Solid oxide fuel cell system combustor flame image processing method and apparatus

Abstract

This invention provides a method and apparatus for processing combustion chamber flame images in a solid oxide fuel cell system, belonging to the field of image processing technology. The method includes: inputting an original flame image into an image compression model to obtain a compressed image; enhancing the original flame image to obtain an enhanced color image and a grayscale image; fusing the enhanced color image and grayscale image with the original flame image to obtain a fused image; and combining the compressed image and the fused image to obtain a target image. The image compression model is obtained by training a convolutional neural network model. The convolutional neural network model uses a combination of nonlinear transformation and uniform quantization to quantize the input image, and uses a perceptual quality optimization mechanism to construct a loss function during the quantization process. This invention can solve the technical problems of low image compression efficiency and quality, making it difficult to meet real-time and high-precision requirements.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and specifically to a method and apparatus for processing combustion chamber flame images in a solid oxide fuel cell system. Background Technology

[0002] Solid oxide fuel cell (SOC) systems, as efficient and clean energy conversion devices, rely heavily on combustion state monitoring in their combustion chambers for safety and efficiency. Currently, flame image processing in SOC combustion chambers primarily relies on traditional image enhancement and compression techniques. Traditional compression algorithms, such as JPEG and PNG, are based on mathematical transformations and coding techniques, processing images through discrete cosine transform (DCT) or wavelet transform, and then combining this with Huffman coding or LZW coding to achieve efficient compression.

[0003] Traditional compression methods are prone to distortion at low bitrates, affecting subsequent combustion state analysis. For example, JPEG may experience quality degradation at high compression ratios. Image compression methods based on recurrent neural networks use progressive coding, requiring multiple iterations during image compression, thus slowing down the training and compression processes. Image compression methods based on convolutional neural networks require precise probabilistic modeling of latent representation features during estimation, resulting in less than ideal efficiency. Image compression methods based on generative adversarial networks are difficult to control during the training process, potentially damaging the details of the original image.

[0004] While traditional image enhancement algorithms can enhance the details and contrast of flame images, image quality still needs improvement in complex scenes with high brightness and low contrast.

[0005] Furthermore, existing technologies cannot adaptively optimize based on the characteristics of flame images, making it difficult to meet the high-precision image processing requirements of SOC systems.

[0006] In summary, existing image processing methods for combustion chamber flames in solid oxide fuel cell systems suffer from low image compression efficiency and quality, making it difficult to meet real-time and high-precision requirements. Summary of the Invention

[0007] In view of this, it is necessary to provide a method and apparatus for processing combustion chamber flame images in a solid oxide fuel cell system to solve the technical problems of low image compression efficiency and quality, which makes it difficult to meet real-time and high-precision requirements.

[0008] To address the aforementioned problems, in a first aspect, the present invention provides a method for processing combustion chamber flame images in a solid oxide fuel cell system, comprising: Obtain the raw flame image of the combustion chamber; The original flame image is input into the image compression model to obtain a compressed image; The original flame image is enhanced to obtain an enhanced color image and a grayscale image. The enhanced color image and grayscale image are then fused with the original flame image to obtain a fused image. The compressed image is combined with the fused image to obtain the target image; The image compression model is obtained by training a convolutional neural network model using combustion chamber flame images as training samples and corresponding flame image features as sample labels. The convolutional neural network model uses a combination of nonlinear transformation and uniform quantization to quantize the input image, and uses a perceptual quality optimization mechanism to construct a loss function during the quantization process.

[0009] In one possible implementation, the convolutional neural network model includes: an input layer, an intermediate layer, and an output layer; The input layer includes: a first group of convolutional layers and a first group of normalization layers, wherein each convolutional layer in the first group of convolutional layers and each normalization layer in the first group of normalization layers are sequentially cross-connected; The intermediate layer includes: a Q network layer, an AE layer, and an AD layer connected in sequence; The output layer includes: a second group of convolutional layers and a second group of normalized layers, wherein each convolutional layer in the second group of convolutional layers and each normalized layer in the second group of normalized layers are sequentially cross-connected.

[0010] In one possible implementation, both the first set of normalization layers and the second set of normalization layers contain multiple layers of generalized divisible normalization layers.

[0011] In one possible implementation, the generalized decomposition normalization layer includes a linear decomposition layer and a joint nonlinear layer connected in sequence.

[0012] In one possible implementation, the original flame image is enhanced to obtain enhanced color and grayscale images, including: The original flame image is enhanced using a perceptual variational framework and a multi-scale homomorphic filtering method to obtain an enhanced color image. The original flame image is converted to grayscale, and brightness information is extracted to obtain an enhanced grayscale image.

[0013] In one possible implementation, the perceptual variational framework employs gradient descent to enhance the original flame image.

[0014] In one possible implementation, the convolutional neural network model constructs a loss function using a multi-scale structural similarity evaluation metric during the quantization process.

[0015] In a second aspect, the present invention also provides a flame image processing device for a combustion chamber of a solid oxide fuel cell system, comprising: The acquisition module is used to acquire raw flame images of the combustion chamber; The compression module is used to input the original flame image into the image compression model to obtain a compressed image; The first fusion module is used to enhance the original flame image to obtain an enhanced color image and a grayscale image, and then fuse the enhanced color image and grayscale image with the original flame image to obtain a fused image. The second fusion module is used to combine the compressed image with the fused image to obtain the target image; The image compression model is obtained by training a convolutional neural network model using combustion chamber flame images as training samples and corresponding flame image features as sample labels. The convolutional neural network model uses a combination of nonlinear transformation and uniform quantization to quantize the input image, and uses a perceptual quality optimization mechanism to construct a loss function during the quantization process.

[0016] Thirdly, the present invention also provides an electronic device, including a memory and a processor, wherein, The memory is used to store programs; The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps of the solid oxide fuel cell system combustion chamber flame image processing method as described in any of the preceding claims.

[0017] Fourthly, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the solid oxide fuel cell system combustion chamber flame image processing method as described in any of the preceding claims.

[0018] The beneficial effects of adopting the above implementation method are as follows: The solid oxide fuel cell system combustion chamber flame image processing method and apparatus provided by the present invention utilizes a convolutional neural network model to construct an efficient image compression model, which does not require multiple iterations like a recurrent neural network, thus meeting the real-time requirements; and by introducing a perception quality evaluation index as part of the loss function, it achieves visual perception quality optimization in the flame image compression process, realizes adaptive optimization of the image, and balances compression efficiency and image quality, which can provide more accurate and reliable input data for image-based combustion state analysis, thereby improving the accuracy and efficiency of combustion state analysis.

[0019] Furthermore, this invention enhances and grayscales the original flame image, and then combines the processed image with the compressed image using an image fusion method to form the final output image. This invention, by combining the image fusion method, further optimizes the quality of the flame image, better highlights the characteristics of the flame image, and improves the visual effect and analytical value of the image.

[0020] Therefore, the present invention can solve the technical problem of low image compression efficiency and quality, which makes it difficult to meet real-time and high-precision requirements. Attached Figure Description

[0021] 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.

[0022] Figure 1 A flowchart of an embodiment of the solid oxide fuel cell system combustion chamber flame image processing method provided by the present invention; Figure 2 This is a schematic diagram of the structure of the convolutional neural network model provided by the present invention; Figure 3 A schematic block diagram of an embodiment of the flame image processing device for the combustion chamber of a solid oxide fuel cell system provided by the present invention; Figure 4 A schematic diagram of an embodiment of the electronic device provided by the present invention. Detailed Implementation

[0023] 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.

[0024] In the description of the embodiments of this application, unless otherwise stated, "a plurality of" means two or more.

[0025] In this embodiment of the invention, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, apparatus, product or device that includes a series of steps or modules is not necessarily limited to those steps or modules that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to such process, method, product or device.

[0026] The naming or numbering of steps in the embodiments of the present invention does not mean that the steps in the method flow must be executed in the time / logical order indicated by the naming or numbering. The execution order of the named or numbered process steps can be changed according to the technical purpose to be achieved, as long as the same or similar technical effect can be achieved.

[0027] 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 mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0028] This invention provides a method and apparatus for processing combustion chamber flame images in a solid oxide fuel cell system, which will be described below.

[0029] This invention provides a method for processing combustion chamber flame images in a solid oxide fuel cell system. The method can be implemented by executing an application on a terminal or server. The terminal can be a mobile phone or tablet, and the server can be a cloud server or an edge server. Figure 1 As shown, the method includes: S101. Obtain the original flame image of the combustion chamber.

[0030] Understandably, raw flame images can be obtained by photographing the flame in the combustion chamber of a solid oxide fuel cell (SOC) system using an industrial camera, thus acquiring multiple raw flame images of the combustion chamber.

[0031] S102. The original flame image is input into the image compression model to obtain a compressed image. The image compression model is obtained by training a convolutional neural network model using combustion chamber flame images as training samples and corresponding flame image features as sample labels. The convolutional neural network model uses a combination of nonlinear transformation and uniform quantization to quantize the input image, and employs a perceptual quality optimization mechanism to construct a loss function during the quantization process.

[0032] Understandably, this embodiment utilizes a Convolutional Neural Network (CNN) from deep learning to construct an image compression model. This model learns the feature representation of a flame image, mapping the original image to a low-dimensional latent space, thereby achieving efficient image compression. During compression, the model employs a combination of nonlinear transformation and uniform quantization to quantize the latent representation of the image, reducing information redundancy and storage requirements. Simultaneously, to further optimize compression performance, this invention introduces a perceptual quality optimization mechanism, guiding the model to focus more on the visual perceptual quality of the image during compression, making the compressed image visually closer to the original image and reducing visual distortion caused by compression.

[0033] S103. Enhance the original flame image to obtain an enhanced color image and a grayscale image, and then fuse the enhanced color image and grayscale image with the original flame image to obtain a fused image.

[0034] Understandably, the original flame image is enhanced by extracting the main color information, enhancing the image's detail features, and further improving the image's color contrast to make the flame image more vivid. The flame image is also grayscaled to extract brightness information. The enhanced color image, grayscale image, and original flame image are then fused together to obtain a fused image that integrates multiple types of information.

[0035] S104. Combine the compressed image with the fused image to obtain the target image; It is understood that the compressed image is combined with the fused image to form the final output image. This output image has both high compression efficiency and retains the key details and visual quality of the flame image, thus better meeting the needs of combustion state analysis in the SOC system combustion chamber.

[0036] This invention aims to address the problems of low efficiency in flame image processing and compression, insufficient image quality, and lack of intelligent processing in existing technologies. Specific objectives are as follows: to improve the compression efficiency of flame images, reduce the bandwidth required for storage and transmission, and meet real-time requirements; and to optimize image quality, especially maintaining image detail and contrast at low bit rates.

[0037] In some embodiments, the convolutional neural network model includes: an input layer, an intermediate layer, and an output layer; The input layer includes: a first group of convolutional layers and a first group of normalization layers, wherein each convolutional layer in the first group of convolutional layers and each normalization layer in the first group of normalization layers are sequentially cross-connected; The intermediate layer includes: a Q network layer, an AE layer, and an AD layer connected in sequence; The output layer includes: a second group of convolutional layers and a second group of normalized layers, wherein each convolutional layer in the second group of convolutional layers and each normalized layer in the second group of normalized layers are sequentially cross-connected.

[0038] The first group of normalization layers and the second group of normalization layers both contain multiple generalized divisional normalization layers.

[0039] It is understandable that the structure of the convolutional neural network model is as follows: Figure 2 As shown, Conv is a convolutional layer and GDN is a generalized normalization layer. Multiple convolutions and normalizations can effectively compress the image size, which is convenient for subsequent model training.

[0040] The Q-Network is a core component of the Deep Q-Network (DQN). It is a deep neural network used to approximate the Q-value function, which estimates the cumulative future reward that can be obtained by taking each possible action in a given state.

[0041] In the context of CNNs (Convolutional Neural Networks), the AE layer usually refers to an autoencoder. An autoencoder is an unsupervised learning neural network structure whose core goal is to learn an efficient representation (encoding) of the input data and then reconstruct the original input as accurately as possible (decoding). When a convolutional neural network is used as the main encoder and decoder, it forms a CNN-AE, a structure that is particularly good at processing spatially structured data such as images.

[0042] The AD layer, or ADNet (Attention-guided Denoising CNN), is an attention-guided image denoising network.

[0043] In some embodiments, the generalized decomposition normalization layer includes a linear decomposition layer and a joint nonlinear layer connected in sequence.

[0044] Understandably, the linear decomposition layer is used to perform linear transformations on the input data, usually implemented by convolution operations, to extract or transform features. This step can be viewed as a linear projection or filtering of the input data.

[0045] The joint nonlinear layer employs a nonlinear normalization function that depends not only on the activation value of the current channel but also on the response intensity of neighboring channels, thereby achieving mutual inhibition between channels.

[0046] The GDN transform consists of a linear decomposition H followed by a joint nonlinearity that divides the output of each linear filter by a measure of the overall filter activity.

[0047] In some embodiments, the original flame image is enhanced to obtain enhanced color and grayscale images, including: The original flame image is enhanced using a perceptual variational framework and a multi-scale homomorphic filtering method to obtain an enhanced color image. The original flame image is converted to grayscale, and brightness information is extracted to obtain an enhanced grayscale image.

[0048] The perceptual variational framework employs gradient descent to enhance the original flame image.

[0049] Understandably, in terms of image processing, this embodiment combines image enhancement techniques such as Perception-Based Variational Framework (PIVF) and Multi-Scale Homomorphic Filtering (MSRCP). PIVF processing enhances the flame image through gradient descent, extracting key color information and enhancing image detail. MSRCP processing further improves the image's color contrast, making the flame image more vivid. Furthermore, this invention performs grayscale processing on the flame image, extracting brightness information, and then fuses the enhanced color image, the grayscale image, and the original image to obtain a fused image that integrates multiple pieces of information. This fusion method not only preserves the color details of the flame image but also enhances its brightness information, providing richer image features for subsequent combustion state analysis.

[0050] In some embodiments, the convolutional neural network model uses a multi-scale structural similarity evaluation metric to construct a loss function during the quantization process.

[0051] Understandably, the multi-scale structural similarity evaluation metric, namely multi-scale SSIM (MS-SSIM), is an extension of SSIM. By calculating SSIM values ​​at multiple scales (resolutions) and performing weighted fusion, it can more comprehensively evaluate the structural similarity of an image at different levels of detail.

[0052] In some embodiments, the core of the present invention is to combine deep learning technology and image processing methods to compress and process flame images acquired by the combustion chamber of the SOC system, so as to achieve efficient and high-quality image storage and transmission, while providing better image input for combustion state analysis.

[0053] First, this invention utilizes a Convolutional Neural Network (CNN) from deep learning to construct an image compression model. This model learns the feature representation of a flame image, mapping the original image to a low-dimensional latent space, thereby achieving efficient image compression. During compression, the model employs a combination of nonlinear transformation and uniform quantization to quantize the latent representation of the image, reducing information redundancy and storage requirements. Simultaneously, to further optimize compression performance, this invention introduces a perceptual quality optimization mechanism. By using perceptual quality evaluation metrics such as Multi-Scale Structural Similarity (MS-SSIM) as part of the loss function, the model is guided to focus more on the visual perceptual quality of the image during compression, making the compressed image visually closer to the original image and reducing visual distortion caused by compression.

[0054] A schematic diagram of the structure of a convolutional neural network is shown below. Figure 2 As shown, Conv is a convolutional layer, and GDN is a generalized normalization layer. Multiple convolutions and normalizations effectively compress the image size, facilitating subsequent model training. The GDN transform consists of a linear decomposition H followed by a joint nonlinearity, which divides the output of each linear filter by a metric of the overall filter activity. (1) In the above transform coding framework, the rate-distortion functional is minimized by adjusting the analysis and synthesis transforms ga and gs: (2) The first term represents the discrete entropy of the quantization index vector q. The second term measures the distortion of the reference image z and its reconstruction in the perceptual representation; both terms are expectations over the image set.

[0055] In continuous parameters When minimizing this objective, most optimization methods rely on differentiability. However, both terms in the objective depend on the quantization value q, and the derivative of the quantizer is discontinuous. Therefore, this invention uses a continuously differentiable objective function to approximate the objective function by replacing the deterministic quantizer with an additive uniform noise source. The uniform scalar quantizer is a piecewise constant function applied to each element of y: The marginal density of quantized values ​​is given by the following formula: (3) in: (4) It is the probability mass in the nth quantization bin. Here, ' ' represents continuous convolution, and rect is ( A uniform distribution on 1 / 2, 1 / 2).

[0056] Secondly, in terms of image processing, this invention combines image enhancement techniques such as Perception-Based Variational Framework (PIVF) and Multi-Scale Homomorphic Filtering (MSRCP). PIVF processing enhances the flame image through gradient descent, extracting key color information and enhancing image detail. MSRCP processing further improves the color contrast, making the flame image more vivid. Furthermore, this invention performs grayscale processing on the flame image, extracting brightness information, and then fuses the enhanced color image, the grayscale image, and the original image to obtain a fused image that integrates multiple pieces of information. This fusion method not only preserves the color details of the flame image but also enhances its brightness information, providing richer image features for subsequent combustion state analysis.

[0057] Finally, the present invention combines the image compressed in the first step with the fused image processed in the second step to form the final output image. This output image has both high compression efficiency and retains the key details and visual quality of the flame image, which can better meet the needs of combustion state analysis in the combustion chamber of the SOC system.

[0058] In summary, the key technical solutions provided by this invention are as follows: (1) Deep learning-based image compression model and perception quality optimization: The key to this invention is to use deep learning to build an efficient image compression model and to optimize the visual perception quality in the process of flame image compression by introducing a perception quality evaluation index as part of the loss function, thus taking into account both compression efficiency and image quality.

[0059] (2) Flame Image Enhancement and Fusion Method Combining Multiple Image Processing Techniques: This invention processes flame images using multiple image enhancement techniques such as PIVF, MSRCP, and grayscale processing, and combines the processed image with the compressed image using an image fusion method to form the final output image. This comprehensive image processing and fusion strategy can effectively improve the quality of flame images and provide higher-quality image input for combustion state analysis.

[0060] Compared with existing technical solutions, the technical solution of the present invention has the following advantages: Efficient compression and high-quality image preservation: This invention combines a deep learning-based image compression model to achieve efficient image compression at low bit rates. At the same time, through a perceptual quality optimization mechanism, it effectively preserves key details and visual quality of flame images, avoiding the serious distortion problem that occurs in traditional compression methods at low bit rates, and providing a higher quality image foundation for subsequent combustion state analysis.

[0061] Targeted image processing optimization: Considering the characteristics of flame images in the SOC combustion chamber, this invention employs multiple image enhancement techniques, including PIVF, MSRCP, and grayscale processing, combined with image fusion methods, to further optimize the quality of the flame images. This targeted processing approach better highlights the features of the flame images, improving their visual appeal and analytical value.

[0062] Improving the accuracy and efficiency of combustion state analysis: The compression and processing method of this invention yields images with superior visual quality and detail preservation, providing more accurate and reliable input data for image-based combustion state analysis. This improves the accuracy and efficiency of combustion state analysis, helps to promptly identify potential problems in the SOC system, and ensures the safe and stable operation of the system.

[0063] like Figure 3 As shown, the present invention also provides a flame image processing device 300 for a combustion chamber of a solid oxide fuel cell system, comprising: The acquisition module 301 is used to acquire the original flame image of the combustion chamber; Compression module 302 is used to input the original flame image into the image compression model to obtain a compressed image; The first fusion module 303 is used to enhance the original flame image to obtain an enhanced color image and a grayscale image, and to fuse the enhanced color image and grayscale image with the original flame image to obtain a fused image; The second fusion module 304 is used to combine the compressed image with the fused image to obtain the target image; The image compression model is obtained by training a convolutional neural network model using combustion chamber flame images as training samples and corresponding flame image features as sample labels. The convolutional neural network model uses a combination of nonlinear transformation and uniform quantization to quantize the input image, and uses a perceptual quality optimization mechanism to construct a loss function during the quantization process.

[0064] The solid oxide fuel cell system combustion chamber flame image processing device provided in the above embodiments can realize the technical solutions described in the above embodiments of the solid oxide fuel cell system combustion chamber flame image processing method. The specific implementation principles of each module or unit can be found in the corresponding content in the above embodiments of the solid oxide fuel cell system combustion chamber flame image processing method, which will not be repeated here.

[0065] like Figure 4 As shown, the present invention also provides an electronic device 400. The electronic device 400 includes a processor 401, a memory 402, and a display 403. Figure 4 Only some components of the electronic device 400 are shown, but it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead.

[0066] In some embodiments, memory 402 may be an internal storage unit of electronic device 400, such as a hard disk or memory of electronic device 400. In other embodiments, memory 402 may also be an external storage device of electronic device 400, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on electronic device 400.

[0067] Furthermore, the memory 402 may include both internal storage units of the electronic device 400 and external storage devices. The memory 402 is used to store application software and various types of data installed on the electronic device 400.

[0068] In some embodiments, processor 401 may be a central processing unit (CPU), microprocessor, or other data processing chip, used to run program code stored in memory 402 or process data, such as the solid oxide fuel cell system combustion chamber flame image processing method of the present invention.

[0069] In some embodiments, display 403 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. Display 403 is used to display information from electronic device 400 and to display a visual user interface. Components 401-403 of electronic device 400 communicate with each other via a system bus.

[0070] In some embodiments of the present invention, when the processor 401 executes the solid oxide fuel cell system combustion chamber flame image processing program in the memory 402, the following steps can be implemented: Obtain the raw flame image of the combustion chamber; The original flame image is input into the image compression model to obtain a compressed image; The original flame image is enhanced to obtain an enhanced color image and a grayscale image. The enhanced color image and grayscale image are then fused with the original flame image to obtain a fused image. The compressed image is combined with the fused image to obtain the target image; The image compression model is obtained by training a convolutional neural network model using combustion chamber flame images as training samples and corresponding flame image features as sample labels. The convolutional neural network model uses a combination of nonlinear transformation and uniform quantization to quantize the input image, and uses a perceptual quality optimization mechanism to construct a loss function during the quantization process.

[0071] It should be understood that when the processor 401 executes the solid oxide fuel cell system combustion chamber flame image processing program in the memory 402, in addition to the functions mentioned above, it can also perform other functions, as can be found in the description of the corresponding method embodiments above.

[0072] Furthermore, the embodiments of the present invention do not specifically limit the type of electronic device 400 mentioned. Electronic device 400 can be a mobile phone, tablet computer, personal digital assistant (PDA), wearable device, laptop computer, or other portable electronic device. Exemplary embodiments of portable electronic devices include, but are not limited to, portable electronic devices running iOS, Android, Microsoft, or other operating systems. The aforementioned portable electronic device can also be other portable electronic devices, such as a laptop computer with a touch-sensitive surface (e.g., a touch panel). It should also be understood that in some other embodiments of the present invention, electronic device 400 may not be a portable electronic device, but rather a desktop computer with a touch-sensitive surface (e.g., a touch panel).

[0073] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the solid oxide fuel cell system combustion chamber flame image processing method provided by the methods described above, the method comprising: Obtain the raw flame image of the combustion chamber; The original flame image is input into the image compression model to obtain a compressed image; The original flame image is enhanced to obtain an enhanced color image and a grayscale image. The enhanced color image and grayscale image are then fused with the original flame image to obtain a fused image. The compressed image is combined with the fused image to obtain the target image; The image compression model is obtained by training a convolutional neural network model using combustion chamber flame images as training samples and corresponding flame image features as sample labels. The convolutional neural network model uses a combination of nonlinear transformation and uniform quantization to quantize the input image, and uses a perceptual quality optimization mechanism to construct a loss function during the quantization process.

[0074] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0075] The above provides a detailed description of the combustion chamber flame image processing method and apparatus for a solid oxide fuel cell system provided by the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method of processing a flame image of a combustion chamber of a solid oxide fuel cell system, characterized by, include: Obtain the raw flame image of the combustion chamber; The original flame image is input into the image compression model to obtain a compressed image; The original flame image is enhanced to obtain an enhanced color image and a grayscale image. The enhanced color image and grayscale image are then fused with the original flame image to obtain a fused image. The compressed image is combined with the fused image to obtain the target image; The image compression model is obtained by training a convolutional neural network model using combustion chamber flame images as training samples and corresponding flame image features as sample labels. The convolutional neural network model uses a combination of nonlinear transformation and uniform quantization to quantize the input image, and uses a perceptual quality optimization mechanism to construct a loss function during the quantization process.

2. The solid oxide fuel cell system combustor flame image processing method according to claim 1, characterized by, The convolutional neural network model includes: an input layer, an intermediate layer, and an output layer; The input layer includes: a first group of convolutional layers and a first group of normalization layers, wherein each convolutional layer in the first group of convolutional layers and each normalization layer in the first group of normalization layers are sequentially cross-connected; The intermediate layer includes: a Q network layer, an AE layer, and an AD layer connected in sequence; The output layer includes: a second group of convolutional layers and a second group of normalized layers, wherein each convolutional layer in the second group of convolutional layers and each normalized layer in the second group of normalized layers are sequentially cross-connected.

3. The solid oxide fuel cell system combustor flame image processing method according to claim 2, characterized by, Both the first group of normalization layers and the second group of normalization layers contain multiple generalized divisional normalization layers.

4. The solid oxide fuel cell system combustor flame image processing method according to claim 3, characterized by, The generalized decomposition normalization layer includes a linear decomposition layer and a joint nonlinear layer connected in sequence.

5. The method of claim 1, wherein the method further comprises: determining a plurality of flame images of the combustion chamber; and determining a plurality of flame images of the combustion chamber. The original flame image is enhanced to obtain enhanced color and grayscale images, including: The original flame image is enhanced using a perceptual variational framework and a multi-scale homomorphic filtering method to obtain an enhanced color image. The original flame image is converted to grayscale, and brightness information is extracted to obtain an enhanced grayscale image.

6. The solid oxide fuel cell system combustor flame image processing method according to claim 5, characterized by, The perceptual variational framework uses gradient descent to enhance the original flame image.

7. The method of claim 1-6, wherein The convolutional neural network model uses a multi-scale structural similarity evaluation index to construct a loss function during the quantization process.

8. A flame image processing device for a combustion chamber of a solid oxide fuel cell system, characterized in that, include: The acquisition module is used to acquire raw flame images of the combustion chamber; The compression module is used to input the original flame image into the image compression model to obtain a compressed image; The first fusion module is used to enhance the original flame image to obtain an enhanced color image and a grayscale image, and then fuse the enhanced color image and grayscale image with the original flame image to obtain a fused image. The second fusion module is used to combine the compressed image with the fused image to obtain the target image; The image compression model is obtained by training a convolutional neural network model using combustion chamber flame images as training samples and corresponding flame image features as sample labels. The convolutional neural network model uses a combination of nonlinear transformation and uniform quantization to quantize the input image, and uses a perceptual quality optimization mechanism to construct a loss function during the quantization process.

9. An electronic device, comprising: Including memory and processor, among which, The memory is used to store programs; The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps of the combustion chamber flame image processing method for a solid oxide fuel cell system as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the combustion chamber flame image processing method for a solid oxide fuel cell system as described in any one of claims 1 to 7.

Images