge port. The alignment accuracy between the tank opening and the discharge port directly affects loading efficiency and operational safety—excessive alignment deviation will lead to fly ash spillage, while excessive alignment time will restrict vehicle turnaround efficiency.

[0003] Currently, thermal power plants generally employ a work mode that relies on on-site manual command and driver experience. Commanders visually assess the relative positions of the tank opening and the ash discharge port, then guide the driver through gestures or verbal commands, repeatedly moving forward and backward, making minor adjustments until alignment is achieved. This mode has the following drawbacks: High dust concentrations in the ash discharge area make the tank opening outline easily obscured by dust, and changes in lighting and viewing angles make it difficult for commanders to accurately determine the true relative position. Alignment accuracy depends on personal experience, requiring drivers to make multiple adjustments, resulting in long alignment times per vehicle. Commanders must remain in the high-dust environment near the ash discharge port for extended periods, facing safety risks from dust inhalation and blind spots caused by vehicle movement. This method lacks quantitative measurement methods for the tank opening position; instructions are based on subjective judgment rather than objective data, leading to poor consistency in work quality.

[0004] Therefore, how to achieve precise alignment between the ash truck tank opening and the ash discharge port in the high-dust environment of thermal power plants, reduce reliance on manual visual inspection and experience judgment, shorten the alignment time of a single truck and eliminate safety hazards for on-site personnel, and at the same time provide objective and unified quantitative guidance for the alignment process has become an urgent technical problem to be solved. Summary of the Invention

[0005] In view of this, it is necessary to provide a visual alignment guidance method and system for ash truck openings in thermal power plants to solve the technical problems of low efficiency and safety hazards of manual alignment of ash truck openings in high dust environments in the existing technology.

[0006] To address the aforementioned technical problems, in a first aspect, the present invention provides a visual alignment guidance method for ash truck openings in thermal power plants, the method comprising: Acquire images of the work area captured by a camera positioned above the ash discharge port, the images of the work area including the ash truck opening; The image of the work area is subjected to anti-dust image enhancement preprocessing to obtain the enhanced image; The enhanced image is input into a trained target detection model to detect the position of the gray tank opening in the enhanced image and output the center pixel coordinates of the gray tank opening. Using the position of the center of the ash discharge port in the image as a reference point, calculate the pixel deviation of the center pixel coordinates of the ash tank opening relative to the reference point; Based on the calibration parameters of the camera, the pixel deviation is converted into the physical deviation of the ash tank opening relative to the ash discharge opening; Alignment guidance information is generated based on the physical deviation, and the alignment guidance information is output to the operator of the gray vehicle.

[0007] In one possible implementation, the step of performing anti-dust image enhancement preprocessing on the image of the work area to obtain the enhanced image includes: The image of the work area is subjected to dark channel prior dehazing to obtain a dehazed image; The dehazed image is subjected to multi-scale Retinex image enhancement processing to obtain an initial enhanced image; The initial enhanced image is subjected to Gaussian filtering to obtain the enhanced image.

[0008] In one possible implementation, performing dark channel prior dehazing on the image of the work area to obtain a dehazed image includes: Select a local window of the dark channel; Based on the dark channel local window, the minimum channel value of each pixel in the working area image is determined in the three RGB channels, and the minimum channel value is used as the dark channel value of each pixel to generate a dark channel image. Obtain the pixels with the highest brightness values ​​in the dark channel image, representing a predetermined percentage. Map these pixels to the RGB components of the working area image and calculate their average value as the atmospheric light value. The initial transmittance map is calculated based on the atmospheric light value, and the initial transmittance map is optimized by guided filtering to obtain the optimized transmittance map. Based on the atmospheric light value and the optimized transmittance map, the pixel values ​​of each pixel in the image of the working area are recovered and calculated to obtain the dehazed image.

[0009] In one possible implementation, performing dark channel prior dehazing on the image of the work area to obtain a dehazed image includes: Select a local window of the dark channel; Based on the dark channel local window, the minimum channel value of each pixel in the working area image is determined in the three RGB channels, and the minimum channel value is used as the dark channel value of each pixel to generate a dark channel image. Obtain the pixels with the highest brightness values ​​in the dark channel image, representing a predetermined percentage. Map these pixels to the RGB components of the working area image and calculate their average value as the atmospheric light value. The initial transmittance map is calculated based on the atmospheric light value, and the initial transmittance map is optimized by guided filtering to obtain the optimized transmittance map. Set multiple defogging intensity parameters corresponding to dust concentration; Obtain the global contrast of the image of the working area, and / or obtain the dark channel mean of the dark channel image; Based on the global contrast and / or the dark channel mean, determine the concentration level to which the current dust concentration belongs; Based on the concentration level, select the target defogging intensity parameter corresponding to the current dust concentration from the multiple defogging intensity parameters; Based on the atmospheric light value and the optimized transmittance map, the pixel values ​​of each pixel in the image of the working area are recovered using the target defogging intensity parameter to obtain the defogging image.

[0010] In one possible implementation, the target detection model is a YOLOv8 model, and the outputs of the last two C2f modules in the backbone network of the YOLOv8 model are respectively inserted with CBAM attention modules. The CBAM attention module includes a channel attention submodule and a spatial attention submodule connected in series. The loss function of the YOLOv8 model is the CIoU loss function.

[0011] In one possible implementation, the calibration parameters include resolution and installation height; the step of converting the pixel deviation into a physical deviation of the ash tank opening relative to the ash discharge opening based on the camera's calibration parameters includes: Based on the resolution and installation height of the camera, determine the physical width and physical height corresponding to a single pixel in the image of the work area; The horizontal pixel deviation is obtained by multiplying the horizontal pixel deviation by the physical width corresponding to the single pixel. The vertical pixel deviation is obtained by multiplying the physical height corresponding to the single pixel. Based on a pre-calibrated perspective error correction coefficient, the physical deviations in the horizontal direction and the vertical direction are corrected to obtain corrected physical deviations in the horizontal and vertical directions. These corrected physical deviations in the horizontal and vertical directions constitute the physical deviation of the ash tank opening relative to the ash discharge opening.

[0012] In one possible implementation, the alignment guidance information includes voice command instructions and / or visual guidance information; The voice command instructions are generated based on TTS speech synthesis technology, and include deviation broadcast instructions, step-by-step operation guidance instructions, alignment success prompt instructions, and abnormal situation warning instructions. The visual guidance information is displayed through the operator's mobile terminal, which shows the relative position, offset direction, offset distance, and operation status information of the ash truck opening and the ash discharge opening.

[0013] One possible implementation also includes: When the alignment of the ash truck's tank opening is detected, the vehicle information, ash discharge task information, alignment result, and operation record of the ash truck are uploaded to the power plant management system.

[0014] Secondly, the present invention also provides a visual alignment guidance system for ash truck openings in thermal power plants, the system comprising: The image acquisition layer includes a camera deployed above the ash discharge port for real-time acquisition of images of the work area, including the ash truck opening; The AI ​​recognition layer includes an edge computing gateway, which runs a target detection model to receive the work area image and perform anti-dust image enhancement preprocessing on the work area image, detect the position of the ash truck can opening in the image, output the center pixel coordinates of the ash truck can opening, calculate the pixel deviation with the position of the ash outlet center in the image as the reference point, and convert the pixel deviation into physical deviation according to the camera calibration parameters. The data service layer is used to generate alignment guidance information based on the physical deviation; The terminal application layer is used to output the alignment guidance information to the operator of the gray vehicle.

[0015] In one possible implementation, a power plant management system that is communicatively connected to the data service layer is also included; The power plant management system is used to acquire vehicle information and ash discharge task information, as well as to transmit alignment results and operation records.

[0016] The beneficial effects of this invention are: The present invention provides a visual alignment guidance method for ash truck canister openings in thermal power plants. This method acquires images of the work area, including the ash truck canister openings, from a camera positioned above the ash discharge opening. By fixing the camera above the opening, the stability of the opening center as a constant reference point in the image is ensured, providing objective image acquisition conditions for subsequent deviation calculations. The work area image undergoes anti-dust image enhancement preprocessing to obtain an enhanced image, which improves the accuracy of canister opening recognition under high dust conditions. The enhanced image is then input into a trained target detection model to detect the position of the ash truck canister opening in the enhanced image and outputs the center pixel coordinates of the canister opening. This achieves automation, objectivity, and standardization of canister opening position detection, ensuring the accuracy and repeatability of the positioning results. The method uses the position of the ash discharge opening center in the image as a basis... The system accurately quantifies the deviation of the center pixel coordinates of the ash truck's tank opening from the reference point. Based on the camera's calibration parameters, the pixel deviation is converted into a physical deviation between the ash truck's tank opening and the ash discharge port. This ensures that the alignment guidance information includes specific physical deviations, allowing operators to perform adjustments based on quantified distance values, thus improving the accuracy and efficiency of adjustments. Alignment guidance information is generated based on the physical deviations and output to the ash truck operators, ensuring standardized and quantifiable alignment instructions. This effectively shortens the alignment time per vehicle, improves vehicle turnaround efficiency, and allows operators to independently complete precise alignment without relying on on-site manual guidance. This eliminates safety hazards for personnel working in high-dust areas, enhancing alignment accuracy and operational safety. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic flowchart of an embodiment of the visual alignment guidance method for ash truck openings in thermal power plants provided by the present invention. Figure 2 For the present invention Figure 1 A schematic diagram of an embodiment of S102; Figure 3 A schematic diagram of the layout of the visual alignment guidance system for the ash truck opening of a thermal power plant provided by the present invention. Figure 4 This is a schematic diagram illustrating the pixel deviation between the center of the ash tank opening and the center of the ash discharge opening provided by the present invention. Figure 5This is a schematic diagram of an embodiment of the visual alignment guidance system for ash truck openings in thermal power plants provided by the present invention. Detailed Implementation

[0019] 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 a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0020] In the description of the embodiments of the present invention, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.

[0021] The terms "first," "second," etc., used in the embodiments of this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a technical feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature.

[0022] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0023] This invention provides a visual alignment guidance method and system for ash truck openings in thermal power plants, which will be described below.

[0024] The execution entity of the visual alignment guidance method for ash truck openings in thermal power plants in this application embodiment can be the visual alignment guidance system for ash truck openings in thermal power plants provided in this application embodiment, or different types of electronic devices such as server equipment, physical host, or user equipment (UE) that integrate the visual alignment guidance system for ash truck openings in thermal power plants. The visual alignment guidance system for ash truck openings in thermal power plants can be implemented in hardware or software. The UE can specifically be a terminal device such as a smartphone, tablet computer, laptop computer, handheld computer, desktop computer, or personal digital assistant (PDA).

[0025] The embodiments of this application provide a visual alignment guidance method for the ash truck opening of a thermal power plant, which can be applied to the ash discharge and loading operation of a thermal power plant.

[0026] Figure 1 This is a schematic flowchart of an embodiment of the visual alignment guidance method for ash truck openings in thermal power plants provided by the present invention. The visual alignment guidance method for ash truck openings in thermal power plants includes: S101. Obtain an image of the work area captured by a camera positioned above the ash discharge port, the image of the work area including the ash truck opening.

[0027] The camera can be an industrial-grade high-definition camera, deployed at an appropriate position above the ash discharge port, with its field of view covering the ash truck parking and operation area.

[0028] The working area image is a digital image that includes the area around the ash discharge port and the area of ​​the ash truck tank opening.

[0029] Specifically, after the ash truck enters the work area, the camera captures real-time images of the work area, including the ash truck's outlet. The camera is communicatively connected to an edge computing gateway, and the captured image data is transmitted to the edge computing gateway for subsequent processing via wired or wireless means. In this embodiment, by fixing the camera above the ash outlet, the center of the ash outlet remains fixed in the image, serving as a constant reference point for subsequent deviation calculations.

[0030] S102. Perform anti-dust image enhancement preprocessing on the image of the work area to obtain the enhanced image.

[0031] Among them, anti-dust image enhancement preprocessing refers to digital image processing operations performed on images acquired in high-dust environments, which can suppress the degradation effect of dust particles on image quality and restore the clarity of the tank opening outline.

[0032] Specifically, the air in the ash discharge area of ​​a thermal power plant contains a large amount of suspended fly ash particles, which can easily cause degradation in images captured by cameras, such as fogging, reduced contrast, and weakened edge features of the tank opening. Dust-resistant image enhancement preprocessing includes at least one or more combinations of image dehazing, contrast enhancement, and noise reduction. Image dehazing removes the atmospheric scattering effect caused by dust particles, restoring the true color and brightness of the target in the image; contrast enhancement stretches the grayscale difference between the tank opening area and the background area, making the target at the tank opening more visually prominent; noise reduction suppresses noise that may be introduced during the enhancement process, maintaining the smoothness of the image.

[0033] The dust-resistant image enhancement preprocessing in this embodiment processes the entire working area image, making key visual features such as the tank opening edge and contour stand out in the image. This effectively restores the edge clarity and contrast of the tank opening target in the degraded image, which is beneficial to improving the accuracy of tank opening recognition under high dust conditions.

[0034] S103. Input the enhanced image into the trained target detection model to detect the position of the gray tank opening in the enhanced image and output the center pixel coordinates of the gray tank opening.

[0035] The target detection model is a deep learning-based computer vision model that learns from a large number of image samples labeled with the locations of tank openings. It is capable of automatically identifying tank opening targets and locating their center points in input images. During the training phase, the model uses a set of image samples containing tank openings under different lighting conditions, dust concentrations, and vehicle models. By labeling the location information of the tank opening bounding boxes in the samples, the model learns the visual feature patterns of the tank opening targets.

[0036] Specifically, an object detection model is used to extract features and locate the target in the enhanced image. This identifies the region containing the can opening within the image and calculates the pixel coordinates of the center of this region. These center pixel coordinates represent the position of the can opening's center in the image's pixel coordinate system. The object detection model can be implemented using various algorithmic architectures, including but not limited to single-stage detection architectures based on anchor boxes and two-stage detection architectures based on candidate regions. All object detection models, regardless of their architecture, have the ability to output target location information after inputting the image.

[0037] This embodiment automatically detects the tank opening position using a trained target detection model, eliminating human interference in the tank opening positioning process and achieving automation, objectivity, and standardization of tank opening position detection. The positioning results are not affected by individual differences among operators. Combined with anti-dust image enhancement preprocessing to compensate for image quality under different dust concentrations, the accuracy and repeatability of the positioning results are ensured.

[0038] S104. Using the position of the center of the ash discharge port in the image as a reference point, calculate the pixel deviation of the center pixel coordinates of the ash tank opening relative to the reference point.

[0039] The position of the ash discharge port center in the image is determined in advance through on-site calibration. For example, after the camera is installed, the pixel coordinates of the ash discharge port center in the image are determined by manual annotation or calibration board method, and these coordinates are stored in the system as a fixed reference point.

[0040] Pixel deviation includes horizontal deviation and vertical deviation, which represent the number of pixels that the center of the tank opening is offset from the center of the ash discharge port in the left-right and up-down directions of the image, respectively.

[0041] Specifically, by comparing the center of the tank opening with the center of the ash discharge port in the same coordinate system, the precise offset of the ash tank opening relative to the ash discharge port can be obtained, thus achieving precise quantification of the degree of deviation of the ash tank opening.

[0042] S105. Based on the calibration parameters of the camera, the pixel deviation is converted into the physical deviation of the ash tank opening relative to the ash discharge opening.

[0043] Camera calibration parameters refer to the set of parameters that establish the mapping relationship between the image pixel coordinate system and the physical operation coordinate system. Examples include the camera's resolution, installation height, and field of view. The image pixel coordinate system describes the position of the tank opening in the image. When adjusting the vehicle, operators need to understand the relative positions of the tank opening and the ash discharge port in the physical operation coordinate system. A definite mapping relationship exists between the two coordinate systems.

[0044] Specifically, based on camera calibration parameters, pixel-level offsets in the image can be converted into physical offsets expressed in units of length in the actual working scenario. This embodiment converts pixel deviations into physically meaningful distance values, ensuring that the alignment guidance information includes specific physical deviations. This guarantees that operators can perform adjustments based on quantified distance values, improving the accuracy and efficiency of the adjustments.

[0045] S106. Generate alignment guidance information based on the physical deviation, and output the alignment guidance information to the operator of the gray vehicle.

[0046] Among them, the alignment guidance information refers to the operation guidance information that includes the offset direction and offset distance of the tank opening relative to the ash discharge opening.

[0047] Specifically, based on the sign and magnitude of the physical deviation, corresponding guidance instructions are generated. These instructions include specific adjustment directions and distances to guide operators in adjusting the ash car until the ash car's tank opening and the ash discharge port are aligned.

[0048] This embodiment generates standardized alignment guidance information based on physical deviations, clearly informing operators of the specific adjustment direction and distance. This allows operators to independently complete precise alignment without relying on visual judgment and experience-based instructions from on-site supervisors, simply by following the guidance information. This ensures the standardization and quantification of alignment operation guidance, improving the efficiency of operators in performing alignment adjustments. It also effectively shortens the alignment time per vehicle, improving vehicle turnaround and operational efficiency. Furthermore, the entire alignment guidance process eliminates the need for on-site supervisors to remain in the high-dust area near the ash discharge port, thus eliminating the safety hazards of personnel exposure in high-risk areas and improving alignment accuracy and operational safety.

[0049] In summary, the visual alignment guidance method for ash truck openings in thermal power plants provided by this invention ensures the stability of the ash truck opening center as a constant reference point in the image by fixing the camera above the ash truck opening, providing objective image acquisition conditions for subsequent deviation calculation. By performing anti-dust image enhancement preprocessing on the work area image, the recognition accuracy of the ash truck opening target in high-dust environments is improved, ensuring that the subsequent target detection model can obtain high-quality input images. The trained target detection model automatically detects the center pixel coordinates of the ash truck opening, achieving automation, objectivity, and standardization of opening position detection, ensuring the consistency and repeatability of the positioning results. By using the position of the ash truck opening center in the image as a reference point for calculation… By calculating pixel deviation, the degree of deviation at the tank opening is precisely quantified, ensuring accurate data input for subsequent physical deviation conversion. Pixel deviation is converted into physical deviation based on camera calibration parameters, ensuring that alignment guidance information includes specific physical deviation values. This allows operators to perform adjustments based on quantified distance values. By generating alignment guidance information based on physical deviation and outputting it to operators, the standardization and quantification of alignment operation instructions are ensured, effectively shortening single-vehicle alignment time and improving vehicle turnaround efficiency. Simultaneously, operators can independently complete precise alignment without relying on on-site manual guidance, eliminating safety hazards for personnel working in high-dust areas and improving alignment accuracy and operational safety.

[0050] In some embodiments of the present invention, such as Figure 2 As shown, step S102 includes: S201. Perform dark channel prior dehazing on the image of the work area to obtain a dehazed image; S202. Perform multi-scale Retinex image enhancement processing on the dehazed image to obtain an initial enhanced image; S203. Perform Gaussian filtering on the initial enhanced image to obtain the enhanced image.

[0051] Dark channel prior dehazing refers to a method for dehazing images based on the dark channel prior law. The dark channel prior law states that in most outdoor fog-free images, outside the sky, at least one color channel has a pixel value that is very low and close to zero. Based on this law, it is possible to inversely recover images fogged by dust.

[0052] Multiscale Retinex image enhancement is a color constancy enhancement method based on Retinex theory. Retinex theory decomposes an image into illumination and reflection components, and restores the reflection properties of the object itself by estimating and eliminating the illumination component, thereby achieving image enhancement.

[0053] Gaussian filtering is a linear smoothing filtering method that reduces noise by weighting an image pixel and its neighboring pixels. The weights are determined by a Gaussian function.

[0054] Specifically, the image of the work area is first subjected to dark channel prior dehazing to obtain the dehazed image. This process uses dark channel priors to estimate atmospheric light values ​​and transmittance in the image, and then calculates the dehazed image in reverse based on the atmospheric scattering model, thereby eliminating the haze caused by dust particle scattering. Next, the dehazed image is subjected to multi-scale Retinex image enhancement to obtain the initial enhanced image. This process estimates and suppresses the illumination components at multiple scales, thereby restoring the reflection details of the image and improving potential color casts and insufficient contrast after dehazing. Finally, the initial enhanced image is subjected to Gaussian filtering to obtain the enhanced image.

[0055] Understandably, this embodiment eliminates the fogging caused by dust scattering through dark channel prior dehazing, improving the global contrast and clarity of the image; multi-scale Retinex enhancement improves the brightness balance and color fidelity of the dehazed image, making details such as the can opening edge more prominent; Gaussian filtering suppresses high-frequency noise that may be introduced during the enhancement process, ensuring that the enhanced image maintains the clarity of the can opening outline while having a low noise level. These three processing steps are sequentially linked, progressively performing dehazing, detail enhancement, and noise reduction on the degraded image, resulting in clearer edge contours, more sufficient contrast, and less noise in the enhanced image, thereby improving the accuracy and positioning precision of the subsequent target detection model for can opening recognition.

[0056] In some embodiments of the present invention, step S201 includes: selecting a dark channel local window; determining the minimum channel value of each pixel in the RGB three channels of the working area image according to the dark channel local window, using the minimum channel value as the dark channel value of each pixel to generate a dark channel image; obtaining the pixels with the highest luminance values ​​in the dark channel image, mapping the pixels with the highest luminance values ​​to the RGB components of the working area image and calculating the average value as the atmospheric light value; calculating an initial transmittance map according to the atmospheric light value, and performing guided filtering optimization on the initial transmittance map to obtain an optimized transmittance map; and performing pixel value recovery calculation on each pixel in the working area image according to the atmospheric light value and the optimized transmittance map to obtain a dehazed image.

[0057] Dark channel prior dehazing is an image processing operation that restores dust-degraded images based on dark channel prior theory. Dark channel prior theory is based on statistical observations of a large number of haze-free natural images: in most local image patches outside the sky, at least one color channel has extremely low pixel values, approaching zero. Based on this prior, atmospheric light values ​​and transmittance can be estimated from a single dust-degraded image, and then the clear image after dehazing can be deduced.

[0058] The pre-selected percentage refers to the proportion of pixels that are at the very beginning after being sorted by brightness value from high to low.

[0059] The size of the dark channel local window is determined based on the pixel diameter of the gray truck's tank opening in the work area image. The pixel diameter of the gray truck's tank opening is obtained through manual measurement and statistical analysis of multiple frames of on-site images captured by the camera. As a specific implementation, the size of the dark channel local window is 1 / 8 to 1 / 5 of the pixel diameter. For example, when the typical pixel diameter of the gray truck's tank opening is 125 pixels, the typical size of the dark channel local window is 15×15 pixels.

[0060] Specifically, a local window for the dark channel is selected. This window can be a rectangular sliding window that slides pixel by pixel across the work area image to determine the minimum value of the RGB channels within the neighborhood of each pixel. Then, based on the local window, the minimum channel value of each pixel in the work area image is determined among the RGB channels. This minimum channel value is used as the dark channel value for each pixel, generating a dark channel image. This dark channel image is a grayscale image, and its pixel values ​​reflect the lowest brightness of the corresponding local area. The pixels with the highest brightness values ​​in the dark channel image are obtained, mapped to the RGB components of the work area image, and their average value is calculated as the atmospheric light value. The atmospheric light value reflects the scattering intensity of scene illumination by dust and is an important parameter for dehazing and restoration calculations. An initial transmittance map is calculated based on the atmospheric light value. Transmittance describes the degree of light attenuation from the scene point to the camera; the higher the dust concentration, the lower the transmittance. Guided filtering optimization is applied to the initial transmittance map to obtain an optimized transmittance map. Guided filtering uses the image of the work area as a guide map, smoothing noise in the transmittance map while preserving the abrupt transmittance changes at the tank opening edge, thus avoiding blurring of the tank opening edge. Based on the atmospheric light value and the optimized transmittance map, the pixel values ​​of each pixel in the work area image are recovered to obtain the dehazed image. The recovery calculation is based on the inverse solution of the atmospheric scattering model, restoring the image pixel values ​​affected by dust scattering to the true pixel values ​​under dust-free conditions.

[0061] Understandably, this embodiment restores dust-degraded images by using dark channel prior dehazing, effectively suppressing the degradation effect of high dust environment on image quality and ensuring that the tank opening target has clear edge features in the dehazed image.

[0062] In some embodiments of the present invention, step S201 includes: selecting a dark channel local window; determining the minimum channel value of each pixel in the RGB three channels of the working area image according to the dark channel local window, using the minimum channel value as the dark channel value of each pixel to generate a dark channel image; obtaining the pixels with the highest luminance values ​​in the dark channel image, mapping the pixels with the highest luminance values ​​to the RGB components of the working area image and calculating the average value as the atmospheric light value; calculating an initial transmittance map based on the atmospheric light value, and performing guided filtering optimization on the initial transmittance map. The process involves: obtaining an optimized transmittance map; setting multiple defogging intensity parameters corresponding to dust concentration; acquiring the global contrast of the image of the work area and / or acquiring the dark channel mean of the dark channel image; determining the concentration level of the current dust concentration based on the global contrast and / or the dark channel mean; selecting a target defogging intensity parameter corresponding to the current dust concentration from the multiple defogging intensity parameters based on the concentration level; and performing pixel value recovery calculations on each pixel in the image of the work area using the target defogging intensity parameter based on the atmospheric light value and the optimized transmittance map to obtain the defogging image.

[0063] In this embodiment, the steps of selecting a local window in the dark channel, generating a dark channel image, estimating atmospheric light values, calculating the initial transmittance map, and performing guided filtering optimization in the dark channel prior dehazing process are the same as in the above embodiments. Please refer to the relevant descriptions in the above embodiments, which will not be repeated here.

[0064] As a specific implementation method, the defogging intensity parameter can be set to three levels: ω = 0.95 when there is no dust, ω = 0.85 when there is normal dust, and ω = 0.75 when there is high dust.

[0065] Specifically, the global contrast of the image of the work area and / or the dark channel mean of the dark channel image are obtained. Global contrast reflects the severity of overall brightness variations in the image, while the dark channel mean reflects the overall degree of dust occlusion. Based on the global contrast and / or dark channel mean, the current dust concentration level is determined. A high dust concentration is defined when the global contrast is below a set threshold or the dark channel mean is above a set threshold; conversely, a low dust concentration is defined. Multiple defogging intensity parameters corresponding to the dust concentration are set. Based on the determined concentration level, a target defogging intensity parameter corresponding to the current dust concentration is selected from the multiple defogging intensity parameters. Then, based on the atmospheric light value estimated using the aforementioned method and the optimized transmittance map, the target defogging intensity parameter is used to recover the pixel values ​​of each pixel in the work area image, resulting in the defogging image. The specific calculation method is as follows: The atmospheric scattering model describes the physical process of image degradation in dusty environments, and its expression is: I(x) = J(x)·t(x) + A·(1 - t(x)); Where I(x) is the pixel value of pixel x in the image of the working area captured by the camera, J(x) is the pixel value of that pixel in the fog-free image to be recovered, A is the atmospheric light value, and t(x) is the transmittance value of that pixel in the optimized transmittance map. In the above formula, I(x), A, and t(x) are known quantities, and J(x) is the unknown quantity to be solved.

[0066] By deforming the above atmospheric scattering model, the formula for recovering pixel values ​​of a haze-free image can be obtained: J(x) = (I(x) - A) / t(x) + A; To prevent unstable division results due to excessively low transmittance, a lower limit t0 for transmittance is set. When the transmittance value t(x) of a pixel in the optimized transmittance map is less than t0, t0 is used instead of t(x) in the restoration calculation. The restoration calculation formula is then: J(x) = (I(x) - A) / max(t(x), t0) + A; The target defogging intensity parameter is incorporated into the above recovery calculation. The target defogging intensity parameter is denoted as ω, and its value ranges from 0 to 1. When using the target defogging intensity parameter for recovery calculation, ω is combined with the transmittance value in the calculation: J(x) = (I(x) - A) / max(ω·t(x), t0) + A; Where ω·t(x) represents the effective transmittance after adjusting the dehazing intensity parameter. The closer ω is to 1, the greater the dehazing intensity and the clearer the image; the smaller ω is, the more residual dust is retained and the lighter the dehazing effect. Using the above formula, the pixel value J(x) of each pixel in the dehazed image can be calculated pixel by pixel, thus obtaining the complete dehazed image.

[0067] Understandably, this embodiment adjusts the transmittance by introducing a defogging intensity parameter ω, allowing the defogging recovery intensity to adaptively change according to the actual dust concentration. A larger ω value is used for high dust concentrations to ensure sufficient defogging, while a smaller ω value is used for low dust concentrations to avoid image distortion caused by over-enhancement, thus ensuring the consistency and stability of the visual effect of the recovered image under different dust conditions. Simultaneously, by setting a lower limit t0 for transmittance, numerical anomalies in the recovery calculation are avoided when the transmittance approaches zero, ensuring the numerical stability of the algorithm.

[0068] In some embodiments of the present invention, the target detection model is a YOLOv8 model, and the outputs of the last two C2f modules in the backbone network of the YOLOv8 model are respectively inserted with CBAM attention modules; the CBAM attention module includes a channel attention submodule and a spatial attention submodule connected in series; the loss function of the YOLOv8 model is the CIoU loss function.

[0069] Among them, the YOLOv8 model is a single-stage object detection network whose backbone network is composed of multiple stacked C2f modules. The C2f modules are used to extract image features at different scales.

[0070] The CBAM attention module is a lightweight attention mechanism module consisting of a serially connected channel attention submodule and a spatial attention submodule. The channel attention submodule is used to assign weights to each channel of the input feature map to enhance important feature channels and suppress unimportant feature channels; the spatial attention submodule is used to assign weights to the spatial location of the feature map to focus on the target region and suppress the background region.

[0071] The CIoU loss function is a bounding box regression loss function, expressed as L_CIoU = 1 - IoU + ρ²(b,b_gt) / c² + αv, where IoU is the intersection-union ratio of the predicted box and the ground truth box, ρ²(b,b_gt) is the Euclidean distance between the center points of the predicted box and the ground truth box, c is the diagonal length of the minimum bounding rectangle of the predicted box and the ground truth box, v is the aspect ratio consistency term, and α is the weight coefficient of the aspect ratio penalty term.

[0072] Specifically, in the backbone network of the YOLOv8 model, the CBAM attention module is inserted into the output of the last two C2f modules. This insertion point is in the high-level feature stage of the backbone network, where the feature map size is moderate, preserving both the spatial location information of the can opening target and strong semantic expressive power. The channel attention submodule performs global average pooling and global max pooling on the input feature map to generate two channel descriptors. After processing by a shared multilayer perceptron, channel weights are generated using a Sigmoid activation function. The channel weights are multiplied by the original feature map to obtain the channel-enhanced feature map. The spatial attention submodule performs global average pooling and global max pooling on the channel-enhanced feature map along the channel dimension. The pooling results are concatenated, dimensionality reduced by convolution, and then generated using a Sigmoid activation function. The spatial weights are multiplied by the input feature map to obtain the spatially enhanced feature map. Through the sequential processing of channel attention and spatial attention, the CBAM attention module enhances the can opening features from both channel and spatial dimensions while suppressing the response of the dust background region.

[0073] The YOLOv8 model uses the CIoU loss function for bounding box regression. For the approximately circular target feature of the gray tank opening, the weight coefficient α of the aspect ratio penalty term in the CIoU loss function is set to 0.5 to avoid excessive aspect ratio penalty for circular targets. The total loss function of the YOLOv8 model consists of a weighted average of the bounding box regression loss, classification loss, and confidence loss. The bounding box regression loss uses the CIoU loss function, while the classification loss and confidence loss use the cross-entropy loss function. The expression for the total loss function is L_total = 7.5 × L_CIoU + 0.5 × L_cls + 1.0 × L_obj. Wherein, L_CIoU is the bounding box regression loss, used to measure the positional deviation between the predicted box and the ground truth box; L_cls is the classification loss, used to measure the accuracy of the target category prediction; L_obj is the confidence loss, used to measure the existence of the target and its confidence level; the weight coefficients 7.5, 0.5, and 1.0 correspond to the weights of the above three losses, with the bounding box regression loss having the highest weight to prioritize ensuring the accuracy of the tank opening positioning.

[0074] Understandably, this embodiment, by inserting the CBAM attention module into the high-level feature stage of the YOLOv8 backbone network, enhances the edge contour features of the can opening from both the channel and spatial dimensions and suppresses the interference of dust background, without significantly increasing inference latency. This allows the model to more accurately focus on the can opening target in complex backgrounds, improving the accuracy of can opening detection in high-dust environments. Simultaneously, by employing the CIoU loss function and setting the aspect ratio penalty term weight coefficient to 0.5, the bounding box regression process considers the overlap area between the predicted and ground truth boxes, the center point distance, and aspect ratio consistency. Furthermore, this weight coefficient matches the circular features of the can opening, effectively improving the accuracy and stability of can opening localization and avoiding problems such as predicted box imbalance or box drift.

[0075] In some embodiments of the present invention, the calibration parameters include resolution and installation height; step S105 includes: determining the physical width and physical height corresponding to a single pixel in the image of the working area based on the resolution and installation height of the camera; multiplying the horizontal pixel deviation by the physical width corresponding to the single pixel to obtain the horizontal physical deviation; multiplying the vertical pixel deviation by the physical height corresponding to the single pixel to obtain the vertical physical deviation; correcting the horizontal physical deviation and the vertical physical deviation based on a pre-calibrated perspective error correction coefficient to obtain corrected horizontal physical deviation and corrected vertical physical deviation, wherein the corrected horizontal physical deviation and the corrected vertical physical deviation constitute the physical deviation of the ash tank opening relative to the ash discharge opening.

[0076] The calibration parameters for a camera include at least its resolution and installation height. Resolution determines the total number of pixels in the image, while installation height directly affects the size of the camera's field of view. Together, they determine the actual physical size of a single pixel in the image.

[0077] The perspective error correction coefficient is a correction parameter used to eliminate the impact of image perspective distortion on measurement accuracy caused by factors such as camera installation tilt angle, lens distortion, and vehicle parking position offset. This coefficient is predetermined through on-site multi-point calibration.

[0078] Specifically, firstly, based on the camera's resolution and installation height, the physical width and height corresponding to a single pixel in the work area image are determined. In this embodiment, an industrial camera with a resolution of 1920×1080 is used, installed at a height of approximately 3.5 meters, with a horizontal field of view of approximately 90°, covering a horizontal width of approximately 7 meters in the work area. Calculations show that the horizontal physical width of a single pixel is approximately 3.65 mm / pixel, and the vertical physical height of a single pixel is approximately 3.65 mm / pixel.

[0079] Then, the horizontal pixel deviation Δu is multiplied by the single-pixel physical width Su to obtain the horizontal physical deviation ΔX = Δu × Su; the vertical pixel deviation Δv is multiplied by the single-pixel physical height Sv to obtain the vertical physical deviation ΔY = Δv × Sv. A positive value for the horizontal physical deviation indicates that the can opening is located to the right of the ash discharge port, and a negative value indicates that it is located to the left; a positive value for the vertical physical deviation indicates that the can opening is located in front of the ash discharge port, and a negative value indicates that it is located behind.

[0080] Furthermore, based on pre-calibrated perspective error correction coefficients, the physical deviations in the horizontal and vertical directions are corrected. Since it's difficult to guarantee a perfectly vertical overhead view during actual camera installation, and the lens itself has slight distortion, and the vehicle's parking position is not always precisely the same, these factors can cause perspective distortion in the image edge areas. Simply relying on single-pixel physical size conversion would introduce measurement errors. To eliminate this error, the horizontal correction coefficient ku and the vertical correction coefficient kv are determined through multi-point calibration on-site, and the initial conversion results are corrected a second time: ΔX correction = ku × ΔX, ΔY correction = kv × ΔY. The corrected horizontal physical deviation ΔX and vertical physical deviation ΔY together constitute the final physical deviation of the ash truck's tank opening relative to the ash discharge opening.

[0081] Understandably, this embodiment compensates for the pixel-to-physical conversion result by introducing a perspective error correction coefficient, thereby eliminating measurement errors caused by camera installation tilt, lens distortion, and vehicle parking offset, ensuring the accuracy of the physical deviation conversion result, and making the deviation value in the final output alignment guidance information accurate and reliable.

[0082] In some embodiments of the present invention, the alignment guidance information includes voice command instructions and / or visual guidance information; the voice command instructions are generated based on TTS speech synthesis technology, and include deviation broadcast instructions, step-by-step operation guidance instructions, alignment success prompt instructions, and abnormal situation warning instructions; the visual guidance information is displayed through the operator's mobile terminal, and the mobile terminal displays the relative position, offset direction, offset distance, and operation status information of the ash truck tank opening and the ash discharge opening.

[0083] Voice commands refer to voice prompts generated by text-to-speech synthesis technology and broadcast to operators. TTS (Text-to-Speech) technology can convert text content into natural and fluent speech output.

[0084] Visual guidance information refers to visual prompts that display the relative position and deviation information between the tank opening and the ash discharge port through a graphical interface.

[0085] Specifically, the voice command instructions are generated based on TTS speech synthesis technology and include deviation broadcast instructions, step-by-step operation guidance instructions, alignment success prompt instructions, and abnormal situation warning instructions. The deviation broadcast instructions inform the operator of the direction and distance of the deviation between the tank opening and the ash discharge port. The step-by-step operation guidance instructions provide sequential prompts for the next step in the adjustment process. The alignment success prompt instructions send a confirmation signal to the operator after the tank opening and ash discharge port are aligned. The abnormal situation warning instructions issue alerts when image anomalies are detected, recognition fails, or the deviation exceeds a preset threshold. All voice instructions are configured according to power plant safety operation specifications, and the instructions are concise and clear.

[0086] Visual guidance information is displayed via the operator's mobile terminal. The mobile terminal has a corresponding application installed, which communicates with the data service layer to receive and display alignment guidance information in real time. The information displayed on the mobile terminal includes the relative position of the ash truck's tank opening and discharge opening, the direction of offset, the offset distance, and the operational status information.

[0087] Furthermore, the alignment guidance information also includes an alignment completion prompt and / or ash discharge permission command generated when the physical deviation is less than a preset threshold. Specifically, when the calculated physical deviation is less than the preset threshold, it is determined that the tank opening and the ash discharge port are aligned. At this time, an alignment completion prompt command is generated and a confirmation signal of alignment completion is sent to the operator via voice broadcast and / or mobile terminal display. Simultaneously, an ash discharge permission command can be generated. This command is used to unlock the ash discharge control valve or send a signal to the power plant management system allowing ash discharge, ensuring that ash discharge operation is only initiated when the tank opening and the ash discharge port are aligned. The ash discharge permission command can specifically notify the operator that ash discharge can begin via voice broadcast, or it can directly send a permission signal to the ash discharge control equipment via signal transmission, or both methods can be used simultaneously.

[0088] Understandably, this embodiment outputs alignment guidance information to operators through a combination of voice commands and visual guidance information. Voice commands allow operators to obtain adjustment guidance in real time during operation without looking at the screen. Visual guidance information enables operators to intuitively understand the relative positional relationship between the tank opening and the ash discharge port. The two output methods complement each other, ensuring that operators can accurately obtain alignment guidance information under different operating conditions. The alignment completion prompt command allows operators to clearly know that alignment is complete, avoiding delays in the operation process due to uncertainty about the alignment status. The ash discharge permission command is triggered only when the physical deviation is less than a preset threshold, ensuring that the tank opening and the ash discharge port are already aligned when the ash discharge operation starts.

[0089] In some embodiments of the present invention, the method further includes: when the alignment of the ash truck tank opening is detected, uploading the vehicle information, ash release task information, alignment result and operation record of the ash truck to the power plant management system.

[0090] The power plant management system refers to an information platform used by thermal power plants to manage the entire process of by-product transportation and sales. This platform must at least have functions such as vehicle information management, ash discharge task scheduling, and operation data recording and statistics. This management system communicates with the aforementioned visual alignment guidance system via a data interface, enabling two-way data interaction. The power plant management system can acquire various types of operation data uploaded by the alignment guidance system and can also send vehicle information and ash discharge task information to the alignment guidance system.

[0091] Specifically, after the ash truck's tank opening and ash discharge port are aligned, the system automatically triggers a data upload operation. The uploaded data includes, but is not limited to: vehicle information for the ash truck, such as license plate number, vehicle model, and affiliated transportation company; ash discharge task information, such as task number, ash type, and planned ash discharge volume; alignment results, such as alignment completion time and final deviation value; and a complete record of the operation, such as start time, end time, and operator information. This data is uploaded in structured data format to the power plant management system's database for storage via the communication interface between the alignment guidance system and the power plant management system.

[0092] Furthermore, the data upload process can be executed automatically at the same time as alignment completion, without manual intervention. Operators do not need to manually record or fill in any alignment operation information; the system automatically completes the entire process of data collection, processing, and uploading.

[0093] Understandably, this embodiment automatically uploads vehicle information, task information, alignment results, and operation records to the power plant management system after alignment is completed. This achieves automatic collection and centralized storage of alignment operation data, ensuring that each alignment operation has complete data records available for traceability and query. Simultaneously, the uploaded alignment results serve as a prerequisite for ash discharge operations. The power plant management system only allows the ash discharge valve to be opened after receiving a valid alignment completion record, thus avoiding spillage problems caused by starting ash discharge before alignment is complete, and improving the standardization and safety of ash loading operations. Furthermore, the automatic uploading of operation data allows power plant managers to monitor the alignment status and operation progress of each ash truck without being physically present on-site, providing objective data support for subsequent vehicle scheduling, sales statistics, and performance evaluation.

[0094] In one specific implementation, such as Figure 3The diagram shows the layout of a visual alignment guidance system for ash truck openings in a thermal power plant. A high-definition industrial camera is positioned above the ash discharge port, its field of view covering the ash truck parking area, used to capture images of the work area including the ash truck opening. Figure 3 The image also shows an edge computing gateway that communicates with a camera to receive image data captured by the camera and run a target detection model to identify the can opening and calculate deviations. Figure 3 The document further illustrates a voice broadcasting device and a mobile app terminal. The voice broadcasting device is used to broadcast voice commands to the ash truck operators, while the mobile app terminal is used to display the relative position and offset information between the tank opening and the ash discharge opening to the operators. Figure 3 The document also illustrates the communication connection between the data service layer and the power plant management system. The data service layer interacts with the edge computing gateway, voice broadcasting equipment, mobile APP terminal, and power plant management system.

[0095] In one specific implementation, such as Figure 4 The figure shows a schematic diagram of the pixel deviation between the center of the ash truck tank opening and the center of the ash discharge port. The figure illustrates the ash discharge operation area below ash storage silo No. 3. Using the position of the ash discharge port center in the image as a reference point, the position of the ash truck tank opening center detected by the target detection model is offset in both the horizontal and vertical directions relative to this reference point. The horizontal offset reflects the degree of deviation of the tank opening center from the ash discharge port center in the left-right direction of the image, while the vertical offset reflects the degree of deviation of the tank opening center from the ash discharge port center in the up-down direction of the image. Figure 4 The diagram also shows the correspondence between the above pixel deviations and physical deviations based on camera calibration parameters. The horizontal physical deviation is obtained by multiplying the horizontal pixel deviation by the physical width of a single pixel, and the vertical physical deviation is obtained by multiplying the vertical pixel deviation by the physical height of a single pixel.

[0096] In one specific implementation, as shown in Table 1, key operational indicators are compared between the traditional manual alignment mode before system commissioning and the AI ​​vision plus voice command alignment mode after system commissioning.

[0097] Table 1. Comparison of operational performance between traditional manual alignment mode and AI vision plus voice command alignment mode

[0098] As shown in Table 1, the average alignment time is approximately 10 minutes using the traditional manual method, while it is reduced to about 3 minutes using the method described in this embodiment, a reduction of about 70%. This significant reduction in alignment time directly improves the efficiency of single-vehicle operations, reduces the waiting time of ash trucks at the ash discharge station, and allows the same ash discharge port to serve more vehicles per unit time.

[0099] Regarding the ash spillage rate, the traditional manual method resulted in a spillage rate of 9.8%, while the method described in this embodiment reduced the spillage rate to 2.1%, a reduction of approximately 78.6%. This significant decrease in spillage rate not only reduces material loss of fly ash but also lowers the frequency of on-site cleanup and environmental treatment costs caused by spillage.

[0100] Regarding on-site command personnel, the traditional model requires dedicated command personnel to be stationed near the ash discharge port. However, with the method of this embodiment, no on-site command personnel are required to complete the alignment, thus completely eliminating the safety hazards of command personnel being exposed in high dust areas from the perspective of personnel allocation.

[0101] In terms of operational efficiency, under the traditional model, a single ash discharge port can complete the alignment and ash discharge operations for 4 to 5 ash trucks per hour. Using the method described in this embodiment, the efficiency can be increased to 10 to 12 trucks per hour, more than doubling the operational efficiency. This improvement allows power plants to complete more fly ash transportation tasks within the same timeframe, effectively releasing the capacity for by-product transportation and sales.

[0102] Regarding driver operational difficulty and data traceability, in the traditional model, drivers need to repeatedly adjust the vehicle position based on the gestures or shouts of the commanding personnel, which is labor-intensive and prone to errors. Furthermore, there is no data recording during the entire alignment process. Using the method in this embodiment, drivers can independently complete alignment based solely on the guidance information output by the system, significantly reducing operational difficulty. Moreover, vehicle information, task information, deviation data, and alignment results for each alignment operation are automatically recorded and uploaded to the power plant management system, achieving complete traceability and intelligent management of operational data.

[0103] Understandably, through the comparison of the above six dimensions, it can be determined that the visual alignment guidance method and system for ash truck openings in thermal power plants provided in this embodiment have made significant technological progress in improving alignment efficiency, reducing material spillage, eliminating safety hazards, increasing vehicle turnover rate, improving operating experience, and realizing closed-loop data management.

[0104] To better implement the visual alignment guidance method for ash truck openings in thermal power plants in this embodiment of the invention, based on the visual alignment guidance method for ash truck openings in thermal power plants, correspondingly, as follows: Figure 5 As shown, this embodiment of the invention also provides a visual alignment guidance system for ash truck openings in thermal power plants. The visual alignment guidance system 500 for ash truck openings in thermal power plants includes: Image acquisition layer 501 includes a camera deployed above the ash discharge port for real-time acquisition of images of the work area including the ash truck opening; AI recognition layer 502 includes an edge computing gateway. The edge computing gateway runs a target detection model, which is used to receive the image of the work area and perform anti-dust image enhancement preprocessing on the image of the work area, detect the position of the ash truck can opening in the image, output the center pixel coordinates of the ash truck can opening, calculate the pixel deviation with the position of the ash opening center in the image as a reference point, and convert the pixel deviation into physical deviation according to the calibration parameters of the camera. Data service layer 503 is used to generate alignment guidance information based on the physical deviation; Terminal application layer 504 is used to output the alignment guidance information to the operator of the gray vehicle.

[0105] In some embodiments of the present invention, a power plant management system communicatively connected to the data service layer is also included; the power plant management system is used to acquire vehicle information and ash release task information, as well as to transmit alignment results and operation records back.

[0106] The visual alignment guidance system 500 for ash truck openings in thermal power plants provided in the above embodiments can realize the technical solutions described in the embodiments of the visual alignment guidance method for ash truck openings in thermal power plants. The specific implementation principles of each module or unit can be found in the corresponding content in the embodiments of the visual alignment guidance method for ash truck openings in thermal power plants, and will not be repeated here.

[0107] The above provides a detailed description of the visual alignment guidance method and system for ash truck openings in thermal power plants provided by this invention. Specific examples have been used to illustrate the principles and implementation methods of this invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this invention. At the same time, those skilled in the art will recognize that there will be changes in the specific implementation methods and application scope based on the ideas of this invention. Therefore, the content of this specification should not be construed as a limitation of this invention.

Claims

1. A visual alignment guiding method for a coal-fired power plant ash car tank mouth, characterized in that, The method includes: Acquire images of the work area captured by a camera positioned above the ash discharge port, the images of the work area including the ash truck opening; The image of the work area is subjected to anti-dust image enhancement preprocessing to obtain the enhanced image; The enhanced image is input into a trained target detection model to detect the position of the gray tank opening in the enhanced image and output the center pixel coordinates of the gray tank opening. Using the position of the center of the ash discharge port in the image as a reference point, calculate the pixel deviation of the center pixel coordinates of the ash tank opening relative to the reference point; Based on the calibration parameters of the camera, the pixel deviation is converted into the physical deviation of the ash tank opening relative to the ash discharge opening; Alignment guidance information is generated based on the physical deviation, and the alignment guidance information is output to the operator of the gray vehicle.

2. The method of claim 1, wherein the method further comprises: The step of performing anti-dust image enhancement preprocessing on the image of the work area to obtain the enhanced image includes: The image of the work area is subjected to dark channel prior dehazing to obtain a dehazed image; The dehazed image is subjected to multi-scale Retinex image enhancement processing to obtain an initial enhanced image; The initial enhanced image is subjected to Gaussian filtering to obtain the enhanced image.

3. The method of claim 2, wherein the method further comprises: The step of performing dark channel prior dehazing on the image of the work area to obtain a dehazed image includes: Select a local window of the dark channel; Based on the dark channel local window, the minimum channel value of each pixel in the working area image is determined in the three RGB channels, and the minimum channel value is used as the dark channel value of each pixel to generate a dark channel image. Obtain the pixels with the highest brightness values ​​in the dark channel image, representing a predetermined percentage. Map these pixels to the RGB components of the working area image and calculate their average value as the atmospheric light value. The initial transmittance map is calculated based on the atmospheric light value, and the initial transmittance map is optimized by guided filtering to obtain the optimized transmittance map. Based on the atmospheric light value and the optimized transmittance map, the pixel values ​​of each pixel in the image of the working area are recovered and calculated to obtain the dehazed image.

4. The method of claim 2, wherein the method further comprises: The step of performing dark channel prior dehazing on the image of the work area to obtain a dehazed image includes: Select a local window of the dark channel; Based on the dark channel local window, the minimum channel value of each pixel in the working area image is determined in the three RGB channels, and the minimum channel value is used as the dark channel value of each pixel to generate a dark channel image. Obtain the pixels with the highest brightness values ​​in the dark channel image, representing a predetermined percentage. Map these pixels to the RGB components of the working area image and calculate their average value as the atmospheric light value. The initial transmittance map is calculated based on the atmospheric light value, and the initial transmittance map is optimized by guided filtering to obtain the optimized transmittance map. Set multiple defogging intensity parameters corresponding to dust concentration; Obtain the global contrast of the image of the working area, and / or obtain the dark channel mean of the dark channel image; Based on the global contrast and / or the dark channel mean, determine the concentration level to which the current dust concentration belongs; Based on the concentration level, select the target defogging intensity parameter corresponding to the current dust concentration from the multiple defogging intensity parameters; Based on the atmospheric light value and the optimized transmittance map, the pixel values ​​of each pixel in the image of the working area are recovered using the target defogging intensity parameter to obtain the defogging image.

5. The method of claim 1, wherein the method further comprises: The target detection model is the YOLOv8 model, and the outputs of the last two C2f modules in the backbone network of the YOLOv8 model are respectively inserted with CBAM attention modules. The CBAM attention module includes a channel attention submodule and a spatial attention submodule connected in series. The loss function of the YOLOv8 model is the CIoU loss function.

6. The method of claim 1, wherein the method further comprises: The calibration parameters include resolution and installation height; the step of converting the pixel deviation into a physical deviation of the ash tank opening relative to the ash discharge opening based on the camera's calibration parameters includes: Based on the resolution and installation height of the camera, determine the physical width and physical height corresponding to a single pixel in the image of the work area; The horizontal pixel deviation is obtained by multiplying the horizontal pixel deviation by the physical width corresponding to the single pixel. The vertical pixel deviation is obtained by multiplying the physical height corresponding to the single pixel. Based on a pre-calibrated perspective error correction coefficient, the physical deviations in the horizontal direction and the vertical direction are corrected to obtain corrected physical deviations in the horizontal and vertical directions. These corrected physical deviations in the horizontal and vertical directions constitute the physical deviation of the ash tank opening relative to the ash discharge opening.

7. The method of claim 1, wherein the method further comprises: determining a distance between the camera and the target; and adjusting the distance between the camera and the target based on the determined distance. The alignment guidance information includes voice command instructions and / or visual guidance information; The voice command instructions are generated based on TTS speech synthesis technology, and include deviation broadcast instructions, step-by-step operation guidance instructions, alignment success prompt instructions, and abnormal situation warning instructions. The visual guidance information is displayed through the operator's mobile terminal, which shows the relative position, offset direction, offset distance, and operation status information of the ash truck opening and the ash discharge opening.

8. The method of claim 1, wherein the method further comprises: determining a distance between the camera and the target; and adjusting the distance between the camera and the target. Also includes: When the alignment of the ash truck's tank opening is detected, the vehicle information, ash discharge task information, alignment result, and operation record of the ash truck are uploaded to the power plant management system.

9. A visual alignment guiding system for a coal-fired power plant ash car hopper, characterized by, The system includes: The image acquisition layer includes a camera deployed above the ash discharge port for real-time acquisition of images of the work area, including the ash truck opening; The AI ​​recognition layer includes an edge computing gateway, which runs a target detection model to receive the work area image and perform anti-dust image enhancement preprocessing on the work area image, detect the position of the ash truck can opening in the image, output the center pixel coordinates of the ash truck can opening, calculate the pixel deviation with the position of the ash outlet center in the image as the reference point, and convert the pixel deviation into physical deviation according to the camera calibration parameters. The data service layer is used to generate alignment guidance information based on the physical deviation; The terminal application layer is used to output the alignment guidance information to the operator of the gray vehicle.

10. The visual guidance system for the ash truck tank mouth alignment of a thermal power plant according to claim 9, characterized in that, It also includes a power plant management system that is communicatively connected to the data service layer; The power plant management system is used to acquire vehicle information and ash discharge task information, as well as to transmit alignment results and operation records.