Remote visual monitoring method and system for running state of collecting ring

Abstract

The invention discloses a remote visual monitoring method and system for the running state of a collecting ring, and belongs to the technical field of monitoring, and the method comprises the steps: obtaining a visible light image and an infrared image of a monitored object in different directions; performing multi-scale feature extraction on the visible light image and the infrared image in each direction to obtain multi-scale visible light image features and multi-scale infrared image features; performing interactive fusion on the visible light image features and the infrared image features to obtain fusion features under each scale; performing multi-scale reconstruction on the fusion features under each scale to obtain a temperature-object fusion image; performing temperature extraction from the temperature-object fusion image of each orientation to obtain temperature data of the monitored object in each orientation; and performing early warning on the temperature data of the monitored object in each direction based on a multi-level alarm mechanism to obtain an early warning result. According to the method, temperature data can be accurately superposed on a clear visual background, so that rapid positioning and identification of temperature anomaly are realized.

Description

Technical Field

[0001] This invention belongs to the field of monitoring technology, specifically relating to a remote visual monitoring method and system for the operating status of a slip ring. Background Technology

[0002] Generator sets are core equipment in modern energy systems, and their safe, stable, and efficient operation is crucial for ensuring power supply. The slip ring-brush system, as a key component connecting the rotor circuit of a large synchronous generator to external stationary equipment, directly affects the reliability of the entire generator set. During operation, the slip rings and brushes inevitably generate heat due to sliding contact. Improper temperature rise control will accelerate oxidation and wear on the brush and slip ring surfaces, leading to serious malfunctions such as brush sparking, surface burns, and overheating deformation of the rings. In severe cases, this can cause unplanned shutdowns of the unit, resulting in significant economic losses.

[0003] Currently, there are three main methods for monitoring the temperature of the slip rings of generator sets:

[0004] 1. Manual handheld infrared thermometer inspection: Maintenance personnel periodically use spot thermometers or handheld infrared thermal imagers to perform measurements. This method has obvious limitations: First, it can only acquire data at discrete time points, making continuous monitoring impossible and easily missing sudden overheating phenomena; second, the measurement results are greatly affected by human operation, and personnel need to be close to high-temperature, high-speed equipment, posing safety risks; third, although the infrared thermal images provided by handheld thermal imagers can display temperature distribution, due to their low resolution and lack of visible light background details, it is difficult to quickly and accurately correlate and locate thermal anomaly areas with the physical structure of the collector ring (such as specific bolts, ring grooves, or corresponding brushes).

[0005] 2. Mounted single-point temperature sensor: such as a pre-embedded Pt100 platinum resistance thermometer or thermocouple. Although this method can achieve continuous monitoring, it can only monitor a limited number of fixed points and cannot fully capture the temperature distribution on the entire circumference of the slip ring, resulting in a monitoring blind zone; secondly, sensor installation requires drilling or pasting into the slip ring, which may damage its dynamic balance structure, and the leads are easily damaged in high-speed rotation environments, resulting in poor reliability.

[0006] 3. Single Infrared Online Monitoring System: This system uses a fixed-installation infrared thermal imager for continuous temperature measurement. This method overcomes the intermittent nature of manual inspections and achieves non-contact area array temperature measurement. However, its core drawback is that the infrared image itself lacks the texture details of visible light. When the system alarms indicating overheating, maintenance personnel find it difficult to accurately determine the specific location of the overheated spot on the collector ring surface, which brush it relates to, or whether it's a false alarm caused by surface contamination, based solely on the blurry infrared thermal image. Manual investigation after system shutdown is still required, reducing the timeliness of early warnings and decision-making efficiency.

[0007] Therefore, existing monitoring methods have limitations such as low accuracy and poor reliability. Summary of the Invention

[0008] The purpose of this invention is to provide a remote visual monitoring method and system for the operating status of a collector ring, in order to solve the limitations of existing monitoring methods, such as low accuracy and poor reliability.

[0009] To achieve the above objectives, the present invention adopts the following technical solution:

[0010] In a first aspect, the present invention provides a method for remote visual monitoring of the operating status of a slip ring, the method comprising:

[0011] Acquire visible light and infrared images of the monitored object from different orientations;

[0012] Multi-scale feature extraction is performed on the visible light and infrared images from each direction to obtain multi-scale visible light image features and multi-scale infrared image features.

[0013] At the same scale, visible light image features and infrared image features are interactively fused to obtain fused features at each scale.

[0014] Multi-scale reconstruction of the fusion features at various scales yields a temperature-object fusion image.

[0015] Temperature is extracted from the temperature-object fusion image from each direction to obtain the temperature data of the monitored object from each direction;

[0016] A multi-level alarm mechanism is used to issue early warnings based on the temperature data of the monitored object in various directions, and the early warning results are obtained.

[0017] Preferably, multi-scale feature extraction is performed on the visible light and infrared images for each orientation, including:

[0018] The visible light image in each direction is enhanced based on an attention mechanism to obtain the enhanced image;

[0019] Multiple convolution operations are performed sequentially on the visible light image in each direction to obtain the output of each convolution operation of the visible light image. The output of each convolution operation of the visible light image is used as a feature of the visible light image at one scale.

[0020] Multiple convolution operations are performed sequentially on the infrared image in each orientation to obtain the output of each convolution operation of the infrared image. The output of each convolution operation of the infrared image is used as a feature of the infrared image at one scale. The parameters of the convolution operation on the visible light image are kept the same as those of the convolution operation on the infrared image.

[0021] Preferably, the attention mechanism includes channel attention and spatial attention. Based on the attention mechanism, the visible light image in each orientation is enhanced to obtain an enhanced image, including:

[0022] For a visible light image from any orientation, feature extraction is performed on the visible light image to obtain a visible light feature map;

[0023] The visible light feature map is enhanced once based on channel attention to obtain a preliminary enhanced feature map;

[0024] The initial enhanced feature map is enhanced a second time based on spatial attention to obtain the final enhanced feature map, which is then used as the enhanced image.

[0025] Preferably, the visible light feature map is enhanced once based on channel attention to obtain a preliminary enhanced feature map, including:

[0026] The visible light feature map is processed by average pooling and max pooling to obtain the average pooling feature map and the max pooling feature map.

[0027] The average pooling feature map and the max pooling feature map are processed by a multilayer perceptron to obtain the perceptron output of the average pooling feature map and the perceptron output of the max pooling feature map.

[0028] The perceptual outputs of the average pooling feature map and the perceptual outputs of the max pooling feature map are successively added and nonlinearly mapped to obtain the channel attention weight matrix.

[0029] The visible light feature map is enhanced based on the channel attention weight matrix to obtain a preliminary enhanced feature map.

[0030] Preferably, the initial enhanced feature map is further enhanced based on spatial attention to obtain the final enhanced feature map, including:

[0031] Average pooling and max pooling are performed on the channel dimension of the initial enhanced feature map to obtain the average pooling output and max pooling output of the initial enhanced feature map.

[0032] The average pooling output and the max pooling output of the preliminary enhanced feature map are concatenated to obtain the concatenated preliminary enhanced feature map.

[0033] The spliced ​​preliminary enhanced feature map is subjected to nonlinear mapping to obtain the spatial attention weight matrix;

[0034] The initial enhanced feature map is enhanced based on the spatial attention weight matrix to obtain the final enhanced feature map.

[0035] Preferably, visible light image features and infrared image features are interactively fused at the same scale to obtain fused features at each scale, including:

[0036] At any scale, channel attention is calculated for visible light image features and infrared image features respectively to obtain visible light attention matrix and infrared attention matrix;

[0037] The visible light image features are processed based on the visible light attention matrix to obtain the first feature; the infrared image features are processed based on the visible light attention matrix to obtain the second feature.

[0038] The third feature is obtained by processing the visible light image features based on the infrared attention matrix; the fourth feature is obtained by processing the infrared image features based on the infrared attention matrix.

[0039] The first feature is fused with the third feature to obtain the visible light fused feature, and the second feature is fused with the fourth feature to obtain the infrared fused feature.

[0040] Spatial attention was calculated for visible light fusion features and infrared fusion features respectively to obtain visible light spatial attention matrix and infrared spatial attention matrix;

[0041] The visible light fusion features are processed based on the visible light spatial attention matrix to obtain the fifth feature, and the infrared fusion features are processed based on the visible light spatial attention matrix to obtain the sixth feature.

[0042] The visible light fusion features are processed based on the infrared spatial attention matrix to obtain the seventh feature, and the infrared fusion features are processed based on the infrared spatial attention matrix to obtain the eighth feature.

[0043] The fifth feature is fused with the seventh feature to obtain the ninth feature, and the sixth feature is fused with the eighth feature to obtain the tenth feature.

[0044] The ninth and tenth features are fused to obtain the fused features at this scale.

[0045] Preferably, multi-scale reconstruction is performed on the fusion features at each scale to obtain a temperature-object fusion image, including:

[0046] Obtain the fusion features and visible light image features at the Nth scale, and use the visible light image features as the object to be fused, where N is the highest scale number;

[0047] The fused features at the Nth scale are concatenated with the object to be fused to obtain the concatenated feature map;

[0048] The spliced ​​feature map is reconstructed to obtain the reconstructed feature map at the Nth scale.

[0049] The reconstructed feature map at the Nth scale is used as the object to be fused at the N-1th scale and is stitched together and reconstructed with the fusion features at the N-1th scale to obtain the reconstructed feature map at the N-1th scale. The object to be fused is updated repeatedly until the fusion features at each scale are reconstructed. The reconstructed feature map at the last scale is used as the temperature-object fusion image.

[0050] Secondly, the present invention provides a remote visual monitoring system for the operating status of a slip ring, the system comprising:

[0051] The server is used to implement the above-mentioned remote visual monitoring method for the operating status of the collector ring;

[0052] Several dual-modal detectors are provided, each of which is connected to the server. Each dual-modal detector is deployed in the slip ring chamber of the generator set and is distributed at circumferential intervals along the slip ring. Each dual-modal detector is used to acquire visible light and infrared images of the monitored object from different orientations.

[0053] The terminal device communicates with the server and is used to visualize the temperature data of the monitored object from various directions and the early warning results.

[0054] Preferably, the system further includes a switch, which is used to establish communication connections between each dual-modal detector and the server.

[0055] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method for remote visual monitoring of the operating status of the slip ring.

[0056] The beneficial effects of this invention are:

[0057] This invention acquires visible light and infrared images of the monitored object from different orientations. Visible light images, with their high resolution and clarity, provide rich texture, edge, and detail features; while infrared thermal images directly reflect the temperature distribution field on the object's surface, revealing thermal anomalies invisible to the human eye. By fusing visible light and infrared images at multiple scales, complementary information captured from the same scene is integrated to generate a temperature-object fusion image containing richer and more comprehensive information than any single-source image. This allows for precise overlay of temperature data onto a clear, visualized background, enabling rapid location and identification of temperature anomalies. Attached Figure Description

[0058] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings:

[0059] Figure 1 This is a flowchart of a remote visual monitoring method for the operating status of a slip ring provided in one embodiment of the present invention;

[0060] Figure 2 This is a block diagram of a remote visual monitoring system for the operating status of a collector ring provided in one embodiment of the present invention. Detailed Implementation

[0061] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the present invention will be briefly introduced below in conjunction with the accompanying drawings and descriptions of the embodiments or the prior art. Obviously, the following description of the structure of the accompanying drawings is 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. It should be noted that the description of these embodiments is for the purpose of helping to understand the present invention, but does not constitute a limitation of the present invention.

[0062] Example 1

[0063] Figure 1 This is a flowchart of a remote visual monitoring method for the operating status of a collector ring according to one embodiment of the present invention. This method can be executed, but is not limited to, by a computer device with certain computing resources, such as a server, a personal computer (PC, referring to a multi-purpose computer of a size, price, and performance suitable for personal use; desktop computers, laptops, mini-laptops, tablets, and ultrabooks all belong to personal computers), and other electronic devices. Figure 1 As shown in the figure, this embodiment provides a method for remote visual monitoring of the operating status of a collector ring, the method including steps S10 to S60.

[0064] Step S10: Acquire visible light and infrared images of the monitored object from different orientations.

[0065] In this embodiment, the monitored object is the slip ring chamber's collector ring. Therefore, multiple dual-mode detectors are installed in the slip ring chamber of the generator set, with the detectors spaced circumferentially along the collector ring. Each dual-mode detector includes an infrared camera and a visible light camera, used to acquire images in two modes: visible light and infrared, from one orientation. Each dual-mode detector is connected to a server, and each detector uploads the acquired images in both visible light and infrared modes to the server, enabling the server to obtain visible light and infrared images of the monitored object from different orientations.

[0066] In this embodiment, visible light images are characterized by high resolution and high definition, providing rich texture, edge, and detail features; while infrared thermal images can intuitively reflect the temperature distribution field of an object's surface, revealing thermal anomalies invisible to the human eye. Therefore, effective fusion of the two is necessary. Traditional fusion methods for visible light and infrared images include multi-scale transformation methods and sparse representation methods. These traditional methods have good image fusion effects, but due to insufficient manually designed feature representation capabilities, spatial distortions such as artifacts and incomplete information are generated in the fused image. With the development of deep learning, such as methods based on convolutional neural networks, these methods automatically extract deep features through convolution and improve the fusion effect through improvements in feature map fusion and interaction. These methods have the following shortcomings in fusing visible light and infrared images of slip rings: First, existing methods cannot effectively learn deep features, and gradient vanishing problems easily occur during image reconstruction; second, due to the dim lighting in the slip ring chamber, current algorithms are easily affected by low-quality visible light images, leading to a decrease in fusion quality. Therefore, in order to effectively fuse the visible light image and the infrared image in the special scenario of this embodiment, the fusion operation includes steps S20 to S40.

[0067] Step S20: Perform multi-scale feature extraction on the visible light and infrared images of each orientation to obtain multi-scale visible light image features and multi-scale infrared image features.

[0068] Specifically, the steps for multi-scale feature extraction of visible light and infrared images from each orientation are as follows:

[0069] Step S201: Enhance the visible light image of each direction based on the attention mechanism to obtain the enhanced image.

[0070] In this embodiment, the dim lighting in the slip ring chamber results in indistinct brightness differences in the image, making it difficult to distinguish the target from the background. Simultaneously, the signal-to-noise ratio decreases, noise increases, and image quality deteriorates. To address this issue, this embodiment utilizes an attention mechanism to enhance the visible light image in each orientation, thereby improving image quality.

[0071] Specifically, the visible light image in each direction is enhanced based on an attention mechanism to obtain the enhanced image, including:

[0072] Step a10: For a visible light image in any orientation, feature extraction is performed on the visible light image to obtain a visible light feature map. In this embodiment, two convolutional layers and one activation layer are used for feature extraction. The two convolutional layers perform convolution operations on the visible light image, and the output of the second convolution operation is used as the input of the activation function of the activation layer. The activation function performs nonlinear mapping to obtain the visible light feature map.

[0073] Step a20: Perform a first enhancement on the visible light feature map based on channel attention to obtain a preliminary enhanced feature map; wherein the first enhancement steps are as follows:

[0074] First, the visible light feature map is processed by average pooling and max pooling to obtain the average pooling feature map and the max pooling feature map.

[0075] Then, the average pooling feature map and the max pooling feature map are processed by the multilayer perceptron to obtain the perceptual output of the average pooling feature map and the perceptual output of the max pooling feature map. The multilayer perceptron can learn complex nonlinear relationships, thereby more accurately capturing the dependencies between feature map channels and generating a more targeted attention weight matrix, thus better enhancing the image detail features.

[0076] Next, the perceptual outputs of the average pooling feature map and the max pooling feature map are successively added and nonlinearly mapped to obtain the channel attention weight matrix. The nonlinear mapping can be performed using the Sigmoid function, with the result of adding the perceptual outputs of the average pooling feature map and the max pooling feature map as the input to the Sigmoid function, which outputs the channel attention weight matrix.

[0077] Finally, the visible light feature map is enhanced based on the channel attention weight matrix to obtain a preliminary enhanced feature map. This is achieved by multiplying each element in the visible light feature map with the corresponding element in the channel attention weight matrix, and the resulting feature map is used as the preliminary enhanced feature map.

[0078] Step a30: Perform secondary enhancement on the initial enhanced feature map based on spatial attention to obtain the final enhanced feature map, and use the final enhanced feature map as the enhanced image; wherein, the secondary enhancement steps are as follows:

[0079] First, average pooling and max pooling are performed on the channel dimension of the preliminary enhanced feature map to obtain the average pooling output and max pooling output of the preliminary enhanced feature map.

[0080] Then, the average pooling output and the max pooling output of the preliminary enhanced feature map are concatenated to obtain the concatenated preliminary enhanced feature map.

[0081] Next, a nonlinear mapping is performed on the concatenated preliminary enhanced feature map to obtain the spatial attention weight matrix. The nonlinear mapping in this step also uses the Sigmoid function, with the concatenated preliminary enhanced feature map as the input to the Sigmoid function, and the Sigmoid function outputting the spatial attention weight matrix.

[0082] Finally, the initial enhanced feature map is enhanced based on the spatial attention weight matrix to obtain the final enhanced feature map; each element in the initial enhanced feature map is multiplied with the corresponding element in the spatial attention weight matrix, and the resulting feature map is used as the final enhanced feature map.

[0083] Therefore, by performing two enhancement operations on the visible light image, this embodiment can effectively improve the quality of the visible light image and significantly reduce the adverse effects of light problems on the fusion of the visible light image and the outer infrared image.

[0084] Step S202: Perform multiple convolution operations sequentially on the visible light image from each direction to obtain the output of each convolution operation. Each convolution operation output serves as a feature at one scale of the visible light image. In this embodiment, 4 to 6 convolution operations can be performed sequentially on the visible light image. Preferably, 5 convolution operations are performed, resulting in 5 convolution operation outputs with features at 5 scales. Furthermore, after each convolution operation, the ReLU activation function is used to nonlinearly enhance the output of each convolution operation, improving the nonlinear expressive power of the features.

[0085] Step S203: Perform multiple convolution operations on the infrared image in each orientation sequentially to obtain the output of each convolution operation of the infrared image. The output of each convolution operation of the infrared image is used as a feature of the infrared image at one scale. The parameters of the convolution operation on the visible light image are kept the same as those of the convolution operation on the infrared image. The same parameters can reduce the computational load of the model and ensure that the features in the two modalities are consistent, which helps to improve the quality of subsequent image fusion.

[0086] Step S30: Perform interactive fusion of visible light image features and infrared image features at the same scale to obtain fused features at each scale.

[0087] Specifically, the steps for interactively fusing visible light image features and infrared image features at the same scale are as follows:

[0088] Step S301: At any scale, channel attention is calculated for visible light image features and infrared image features respectively to obtain visible light attention matrix and infrared attention matrix; In this embodiment, the channel attention calculation steps are as follows: First, global max pooling and average pooling are performed on visible light image features and infrared image features to obtain global information of visible light image features and global information of infrared image features; Then, the visible light attention matrix and infrared attention matrix are calculated using the multilayer perceptron and sigmoid function in step a20.

[0089] Step S302: Process the visible light image features based on the visible light attention matrix to obtain the first feature; process the infrared image features based on the visible light attention matrix to obtain the second feature.

[0090] Specifically, the first feature can be obtained by multiplying the visible light image features element-wise with the visible light attention matrix, and the second feature can be obtained by multiplying the infrared image features element-wise with the visible light attention matrix.

[0091] Step S303: Process the visible light image features based on the infrared attention matrix to obtain the third feature; process the infrared image features based on the infrared attention matrix to obtain the fourth feature.

[0092] Specifically, multiplying the visible light image features element-wise with the infrared attention matrix yields the third feature, and multiplying the infrared image features element-wise with the infrared attention matrix yields the fourth feature.

[0093] Step S304: Fuse the first feature and the third feature to obtain the visible light fused feature, and fuse the second feature and the fourth feature to obtain the infrared fused feature;

[0094] Specifically, the first feature and the third feature are summed element-wise to obtain the visible light fusion feature, and the second feature and the fourth feature are summed element-wise to obtain the infrared fusion feature. At this time, both the infrared fusion feature and the visible light fusion feature can focus on their own feature information and learn feature information from another mode, and can discover the dependencies between channels.

[0095] Step S305: Perform spatial attention calculations on the visible light fusion features and the infrared fusion features respectively to obtain the visible light spatial attention matrix and the infrared spatial attention matrix. In this embodiment, the steps for spatial attention calculation are as follows: First, perform max pooling and average pooling on the visible light fusion features and the infrared fusion features in the channel direction. Then, concatenate and stack the results of the two pooling processes. Then, use a convolution kernel with 2 input channels and 1 output channel to perform convolution processing on the stacked features, adjust the number of output channels to 1, and finally input the convolution-processed features into the Sigmoid function. The Sigmoid function outputs the visible light spatial attention matrix and the infrared spatial attention matrix.

[0096] Step S306: Process the visible light fusion features based on the visible light spatial attention matrix to obtain the fifth feature, and process the infrared fusion features based on the visible light spatial attention matrix to obtain the sixth feature.

[0097] Specifically, the fifth feature can be obtained by multiplying the visible light fusion feature element-wise with the visible light spatial attention matrix, and the sixth feature can be obtained by multiplying the infrared fusion feature element-wise with the visible light spatial attention matrix.

[0098] Step S307: Process the visible light fusion features based on the infrared spatial attention matrix to obtain the seventh feature, and process the infrared fusion features based on the infrared spatial attention matrix to obtain the eighth feature.

[0099] Specifically, multiplying the visible light fusion feature element-wise with the infrared spatial attention matrix yields the seventh feature, and multiplying the infrared fusion feature element-wise with the infrared spatial attention matrix yields the eighth feature.

[0100] Step S308: Fuse the fifth feature with the seventh feature to obtain the ninth feature, and fuse the sixth feature with the eighth feature to obtain the tenth feature.

[0101] Specifically, the fifth feature and the seventh feature are summed element-wise to obtain the ninth feature, and the sixth feature and the eighth feature are summed element-wise to obtain the tenth feature. At this point, the fused feature can not only focus on its own features, but also on the features of the other image.

[0102] Step S309: Fuse the ninth feature and the tenth feature to obtain the fused feature at this scale; In this embodiment, the ninth feature and the tenth feature at each scale are summed element by element to obtain the fused feature at each scale.

[0103] In this embodiment, through steps S301 to S309, infrared images and visible light images of different modalities are complementaryly fused using a channel-spatial attention mechanism, which can enhance the ability of the fused features to capture key features and avoid the loss of details in the fused image.

[0104] Step S40: Perform multi-scale reconstruction on the fusion features at each scale to obtain a temperature-object fusion image.

[0105] In this embodiment, to avoid the gradient vanishing problem that easily occurs when processing features during reconstruction, the following method is used for reconstruction, with the following steps:

[0106] Step S401: Obtain the fusion features and visible light image features at the Nth scale, and use the visible light image features as the object to be fused, where N is the highest scale number. Since this embodiment has features at 5 scales, the value of N is 5.

[0107] Step S402: Concatenate the fused features at the Nth scale with the object to be fused to obtain a concatenated feature map.

[0108] Step S403: Reconstruct the stitched feature map to obtain the reconstructed feature map at the Nth scale; in this embodiment, the size of the stitched feature map is gradually restored through convolution and PixelShuffle operations to achieve reconstruction.

[0109] Step S404: The reconstructed feature map at the Nth scale is used as the object to be fused at the N-1th scale and is stitched and reconstructed with the fusion features at the N-1th scale to obtain the reconstructed feature map at the N-1th scale. The object to be fused is updated repeatedly until the fusion features at each scale are reconstructed. The reconstructed feature map at the last scale is used as the temperature-object fusion image.

[0110] Therefore, this embodiment effectively preserves the detailed features of the image by introducing features at different levels into the reconstruction stage, avoiding information loss; at the same time, it directly transmits high-level features to low-level features, alleviating the gradient vanishing problem, enabling the network to learn deeper features, so that the temperature-object fusion image of this embodiment contains richer and more comprehensive information, and can accurately overlay temperature data on a clear visual background, with advantages such as higher clarity and richer detailed features.

[0111] Step S50: Extract temperature from the temperature-object fusion image of each location to obtain the temperature data of the monitored object in each location; in this embodiment, the temperature data includes the maximum temperature, minimum temperature and average temperature of each location and different areas (e.g., each collector ring, each brush).

[0112] Step S60: Based on the multi-level alarm mechanism, issue early warnings for the temperature data of the monitored object in various directions, and obtain the early warning results.

[0113] In this embodiment, the multi-level alarm mechanism can issue a warning when the temperature of a certain area reaches 80°C, trigger an alarm when the temperature reaches 100°C, and initiate an emergency shutdown when the temperature reaches 120°C. This embodiment can also generate temperature change trend curve reports for each area, so that staff can view the temperature changes of the slip ring.

[0114] Example 2

[0115] Figure 2 This is a block diagram of a remote visual monitoring system for the operating status of a slip ring provided in one embodiment of the present invention. Figure 2 As shown, this embodiment provides a remote visualization monitoring system for the operating status of a collector ring. The system includes: a server, several dual-modal detectors, terminal devices, and a switch. The switch is a PoE switch. Each dual-modal detector is connected to the PoE switch via fiber optic cables or other network cables. The PoE switch is then connected to the server to achieve communication between the dual-modal detectors and the server.

[0116] The server is used to implement the remote visual monitoring method for the operating status of the collector ring in Embodiment 1;

[0117] Each dual-mode detector is deployed in the slip ring chamber of the generator set and is distributed at intervals along the circumference of the slip ring; each dual-mode detector is used to acquire visible light images and infrared images of the monitored object in different directions; the dual-mode detector in this embodiment integrates an infrared camera and a visible light camera to acquire two modes of images, namely visible light images and infrared images, in one direction.

[0118] The terminal device is used to visualize the temperature data of the monitored object from various directions and the early warning results.

[0119] This embodiment also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the remote visualization monitoring method for the operating status of the collector ring in Embodiment 1.

[0120] This embodiment also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the remote visual monitoring method for the operating status of the collector ring in Embodiment 1.

[0121] Therefore, this embodiment acquires visible light and infrared images of the monitored object from different orientations. Visible light images, with their high resolution and high definition, provide rich texture, edge, and detail features; while infrared thermal images can intuitively reflect the temperature distribution field on the object's surface, revealing thermal anomalies invisible to the human eye. Multi-scale fusion of visible light and infrared images integrates complementary information captured from the same scene, generating a temperature-object fusion image that contains richer and more comprehensive information than any single-source image. This allows for precise overlay of temperature data onto a clear, visualized background, enabling rapid location and identification of temperature anomalies.

[0122] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application 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.

[0123] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and 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 system that specifies functions in one or more boxes.

[0124] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for remote visual monitoring of the operating status of a slip ring, characterized in that, The method includes: Acquire visible light and infrared images of the monitored object from different orientations; Multi-scale feature extraction is performed on the visible light and infrared images from each direction to obtain multi-scale visible light image features and multi-scale infrared image features. At the same scale, visible light image features and infrared image features are interactively fused to obtain fused features at each scale. Multi-scale reconstruction of the fusion features at various scales yields a temperature-object fusion image. Temperature is extracted from the temperature-object fusion image from each direction to obtain the temperature data of the monitored object from each direction; A multi-level alarm mechanism is used to issue early warnings based on the temperature data of the monitored object in various directions, and the early warning results are obtained.

2. The remote visual monitoring method for the operating status of the slip ring according to claim 1, characterized in that, Multi-scale feature extraction is performed on the visible light and infrared images for each orientation, including: The visible light image in each direction is enhanced based on an attention mechanism to obtain the enhanced image; Multiple convolution operations are performed sequentially on the visible light image in each direction to obtain the output of each convolution operation of the visible light image. The output of each convolution operation of the visible light image is used as a feature of the visible light image at one scale. Multiple convolution operations are performed sequentially on the infrared image in each orientation to obtain the output of each convolution operation of the infrared image. The output of each convolution operation of the infrared image is used as a feature of the infrared image at one scale. The parameters of the convolution operation on the visible light image are kept the same as those of the convolution operation on the infrared image.

3. The remote visual monitoring method for the operating status of the slip ring according to claim 2, characterized in that, The attention mechanism includes channel attention and spatial attention. Based on the attention mechanism, the visible light image in each direction is enhanced to obtain the enhanced image, including: For a visible light image from any orientation, feature extraction is performed on the visible light image to obtain a visible light feature map; The visible light feature map is enhanced once based on channel attention to obtain a preliminary enhanced feature map; The initial enhanced feature map is enhanced a second time based on spatial attention to obtain the final enhanced feature map, which is then used as the enhanced image.

4. The remote visual monitoring method for the operating status of the slip ring according to claim 3, characterized in that, The visible light feature map is enhanced once based on channel attention to obtain a preliminary enhanced feature map, including: The visible light feature map is processed by average pooling and max pooling to obtain the average pooling feature map and the max pooling feature map. The average pooling feature map and the max pooling feature map are processed by a multilayer perceptron to obtain the perceptron output of the average pooling feature map and the perceptron output of the max pooling feature map. The perceptual outputs of the average pooling feature map and the perceptual outputs of the max pooling feature map are successively added and nonlinearly mapped to obtain the channel attention weight matrix. The visible light feature map is enhanced based on the channel attention weight matrix to obtain a preliminary enhanced feature map.

5. The remote visual monitoring method for the operating status of the slip ring according to claim 3 or 4, characterized in that, The initial enhanced feature map is then enhanced a second time based on spatial attention to obtain the final enhanced feature map, including: Average pooling and max pooling are performed on the channel dimension of the initial enhanced feature map to obtain the average pooling output and max pooling output of the initial enhanced feature map. The average pooling output and the max pooling output of the preliminary enhanced feature map are concatenated to obtain the concatenated preliminary enhanced feature map. The spliced ​​preliminary enhanced feature map is subjected to nonlinear mapping to obtain the spatial attention weight matrix; The initial enhanced feature map is enhanced based on the spatial attention weight matrix to obtain the final enhanced feature map.

6. The remote visual monitoring method for the operating status of the slip ring according to claim 1, characterized in that, At the same scale, visible light image features and infrared image features are interactively fused to obtain fused features at various scales, including: At any scale, channel attention is calculated for visible light image features and infrared image features respectively to obtain visible light attention matrix and infrared attention matrix; The visible light image features are processed based on the visible light attention matrix to obtain the first feature; the infrared image features are processed based on the visible light attention matrix to obtain the second feature. The third feature is obtained by processing the visible light image features based on the infrared attention matrix; the fourth feature is obtained by processing the infrared image features based on the infrared attention matrix. The first feature is fused with the third feature to obtain the visible light fused feature, and the second feature is fused with the fourth feature to obtain the infrared fused feature. Spatial attention was calculated for visible light fusion features and infrared fusion features respectively to obtain visible light spatial attention matrix and infrared spatial attention matrix; The visible light fusion features are processed based on the visible light spatial attention matrix to obtain the fifth feature, and the infrared fusion features are processed based on the visible light spatial attention matrix to obtain the sixth feature. The visible light fusion features are processed based on the infrared spatial attention matrix to obtain the seventh feature, and the infrared fusion features are processed based on the infrared spatial attention matrix to obtain the eighth feature. The fifth feature is fused with the seventh feature to obtain the ninth feature, and the sixth feature is fused with the eighth feature to obtain the tenth feature. The ninth and tenth features are fused to obtain the fused features at this scale.

7. The remote visual monitoring method for the operating status of the slip ring according to claim 1, characterized in that, Multi-scale reconstruction of the fusion features at various scales yields a temperature-object fusion image, including: Obtain the fusion features and visible light image features at the Nth scale, and use the visible light image features as the object to be fused, where N is the highest scale number; The fused features at the Nth scale are concatenated with the object to be fused to obtain the concatenated feature map; The stitched feature map is reconstructed to obtain the reconstructed feature map at the Nth scale. The reconstructed feature map at the Nth scale is used as the object to be fused at the N-1th scale and stitched and reconstructed with the fusion features at the N-1th scale to obtain the reconstructed feature map at the N-1th scale. The object to be fused is updated repeatedly until the fusion features at each scale are reconstructed. The reconstructed feature map at the last scale is used as the temperature-object fusion image.

8. A remote visual monitoring system for the operating status of a slip ring, characterized in that, The system includes: The server is used to implement the remote visual monitoring method for the operating status of the slip ring as described in any one of claims 1-7; Several dual-modal detectors are provided, each of which is connected to the server. Each dual-modal detector is deployed in the slip ring chamber of the generator set and is distributed at circumferential intervals along the slip ring. Each dual-modal detector is used to acquire visible light and infrared images of the monitored object from different orientations. The terminal device communicates with the server and is used to visualize the temperature data of the monitored object from various directions and the early warning results.

9. The remote visual monitoring system for the operating status of the slip ring according to claim 8, characterized in that, The system also includes a switch, which is used to establish communication connections between each dual-modal detector and the server.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the remote visualization monitoring method for the operating status of the collector ring as described in any one of claims 1-7.

Images