Method and device for monitoring running state of inland nuclear power station, terminal and storage medium

Abstract

The invention belongs to the technical field of remote sensing image processing and target identification, and particularly relates to an inland nuclear power station operation state monitoring method and device, a terminal and a storage medium, and the method comprises the steps: obtaining an optical remote sensing image and a thermal infrared image containing an inland nuclear power station region; processing the obtained optical remote sensing image to obtain a sample for training and testing; performing brightness temperature inversion on the obtained thermal infrared image to obtain brightness temperature abnormal data; constructing a cooling tower detection network model; inputting a to-be-detected optical remote sensing image into the cooling tower detection model to obtain initial detection results of the cooling tower in a working state and a non-working state; positioning to the corresponding position of the thermal infrared image according to the preliminary detection result of the cooling tower in the working state, and screening out the cooling tower in the working state according to the thermal infrared characteristics; and determining the operation state of the inland nuclear power station. According to the method, the monitoring timeliness can be improved, the labor cost is reduced, and the monitoring accuracy is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of remote sensing image processing and target recognition, and in particular relates to a monitoring method, device, terminal and storage medium for the operating status of an inland nuclear power plant. Background Art

[0002] The power industry is an important basic industry of the national economy. Electricity comes from a wide range of sources, such as thermal power, hydropower, wind power, nuclear power, solar power, etc. Nuclear power, as a sustainable clean energy, is an important option to help achieve the "dual carbon" goal.

[0003] Nuclear power plants are complex functional facilities that convert nuclear energy into electrical energy and are composed of multiple subunits. Different types of nuclear power plants may contain the same subunits, such as reactors, substations, etc. However, most inland nuclear power plants consider using natural ventilation cooling towers for secondary circulation heat dissipation. During the exchange and cooling process of cold air and hot water inside the cooling tower, a large amount of water vapor will be emitted into the atmosphere. Compared with coastal nuclear power plants, tall cooling tower buildings are one of the key components of inland nuclear power plants. Whether the cooling tower emits water vapor and the temperature anomaly between the cooling tower and the surrounding environment when emitting water vapor can be used as an important indicator of whether the inland nuclear power plant is in operation.

[0004] The traditional method mainly uses field surveys to obtain the operating status data of inland nuclear power plants, which has problems such as poor timeliness, high cost, susceptibility to smoke interference, and low accuracy. Summary of the invention

[0005] The purpose of the present invention is to provide a method, device, terminal and storage medium for monitoring the operating status of an inland nuclear power plant, so as to solve the problems in the prior art of monitoring the operating status of an inland nuclear power plant, such as poor timeliness, high cost, susceptibility to smoke interference and low accuracy.

[0006] The technical solution to achieve the purpose of the present invention is:

[0007] A method for monitoring the operating status of an inland nuclear power plant, the method comprising:

[0008] S00: Acquire optical remote sensing images and thermal infrared images of the inland nuclear power plant area;

[0009] S10: Processing the acquired optical remote sensing images to obtain samples for training and testing the cooling tower detection network model;

[0010] S20: Perform brightness temperature inversion on the acquired thermal infrared image to obtain brightness temperature anomaly data;

[0011] S30: construct cooling tower detection network model;

[0012] S40: inputting the optical remote sensing image to be detected into the cooling tower detection model to obtain preliminary detection results of the cooling tower in working state and non-working state;

[0013] S50: locating the corresponding position of the thermal infrared image according to the preliminary detection results of the cooling towers in working state, and further selecting the cooling towers in working state according to the thermal infrared characteristics;

[0014] S60: Determine the operating status of the inland nuclear power plant according to whether the cooling tower is in working status.

[0015] In the step S00, when obtaining the optical remote sensing image and thermal infrared image of the inland nuclear power plant area, the WorldView-2 high-resolution optical satellite remote sensing image with a spatial resolution of 0.5 meters and three channels of red, green and blue, and the SDGSAT-1 thermal infrared image with a spatial resolution of 30 meters are selected.

[0016] The step S10 also includes cropping the optical remote sensing image. When cropping the optical remote sensing image, a data enhancement method is used to expand the number of samples.

[0017] The step S20 is specifically as follows: converting the grayscale value of the thermal infrared image into thermal radiation brightness according to the sensor's own parameters, and then converting the thermal radiation brightness into brightness temperature; calculating the brightness temperature value of each pixel in the brightness temperature image T and the average value α of the overall brightness temperature image T one by one 1 The difference between 1 ; Then use threshold segmentation to segment the brightness temperature abnormal image Y 1 Perform binarization processing to obtain the overall pixel-by-pixel brightness temperature anomaly binary image Z 1 ; Calculate the brightness temperature value of the center pixel of the window and the average brightness temperature of all pixels in the window α 2 The difference between the brightness and temperature in the moving window is obtained. 2 ; Then use threshold segmentation to segment the brightness temperature abnormal image Y in the moving window 2 Perform binarization processing to obtain the brightness temperature anomaly binarization image Z' of the local moving window 2 ; Binarized image Z of overall pixel-by-pixel brightness and temperature anomaly 1 And the brightness temperature anomaly binary image Z' of the local moving window 2 Perform an intersection operation to obtain the final brightness temperature anomaly binary image Z.

[0018] In step S20, the calculation formula for converting the thermal radiation brightness into the brightness temperature is:

[0019]

[0020] Where T is the brightness temperature, λ 1 and λ 2 is a constant, L 1 , L 2 , L 3 B 1 , B 2 , B 3 Thermal radiation brightness of the band, B 1 , B 2 , B 3 They are the thermal infrared bands of thermal infrared images respectively.

[0021] In the step S30, the cooling tower detection network model includes a feature extraction module, a first feature optimization module, a feature fusion module, a second feature optimization module and a prediction module; the feature extraction module is used to process the input sample image and extract the features of the sample image; the first feature optimization module receives the extraction result from the feature extraction module, optimizes and calculates it to obtain the first optimized feature; the feature fusion module receives the first optimized feature, performs multi-scale fusion calculation on it, and obtains the fused feature; the second optimized feature module receives the fused feature, optimizes and calculates it to obtain the second optimized feature; the prediction module receives the second optimized feature, and performs calculation and prediction on it.

[0022] The step S40 is specifically as follows: using the cooling tower detection model of step S30 to detect the optical remote sensing image test sample of step S10; if a cooling tower target in working state or non-working state is detected, the model outputs the detection frame position information of the cooling tower in working state or non-working state, and uses the position information to calculate the longitude and latitude coordinates of the cooling tower to obtain preliminary detection results of the cooling tower in working state and non-working state.

[0023] The step S50 is specifically as follows: according to the latitude and longitude coordinates of the working cooling tower detected in step S40, locate the brightness temperature anomaly binary image Z obtained in step S20; if the position of the working cooling tower detected in step S40 is consistent with the heating area Z of the thermal infrared image, 1 If there is overlap, it is determined that the cooling tower is in a heating state, that is, the cooling tower is in a working state; if the position of the cooling tower in the working state detected in step S40 is consistent with the heating area Z of the thermal infrared image 1 When there is no overlap, the detection model may misjudge due to interference from thick smoke and fog, and the thermal infrared characteristics may be used to determine that the cooling tower is in a non-heating state, that is, the cooling tower is in a non-working state.

[0024] The step S60 is specifically as follows: the non-working cooling tower detected in step S40 has its corresponding nuclear power plant in a non-operating state; the working cooling tower detected in step S40 is also determined to be in a working state in step S50, and its corresponding nuclear power plant is in an operating state. If the working cooling tower detected in step S40 is determined to be non-working in step S50, its corresponding nuclear power plant is in a non-operating state.

[0025] A device for monitoring the operating status of an inland nuclear power plant, the device comprising:

[0026] An acquisition module is used to acquire optical remote sensing images and thermal infrared images of the inland nuclear power plant area;

[0027] A processing module, used for processing the acquired optical remote sensing images to obtain samples for training and testing the cooling tower detection network model;

[0028] The segmentation module is used to perform brightness temperature inversion on the acquired thermal infrared images to obtain brightness temperature anomaly data;

[0029] A detection module is used to input the optical remote sensing image to be detected into the constructed cooling tower detection model for detection, and obtain preliminary detection results of the cooling tower in working state and non-working state;

[0030] A screening module is used to locate the corresponding position of the thermal infrared image according to the preliminary detection results of the cooling towers in working state, and then further screen out the cooling towers in working state according to the thermal infrared characteristics;

[0031] The determination module is used to determine the operating state of the inland nuclear power plant according to whether the cooling tower is in working state.

[0032] A terminal device comprises a processor and a memory, wherein the memory stores a computer program, and the processor implements the steps of a method for monitoring the operating status of an inland nuclear power plant when running the computer program.

[0033] A storage medium stores a computer program, which implements the steps of a method for monitoring the operating status of an inland nuclear power plant when the computer program is executed.

[0034] The beneficial technical effects of the present invention are:

[0035] 1. The present invention provides a method for monitoring the operating status of an inland nuclear power plant. Based on the obvious geometric texture features of the cooling tower in the remote sensing image, the cooling tower is identified using remote sensing images, and then the operating status of the inland nuclear power plant can be effectively determined based on the water vapor emission characteristics and temperature anomaly characteristics under its working state.

[0036] 2. The present invention provides a method for monitoring the operating status of an inland nuclear power plant, which can improve the timeliness of monitoring, reduce labor costs, effectively eliminate smoke interference, and improve the accuracy of monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a flow chart of a method for monitoring the operating status of an inland nuclear power plant in an embodiment of the present invention;

[0038] Figure 2 The remote sensing images of the cooling tower in working state and non-working state in the embodiment of the present invention;

[0039] Figure 3 This is a flow chart of temperature anomaly detection using thermal infrared imaging in an embodiment of the present invention;

[0040] Figure 4 Schematic diagram of a cooling tower detection network model in an embodiment of the present invention;

[0041] Figure 5 A schematic diagram of a cooling tower detection network model constructed in an embodiment of the present invention;

[0042] Figure 6 It is a schematic diagram of the structure of a monitoring device for the operating status of an inland nuclear power plant in an embodiment of the present invention;

[0043] Figure 7 Schematic diagram of terminal equipment in an embodiment of the present invention.

[0044] In the figure: 1. Sample image; 2. Feature extraction module; 3. First feature optimization module; 4. Feature fusion module; 5. Second feature optimization module; 6. Prediction module; 61. Region proposal network; 62. Classification determination network. DETAILED DESCRIPTION

[0045] The present invention is further described in detail below with reference to the accompanying drawings and embodiments.

[0046] like Figure 1 As shown, the inland nuclear power plant operation status monitoring method provided by the embodiment of the present invention specifically includes the following steps:

[0047] S00: Acquire optical remote sensing images and thermal infrared images of the inland nuclear power plant area;

[0048] S10: Processing the acquired optical remote sensing images to obtain samples for training and testing the cooling tower detection network model;

[0049] S20: Perform brightness temperature inversion on the acquired thermal infrared image to obtain brightness temperature anomaly data;

[0050] S30: construct cooling tower detection network model;

[0051] S40: inputting the optical remote sensing image to be detected into the cooling tower detection model to obtain preliminary detection results of the cooling tower in working state and non-working state;

[0052] S50: locating the corresponding position of the thermal infrared image according to the preliminary detection results of the cooling towers in working state, and further selecting the cooling towers in working state according to the thermal infrared characteristics;

[0053] S60: Determine the operating status of the inland nuclear power plant according to whether the cooling tower is in working status.

[0054] In some embodiments, in step S00, when obtaining optical remote sensing images and thermal infrared images including the inland nuclear power plant area, the WorldView-2 high-resolution optical satellite remote sensing images with a spatial resolution of 0.5 meters and three channels of red, green and blue, and the SDGSAT-1 thermal infrared images with a spatial resolution of 30 meters can be selected.

[0055] In some embodiments, after acquiring the optical remote sensing image including the inland nuclear power plant area, the optical remote sensing image may be processed to obtain a pre-processed optical remote sensing image to facilitate subsequent steps.

[0056] In some embodiments, processing the acquired optical remote sensing image may include using processing software to perform radiometric calibration, geometric correction, and atmospheric correction on the image.

[0057] In some embodiments, in step S10, when using optical remote sensing images to obtain samples for training and testing the cooling tower detection network model, the number of training samples and test samples can be determined according to a predetermined ratio. The training samples can be used to train the established cooling tower detection network model to obtain a model that meets the requirements; the test samples can be used to input into the trained cooling tower detection network model for target detection and accuracy evaluation.

[0058] In some embodiments, when determining the number of training samples and test samples, the predetermined ratio may be 8 to 2, that is, the number of training samples may account for 8 / 10 of the total number of samples, and the number of test samples may account for 2 / 10 of the total number of samples.

[0059] In some embodiments, step S10 may further include cropping the optical remote sensing image, and the cropping size may be set to 1024×1024 pixels to form training samples and test samples.

[0060] In some embodiments, in step S10, when the optical remote sensing image is cropped to form training samples and test samples, a data enhancement method may be used to expand the number of samples.

[0061] In some embodiments, the cropped image may be enhanced by changing brightness, rotating at different angles, flipping in different ways, Mosaic, etc., which may increase the diversity of cooling tower samples.

[0062] In some embodiments, when training samples and test samples are obtained, a labeling tool can be used to label the cooling towers, and the labeling targets are divided into two categories: cooling towers in working state and cooling towers in non-working state, such as Figure 2 shown.

[0063] like Figure 3 As shown, it shows the thermal infrared image temperature anomaly detection process of an embodiment of the present invention. In step S20, the thermal infrared image can be processed before the brightness temperature inversion is performed on the acquired thermal infrared image to obtain a preprocessed thermal infrared image, which is convenient for subsequent steps.

[0064] In some embodiments, processing the acquired thermal infrared image may include performing radiometric calibration, geometric correction, and image registration on the image.

[0065] In some embodiments, when performing image registration processing on the acquired thermal infrared image, the bit depth and resolution of the acquired thermal infrared image may be adjusted to maintain a consistent range with the optical remote sensing image.

[0066] In some embodiments, when using thermal infrared images for brightness temperature inversion, the grayscale value of the thermal infrared image can be first converted into thermal radiation brightness according to the sensor's own parameters, and then the thermal radiation brightness can be converted into brightness temperature according to a calculation formula.

[0067] In some embodiments, the acquired thermal infrared image may be a SDGSAT-1 thermal infrared image, which includes three thermal infrared bands, namely, B 1 , B 2 , B 3 .

[0068] In some embodiments, the calculation formula for converting the thermal radiation brightness of the SDGSAT-1 thermal infrared image into brightness temperature is: Where T is the brightness temperature, λ 1 and λ 2 is a constant, L 1 , L 2 , L 3 B 1 , B 2 , B 3 The thermal radiation brightness of the band.

[0069] In some embodiments, the average value α of the overall brightness temperature image T may be calculated. 1 and standard deviation β 1 , and calculate the brightness temperature value of each pixel in the brightness temperature image T and the average value α of the overall brightness temperature image T one by one 1 The difference between 1 Then, the brightness temperature abnormal image Y is segmented by threshold value. 1 Perform binarization and set the feature threshold to 1.5β 1 When the brightness temperature abnormal image Y 1 The pixel value is less than or equal to 1.5β 1 , the pixel is not a temperature anomaly pixel, and its value is set to 0; when the brightness temperature anomaly image Y 1 The pixel value is greater than 1.5β 1 , the pixel is a temperature anomaly pixel, its value is set to 1, the brightness temperature anomaly image Y 1 After pixel-by-pixel threshold segmentation, the brightness temperature anomaly binary image Z is obtained. 1 .

[0070] In some embodiments, a moving window of 9×9 pixels may be selected to traverse the brightness temperature image T with a step size of 1 pixel, and the average brightness temperature α of all pixels in the window may be calculated. 2 and standard deviation β 2 , and calculate the brightness temperature value of the pixel at the center of the window and the average brightness temperature of all pixels in the window α 2 The difference between the brightness and temperature in the moving window is obtained. 2 Then, the brightness temperature abnormal image Y in the moving window is segmented by threshold value. 2 Perform binarization and set the feature threshold to 1.5β 2 When the brightness temperature abnormal image Y 2 The pixel value is less than or equal to 1.5β 2 , the pixel is not a temperature anomaly pixel, and its value is set to 0; when the brightness temperature anomaly image Y 2 The pixel value is greater than 1.5β 2 , the pixel is a temperature anomaly pixel, its value is set to 1, and the brightness temperature anomaly image Y in the moving window 2 After threshold segmentation, it becomes a black and white binary image Z 2 After the moving window traverses and calculates the entire brightness temperature image T, the brightness temperature anomaly binary image Z' is obtained. 2 .

[0071] The brightness temperature anomaly binary image Z is obtained by overall pixel-by-pixel calculation 1And the brightness temperature anomaly binary image Z' calculated by the local 9×9 moving window 2 The intersection operation is performed to obtain the final brightness temperature anomaly binary image Z, which can reflect the heating area and the non-heating area.

[0072] like Figure 4 As shown, it shows the cooling tower detection network model of the embodiment of the present invention. In step S30, the cooling tower detection network model may include a feature extraction module 2, a first feature optimization module 3, a feature fusion module 4, a second feature optimization module 5 and a prediction module 6. Among them, the feature extraction module 2 is used to process the input sample image 1 and extract the features of the sample image; the first feature optimization module 3 receives the extraction result from the feature extraction module 2, optimizes and calculates it to obtain the first optimized feature; the feature fusion module 4 receives the first optimized feature, performs multi-scale fusion calculation on it, and obtains the fused feature; the second optimized feature module 5 receives the fused feature, optimizes and calculates it to obtain the second optimized feature; the prediction module 6 receives the second optimized feature, and calculates and predicts it.

[0073] The feature extraction module 2 may include j layers, each used to extract multi-scale features of the cooling tower target in the input sample image 1.

[0074] In some embodiments, the 1st to jth layers of the feature extraction module 2 may be composed of convolution blocks of the same structure, and the selected convolution block structure includes a 3×3 convolution layer, a batch normalization layer, and a ReLU activation layer. Among them, an additional maximum pooling layer with a stride of 2 is added between the convolution blocks of the j-1th layer and the jth layer.

[0075] Sample image 1 can be input to the first convolution block of feature extraction module 2. The result extracted by the first convolution block can be input to the maximum pooling layer. The result obtained by the maximum pooling layer can be input to the second convolution block. By analogy, the result extracted by the j-1th convolution block is input to the maximum pooling layer, and the result is input to the jth convolution block. The feature map obtained by sample image 1 after being processed by feature extraction module 2 can be expressed as {A 1 , A 2 , A 3 , A 4 , …A j}.

[0076] The extraction results of the 1st layer convolution block of the feature extraction module 2 are input into the maximum pooling layer, and the results obtained by the maximum pooling layer can also be input into the corresponding 1st layer first feature optimization module 3. And so on, the extraction results of the jth layer convolution block are input into the maximum pooling layer, and the results obtained by the maximum pooling layer can also be input into the corresponding jth layer first feature optimization module 3.

[0077] Each layer in the first feature optimization module 3 of all j layers can screen the input features and enhance the feature representation related to the cooling tower.

[0078] In some embodiments, the first feature optimization module 3 can be coordinate attention (CA), which can enable the first feature optimization module 3 to screen out features that are more useful to the cooling tower from the complex and rich features, optimize the feature representation, effectively remove useless information interference, and reduce the probability of false detection of the cooling tower.

[0079] Feature graph 1 , A 2 , A 3 , A 4 , …A j After CA operation, we can get the feature map after layer-by-layer coordinate attention optimization and enhancement {B 1 , B 2 , B 3 , B 4 , …B j}.

[0080] In some embodiments, the feature fusion module 4 uses dense connection to merge, and realizes dense fusion of features of all j scales by pixel-by-pixel addition, making full use of spatial detail information and abstract semantic information at different levels, and effectively improving problems such as blurred boundaries and lost detection targets of cooling tower detection.

[0081] The result processed by the first feature optimization module 3 of the first layer, the result processed by the feature fusion module 4 of the second layer, the result processed by the feature fusion module 4 of the third layer, and so on until the result processed by the feature fusion module 4 of the jth layer is all input to the feature fusion module 4 of the first layer, and after superposition processing, the first layer fusion feature C is obtained. 1 The result of the first feature optimization module 3 of the first layer, the result of the first feature optimization module 3 of the second layer, the result of the feature fusion module 4 of the third layer, the result of the feature fusion module 4 of the fourth layer, and so on until the result of the feature fusion module 4 of the jth layer is all input to the feature fusion module 4 of the second layer, and after superposition processing, the fusion feature C of the second layer is obtained. 2 ; By analogy, the result processed by the first feature optimization module 3 of the first layer, the result processed by the first feature optimization module 3 of the second layer, and so on until the result processed by the first feature optimization module 3 of the j-1th layer, and the result processed by the feature fusion module 4 of the j-1th layer are all input to the feature fusion module 4 of the j-1th layer, and after superposition processing, the fusion feature C of the j-1th layer is obtained. j-1 The result of the processing of the first feature optimization module 3 of the jth layer is directly input into the feature fusion module 4 of the jth layer, and after processing, the fusion feature C of the jth layer is obtained.j At this point, the feature map {C 1 , C 2 , C 3 , C 4 ,…C j}.

[0082] The result processed by the first-layer feature fusion module 4 can also be input into the corresponding first-layer second feature optimization module 5, and so on. The result processed by the j-th layer feature fusion module 4 can also be input into the corresponding j-th layer second feature optimization module 5.

[0083] In some embodiments, the second feature optimization module 5 can be an Interleaved Sparse Self-Attention (ISSAM), which can enable the second feature optimization module 5 to model long-distance global context information, obtain more discriminative features of the cooling tower, reduce redundant feature interference, and improve the accuracy of cooling tower detection.

[0084] Feature map 1 , C 2 , C 3 , C 4 ,…C j After the ISSAM operation, we can get the enhanced feature map after layer-by-layer optimization. 1 , D 2 , D 3 , D 4 ,…D j}.

[0085] In some embodiments, the prediction module 6 may include a region proposal network 62, and the feature map {D 1 , D 2 , D 3 , D 4 ,…D j} can be input into the regional proposal network 62. After being processed by the regional proposal network 62, the preliminary positioning results of the cooling towers in the working state and the non-working state can be determined; the prediction module 6 can also include a classification determination network 61. The preliminary cooling tower positioning results can be input into the classification determination network 61. After being processed by the classification determination network 61, the model's prediction results for the cooling towers in the working state and the non-working state can be obtained.

[0086] In some embodiments, the region proposal network 62 may include one 3×3 convolution operation, two 1×1 convolution operations, and the anchor point size is set to {32 2 , 64 2 , 128 2 , 256 2} pixels, the aspect ratio of each anchor point includes {1:1, 1:2, 2:1}, and the preliminary positioning results of the cooling towers in working and non-working states can be a series of candidate boxes containing the cooling towers and with the coordinates of the corner points.

[0087] In some embodiments, the classification determination network 61 may include a RoI (Region of Interest) pooling layer and a fully connected layer.

[0088] In some embodiments, the classification determination network 61 can process the results output by the region proposal network 62 to extract the candidate area feature map; then calculate the category and position offset of each candidate box through the fully connected layer, and obtain the final cooling tower detection result through regression.

[0089] like Figure 5 As shown, it shows a schematic diagram of the cooling tower detection network model constructed in the embodiment of the present invention, and j=4 at this time. Figure 5 , details the process of building a cooling tower detection network model, which can include the following steps:

[0090] The sample image 1 used for training is input into the feature extraction module 2, and the feature map A is obtained through the first layer feature extraction module 2 1 , obtain feature map A through the second layer feature extraction module 2 2 , and so on, the feature map {A 1 , A 2 , A 3 , A 4}; The feature map {A 1 , A 2 , A 3 , A 4} Input to the first feature optimization module 3, and obtain the feature map B through the first feature optimization module 3 of the first layer 1 , obtain feature map B through the first feature optimization module 3 of the second layer 2 , and so on, the feature map {B 1 , B 2 , B 3 , B 4}; The feature map {B 1 , B 2 , B 3 , B 4} is input to feature fusion module 4, and B 4 Input to the 4th layer feature fusion module 4 to get C 4 , C 4 With B 1 , B 2 , B 3After all four scale features are superimposed, they are input into the third layer feature fusion module 4, and after processing, C 3 , C 3 With C 4 , B 1 , B 2 After all four scale features are superimposed, they are input into the second-layer feature fusion module 4, and C is obtained after processing. 2 , C 2 With C 3 , C 4 , B 1 After all four scale features are superimposed, they are input into the first-layer feature fusion module 4, and C is obtained after processing. 1 , the feature map {C 1 , C 2 , C 3 , C 4}.

[0091] Feature map 1 , C 2 , C 3 , C 4} can also be input to the second feature optimization module 5, C 4 Input to the second feature optimization module 5 of the 4th layer to obtain D 4 , C 3 Input to the second feature optimization module 5 of the third layer to obtain D 3 , C 2 Input to the second feature optimization module 5 of the second layer to obtain D 2 , C 1 Input to the second feature optimization module 5 of the first layer to obtain D 1 , the feature map is obtained through the second feature optimization module 5

[0092] {D 1 , D 2 , D 3 , D 4}.

[0093] The feature map {D 1 , D 2 , D 3 , D 4 The preliminary cooling tower positioning result is input into the regional proposal network 62 of the prediction module 6 to determine the preliminary cooling tower positioning result; the preliminary cooling tower positioning result is then input into the classification determination network 61 to obtain the final prediction results of the cooling tower in the working state and the non-working state.

[0094] In some embodiments, in step S40, the cooling tower detection model of step S30 is used to detect the optical remote sensing image test sample of step S10. If a cooling tower target in a working state or a non-working state is detected, the model outputs the detection frame position information of the cooling tower in a working state or a non-working state, such as the position values ​​v1, v2, v3, and v4 of the four corner points of the detection frame. The longitude and latitude coordinates of the four corner points of the cooling tower can be calculated using the four corner point position values ​​to obtain preliminary detection results of the cooling tower in a working state and a non-working state.

[0095] In some embodiments, in step S50, the brightness temperature anomaly binary image Z obtained in step S20 can be located according to the latitude and longitude coordinates of the working cooling tower detected in step S40. If the position of the working cooling tower detected in step S40 is consistent with the heating area Z of the thermal infrared image, 1 If there is overlap, it is determined that the cooling tower is in a heating state, that is, the cooling tower is in a working state; if the position of the cooling tower in the working state detected in step S40 is consistent with the heating area Z of the thermal infrared image 1 When there is no overlap, the detection model may misjudge due to interference from thick smoke and fog, and the thermal infrared characteristics may be used to determine that the cooling tower is in a non-heating state, that is, the cooling tower is in a non-working state.

[0096] In some embodiments, in step S60, the corresponding nuclear power plant operation state is determined according to the preliminary detection result of whether the cooling tower is in the working state in step S40 and the result of the cooling tower working state further determined in step S50. Specifically, if the cooling tower in the non-working state detected in step S40 is in the non-operating state, the corresponding nuclear power plant is in the non-operating state; if the cooling tower in the working state detected in step S40 is also determined to be in the working state in step S50, the corresponding nuclear power plant is in the operating state. If the cooling tower in the working state detected in step S40 is determined to be in the non-operating state in step S50, the corresponding nuclear power plant is in the non-operating state.

[0097] The following is an embodiment of the device of the present invention. For details not described in detail therein, reference may be made to the corresponding method embodiment described above.

[0098] like Figure 6 As shown, the inland nuclear power plant operation status monitoring device provided by the embodiment of the present invention includes an acquisition module 601, a processing module 602, a segmentation module 603, a detection module 604, a screening module 605 and a determination module 606.

[0099] The acquisition module 601 is used to acquire optical remote sensing images and thermal infrared images of the inland nuclear power plant area.

[0100] The processing module 602 is used to process the acquired optical remote sensing images to obtain samples for training and testing the cooling tower detection network model.

[0101] The segmentation module 603 is used to perform brightness temperature inversion on the acquired thermal infrared image to obtain brightness temperature anomaly data.

[0102] The detection module 604 is used to input the optical remote sensing image to be detected into the constructed cooling tower detection model for detection, and obtain preliminary detection results of the cooling tower in the working state and the non-working state.

[0103] The screening module 605 is used to locate the corresponding position of the thermal infrared image according to the preliminary detection results of the cooling towers in working state, and then further screen out the cooling towers in working state according to the thermal infrared characteristics.

[0104] The determination module 606 is used to determine the operating state of the inland nuclear power plant according to whether the cooling tower is in working state.

[0105] The inland nuclear power plant operation status monitoring device provided by the embodiment of the present invention obtains optical remote sensing images and thermal infrared images including the inland nuclear power plant area through the acquisition module 601; the processing module 602 processes the acquired optical remote sensing images to obtain samples for training and testing the cooling tower detection network model; the segmentation module 603 performs brightness temperature inversion on the acquired thermal infrared images to obtain brightness temperature anomaly data; the detection module 604 inputs the optical remote sensing image to be detected into the constructed cooling tower detection model for detection, and obtains preliminary detection results of cooling towers in working state and non-working state; the screening module 605 locates the cooling towers in working state to the corresponding position of the thermal infrared image according to the preliminary detection results of the cooling towers, and further screens out the cooling towers in working state according to the thermal infrared characteristics; the determination module 606 determines the operation status of the inland nuclear power plant according to whether the cooling towers are in working state. The present invention solves the problems of poor timeliness, high cost, susceptibility to smoke interference and low accuracy in the prior art for identifying the operating status of inland nuclear power plants. The preliminary detection results of cooling towers in working and non-working states in optical remote sensing images are obtained through a cooling tower detection model, and then the cooling towers in working state are secondary screened and determined based on the abnormal thermal infrared temperature characteristics of the cooling towers in working state in the preliminary detection results. The operating status of the inland nuclear power plant is then determined based on the working status of the cooling towers, thereby achieving the effect of efficiently, accurately and low-costly determining the operating status of the inland nuclear power plant.

[0106] In the inland nuclear power plant operation status monitoring device provided in an embodiment of the present invention, the processing module 602 is used to process the acquired optical remote sensing images to obtain samples for training and testing the cooling tower detection network model. In some embodiments, the processing process may include radiation calibration, geometric correction, and atmospheric correction. When using optical remote sensing images to obtain samples for training and testing the cooling tower detection network model, the optical remote sensing images can be cropped to form training samples and test samples, and data enhancement methods can be used to expand the number of samples. In some embodiments, data enhancement methods may include changing brightness, rotating at different angles, flipping in different ways, Mosaic, etc.

[0107] In the inland nuclear power plant operation status monitoring device provided in an embodiment of the present invention, the segmentation module 603 is used to perform brightness temperature inversion on the acquired thermal infrared image to obtain brightness temperature anomaly data. The first step is to process the thermal infrared image. In some embodiments, the processing process may include radiation calibration, geometric correction and image registration. Among them, when performing image registration processing on the acquired thermal infrared image, the bit depth and resolution of the acquired thermal infrared image can be adjusted to keep it in the same range as the optical remote sensing image. The second step is to perform brightness temperature inversion of the thermal infrared image. The grayscale value of the thermal infrared image can be converted into thermal radiation brightness according to the sensor's own parameters, and then the thermal radiation brightness can be converted into brightness temperature according to the calculation formula. In some embodiments, the acquired thermal infrared image can select SDGSAT-1, and SDGSAT-1 contains B 1 , B 2 , B 3 The brightness temperature of the three thermal infrared bands is calculated as follows: Where T is the brightness temperature, λ 1 and λ 2 is a constant, L 1 , L 2 , L 3 B 1 , B 2 , B 3 The third step is to calculate the brightness temperature anomaly data and binarize it. In some embodiments, the average value α of the overall brightness temperature image T can be calculated. 1 and standard deviation β 1 , and calculate the brightness temperature value of each pixel in the brightness temperature image T and the average value α of the overall brightness temperature image T one by one 1 The difference between 1 Then, the brightness temperature abnormal image Y is segmented by threshold value. 1 Perform binarization and set the feature threshold to 1.5β 1 When the brightness temperature abnormal image Y1 The pixel value is less than or equal to 1.5β 1 , the pixel is not a temperature anomaly pixel, and its value is set to 0; when the brightness temperature anomaly image Y 1 The pixel value is greater than 1.5β 1 , the pixel is a temperature anomaly pixel, its value is set to 1, the brightness temperature anomaly image Y 1 After pixel-by-pixel threshold segmentation, the brightness temperature anomaly binary image Z is obtained. 1 In some embodiments, a moving window of 9×9 pixels may be selected to traverse the brightness temperature image T with a step size of 1 pixel, and the average brightness temperature α of all pixels in the window may be calculated. 2 and standard deviation β 2 , and calculate the brightness temperature value of the pixel at the center of the window and the average brightness temperature of all pixels in the window α 2 The difference between the brightness and temperature in the moving window is obtained. 2 Then, the brightness temperature abnormal image Y in the moving window is segmented by threshold value. 2 Perform binarization and set the feature threshold to 1.5β 2 When the brightness temperature abnormal image Y 2 The pixel value is less than or equal to 1.5β 2 , the pixel is not a temperature anomaly pixel, and its value is set to 0; when the brightness temperature anomaly image Y 2 The pixel value is greater than 1.5β 2 , the pixel is a temperature anomaly pixel, its value is set to 1, and the brightness temperature anomaly image Y in the moving window 2 After threshold segmentation, it becomes a black and white binary image Z 2 After the moving window traverses and calculates the entire brightness temperature image T, the brightness temperature anomaly binary image Z' is obtained. 2 . Binarize the brightness temperature anomaly image Z 1 and Z' 2 The intersection operation is performed to obtain the final brightness temperature anomaly binary image Z, which can reflect the heating area and the non-heating area.

[0108] In the inland nuclear power plant operation status monitoring device provided in an embodiment of the present invention, the detection module 604 is used to input the optical remote sensing image to be detected into the constructed cooling tower detection model for detection. If a cooling tower target in a working state or a non-working state is detected, the model outputs the detection frame position information of the cooling tower in a working state or a non-working state. The longitude and latitude coordinates of the four corner points of the cooling tower can be calculated using the position information of the four corner points of the detection frame to obtain preliminary detection results of the cooling tower in a working state and a non-working state.

[0109] The cooling tower detection model includes a feature extraction module 2, a first feature optimization module 3, a feature fusion module 4, a second optimized feature module 5 and a prediction module 6.

[0110] The feature extraction module 2 is used to process the input sample image 1 and extract the multi-scale features of the cooling tower target in the input sample image 1. In some embodiments, the feature extraction module 2 may include j layers, and the 1st to jth layers may be composed of convolution blocks of the same structure, and the convolution blocks include a 3×3 convolution layer, a batch normalization layer, and a ReLU activation layer, and an additional maximum pooling layer with a step size of 2 is added between the convolution blocks of the j-1th layer and the jth layer.

[0111] The first feature optimization module 3 receives the extraction results from the feature extraction module 2, and screens the features of each layer input to enhance the feature representation related to the cooling tower. In some embodiments, the first feature optimization module 3 can be coordinate attention.

[0112] The feature fusion module 4 receives the optimization results from the first feature optimization module 3, merges them in a dense connection manner, and realizes feature fusion of all scales by pixel-by-pixel addition. In some embodiments, the results processed by the first feature optimization module 3 of the first layer, the results processed by the second layer feature fusion module 4, the results processed by the third layer feature fusion module 4, and so on until the results processed by the jth layer feature fusion module 4 are all input to the first layer feature fusion module 4, and after superposition processing, the first layer fusion feature C is obtained. 1 The result of the first feature optimization module 3 of the first layer, the result of the first feature optimization module 3 of the second layer, the result of the feature fusion module 4 of the third layer, the result of the feature fusion module 4 of the fourth layer, and so on until the result of the feature fusion module 4 of the jth layer is all input to the feature fusion module 4 of the second layer, and after superposition processing, the fusion feature C of the second layer is obtained. 2 ; By analogy, the result processed by the first feature optimization module 3 of the first layer, the result processed by the first feature optimization module 3 of the second layer, and so on until the result processed by the first feature optimization module 3 of the j-1th layer, and the result processed by the feature fusion module 4 of the j-1th layer are all input to the feature fusion module 4 of the j-1th layer, and after superposition processing, the fusion feature C of the j-1th layer is obtained. j-1 The result of the processing of the first feature optimization module 3 of the jth layer is directly input into the feature fusion module 4 of the jth layer, and after processing, the fusion feature C of the jth layer is obtained. j .

[0113] The second feature optimization module 5 receives the fusion features from the feature fusion module 4 and further optimizes and enhances them. Among them, the first-layer second feature optimization module 5 receives the results processed by the corresponding first-layer feature fusion module 4, the second-layer second feature optimization module 5 receives the results processed by the corresponding second-layer feature fusion module 4, and so on, the j-th layer second feature optimization module 5 receives the results processed by the corresponding j-th layer feature fusion module 4. In some embodiments, the second feature optimization module 5 can be a cross-sparse self-attention, which helps the second feature optimization module 5 to model the long-distance global context information, obtain more discriminative features of the cooling tower, reduce redundant feature interference, and improve the accuracy of cooling tower detection.

[0114] The prediction module 6 receives the optimization result from the second optimization feature module 5 and performs calculation and prediction on it. The prediction module 6 may include a region proposal network 62 and a classification determination network 61. After the second optimization feature is processed by the region proposal network 62, the preliminary positioning results of the cooling tower in the working state and the non-working state can be determined. After being processed by the classification determination network 61, the prediction results of the model for the cooling tower in the working state and the non-working state can be obtained.

[0115] In the inland nuclear power plant operation status monitoring device provided by the embodiment of the present invention, the screening module 605 is used to locate the corresponding position of the thermal infrared image according to the preliminary detection result of the working cooling tower, and then further screen out the cooling tower in working state according to the thermal infrared characteristics. Specifically, according to the longitude and latitude coordinates of the working cooling tower preliminarily detected by the detection module 604, locate the brightness temperature anomaly binary image Z obtained by the segmentation module 603. If the position of the working cooling tower preliminarily detected is consistent with the heating area Z of the thermal infrared image, 1 If there is overlap, it is determined that the cooling tower is in a heating state, that is, the cooling tower is in a working state; if the initially detected working state cooling tower position is consistent with the heating area Z of the thermal infrared image 1 When there is no overlap, the detection model may misjudge due to interference from thick smoke and fog, and the thermal infrared characteristics may be used to determine that the cooling tower is in a non-heating state, that is, the cooling tower is in a non-working state.

[0116] In the inland nuclear power plant operation status monitoring device provided by the embodiment of the present invention, the determination module 606 is used to determine the operation status of the inland nuclear power plant according to whether the cooling tower is in working state. It includes determining the corresponding nuclear power plant operation status according to the preliminary detection result of whether the cooling tower is in working state by the detection module 604 and the cooling tower working state result further determined by the screening module 605. Specifically, the non-working cooling tower detected by the detection module 604 has a corresponding nuclear power plant in a non-operating state; the working cooling tower detected by the detection module 604 is also determined to be in working state in the screening module 605, and the corresponding nuclear power plant is in an operating state; the working cooling tower detected by the detection module 604 is determined to be in a non-working state in the screening module 605, and the corresponding nuclear power plant is in a non-operating state.

[0117] like Figure 7 , which shows a schematic diagram of a terminal device provided in an embodiment of the present application, wherein the terminal device 70 includes a processor 701, a memory 702, and a computer program 703 located in the memory 702 and executable on the processor 701. The processor 701 executes the computer program 703 to implement the various steps in the above-mentioned inland nuclear power plant operation status monitoring method embodiment.

[0118] The computer program 703 is composed of one or more modules, which are stored in the memory 702 and can be executed by the processor 701 to complete the present invention. The one or more modules can be a series of computer code instructions for implementing a certain function, and the code instructions are used to describe the operation process of the computer program 703 in the terminal device 70.

[0119] The terminal device 70 may be a server, a PC, a PDA, etc. The terminal device 70 includes but is not limited to a processor 701 and a memory 702. Those skilled in the art will appreciate that Figure 7 The terminal device 70 is merely an example and does not constitute a limitation on the terminal device 70 . The terminal device 70 may include more or fewer components than those in the example, or a combination of different components. For example, the terminal device 70 may also include input and output devices, network devices, computing acceleration devices, etc.

[0120] The processor 701 may be a central processing unit, a microprocessor, a digital signal processor, a programmable logic device, etc.

[0121] The memory 702 may be an internal storage device of the terminal device 70, such as a hard disk, a memory, etc. The memory 702 may also be an external storage device of the terminal device 70, such as an external hard disk, a flash memory card, etc. The memory 702 is used to store the computer program 703 and other programs or data required by the terminal device 70.

[0122] Those skilled in the art can realize that, for the convenience of description, only the division of the above-mentioned functional modules is used as an example. In practical applications, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The functional modules in the embodiment can be integrated in the same processing unit, or integrated in different processing units individually or in different combinations. The names of the above-mentioned functional modules are only for distinguishing each other and are not used to limit the scope of protection of this application. The specific working process of the above-mentioned functional modules can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0123] Those skilled in the art will appreciate that the example modules and algorithm steps described in conjunction with the embodiments disclosed in the present invention can be implemented in hardware or in a combination of computer software and hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0124] The present invention is described in detail above with reference to the accompanying drawings and embodiments, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge of ordinary technicians in the field without departing from the purpose of the present invention. The contents not described in detail in the present invention can adopt the existing technology.

Claims

1. A method for monitoring the operating status of an inland nuclear power plant, characterized in that: The method comprises: S00: Acquire optical remote sensing images and thermal infrared images of the inland nuclear power plant area; S10: Processing the acquired optical remote sensing images to obtain samples for training and testing the cooling tower detection network model; S20: Perform brightness temperature inversion on the acquired thermal infrared image to obtain brightness temperature anomaly data; S30: construct cooling tower detection network model; S40: inputting the optical remote sensing image to be detected into the cooling tower detection model to obtain preliminary detection results of the cooling tower in working state and non-working state; S50: locating the corresponding position of the thermal infrared image according to the preliminary detection results of the cooling towers in working state, and further selecting the cooling towers in working state according to the thermal infrared characteristics; S60: Determine the operating status of the inland nuclear power plant according to whether the cooling tower is in working status.

2. A method for monitoring the operating status of an inland nuclear power plant according to claim 1, characterized in that: In the step S00, when obtaining the optical remote sensing image and thermal infrared image of the inland nuclear power plant area, the WorldView-2 high-resolution optical satellite remote sensing image with a spatial resolution of 0.5 meters and three channels of red, green and blue, and the SDGSAT-1 thermal infrared image with a spatial resolution of 30 meters are selected.

3. The method for monitoring the operating status of an inland nuclear power plant according to claim 1, characterized in that: The step S10 also includes cropping the optical remote sensing image. When cropping the optical remote sensing image, a data enhancement method is used to expand the number of samples.

4. The method for monitoring the operating status of an inland nuclear power plant according to claim 1, characterized in that: The step S20 is specifically as follows: converting the gray value of the thermal infrared image into thermal radiation brightness according to the parameters of the sensor, and then converting the thermal radiation brightness into brightness temperature; calculating the difference between the brightness temperature value of each pixel in the brightness temperature image T and the average value α1 of the overall brightness temperature image T one by one, thereby obtaining a brightness temperature abnormality image Y1; Then, the brightness temperature anomaly image Y1 is binarized by threshold segmentation to obtain the overall pixel-by-pixel brightness temperature anomaly binarized image Z1; The difference between the brightness temperature value of the central pixel of the window and the average value α2 of the brightness temperature of all pixels in the window is calculated to obtain the brightness temperature anomaly image Y2 in the moving window; the brightness temperature anomaly image Y2 in the moving window is then binarized by threshold segmentation to obtain the brightness temperature anomaly binary image Z'2 of the local moving window; the overall pixel-by-pixel brightness temperature anomaly binary image Z1 and the brightness temperature anomaly binary image Z'2 of the local moving window are intersected to obtain the final brightness temperature anomaly binary image Z.

5. A method for monitoring the operating status of an inland nuclear power plant according to claim 4, characterized in that: In step S20, the calculation formula for converting the thermal radiation brightness into the brightness temperature is: Where T is the brightness temperature, λ1 and λ2 are constants, L1, L2, and L3 are the thermal radiation brightness of the B1, B2, and B3 bands, respectively, and B1, B2, and B3 are the thermal infrared bands of thermal infrared images, respectively.

6. The method for monitoring the operating status of an inland nuclear power plant according to claim 1, characterized in that: In the step S30, the cooling tower detection network model includes a feature extraction module, a first feature optimization module, a feature fusion module, a second feature optimization module and a prediction module; the feature extraction module is used to process the input sample image and extract the features of the sample image; the first feature optimization module receives the extraction result from the feature extraction module, and optimizes and calculates it to obtain the first optimized feature; feature The fusion module receives the first optimized feature, performs multi-scale fusion calculation on it, and obtains a fusion feature; The second optimized feature module receives the fused feature and optimizes and calculates it to obtain the second optimized feature; The prediction module receives the second optimized feature and performs calculation prediction on it.

7. A method for monitoring the operating status of an inland nuclear power plant according to claim 1, characterized in that: The step S40 is specifically as follows: using the cooling tower detection model of step S30 to detect the optical remote sensing image test sample of step S10; if a cooling tower target in working state or non-working state is detected, the model outputs the detection frame position information of the cooling tower in working state or non-working state, and uses the position information to calculate the longitude and latitude coordinates of the cooling tower to obtain preliminary detection results of the cooling tower in working state and non-working state.

8. The method for monitoring the operating status of an inland nuclear power plant according to claim 1, characterized in that: The step S50 is specifically as follows: according to the latitude and longitude coordinates of the working cooling tower detected in step S40, positioning to the brightness temperature anomaly binary image Z obtained in step S20; if the position of the working cooling tower detected in step S40 overlaps with the heating area Z1 of the thermal infrared image, it is determined that the cooling tower is in a heating state, that is, the cooling tower is in a working state; if the position of the working cooling tower detected in step S40 does not overlap with the heating area Z1 of the thermal infrared image, the detection model may be misjudged due to interference from thick smoke and fog, and the thermal infrared characteristics are used to determine that the cooling tower is in a non-heating state, that is, the cooling tower is in a non-working state.

9. The method for monitoring the operating status of an inland nuclear power plant according to claim 1, characterized in that: The step S60 is specifically as follows: the non-working cooling tower detected in step S40 has its corresponding nuclear power plant in a non-operating state; the working cooling tower detected in step S40 is also determined to be in a working state in step S50, and its corresponding nuclear power plant is in an operating state. If the working cooling tower detected in step S40 is determined to be non-working in step S50, its corresponding nuclear power plant is in a non-operating state.

10. A monitoring device for the operating status of an inland nuclear power plant, characterized in that: The device comprises: An acquisition module is used to acquire optical remote sensing images and thermal infrared images of the inland nuclear power plant area; A processing module, used for processing the acquired optical remote sensing images to obtain samples for training and testing the cooling tower detection network model; The segmentation module is used to perform brightness temperature inversion on the acquired thermal infrared images to obtain brightness temperature anomaly data; A detection module is used to input the optical remote sensing image to be detected into the constructed cooling tower detection model for detection, and obtain preliminary detection results of the cooling tower in working state and non-working state; A screening module is used to locate the corresponding position of the thermal infrared image according to the preliminary detection results of the cooling towers in working state, and then further screen out the cooling towers in working state according to the thermal infrared characteristics; The determination module is used to determine the operating state of the inland nuclear power plant according to whether the cooling tower is in working state.

11. A terminal device, comprising a processor and a memory, wherein a computer program is stored in the memory, wherein: When the processor runs the computer program, the steps of the method for monitoring the operating status of an inland nuclear power plant according to any one of claims 1-9 are implemented.

12. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed, the steps of the method for monitoring the operating status of an inland nuclear power plant according to any one of claims 1 to 9 are implemented.

Images