A power transmission line image defogging method and system adaptive to fog concentration
By simultaneously collecting high-definition camera and micro-meteorological data using drones and combining deep learning technology, an adaptive fog defogging method was designed. This solved the problem of unstable defogging effect in power transmission line inspection, and achieved clear image restoration and efficient inspection of power transmission lines.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHUHAI YOUKUO MICROPOWER TECHNOLOGY CO LTD
- Filing Date
- 2026-04-23
- Publication Date
- 2026-07-21
Smart Images
Figure CN122434776A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of image processing technology, specifically relating to an adaptive fog concentration method and system for defogging images of power transmission lines. Background Technology
[0002] Transmission lines are mostly distributed in complex outdoor environments. In foggy weather, water vapor and suspended particulate matter in the atmosphere will scatter and absorb light, causing problems such as reduced contrast, blurred details, and color distortion in the images of transmission lines collected by drones. In severe cases, it may even be impossible to identify defects in key components such as conductors and insulators, which directly affects the quality of inspections, may lead to missed hidden dangers, and may even cause transmission line failures.
[0003] Currently, image dehazing technology has become a key means to solve the problem of image quality in foggy weather inspections. However, when existing dehazing technologies are applied to power transmission line inspection scenarios, the following prominent pain points still exist: First, they do not fully utilize the micro-meteorological data that can be acquired synchronously during drone inspections, relying solely on the image's own information to estimate fog concentration and transmittance. This results in the dehazing effect being greatly affected by weather changes, making it difficult to adapt to complex foggy environments. Second, they do not design targeted transmittance correction strategies for the characteristics of power transmission line images, such as a large proportion of sky, thin conductors, and complex backgrounds. This makes it impossible to balance the dehazing effect of the sky area and the line area, easily leading to overexposure of the sky or blurring of line details. Third, the estimation of atmospheric light values relies heavily on local image features without combining spatiotemporal information such as shooting location and shooting time, resulting in inaccurate atmospheric light value matching, which in turn affects the color reproduction of fog-free images.
[0004] To address the aforementioned issues, this application presents a method and system for defogging transmission line images based on adaptive fog concentration. Summary of the Invention
[0005] To address the shortcomings of the prior art mentioned in the background section, this application proposes an adaptive fog concentration defogging method and system for power transmission line images. This invention integrates foggy images captured by a high-definition visible light camera mounted on a UAV with micro-meteorological measurement data synchronously collected by a micro-meteorological monitoring module, along with deep learning technology. The micro-meteorological measurement data is used to correct and compensate the initial global fog concentration parameters output by the dual-branch feature extraction network. A region-aware refined convolutional network is designed to differentially optimize the initial transmittance map, adapting to the characteristics of the sky and line areas in the power transmission line images. An atmospheric light value spatiotemporal calculation model is invoked, and accurate global atmospheric light values are obtained by combining shooting location and shooting time information, forming a collaborative innovative design to solve the problems in the background section.
[0006] Firstly, to achieve the above objectives, this application provides an adaptive fog concentration method for dehazing transmission line images, comprising the following specific steps: Step S1: Collect foggy images using a high-definition visible light camera mounted on the drone. Simultaneously, collect measured micro-meteorological data at the moment of shooting using a micro-meteorological monitoring module triggered synchronously with the high-definition visible light camera. Obtain shooting location and shooting time information. Perform size adjustment and normalization preprocessing on the foggy images to obtain normalized foggy images. Step S2: Construct a two-branch feature extraction network including a shared feature extraction layer, a depth estimation branch, and a fog density mapping branch. Input the normalized foggy image into the two-branch feature extraction network. The depth estimation branch outputs the scene depth map D(x), and the fog density mapping branch outputs the initial global fog density parameters. ; Step S3: Utilize micrometeorological measurement data to determine the initial global fog concentration parameters. Correction and compensation are performed to obtain the accurate global fog concentration parameter β, and the adaptive atmospheric scattering coefficient k is determined based on the global fog concentration parameter β. Step S4: Calculate the initial transmittance map based on the adaptive atmospheric scattering coefficient k and the scene depth map D(x). and the initial transmittance map After being stitched with a normalized hazy image along the channel dimension, it is input into a region-aware refined convolutional network to obtain a refined transmittance map t(x); Step S5: First, based on the shooting location information and shooting time information, call the pre-built atmospheric light value spatiotemporal calculation model to obtain the global atmospheric light value A that is precisely matched one-to-one with the current foggy image. Then, based on the atmospheric scattering model, use the fine transmittance map t(x) and the global atmospheric light value A to recover the fog-free image J(x). Step S6: Perform local contrast adaptive detail enhancement processing on the haze-free image J(x) to output the final dehaze-enhanced image.
[0007] Based on the above scheme, the preferred embodiment of the micrometeorological data includes visibility V and relative humidity RH. Step S3 specifically includes the following steps: Step S31: Convert the micrometeorological measurement data into a physical domain equivalent fog concentration factor. A negative correlation mapping was applied to visibility V, and a positive correlation mapping was applied to relative humidity RH. After normalizing the two types of parameters to the [0,1] interval, the physical domain equivalent fog concentration factor was obtained by weighted fusion. The sum of the weight coefficients of the two types of parameters is 1; Step S32: Based on the initial global fog concentration parameters Equivalent fog concentration factor in the physical domain Differences The fusion strategy is dynamically determined to obtain the accurate global fog concentration parameter β: when At that time, a weighted fusion method is used: ; when At that time, correction fusion is used: ; Where δ is the preset difference threshold, λ∈[0,1] is the fusion coefficient, and α∈[0,1] is the correction intensity coefficient; Step S33: Determine the adaptive atmospheric scattering coefficient k based on the global fog concentration parameter β. The formula for determining the adaptive atmospheric scattering coefficient k is:
[0008] in and These are the preset minimum and maximum scattering coefficients, respectively.
[0009] Preferably, based on the above scheme, the correction intensity coefficient α is based on the physical domain equivalent fog concentration factor. The difference Δ is adaptively adjusted, and the adjustment formula is:
[0010] in, This is the minimum correction strength coefficient, a preset parameter used to limit the lower limit of the correction strength, preventing insufficient compensation due to an excessively small correction amplitude. The maximum value of the correction strength coefficient is a preset parameter used to limit the upper limit of the correction strength, preventing overcorrection due to excessive correction amplitude. The maximum threshold for the difference is a preset parameter used to limit the normalization range of the difference Δ, avoiding abnormal calculation of the correction intensity due to an excessively large denominator.
[0011] Based on the above scheme, the preferred option is the initial transmittance map described in step S4. The calculation formula is:
[0012] in, is a natural exponential function used to describe the exponential decay law of atmospheric scattering, and k is an adaptive atmospheric scattering coefficient, which is adaptively determined by the global fog concentration parameter β, reflecting the degree of scattering of light by fog in the current environment; The region-aware refined convolutional network described in step S4 learns, through a data-driven approach, to differentiate the transmittance of bright sky regions and darker non-sky regions in the image. This ensures that the refined transmittance map t(x) automatically acquires a higher transmittance value in the sky region than the initial transmittance, while maintaining a continuous and smooth transition at the region boundaries. Furthermore, the training loss function of the region-aware refined convolutional network includes a region-aware regularization term, based on the semantic mask of the sky region. The network is guided to enhance and correct the transmittance of the sky region.
[0013] Based on the above scheme, the preferred embodiment of the dual-branch feature extraction network is that the depth estimation branch consists of multiple residual modules and an output convolutional layer, which outputs a pixel-level normalized scene depth map D(x). The fog concentration mapping branch consists of a global average pooling layer and at least one fully connected layer, and outputs the initial global fog concentration parameters. ,and .
[0014] Based on the above scheme, the preferred method is to recover the haze-free image J(x) in step S5 pixel by pixel using the following formula:
[0015] Where x represents the spatial coordinates of a pixel in the image. The original brightness value of the c-th color channel at pixel position x in the captured hazy image. The brightness value of the c-th color channel at pixel position x in the restored haze-free image. The color channels are represented by red, green, and blue channels respectively. t(x) is the fine transmittance shared by all channels. Ac is the c-th channel component of the global atmospheric light value, representing the intensity of ambient light in the fog. t0 is the lower limit threshold of transmittance, usually taken as 0.05 to 0.1, to prevent the denominator from approaching 0, which would lead to calculation overflow and noise amplification.
[0016] Based on the above scheme, the preferred local contrast adaptive detail enhancement processing in step S6 specifically involves using the CLAHE algorithm to perform adaptive histogram equalization on the brightness channel of the haze-free image, while combining guided filtering to suppress noise amplification, thereby achieving a balance between detail enhancement and noise suppression.
[0017] Secondly, this application provides an adaptive fog concentration transmission line image defogging system, which specifically includes: a UAV image and micro-meteorological synchronous acquisition module, an image preprocessing module, a dual-branch feature extraction module, a micro-meteorological measured compensation module, a regional adaptive transmittance estimation module, a spatiotemporal accurate atmospheric light estimation module, an image restoration module, a detail enhancement module, and an output module. The UAV image and micro-meteorological synchronous acquisition module includes a high-definition visible light camera and a micro-meteorological monitoring module, which are used to synchronously acquire foggy images, micro-meteorological measured data at the moment of shooting, shooting location information and shooting time information; The image preprocessing module is used to resize and normalize the foggy image; The dual-branch feature extraction module includes a shared feature extraction layer, a depth estimation branch, and a fog concentration mapping branch, which are used to output a scene depth map and initial global fog concentration parameters. The micro-meteorological measurement compensation module includes a physical domain equivalent calculation unit and an adaptive correction compensation unit, which are used to perform correction compensation using the micro-meteorological measurement data and output accurate global fog concentration parameters. The region adaptive transmittance estimation module includes a scattering coefficient calculation unit, an initial transmittance calculation unit, and a region perception fine-tuning network unit, which are used to generate a fine transmittance map based on global fog concentration parameters and scene depth map. The spatiotemporal precise atmospheric light estimation module is used to obtain a one-to-one precise matching global atmospheric light value A based on the shooting location and time information; The image restoration module is used to restore the haze-free image J(x) based on the atmospheric scattering model, using fine transmittance t(x) and A. The detail enhancement module is used to perform local contrast adaptive detail enhancement processing on the fog-free image J(x); The output module is used to output the final dehazing and enhanced image; The physical domain equivalent calculation unit in the micro-meteorological measurement compensation module is used to convert micro-meteorological measurement data into a physical domain equivalent fog concentration factor. The adaptive correction compensation unit is used to dynamically select a fusion strategy based on the difference between the initial global fog concentration parameter and the physical domain equivalent fog concentration factor to generate an accurate global fog concentration parameter β. The region-aware refined network unit introduces region-aware regularization loss during training to guide the network to make differential corrections to the transmittance of sky regions and non-sky regions.
[0018] Thirdly, this application provides an electronic device, including: a processor and a memory, wherein the memory stores a computer program that can be called by the processor; The processor executes the aforementioned method for defogging transmission line images based on adaptive fog concentration by calling a computer program stored in the memory.
[0019] Fourthly, this application provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the aforementioned method for dehazing transmission line images with adaptive fog concentration.
[0020] Compared with the prior art, the beneficial effects of the present invention are: I. This invention introduces micro-meteorological measurement data to correct and compensate the initial fog concentration parameters, combining image perception with physical measurement. This solves the problem that existing methods rely solely on image information to estimate fog concentration and transmittance, leading to unstable defogging effects in extreme foggy weather. It makes the estimation of fog concentration and transmittance more consistent with the real environment, significantly improving the defogging robustness under different foggy weather scenarios and ensuring that the details of key components of power transmission lines are clearly distinguishable.
[0021] Second, this invention designs a refined convolutional network for region awareness. Addressing the characteristics of large sky areas and dark line areas in transmission line images, it differentially corrects the initial transmittance map, ensuring the sky area obtains a reasonable transmittance value and avoiding overexposure. Simultaneously, it ensures thorough defogging of the line area, preserving the detailed features of small components such as conductors and fittings. This overcomes the technical bottleneck of existing methods that cannot simultaneously achieve defogging effects for both types of areas. Furthermore, based on the shooting location and time information, it invokes a spatiotemporal calculation model of atmospheric light values to obtain a global atmospheric light value uniquely corresponding to the current shooting environment. This avoids the inaccurate estimation problems caused by bright spots, white conductors, etc., when existing methods rely on local image features to estimate atmospheric light values. This ensures that the restored fog-free image has natural colors and high fidelity, meeting the visual inspection requirements of transmission line inspections.
[0022] Third, the collection of micro-meteorological data is achieved by using a drone equipped with a conventional micro-meteorological monitoring module. The training of the network model can be completed based on the existing foggy image dataset of transmission lines, without the need to build additional complex hardware equipment. At the same time, through the design of adaptive scattering coefficient and dynamic fog concentration compensation, the solution can be adapted to transmission line inspection scenarios in different regions and seasons, which is significantly better than the existing single defogging method. It can be deployed on the edge computing platform of the drone to realize real-time defogging of inspection images without manual intervention, which can greatly improve the efficiency of foggy inspection and meet the engineering application requirements of intelligent inspection of transmission lines. Attached Figure Description
[0023] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a schematic diagram of the overall process of a method for defogging transmission line images with adaptive fog concentration according to the present invention. Figure 2 This is a flowchart of a method for defogging transmission line images based on adaptive fog concentration according to the present invention. Figure 3 This is an architectural diagram of an adaptive fog concentration image defogging system for power transmission lines according to the present invention. Detailed Implementation
[0024] 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 embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.
[0025] Example 1: To solve the technical problems mentioned in the background art, this application provides a preferred embodiment: as follows Figures 1-3 As shown, an adaptive fog concentration method for dehazing transmission line images includes the following specific steps: Step S1: Collect foggy images using a high-definition visible light camera mounted on the drone. Simultaneously, collect measured micro-meteorological data at the moment of shooting using a micro-meteorological monitoring module triggered synchronously with the high-definition visible light camera. Obtain shooting location and shooting time information. Perform size adjustment and normalization preprocessing on the foggy images to obtain normalized foggy images. Step S2: Construct a two-branch feature extraction network including a shared feature extraction layer, a depth estimation branch, and a fog density mapping branch. Input the normalized foggy image into the two-branch feature extraction network. The depth estimation branch outputs the scene depth map D(x), and the fog density mapping branch outputs the initial global fog density parameters. ; Step S3: Utilize micrometeorological measurement data to determine the initial global fog concentration parameters. Correction and compensation are performed to obtain the accurate global fog concentration parameter β, and the adaptive atmospheric scattering coefficient k is determined based on the global fog concentration parameter β. Step S4: Calculate the initial transmittance map based on the adaptive atmospheric scattering coefficient k and the scene depth map D(x). and the initial transmittance map After being stitched with a normalized hazy image along the channel dimension, it is input into a region-aware refined convolutional network to obtain a refined transmittance map t(x); Step S5: First, based on the shooting location information and shooting time information, call the pre-built atmospheric light value spatiotemporal calculation model to obtain the global atmospheric light value A that is precisely matched one-to-one with the current foggy image. Then, based on the atmospheric scattering model, use the fine transmittance map t(x) and the global atmospheric light value A to recover the fog-free image J(x). Step S6: Perform local contrast adaptive detail enhancement processing on the haze-free image J(x) to output the final dehaze-enhanced image.
[0026] The advantages of this embodiment compared to existing technologies are as follows: By integrating foggy images captured by a high-definition visible light camera mounted on a drone with micro-meteorological measurement data synchronously collected by a micro-meteorological monitoring module and deep learning technology, the initial global fog concentration parameters output by the dual-branch feature extraction network are corrected and compensated using the micro-meteorological measurement data. A refined convolutional network for region perception is designed to perform differential optimization of the initial transmittance map, adapting to the characteristics of the sky and line areas in the transmission line images. An atmospheric light value spatiotemporal calculation model is invoked, and accurate global atmospheric light values are obtained by combining the shooting location information and shooting time information, forming a collaborative and innovative design. This effectively solves the problems of blurred and detail-loss images in foggy weather inspections, enabling inspection personnel to clearly identify defects in transmission lines such as broken conductor strands and damaged insulators. It can effectively reduce the rate of missed detection of hidden dangers and provide reliable technical support for the safe operation and maintenance of transmission lines, with significant engineering application value and economic benefits.
[0027] Furthermore: In an optional embodiment, the micrometeorological data includes visibility V and relative humidity RH, and step S3 specifically includes the following steps: Step S31: Convert the micrometeorological measurement data into a physical domain equivalent fog concentration factor. A negative correlation mapping was applied to visibility V, and a positive correlation mapping was applied to relative humidity RH. After normalizing the two types of parameters to the [0,1] interval, the physical domain equivalent fog concentration factor was obtained by weighted fusion. The sum of the weight coefficients of the two types of parameters is 1; Step S32: Based on the initial global fog concentration parameters Equivalent fog concentration factor in the physical domain Differences The fusion strategy is dynamically determined to obtain the accurate global fog concentration parameter β: when At that time, a weighted fusion method is used: ; when At that time, correction fusion is used: ; Where δ is the preset difference threshold, λ∈[0,1] is the fusion coefficient, and α∈[0,1] is the correction intensity coefficient; It should be noted that: when At the same time, by using weighted fusion to take into account the advantages of both types of data, it preserves the scene details of the image while incorporating the environmental constraints of physical measurements, thus improving the stability of fog concentration estimation. The weights of image data and measured data can be flexibly adjusted through the fusion coefficient λ to adapt to the trust requirements of different scenarios. At the same time, the image estimation is used as the benchmark, and the measured data is used for correction to avoid the distortion of pure image estimation. At the same time, the scene adaptability of the image is preserved. The correction intensity coefficient α is used to control the correction amplitude to avoid the fog concentration estimation deviating from the actual scene due to overcorrection. Step S33: Determine the adaptive atmospheric scattering coefficient k based on the global fog concentration parameter β. The formula for determining the adaptive atmospheric scattering coefficient k is:
[0028] in and These are the preset minimum and maximum scattering coefficients, respectively.
[0029] It should be noted that the negative correlation mapping formula in step S31 is:
[0030] Where V is the visibility value measured by the micro-meteorological monitoring module, Vmax is the preset upper limit of visibility (10,000 meters, corresponding to extremely clear weather), and Vmin is the preset lower limit of visibility (50 meters, corresponding to dense fog weather). The normalized visibility concentration factor has its output value strictly limited to [0, 1]. The lower the visibility, the closer the value is to 1. The positive correlation mapping formula in step S31 is:
[0031] Wherein, RH is the relative humidity value measured by the micro-meteorological monitoring module, in %, RHmax is the preset upper limit of relative humidity, preferably 100%, and RHmin is the preset lower limit of relative humidity, preferably 40%. Humidity below this value is considered to have no significant contribution to fog formation. The normalized fog concentration factor is calculated from relative humidity, and its value ranges from [0, 1]. The higher the humidity, the closer its value is to 1.
[0032] In an optional embodiment, the correction intensity coefficient α is based on the physical domain equivalent fog concentration factor. The difference Δ is adaptively adjusted, and the adjustment formula is:
[0033] in, This is the minimum correction strength coefficient, a preset parameter used to limit the lower limit of the correction strength, preventing insufficient compensation due to an excessively small correction amplitude. The maximum value of the correction strength coefficient is a preset parameter used to limit the upper limit of the correction strength, preventing overcorrection due to excessive correction amplitude. The maximum threshold for the difference is a preset parameter used to limit the normalization range of the difference Δ, avoiding abnormal calculation of the correction intensity due to an excessively large denominator.
[0034] It should be noted that the adjustment formula for the correction strength coefficient α is based on the logic that the larger the deviation, the higher the correction strength. When Δ is small, the correction strength is close to... This avoids excessive intervention; when Δ is large, the correction intensity is gradually increased. This allows for full utilization of the correction function of measured data. This is used to map the actual difference Δ to the [0, 1] interval, ensuring that the change in the correction intensity coefficient adaptively adjusts with the difference Δ. It is used to reflect the degree of deviation between the image-estimated fog concentration and the measured physical fog concentration.
[0035] In an optional embodiment, the initial transmittance map in step S4 The calculation formula is:
[0036] in, is a natural exponential function used to describe the exponential decay law of atmospheric scattering, and k is an adaptive atmospheric scattering coefficient, which is adaptively determined by the global fog concentration parameter β, reflecting the degree of scattering of light by fog in the current environment; It should be noted that: The calculation formula adopts an exponential decay model to accurately depict the objective law of light propagation in fog. The farther the transmission distance and the higher the fog concentration, the lower the light transmittance, ensuring that the initial transmittance estimate conforms to the real physical characteristics. At the same time, two key variables, adaptive atmospheric scattering coefficient k and scene depth D(x), are introduced to achieve joint constraints of fog concentration and spatial distance, making the transmittance estimate closer to the real fog scene.
[0037] In step S4, the region-aware refined convolutional network learns, through data-driven methods, to differentiate the transmittance of bright sky regions and darker non-sky regions in the image. This ensures that the refined transmittance map t(x) automatically acquires a higher transmittance value in the sky region than the initial transmittance, while maintaining a continuous and smooth transition at the region boundaries. Furthermore, the training loss function of the region-aware refined convolutional network includes a region-aware regularization term, based on the semantic mask of the sky region. The network is guided to enhance and correct the transmittance of the sky region.
[0038] In an optional embodiment, the depth estimation branch in the dual-branch feature extraction network consists of multiple residual modules and an output convolutional layer, outputting a pixel-level normalized scene depth map D(x), and The fog concentration mapping branch consists of a global average pooling layer and at least one fully connected layer, outputting the initial global fog concentration parameters. ,and .
[0039] Furthermore: In an optional embodiment, the haze-free image J(x) described in step S5 is recovered pixel-by-pixel using the following formula:
[0040] Where x represents the spatial coordinates of a pixel in the image. The original brightness value of the c-th color channel at pixel position x in the captured hazy image. The brightness value of the c-th color channel at pixel position x in the restored haze-free image. The color channels are represented by red, green, and blue channels respectively. t(x) is the fine transmittance shared by all channels. Ac is the c-th channel component of the global atmospheric light value, representing the intensity of ambient light in the fog. t0 is the lower limit threshold of transmittance, usually taken as 0.05 to 0.1, to prevent the denominator from approaching 0, which would lead to calculation overflow and noise amplification.
[0041] It should be noted that the above pixel-by-pixel restoration formula strictly follows the physical model of fog imaging, and uses inverse transformation to solve for the fog-free image to ensure the physical correctness of the restoration result. At the same time, the R, G, and B channels are calculated separately to ensure the accuracy of color reproduction of the fog-free image, which can avoid color deviation problems and is more suitable for the visual inspection needs of power transmission line inspection.
[0042] In an optional embodiment, the local contrast adaptive detail enhancement processing in step S6 specifically involves using the CLAHE algorithm to perform adaptive histogram equalization on the brightness channel of the haze-free image, while combining guided filtering to suppress noise amplification, thereby achieving a balance between detail enhancement and noise suppression.
[0043] Example 2: Based on the same inventive concept as Example 1, such as Figure 3 As shown, this embodiment provides an adaptive fog concentration transmission line image defogging system, which specifically includes: a UAV image and micro-meteorological synchronous acquisition module, an image preprocessing module, a dual-branch feature extraction module, a micro-meteorological measured compensation module, a regional adaptive transmittance estimation module, a spatiotemporally accurate atmospheric light estimation module, an image restoration module, a detail enhancement module, and an output module. The UAV image and micro-meteorological synchronous acquisition module includes a high-definition visible light camera and a micro-meteorological monitoring module, which are used to simultaneously acquire foggy images, measured micro-meteorological data at the moment of shooting, shooting location information, and shooting time information; The image preprocessing module is used to resize and normalize the foggy image; The dual-branch feature extraction module includes a shared feature extraction layer, a depth estimation branch, and a fog concentration mapping branch, which are used to output scene depth maps and initial global fog concentration parameters. The micro-meteorological measurement compensation module includes a physical domain equivalent calculation unit and an adaptive correction compensation unit, which are used to perform correction compensation using the micro-meteorological measurement data and output accurate global fog concentration parameters. The region adaptive transmittance estimation module includes a scattering coefficient calculation unit, an initial transmittance calculation unit, and a region-aware fine-tuning network unit, which is used to generate a fine transmittance map based on global fog concentration parameters and scene depth map. The spatiotemporal precise atmospheric light estimation module is used to obtain a one-to-one precise matching of the global atmospheric light value A based on the shooting location and time information; The image restoration module is used to restore the haze-free image J(x) based on the atmospheric scattering model, using fine transmittance t(x) and A. The detail enhancement module is used to perform local contrast-adaptive detail enhancement processing on the haze-free image J(x); The output module is used to output the final dehazed and enhanced image; The physical domain equivalent calculation unit in the micro-meteorological measurement compensation module is used to convert micro-meteorological measurement data into physical domain equivalent fog concentration factors, and the adaptive correction compensation unit is used to dynamically select the fusion strategy based on the difference between the initial global fog concentration parameter and the physical domain equivalent fog concentration factor to generate accurate global fog concentration parameter β. The region-aware refinement network unit introduces region-aware regularization loss during training to guide the network to differentiate the transmittance of sky regions and non-sky regions.
[0044] The parameters and steps for implementing the corresponding functions of each unit module in the above-described adaptive fog concentration transmission line image defogging system of the present invention can be referred to the parameters and steps in the embodiments of the adaptive fog concentration transmission line image defogging method described above, and will not be repeated here.
[0045] Example 3: Based on the same inventive concept as Example 1, this example provides an electronic device, including: a processor and a memory, wherein the memory stores a computer program that can be called by the processor; The processor executes the aforementioned method for defogging transmission line images based on adaptive fog concentration by calling a computer program stored in the memory.
[0046] It should be noted that all computer programs for an adaptive fog concentration method for defogging transmission line images are implemented in C language.
[0047] Example 4: Based on the same inventive concept as Example 1, this example proposes a computer-readable storage medium storing an erasable and rewritable computer program thereon; When the computer program runs on the computer device, it causes the computer device to execute the aforementioned method for dehazing transmission line images with adaptive fog concentration.
[0048] For example, computer-readable storage media can be read-only memory, random access memory, read-only optical disc, magnetic tape, floppy disk, and optical data storage devices.
[0049] The various embodiments in this invention are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on its differences from other embodiments. In particular, the embodiments for IoT devices and media are relatively simple in description because they are fundamentally similar to the method embodiments; relevant parts can be referred to the descriptions in the method embodiments.
[0050] The systems, media, and methods provided in the embodiments of the present invention are in one-to-one correspondence. Therefore, the systems and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the systems and media will not be repeated here.
[0051] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0052] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0053] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0054] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0055] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.
[0056] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.
[0057] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.
[0058] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. The above are merely embodiments of the present invention and are not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.
Claims
1. A method for dehazing transmission line images with adaptive fog concentration, characterized in that, Includes the following steps: Step S1: Collect foggy images using a high-definition visible light camera mounted on the drone. Simultaneously, collect measured micro-meteorological data at the moment of shooting using a micro-meteorological monitoring module triggered synchronously with the high-definition visible light camera. Obtain shooting location and shooting time information. Perform size adjustment and normalization preprocessing on the foggy images to obtain normalized foggy images. Step S2: Construct a two-branch feature extraction network including a shared feature extraction layer, a depth estimation branch, and a fog density mapping branch. Input the normalized foggy image into the two-branch feature extraction network. The depth estimation branch outputs the scene depth map D(x), and the fog density mapping branch outputs the initial global fog density parameters. ; Step S3: Utilize micrometeorological measurement data to determine the initial global fog concentration parameters. Correction and compensation are performed to obtain the accurate global fog concentration parameter β, and the adaptive atmospheric scattering coefficient k is determined based on the global fog concentration parameter β. Step S4: Calculate the initial transmittance map based on the adaptive atmospheric scattering coefficient k and the scene depth map D(x). and the initial transmittance map After being stitched with a normalized hazy image along the channel dimension, it is input into a region-aware refined convolutional network to obtain a refined transmittance map t(x); Step S5: First, based on the shooting location information and shooting time information, call the pre-built atmospheric light value spatiotemporal calculation model to obtain the global atmospheric light value A that is precisely matched one-to-one with the current foggy image. Then, based on the atmospheric scattering model, use the fine transmittance map t(x) and the global atmospheric light value A to recover the fog-free image J(x). Step S6: Perform local contrast adaptive detail enhancement processing on the haze-free image J(x) to output the final dehaze-enhanced image.
2. The method for dehazing transmission line images with adaptive fog concentration according to claim 1, characterized in that: The measured micrometeorological data includes visibility V and relative humidity RH. Step S3 specifically includes the following steps: Step S31: Convert the micrometeorological measurement data into a physical domain equivalent fog concentration factor. A negative correlation mapping was applied to visibility V, and a positive correlation mapping was applied to relative humidity RH. After normalizing the two types of parameters to the [0,1] interval, the physical domain equivalent fog concentration factor was obtained by weighted fusion. The sum of the weight coefficients of the two types of parameters is 1; Step S32: Based on the initial global fog concentration parameters Equivalent fog concentration factor in the physical domain Differences The fusion strategy is dynamically determined to obtain the accurate global fog concentration parameter β: when At that time, a weighted fusion method is used: ; when At that time, correction fusion is used: ; Where δ is the preset difference threshold, λ∈[0,1] is the fusion coefficient, and α∈[0,1] is the correction intensity coefficient; Step S33: Determine the adaptive atmospheric scattering coefficient k based on the global fog concentration parameter β. The formula for determining the adaptive atmospheric scattering coefficient k is: ; in and These are the preset minimum and maximum scattering coefficients, respectively.
3. The method for dehazing transmission line images with adaptive fog concentration according to claim 2, characterized in that: The correction intensity coefficient α is based on the physical domain equivalent fog concentration factor. The difference Δ is adaptively adjusted, and the adjustment formula is: ; in, This is the minimum correction strength coefficient, a preset parameter used to limit the lower limit of the correction strength, preventing insufficient compensation due to an excessively small correction amplitude. The maximum value of the correction strength coefficient is a preset parameter used to limit the upper limit of the correction strength, preventing overcorrection due to excessive correction amplitude. The maximum threshold for the difference is a preset parameter used to limit the normalization range of the difference Δ, avoiding abnormal calculation of the correction intensity due to an excessively large denominator.
4. The method for dehazing transmission line images with adaptive fog concentration according to claim 1, characterized in that: The initial transmittance map described in step S4 The calculation formula is: ; in, is a natural exponential function used to describe the exponential decay law of atmospheric scattering, and k is an adaptive atmospheric scattering coefficient, which is adaptively determined by the global fog concentration parameter β, reflecting the degree of scattering of light by fog in the current environment; The region-aware refined convolutional network described in step S4 learns, through a data-driven approach, to differentiate the transmittance of bright sky regions and darker non-sky regions in the image. This ensures that the refined transmittance map t(x) automatically acquires a higher transmittance value in the sky region than the initial transmittance, while maintaining a continuous and smooth transition at the region boundaries. Furthermore, the training loss function of the region-aware refined convolutional network includes a region-aware regularization term, based on the semantic mask of the sky region. The network is guided to enhance and correct the transmittance of the sky region.
5. The method for dehazing transmission line images with adaptive fog concentration according to claim 1, characterized in that: The depth estimation branch in the dual-branch feature extraction network consists of multiple residual modules and an output convolutional layer, outputting a pixel-level normalized scene depth map D(x), and The fog concentration mapping branch consists of a global average pooling layer and at least one fully connected layer, and outputs the initial global fog concentration parameters. ,and .
6. The method for dehazing transmission line images with adaptive fog concentration according to claim 1, characterized in that: The haze-free image J(x) described in step S5 is recovered pixel-by-pixel using the following formula: ; Where x represents the spatial coordinates of a pixel in the image. The original brightness value of the c-th color channel at pixel position x in the captured hazy image. The brightness value of the c-th color channel at pixel position x in the restored haze-free image. The color channels are represented by red, green, and blue channels respectively. t(x) is the fine transmittance shared by all channels. Ac is the c-th channel component of the global atmospheric light value, representing the intensity of ambient light in the fog. t0 is the lower limit threshold of transmittance, usually taken as 0.05 to 0.1, to prevent the denominator from approaching 0, which would lead to calculation overflow and noise amplification.
7. The method for dehazing transmission line images based on adaptive fog concentration according to claim 1, characterized in that: The local contrast adaptive detail enhancement process described in step S6 specifically involves using the CLAHE algorithm to perform adaptive histogram equalization on the brightness channel of the haze-free image, while combining guided filtering to suppress noise amplification, thereby achieving a balance between detail enhancement and noise suppression.
8. An adaptive fog concentration image dehazing system for transmission line images, implemented based on the adaptive fog concentration image dehazing method for transmission line images as described in any one of claims 1-7, characterized in that, Specifically, it includes: a UAV image and micro-meteorological synchronous acquisition module, an image preprocessing module, a bi-branch feature extraction module, a micro-meteorological measured compensation module, a regional adaptive transmittance estimation module, a spatiotemporally accurate atmospheric light estimation module, an image restoration module, a detail enhancement module, and an output module; The UAV image and micro-meteorological synchronous acquisition module includes a high-definition visible light camera and a micro-meteorological monitoring module, which are used to synchronously acquire foggy images, micro-meteorological measured data at the moment of shooting, shooting location information and shooting time information; The image preprocessing module is used to resize and normalize the foggy image; The dual-branch feature extraction module includes a shared feature extraction layer, a depth estimation branch, and a fog concentration mapping branch, which are used to output a scene depth map and initial global fog concentration parameters. The micro-meteorological measurement compensation module includes a physical domain equivalent calculation unit and an adaptive correction compensation unit, which are used to perform correction compensation using the micro-meteorological measurement data and output accurate global fog concentration parameters. The region adaptive transmittance estimation module includes a scattering coefficient calculation unit, an initial transmittance calculation unit, and a region perception fine-tuning network unit, which are used to generate a fine transmittance map based on global fog concentration parameters and scene depth map. The spatiotemporal precise atmospheric light estimation module is used to obtain a one-to-one precise matching global atmospheric light value A based on the shooting location and time information; The image restoration module is used to restore the haze-free image J(x) based on the atmospheric scattering model, using fine transmittance t(x) and A. The detail enhancement module is used to perform local contrast adaptive detail enhancement processing on the fog-free image J(x); The output module is used to output the final dehazing and enhanced image; The physical domain equivalent calculation unit in the micro-meteorological measurement compensation module is used to convert micro-meteorological measurement data into a physical domain equivalent fog concentration factor. The adaptive correction compensation unit is used to dynamically select a fusion strategy based on the difference between the initial global fog concentration parameter and the physical domain equivalent fog concentration factor to generate an accurate global fog concentration parameter β. The region-aware refined network unit introduces region-aware regularization loss during training to guide the network to make differential corrections to the transmittance of sky regions and non-sky regions.
9. An electronic device, comprising: A processor and a memory, wherein the memory stores a computer program that can be called by the processor, characterized in that: the processor executes a method for defogging transmission line images with adaptive fog concentration as described in any one of claims 1-7 by calling the computer program stored in the memory.
10. A computer-readable storage medium, characterized in that: The system stores instructions that, when executed on a computer, cause the computer to perform an adaptive fog concentration method for defogging transmission line images as described in any one of claims 1-7.