Oil droplet leakage detection method and device, computer device and storage medium

Abstract

The application relates to an oil droplet leakage detection method and device, computer equipment and a storage medium. The method comprises the following steps: collecting a current polarized light image and a current natural light image of a to-be-detected device, superimposing the current polarized light image and the current natural light image to obtain a superimposed image, detecting oil droplets in the superimposed image by using an oil droplet leakage detection model, and determining the oil droplet leakage position of the to-be-detected device. The method can acquire the polarization and color characteristics of oil droplets on the surface of the to-be-detected device by collecting a polarized light image, and can comprehensively acquire the oil droplet condition on the surface of the to-be-detected device by superimposing the polarized light image and the natural light image, thereby further improving the rationality, accuracy and timeliness of oil droplet leakage detection. Meanwhile, the existing technology has the problem of high cost of an infrared camera, and the polarization camera used in the application has a simple polarization system structure, so the price is lower, and the cost of the acquisition equipment is reduced.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method, apparatus, computer device, and storage medium for detecting oil droplet leaks. Background Technology

[0002] Substations contain a large number of devices that use oil droplets for testing. If the oil droplets leak, it may cause the devices to malfunction, leading to problems such as fires, and also polluting the surrounding environment. Therefore, it is crucial to conduct safety inspections on these devices.

[0003] In existing technologies, infrared cameras are typically used to detect oil droplet leaks in the equipment under test. However, since the infrared images captured by the infrared cameras only contain temperature changes on the surface of the equipment under test, the infrared images contain limited image information about the equipment under test, which can lead to misjudgments of oil droplet leaks. Therefore, improvements are urgently needed. Summary of the Invention

[0004] Therefore, it is necessary to provide an oil droplet leakage detection method, apparatus, computer equipment, computer-readable storage medium, and computer program product that can accurately determine the location of oil droplet leakage, in order to address the above-mentioned technical problems.

[0005] Firstly, this application provides a method for detecting oil droplet leakage. The method includes:

[0006] Acquire the current polarized light image and the current natural light image of the device under test;

[0007] The current polarized light image and the current natural light image are superimposed to obtain a superimposed image;

[0008] By using an oil droplet leakage detection model, oil droplet detection is performed on the superimposed images to determine the location of oil droplet leakage in the device under test.

[0009] In one embodiment, an oil droplet leakage detection model is used to detect oil droplets in the overlay image to determine the location of the oil droplet leakage in the device under test, including:

[0010] By using a single-layer convolutional network in the oil droplet leakage detection model, channel dimensionality reduction processing is performed on the superimposed image to obtain the channel dimensionality reduction feature map corresponding to the superimposed image.

[0011] The encoder network in the oil droplet leakage detection model encodes the channel dimension reduction feature map to determine the global feature information of the superimposed image;

[0012] By using a multi-layer convolutional network in the oil droplet leakage detection model, the global feature information of the superimposed image is extracted to obtain the oil droplet features of the superimposed image;

[0013] The oil droplet features are decoded by the decoder network in the oil droplet leakage detection model to obtain the location of the oil droplet leakage in the device to be detected.

[0014] In one embodiment, the channel dimensionality-reduced feature map is encoded using an encoder network in the oil droplet leakage detection model to determine the global feature information of the overlay image, including:

[0015] By using the multi-head attention mechanism of the oil droplet leakage detection model, the channel dimensionality reduction feature map is encoded to obtain the weighted features of the channel dimensionality reduction feature map;

[0016] By using the gated linear unit of the oil droplet leakage detection model, the weighted features are adjusted to obtain the global feature information of the superimposed image.

[0017] In one embodiment, the multilayer convolutional network includes at least two grouped convolutions;

[0018] By using a multi-layer convolutional network in the oil droplet leakage detection model, feature extraction is performed on the global feature information of the superimposed image to obtain the oil droplet features of the superimposed image, including:

[0019] By using at least two grouped convolutions in the oil droplet leakage detection model, the global feature information of the superimposed image is extracted to obtain the oil droplet features of the superimposed image; the convolution kernel of each group convolution is a dynamic convolution kernel, and the size and shape of the dynamic convolution kernel are determined according to the global feature information input to each group convolution.

[0020] In one embodiment, after determining the location of the oil droplet leak in the device to be detected, the method further includes:

[0021] Based on the location of the oil droplet leak in the device under test, output early warning information for the device under test.

[0022] In one embodiment, the training method for the oil droplet leakage detection model includes:

[0023] Acquire historical polarized light images and historical natural light images of the sample equipment; the sample equipment includes oil droplet leakage equipment and oil droplet non-leakage equipment;

[0024] Historical polarized light images and historical natural light images are overlaid using channel overlay processing to obtain a historical overlay image;

[0025] The historical overlay images corresponding to the oil droplet leakage device are augmented to obtain an augmented overlay image;

[0026] The oil droplet leakage detection model is trained based on historical overlay images and expanded overlay images.

[0027] Secondly, this application also provides an oil droplet leakage detection device. The device includes:

[0028] The image acquisition module is used to acquire the current polarized light image and the current natural light image of the device under test;

[0029] The image overlay module is used to overlay the current polarized light image and the current natural light image to obtain an overlay image;

[0030] The location determination module is used to detect oil droplets in the superimposed image using an oil droplet leakage detection model, and to determine the location of the oil droplet leakage in the device under test.

[0031] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:

[0032] Acquire the current polarized light image and the current natural light image of the device under test;

[0033] The current polarized light image and the current natural light image are superimposed to obtain a superimposed image;

[0034] By using an oil droplet leakage detection model, oil droplet detection is performed on the superimposed images to determine the location of oil droplet leakage in the device under test.

[0035] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, performs the following steps:

[0036] Acquire the current polarized light image and the current natural light image of the device under test;

[0037] The current polarized light image and the current natural light image are superimposed to obtain a superimposed image;

[0038] By using an oil droplet leakage detection model, oil droplet detection is performed on the superimposed images to determine the location of oil droplet leakage in the device under test.

[0039] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, performs the following steps:

[0040] Acquire the current polarized light image and the current natural light image of the device under test;

[0041] The current polarized light image and the current natural light image are superimposed to obtain a superimposed image;

[0042] By using an oil droplet leakage detection model, oil droplet detection is performed on the superimposed images to determine the location of oil droplet leakage in the device under test.

[0043] The aforementioned oil droplet leakage detection method, apparatus, computer equipment, and storage medium acquire current polarized light images and current natural light images of the device under test. These images are then superimposed to obtain a superimposed image. An oil droplet leakage detection model is used to detect oil droplets in the superimposed image to determine the location of the oil droplet leakage on the device under test. By acquiring polarized light images, the polarization and color characteristics of oil droplets on the surface of the device under test can be obtained. Superimposing the polarized light and natural light images provides a more comprehensive understanding of the oil droplet situation on the surface of the device under test, further improving the rationality, accuracy, and timeliness of oil droplet leakage detection. Furthermore, existing technologies typically use infrared cameras to detect oil droplet leakage locations, which suffer from high costs. This application, however, uses a polarized camera to acquire surface images of the device under test. The single polarization system in a polarized camera has a simple structure; therefore, it is cheaper than an infrared camera, reducing the cost of acquisition equipment. Attached Figure Description

[0044] Figure 1 This is an application environment diagram of an oil droplet leakage detection method provided in this embodiment;

[0045] Figure 2 This is a flowchart illustrating the first oil droplet leakage detection method provided in this embodiment;

[0046] Figure 3 This embodiment provides a flowchart for determining the location of an oil droplet leak in a device under test.

[0047] Figure 4 This is a flowchart illustrating a training model for detecting oil droplet leakage provided in this embodiment;

[0048] Figure 5 This is a flowchart illustrating the second oil droplet leakage detection method provided in this embodiment;

[0049] Figure 6 This is a structural block diagram of the first type of oil droplet leakage detection device provided in this embodiment;

[0050] Figure 7 This is a structural block diagram of the second type of oil droplet leakage detection device provided in this embodiment;

[0051] Figure 8 This is a structural block diagram of the third type of oil droplet leakage detection device provided in this embodiment;

[0052] Figure 9This is a structural block diagram of the fourth type of oil droplet leakage detection device provided in this embodiment;

[0053] Figure 10 This is an internal structural diagram of a computer device provided in this embodiment. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0055] The oil droplet leakage detection method provided in this application embodiment can be applied to, for example... Figure 1 In the application environment shown, in one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows. Figure 1 As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores data such as polarized light images, current natural light images, and overlaid images. The network interface communicates with external terminals via a network connection. When executed by the processor, the computer program implements an oil droplet leakage detection method.

[0056] This embodiment provides a method for detecting oil droplet leakage. It acquires the current polarized light image and the current natural light image of the device under test, and then overlays the current polarized light image and the current natural light image to obtain an overlaid image. Then, it uses an oil droplet leakage detection model to detect oil droplets in the overlaid image and determine the location of the oil droplet leakage in the device under test.

[0057] In one embodiment, such as Figure 2 As shown, an oil droplet leakage detection method is provided, which is applied to... Figure 1 Taking the server in the example, the following steps are included:

[0058] S201 acquires the current polarized light image and the current natural light image of the device under test.

[0059] The device to be tested can be equipment that needs to detect whether there is an oil droplet leak. Optionally, the device to be tested can be substation equipment that uses oil.

[0060] The current polarized light image can be an image of reflected light from the surface of the device under test, captured by a polarizing filter. Optionally, in this current polarized light image, the reflected light at the location of the oil droplet leak will exhibit a different polarization state than the surrounding environment, thereby allowing the location of the oil droplet leak to be detected by comparing the intensity of reflected light under different polarization states.

[0061] The current natural light image can be an image of light reflected from the surface of the device under test, captured by a natural light source. Optionally, since the current natural light image can accurately determine the location and shape of the oil droplet leak, it can be used in conjunction with the current polarized light image to more accurately detect the location of the oil droplet leak.

[0062] Optionally, when the preset acquisition cycle is reached at the current time, the surface of the device under test is captured by an acquisition device (such as a polarizing camera) around the device under test, thereby obtaining the current natural light image of the device under test. At the same time, the acquisition device automatically switches to a polarizing filter, and then the current polarized light image of the device under test is obtained by the acquisition device with the polarizing filter switched.

[0063] S202 overlays the current polarized light image and the current natural light image to obtain an overlay image.

[0064] Among them, overlay processing can be a method of overlaying channels of pixels in different images.

[0065] Among them, the superimposed image can be an image obtained by superimposing different images.

[0066] Optionally, the pixel values ​​of pixels at the same position in the current polarized light image and the current natural light image are superimposed to obtain an image with superimposed pixels, which is then used as the superimposed image.

[0067] S203 uses an oil droplet leakage detection model to detect oil droplets in superimposed images and determine the location of oil droplet leakage in the device under test.

[0068] Among them, the oil droplet leakage detection model can be a model used to identify the location of oil droplet leakage in the equipment through equipment images.

[0069] Optionally, the acquired superimposed image is input into the oil droplet leakage detection model, which will identify the superimposed image to obtain the mask image corresponding to the superimposed image, determine the oil droplet leakage location in the mask image, and determine the oil droplet leakage location corresponding to the mask image as the oil droplet leakage location of the device to be detected.

[0070] It should be noted that after determining the location of the oil droplet leak in the device under test, the process also includes: outputting a warning message based on the location of the oil droplet leak. Specifically, when the server detects the location of the oil droplet leak, it can generate a warning message containing information such as the device number and the location of the oil droplet leak. This warning message can be displayed to staff directly through pop-ups or audio / video, or it can be sent to staff's terminal devices so that staff can go to the area where the device under test is located to promptly clean up the oil droplet leak and investigate the cause of the leak.

[0071] The aforementioned oil droplet leakage detection method, apparatus, computer equipment, and storage medium acquire current polarized light images and current natural light images of the device under test. These images are then superimposed to obtain a superimposed image. An oil droplet leakage detection model is used to detect oil droplets in the superimposed image to determine the location of the oil droplet leakage on the device under test. By acquiring polarized light images, the polarization and color characteristics of oil droplets on the surface of the device under test can be obtained. Superimposing the polarized light and natural light images provides a more comprehensive understanding of the oil droplet situation on the surface of the device under test, further improving the rationality, accuracy, and timeliness of oil droplet leakage detection. Furthermore, existing technologies typically use infrared cameras to detect oil droplet leakage locations, which suffer from high costs. This application, however, uses a polarized camera to acquire surface images of the device under test. The single polarization system in a polarized camera has a simple structure; therefore, it is cheaper than an infrared camera, reducing the cost of acquisition equipment.

[0072] Figure 3 This is a flowchart illustrating the process of determining the location of an oil droplet leak in a device under test, as described in one embodiment. In the prior art, the detection of oil droplet leak locations is typically done manually, which often results in delays and inaccuracies. Therefore, to ensure that the location of an oil droplet leak in the device under test can be detected promptly and accurately, this embodiment provides an optional method for determining the location of an oil droplet leak in the device under test, including the following steps:

[0073] S301 uses a single-layer convolutional network in the oil droplet leakage detection model to perform channel dimensionality reduction on the superimposed image, obtaining the channel dimensionality reduction feature map corresponding to the superimposed image.

[0074] Among them, the channel dimension reduction feature map can be a superimposed image after reducing the channel dimension.

[0075] Optionally, the superimposed image is input into the oil droplet leakage detection model, where a single-layer convolutional network performs channel dimensionality reduction on the superimposed image to obtain the superimposed image with reduced channel dimensions, which is then used as the channel dimensionality reduction feature map corresponding to the superimposed image.

[0076] S302 uses the encoder network in the oil droplet leakage detection model to encode the channel dimensionality reduction feature map and determine the global feature information of the superimposed image.

[0077] The global feature information can be all feature information in the overlaid image. Optionally, the global feature information can include feature information such as color, shape, and position.

[0078] Optionally, there are various ways to encode the channel dimensionality reduction feature map through the encoder network in the oil droplet leakage detection model to determine the global feature information of the superimposed image, and this application does not limit this one.

[0079] One possible implementation is to encode the obtained channel dimensionality reduction feature map using the multi-head attention mechanism of the oil droplet leakage detection model, and then input the encoded channel dimensionality reduction feature map into the feedforward neural network. The feedforward neural network extracts features from the encoded channel dimensionality reduction feature map to obtain the global feature information corresponding to the superimposed image.

[0080] Optionally, in order to improve the computational efficiency and computational power of the encoder, this embodiment can also improve the encoder in the oil droplet leakage detection model by inserting a gated linear unit. Therefore, another optional implementation method is to encode the channel dimensionality reduction feature map through the multi-head attention mechanism of the oil droplet leakage detection model to obtain the weighted features of the channel dimensionality reduction feature map; and to adjust the weighted features through the gated linear unit of the oil droplet leakage detection model to obtain the global feature information of the superimposed image.

[0081] Specifically, based on the obtained channel dimensionality reduction feature map, the multi-head attention mechanism of the oil droplet leakage detection model is used to encode the channel dimensionality reduction feature map to obtain the weighted features of the channel dimensionality reduction feature map. Then, the weighted features are input into the gated linear unit of the oil droplet leakage detection model. The gated linear unit will adjust the weighted features to control the weighted features within a fixed range, thereby obtaining the global feature information of the superimposed image.

[0082] S303 uses a multi-layer convolutional network in the oil droplet leakage detection model to extract global feature information from the superimposed image, thereby obtaining the oil droplet features of the superimposed image.

[0083] A multi-layer convolutional network may include at least two grouped convolutions.

[0084] Optionally, there are various ways to extract the global feature information of the superimposed image by using a multi-layer convolutional network in the oil droplet leakage detection model to obtain the oil droplet features of the superimposed image, and this application does not limit this one.

[0085] One possible implementation is to input the obtained global feature information into a multi-layer convolutional network. This multi-layer convolutional network will filter the global feature information and extract some global feature information related to the location of the oil droplet leakage, which will be used as the oil droplet features of the superimposed image.

[0086] Optionally, to further improve the efficiency and accuracy of feature extraction, the multi-layer convolutional network can be optimized by adjusting the original conventional multi-layer convolutional network to a grouped convolutional network. This allows the oil droplet leakage model to analyze and extract multiple feature information simultaneously. Therefore, another optional implementation method is to divide the obtained global feature information into at least two component feature information based on the global feature information, and input each component feature information into at least two grouped convolutions. The grouped convolutions will extract features from the component feature information to obtain the oil droplet features corresponding to the component feature information. The oil droplet features corresponding to each component feature information are then summarized to obtain the oil droplet features corresponding to the global feature information, which are used as the oil droplet features of the overlaid image.

[0087] Optionally, since the scale of each feature in the acquired global feature information is different, a dynamic convolution kernel is introduced in this embodiment to better adapt to different feature information. Therefore, another optional implementation is to extract features from the global feature information of the superimposed image through at least two grouped convolutions in the oil droplet leakage detection model to obtain the oil droplet features of the superimposed image. Here, the convolution kernel of each grouped convolution is a dynamic convolution kernel, and the size and shape of the dynamic convolution kernel are determined according to the global feature information input to each grouped convolution.

[0088] Specifically, based on the acquired global feature information, the global feature information is divided into at least two component feature information, and each component feature information is input into at least two group convolutions. The convolution kernel in the group convolution will adjust itself according to the component feature information. The adjusted convolution kernel will extract features from the component feature information to obtain the oil drop features corresponding to the component feature information. The oil drop features corresponding to each component feature information are summarized to obtain the oil drop features corresponding to the global feature information, which are used as the oil drop features of the overlaid image.

[0089] S304 decodes the oil droplet features through the decoder network in the oil droplet leakage detection model to obtain the location of the oil droplet leakage in the device under test.

[0090] Optionally, based on the oil droplet features of the acquired overlay image, the decoder network in the oil droplet leakage detection model decodes the oil droplet features to obtain a decoded image with oil droplet features. This decoded image is then used as a mask image output by the oil droplet leakage detection model. Simultaneously, the oil droplet leakage location in the mask image is determined, and the oil droplet leakage location corresponding to the mask image is identified as the oil droplet leakage location of the device to be detected.

[0091] The aforementioned method for determining the location of oil droplet leaks in the device under test involves using a single-layer convolutional network in the oil droplet leak detection model to perform channel dimensionality reduction on the superimposed image, obtaining the corresponding channel dimensionality-reduced feature map. The encoder network in the oil droplet leak detection model then encodes the channel dimensionality-reduced feature map to determine the global feature information of the superimposed image. A multi-layer convolutional network in the oil droplet leak detection model extracts features from this global feature information to obtain the oil droplet features of the superimposed image. Finally, a decoder network in the oil droplet leak detection model decodes these oil droplet features to determine the location of the oil droplet leak in the device under test. By using an oil droplet leak detection model composed of a single-layer convolutional network, an encoder network, a multi-layer convolutional network, and a decoder network to detect superimposed images, manual monitoring is no longer required. This automates the monitoring of oil droplet-prone devices, reduces labor costs, and further improves the accuracy and timeliness of oil droplet leak detection.

[0092] Figure 4 This is a flowchart illustrating the training process for an oil droplet leakage detection model in one embodiment. Since oil droplet leakage is relatively rare, the number of image samples containing such leakage is limited, making it difficult to train the oil droplet leakage detection model. Therefore, to ensure successful training, this embodiment provides an optional method for training the oil droplet leakage detection model, including the following steps:

[0093] S401 acquires historical polarized light images and historical natural light images of the sample device.

[0094] The sample equipment includes oil droplet leakage equipment and oil droplet non-leakage equipment;

[0095] Among them, the historical polarized light image can be an image of light reflected from the surface of the sample device captured through a polarizing filter.

[0096] Among them, historical natural light images can be images of light reflected from the surface of a sample device captured by natural light sources.

[0097] Optionally, a sample device exhibiting oil droplet leakage is designated as an oil droplet leakage device, and its surface is photographed using acquisition devices installed around it to obtain historical natural light images and historical polarized light images. Conversely, a sample device without oil droplet leakage is designated as an oil droplet non-leakage device, and its surface is photographed using acquisition devices installed around it to obtain historical natural light images and historical polarized light images.

[0098] S402 performs channel overlay processing on historical polarized light images and historical natural light images to obtain a historical overlay image.

[0099] Optionally, the pixel values ​​of pixels at the same position in the historical polarized light image and the historical natural light image are superimposed to obtain an image with superimposed pixels, which is used as the historical superimposed image.

[0100] S403 performs augmentation processing on the historical overlay images corresponding to the oil droplet leakage equipment to obtain augmented overlay images.

[0101] Optionally, since there are relatively few oil droplet leakage devices, there are also relatively few historical overlay images corresponding to these devices. This results in a shortage of historical overlay images containing the location of oil droplet leakage, making it difficult to accurately train the oil droplet leakage detection model. Therefore, this embodiment expands the historical overlay images corresponding to the oil droplet leakage devices. Specifically, based on the collected historical overlay images corresponding to the oil droplet leakage devices, the historical overlay images are expanded (e.g., by adjusting the size of the historical overlay images, cropping and rotating them, etc.) to obtain expanded overlay images.

[0102] S404 trains an oil droplet leakage detection model based on historical overlay images and expanded overlay images.

[0103] Optionally, the acquired historical overlay images and expanded overlay images can be input into the oil droplet leakage detection model to train the oil droplet leakage detection model.

[0104] It should be noted that the difference between the predicted oil droplet leakage location and the actual oil droplet leakage location can be measured using the cross-entropy loss function, and the model parameters of the oil droplet leakage detection model can be adjusted to improve the accuracy of the oil droplet leakage detection model.

[0105] The aforementioned method for training the oil droplet leakage detection model involves acquiring historical polarized light images and historical natural light images of the sample devices. These images are then overlaid to obtain a historical overlay image. The historical overlay image corresponding to the oil droplet leakage device is then augmented to obtain an augmented overlay image. Based on both the historical overlay image and the augmented overlay image, the oil droplet leakage detection model is trained. Augmenting the historical overlay image corresponding to the oil droplet leakage device avoids the problem of insufficient historical overlay images containing the oil droplet leakage location due to a limited number of oil droplet leakage devices in the sample devices, thus preventing issues that could hinder the training of the oil droplet leakage detection model.

[0106] In one embodiment, this embodiment provides an optional method for configuring business parameters, and the method is illustrated using an application to a server as an example. For example... Figure 5 As shown, the method includes the following steps:

[0107] S501 acquires historical polarized light images and historical natural light images of the sample equipment. The sample equipment includes both oil droplet leakage devices and oil droplet non-leakage devices.

[0108] S502 performs channel overlay processing on historical polarized light images and historical natural light images to obtain a historical overlay image.

[0109] S503 performs augmentation processing on the historical overlay images corresponding to the oil droplet leakage equipment to obtain augmented overlay images.

[0110] S504 trains an oil droplet leakage detection model based on historical overlay images and expanded overlay images.

[0111] The S505 acquires the current polarized light image and the current natural light image of the device under test.

[0112] S506 overlays the current polarized light image and the current natural light image to obtain an overlaid image.

[0113] S507 uses a single-layer convolutional network in the oil droplet leakage detection model to perform channel dimensionality reduction on the superimposed image, obtaining the channel dimensionality reduction feature map corresponding to the superimposed image.

[0114] S508 uses the multi-head attention mechanism of the oil droplet leakage detection model to encode the channel dimensionality reduction feature map, thereby obtaining the weighted features of the channel dimensionality reduction feature map.

[0115] The S509 uses the gated linear unit of the oil droplet leakage detection model to adjust the weighted features and obtain the global feature information of the superimposed image.

[0116] S510 extracts features from the global feature information of the superimposed image through at least two grouped convolutions in the oil droplet leakage detection model, thus obtaining the oil droplet features of the superimposed image. The convolution kernel of each grouped convolution is a dynamic convolution kernel, and the size and shape of the dynamic convolution kernel are determined according to the global feature information input to each grouped convolution.

[0117] S511 decodes the oil droplet features through the decoder network in the oil droplet leakage detection model to obtain the location of the oil droplet leakage in the device under test.

[0118] Based on the location of the oil droplet leak in the device under test, the S512 outputs a warning message for the device under test.

[0119] It should be noted that steps S501-S504 are the process of training the oil droplet leakage detection model, and steps S504-S512 are the process of using the oil droplet leakage detection model to determine the location of the oil droplet leakage.

[0120] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0121] Based on the same inventive concept, this application also provides an oil droplet leakage detection device for implementing the oil droplet leakage detection method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations of one or more embodiments of the oil droplet leakage detection device provided below can be found in the limitations of the oil droplet leakage detection method described above, and will not be repeated here.

[0122] In one embodiment, such as Figure 6 As shown, an oil droplet leakage detection device 1 is provided, comprising: an image acquisition module 10, an image overlay module 11, and a position determination module 12, wherein:

[0123] Image acquisition module 10 is used to acquire the current polarized light image and the current natural light image of the device under test;

[0124] Image overlay module 11 is used to overlay the current polarized light image and the current natural light image to obtain an overlay image;

[0125] The location determination module 12 is used to perform oil droplet detection on the superimposed image using an oil droplet leakage detection model to determine the location of the oil droplet leakage in the device under test.

[0126] In one embodiment, such as Figure 7 As shown, Figure 6 The position determination module 12 in the middle includes:

[0127] Image dimensionality reduction unit 120 is used to perform channel dimensionality reduction on the superimposed image through a single-layer convolutional network in the oil droplet leakage detection model to obtain the channel dimensionality reduction feature map corresponding to the superimposed image.

[0128] The global feature extraction unit 121 is used to encode the channel dimension reduction feature map through the encoder network in the oil droplet leakage detection model to determine the global feature information of the superimposed image;

[0129] The oil drop feature extraction unit 122 is used to extract global feature information of the superimposed image through the multi-layer convolutional network in the oil drop leakage detection model to obtain the oil drop features of the superimposed image.

[0130] The oil droplet location determination unit 123 is used to decode the oil droplet features through the decoder network in the oil droplet leakage detection model to obtain the oil droplet leakage location of the device to be detected.

[0131] In one embodiment, Figure 7 The global feature extraction unit 121 includes:

[0132] The feature weighting subunit is used to encode the channel dimensionality reduction feature map through the multi-head attention mechanism of the oil droplet leakage detection model to obtain the weighted features of the channel dimensionality reduction feature map;

[0133] The weighted feature adjustment subunit is used to adjust the weighted features through the gated linear unit of the oil droplet leakage detection model to obtain the global feature information of the superimposed image.

[0134] In one embodiment, Figure 7 The oil droplet feature extraction unit 122 is also used to extract features from the global feature information of the superimposed image through at least two grouped convolutions in the oil droplet leakage detection model, so as to obtain the oil droplet features of the superimposed image.

[0135] In one embodiment, such as Figure 8 As shown, Figure 6 The oil droplet leakage detection device 1 in the middle also includes:

[0136] The early warning module 13 is used to output early warning information for the device under test based on the location of the oil droplet leakage.

[0137] In one embodiment, such as Figure 9 As shown, Figure 6 The oil droplet leakage detection device 1 in the middle also includes:

[0138] The historical image acquisition module 14 is used to acquire historical polarized light images and historical natural light images of the sample device; wherein, the sample device includes an oil droplet leakage device and an oil droplet non-leakage device;

[0139] The historical image overlay module 15 is used to perform channel overlay processing on historical polarized light images and historical natural light images to obtain a historical overlay image.

[0140] The historical image augmentation module 16 is used to augment the historical overlay image corresponding to the oil droplet leakage device to obtain an augmented overlay image;

[0141] Model training module 17 is used to train the oil droplet leakage detection model based on historical overlay images and expanded overlay images.

[0142] Each module in the aforementioned oil droplet leakage detection device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0143] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 10 As shown, the computer device includes a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When executed by the processor, the computer program implements an oil droplet leakage detection method. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0144] Those skilled in the art will understand that Figure 10 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0145] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0146] Acquire the current polarized light image and the current natural light image of the device under test;

[0147] The current polarized light image and the current natural light image are superimposed to obtain a superimposed image;

[0148] By using an oil droplet leakage detection model, oil droplet detection is performed on the superimposed images to determine the location of oil droplet leakage in the device under test.

[0149] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0150] By using a single-layer convolutional network in the oil droplet leakage detection model, channel dimensionality reduction processing is performed on the superimposed image to obtain the channel dimensionality reduction feature map corresponding to the superimposed image.

[0151] The encoder network in the oil droplet leakage detection model encodes the channel dimension reduction feature map to determine the global feature information of the superimposed image;

[0152] By using a multi-layer convolutional network in the oil droplet leakage detection model, the global feature information of the superimposed image is extracted to obtain the oil droplet features of the superimposed image;

[0153] The oil droplet features are decoded by the decoder network in the oil droplet leakage detection model to obtain the location of the oil droplet leakage in the device to be detected.

[0154] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0155] By using the multi-head attention mechanism of the oil droplet leakage detection model, the channel dimensionality reduction feature map is encoded to obtain the weighted features of the channel dimensionality reduction feature map;

[0156] By using the gated linear unit of the oil droplet leakage detection model, the weighted features are adjusted to obtain the global feature information of the superimposed image.

[0157] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0158] By using a multi-layer convolutional network in the oil droplet leakage detection model, feature extraction is performed on the global feature information of the superimposed image to obtain the oil droplet features of the superimposed image, including:

[0159] By using at least two grouped convolutions in the oil droplet leakage detection model, the global feature information of the superimposed image is extracted to obtain the oil droplet features of the superimposed image; the convolution kernel of each group convolution is a dynamic convolution kernel, and the size and shape of the dynamic convolution kernel are determined according to the global feature information input to each group convolution.

[0160] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0161] Based on the location of the oil droplet leak in the device under test, output early warning information for the device under test.

[0162] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0163] Acquire historical polarized light images and historical natural light images of the sample equipment; the sample equipment includes oil droplet leakage equipment and oil droplet non-leakage equipment;

[0164] Historical polarized light images and historical natural light images are overlaid using channel overlay processing to obtain a historical overlay image;

[0165] The historical overlay images corresponding to the oil droplet leakage device are augmented to obtain an augmented overlay image;

[0166] The oil droplet leakage detection model is trained based on historical overlay images and expanded overlay images.

[0167] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0168] Acquire the current polarized light image and the current natural light image of the device under test;

[0169] The current polarized light image and the current natural light image are superimposed to obtain a superimposed image;

[0170] By using an oil droplet leakage detection model, oil droplet detection is performed on the superimposed images to determine the location of oil droplet leakage in the device under test.

[0171] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:

[0172] Acquire the current polarized light image and the current natural light image of the device under test;

[0173] The current polarized light image and the current natural light image are superimposed to obtain a superimposed image;

[0174] By using an oil droplet leakage detection model, oil droplet detection is performed on the superimposed images to determine the location of oil droplet leakage in the device under test.

[0175] It should be noted that the data involved in this application (including but not limited to data used for analysis, stored data, and displayed data) are all information and data authorized by the user or fully authorized by all parties.

[0176] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0177] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0178] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. An oil droplet leakage detection method characterized by comprising: The method includes: Acquire the current polarized light image and the current natural light image of the device under test; The current polarized light image and the current natural light image are superimposed to obtain a superimposed image; The superimposed image is subjected to channel dimensionality reduction processing by a single-layer convolutional network in the oil droplet leakage detection model to obtain the channel dimensionality reduction feature map corresponding to the superimposed image. The channel dimensionality reduction feature map is encoded using the multi-head attention mechanism of the oil droplet leakage detection model to obtain the weighted features of the channel dimensionality reduction feature map; By using the gated linear unit of the oil droplet leakage detection model, the weighted features are adjusted to obtain the global feature information of the superimposed image; By using at least two grouped convolutions in the oil droplet leakage detection model, feature extraction is performed on the global feature information of the superimposed image to obtain the oil droplet features of the superimposed image; the convolution kernel of each grouped convolution is a dynamic convolution kernel, and the size and shape of the dynamic convolution kernel are determined according to the global feature information input to each grouped convolution; The oil droplet features are decoded by the decoder network in the oil droplet leakage detection model, and a mask image containing the oil droplet leakage location is output. The oil droplet leakage location corresponding to the mask image is then determined as the oil droplet leakage location of the device to be detected.

2. The method of claim 1, wherein, The device to be tested is one that needs to detect whether there is an oil droplet leak.

3. The method of claim 2, wherein, The equipment to be tested is substation equipment that uses oil.

4. The method of claim 1, wherein, The current polarized light image is an image of light reflected from the surface of the device under test, captured by a polarizing filter.

5. The method of claim 1, wherein, After determining the location of the oil droplet leak in the device under test, the method further includes: Based on the location of the oil droplet leakage in the device under test, an early warning message is output for the device under test.

6. The method according to any one of claims 1 to 5, characterized in that, The training methods for the oil droplet leakage detection model include: Acquire historical polarized light images and historical natural light images of the sample device; wherein, the sample device includes an oil droplet leakage device and an oil droplet non-leakage device; Historical polarized light images and historical natural light images are overlaid using channel overlay processing to obtain a historical overlay image; The historical overlay images corresponding to the oil droplet leakage device are augmented to obtain an augmented overlay image; The oil droplet leakage detection model is trained based on the historical overlay image and the expanded overlay image.

7. An oil droplet leakage detection device characterized by comprising: The device includes: The image acquisition module is used to acquire the current polarized light image and the current natural light image of the device under test; An image overlay module is used to overlay the current polarized light image and the current natural light image to obtain an overlay image; The location determination module is used to perform channel dimensionality reduction processing on the superimposed image through a single-layer convolutional network in the oil droplet leakage detection model to obtain a channel dimensionality reduction feature map corresponding to the superimposed image; to encode the channel dimensionality reduction feature map through the multi-head attention mechanism of the oil droplet leakage detection model to obtain weighted features of the channel dimensionality reduction feature map; to adjust the weighted features through the gated linear unit of the oil droplet leakage detection model to obtain global feature information of the superimposed image; to extract features from the global feature information of the superimposed image through at least two group convolutions in the oil droplet leakage detection model to obtain oil droplet features of the superimposed image; the convolution kernel of each group convolution is a dynamic convolution kernel, and the size and shape of the dynamic convolution kernel are determined according to the global feature information input to each group convolution; to decode the oil droplet features through the decoder network in the oil droplet leakage detection model, output a mask image containing the oil droplet leakage location, and determine the oil droplet leakage location corresponding to the mask image as the oil droplet leakage location of the device to be detected. 8.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-7. When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images