Effectiveness evaluation method and device for ultraviolet optical detection equipment in low-temperature environment
Through the compensation network and effectiveness evaluation network based on the pix2pix model, the problem of low image monitoring quality of ultraviolet optical sensors in low temperature environments is solved, and the effectiveness evaluation and screening of ultraviolet optical sensors is realized, which improves identification accuracy and screening efficiency.
Patent Information
- Application Number
- CN202411198453.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-29
- Publication Date
- 2024-12-27
AI Technical Summary
Under low temperature and icy weather conditions, the image monitoring quality of ultraviolet optical sensors is affected, and the lack of effective evaluation and screening methods leads to low recognition and poor imaging quality.
The compensation network based on the pix2pix model is adopted to process and integrate data through image compensation and effectiveness evaluation networks to ignore the impact of changes in extremely low temperature and large temperature difference, and improve evaluation accuracy and screening universality.
The effectiveness evaluation of ultraviolet optical sensors in low temperature environments is achieved, identification accuracy and screening efficiency are improved, and a reliable basis for subsequent data analysis and detection is provided.
Smart Images
Figure CN119206407A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of image processing, and in particular to a method and a device for evaluating the effectiveness of ultraviolet optical detection equipment in a low-temperature environment. Background Art
[0002] When line insulators, hardware and other equipment fail, they will discharge and radiate ultraviolet light at the surface fault site, and the discharge corona characteristic quantity of the fault can be effectively detected by the sensing equipment. The UAV system has the advantages of strong mobility, agility and safety, and can complete the detection task of corona discharge by carrying ultraviolet optical detection equipment. However, under severe weather or temperature conditions such as low temperature and ice and snow, the image of the ultraviolet optical sensor will be disturbed by snowflakes, and the recognition degree is low. Extremely low temperature or large temperature difference changes indirectly affect the final imaging quality of the ultraviolet optical sensor by affecting the performance of sensor materials and battery performance, etc., which brings a series of problems to the image monitoring of ultraviolet optical sensors. Therefore, evaluating the effectiveness of ultraviolet optical detection equipment and screening out ultraviolet optical detection equipment that works properly is a necessary technical means before applying ultraviolet optical sensors. However, there is currently a lack of evaluation and screening methods for the effectiveness of ultraviolet optical sensors under severe weather conditions such as ice and snow and extremely low temperatures or large temperature differences. Summary of the invention
[0003] The present invention provides a method and device for evaluating the effectiveness of ultraviolet optical detection equipment in a low-temperature environment. The compensation network based on the pix2pix model is used to compensate for the differences in ultraviolet images caused by different sensor types, severe weather conditions such as ice and snow, and extremely low temperatures or large temperature differences. The above-mentioned influence can be ignored when the compensated unified ultraviolet image is used for effectiveness evaluation. In contrast, the traditional method needs to adjust the evaluation parameters according to the sensor type and temperature range to complete the effectiveness evaluation. This method can effectively improve the evaluation accuracy and screening versatility in severe weather conditions such as ice and snow, and extremely low temperatures or large temperature differences, and can quickly and accurately screen out effective ultraviolet optical sensors, providing a basis for subsequent data analysis and processing, and providing a reference for further detection.
[0004] The present invention solves the above technical problems with the following solution: A method for evaluating the effectiveness of ultraviolet optical detection equipment in a low temperature environment comprises the following steps:
[0005] Acquire the ambient temperature of multiple ultraviolet optical sensors under working conditions, image data collected at corresponding temperatures, working parameters of corresponding ultraviolet optical sensors, and sensor effectiveness data after each operation, perform data preprocessing, and divide the processed data into a first data set and a second data set;
[0006] Construct image compensation network;
[0007] Using the first data set to train the image compensation network to obtain a trained image compensation network;
[0008] The second data set is input into the trained image compensation network to obtain the compensated image and the corresponding sensor device validity data, and then integrated into the third data set; specifically, the compensated image and the corresponding sensor device validity data are named with the same number (the image can be in jpg, png, mat and other formats, and the validity data can be in txt and other formats), and then all the compensated images and sensor device validity data are placed in the same folder as the third data set;
[0009] Constructing effectiveness evaluation network;
[0010] The effectiveness evaluation network is trained using the third data set to obtain a trained effectiveness evaluation network;
[0011] After the ambient temperature under the working state of the target ultraviolet optical sensor, the image data collected at the corresponding temperature, and the corresponding working parameters of the ultraviolet optical sensor are input into the trained image compensation network, the compensated image of the ultraviolet optical captured image is input into the trained effectiveness evaluation network to obtain the equipment effectiveness evaluation result of the target ultraviolet optical sensor.
[0012] Preferably, the data preprocessing includes image grayscale conversion and image size uniformity processing.
[0013] Preferably, the step of converting the image grayscale is specifically as follows: Gray =
[0014] 0.299*R+0.578*G+0.114*B
[0015] Among them: R is the red channel component of the image; G is the green channel component of the image; B is the blue channel component of the image; Gray is the grayscale image.
[0016] Preferably, the image size unification step is performed by bicubic interpolation to match the data input shape of the subsequent neural network model, as follows:
[0017]
[0018] Where: (x, y) represents the coordinates of the pixel to be interpolated, f(x, y) represents the value that should be inserted after calculation of the pixel to be interpolated, (x i ,y j ) represents the pixels in the 4x4 region around the pixel to be interpolated, i.e., i, j = 0, 1, 2, 3, and W(*) is a BiCubic function used to determine the influence of the surrounding pixels on the value to be interpolated at the pixel to be interpolated;
[0019] W(*) can take the following forms:
[0020]
[0021] Where a = -0.5, x is the input distance difference, that is, xx i or yy j .
[0022] Preferably, the image compensation network is constructed based on a pix2pix model, including a generator module and a discriminator module;
[0023] The generator module adopts a U-Net network structure, including: an encoding stage and a decoding stage; used to convert an input image into a feature map and then into a compensated image output; in the encoding stage, the size of the input image is gradually reduced, image features are extracted to form a feature map, and image features at various levels are provided for learning for image compensation; in the decoding stage, the image size is gradually restored, the feature map of the encoding stage is spliced with the feature map of the decoding stage, the feature map is converted into an image output, and the encoded features are converted into a compensated image and output; in the two stages, image compensation is performed as parameters are iteratively optimized, so that the compensated image finally output is closer to the ultraviolet image of a normal sensor at a standard temperature.
[0024] The discriminator module adopts the patchGAN network structure, including: convolution layer, batch normalization layer and sigmoid activation function, which are used to eliminate the noise and blur of the input image. Convolution layer, batch normalization layer and sigmoid activation function, the convolution layer extracts the features of the input image for neural network learning; the features are sent to the normalization layer to eliminate the errors caused by the non-uniform range of image feature values, and finally the activation function introduces nonlinearity so that the neural network can better fit the target image. Through multi-layer superposition, the output image of the neural network can be continuously close to the output image of the standard temperature normal sensor during the iteration of network parameters.
[0025] Preferably, the first data set is used to train the image compensation network to obtain a trained image compensation network, which specifically includes:
[0026] The joint loss function is used as the training loss function of the image compensation network, as follows:
[0027]
[0028] Where, L cGAN represents the loss function of the conditional generative adversarial network, L1 represents the L1 distance loss function, and λ is a weight parameter;
[0029] The first data set is randomly divided into k parts, k-1 parts are used as training sets, and 1 part is used as a validation set. The training set and the validation set are rotated k times in the image compensation network model, and the joint loss function is used as the loss function of the compensation network to participate in training and verification, and a trained image compensation network is obtained.
[0030] Preferably, the L Cgan The loss function of the conditional generative adversarial network is expressed as:
[0031] L cGAN (G, D) = E x，y [logD(x,y)]+E x，z [log(1-D(x,G(x,z)))]
[0032] In the formula, G represents the generator, D represents the discriminator, x is the input UV image, y is the standard UV image, z is the random noise vector, and E x，y represents the expectation of the real data distribution, E x，z represents the expectation of the generated data distribution, G(x, z) represents the fake samples generated by the generator G, and D(x, y) represents the probability estimate of the similarity between the inputs x and y by the discriminator D;
[0033] The L1 is the distance loss function, expressed as:
[0034]
[0035] In the formula, G represents the generator, D represents the discriminator, x is the input ultraviolet image, and y is the standard ultraviolet image.
[0036] Preferably, the validity evaluation network based on the ResNet model is constructed based on identity block units and convolution block units. Specifically, different numbers of identity block units and convolution block units are connected to form different residual blocks, and the stacking of different residual blocks forms the validity evaluation network of the ResNet model.
[0037] The compensated image data is extracted through the multi-layer convolutional layers in ResNet to form a feature map. The feature map can be converted into a 1xN value output after the final fully connected layer and other structures, where each value represents the probability that the image belongs to the corresponding category.
[0038] Among them, the ath value represents the probability that the input image data belongs to the ath category (between 0 and 1). For example, N can be set to 2, that is, 0 represents an invalid sensor and 1 represents a valid sensor. Then the output of the evaluation network is a 1x2 matrix. The number of N needs to be the same as the sensor validity label category in the third training set and have the same meaning. The residual block in the neural network is used to prevent the problem of being unable to train due to gradient explosion or disappearance, and the use of residual blocks allows ResNet to have many layers. During the training process, the compensated image data in the third data set is used to input the evaluation network to obtain a 1xN output. Then, the sensor validity label in the third training set is compared with the class with the highest probability in the output. If they are not the same class, it means that the model prediction is wrong, otherwise it is correct. Finally, the loss function is calculated based on the above correct or incorrect results to update the parameters of the neural network.
[0039] Further preferably, the ResNet model specifically adopts the ResNet-101 network and introduces a residual block.
[0040] The ResNet-101 deep residual network is composed of conventional neural network modules such as an input layer, a convolutional layer, a residual block, a pooling layer, a fully connected layer, and a softmax layer connected in sequence. Its main feature is the introduction of residual connections, that is, the output of a shallower neural network layer is used as the input of a deeper neural network layer, thereby solving the gradient vanishing problem and allowing more neural network layers to be stacked.
[0041] Preferably, the effectiveness evaluation network is trained using the third data set to obtain a effectiveness evaluation network, as follows:
[0042] The cross entropy loss function (the conventional loss function for classification evaluation tasks) is used as the cost function to train the ResNet-101 deep residual network model, specifically:
[0043]
[0044] In the formula, i represents the i-th picture, g i Represents the real UV discharge image; s i represents the predicted UV discharge image output of the network, k is the number of classes in the UV discharge image dataset; N represents the spatial resolution of the specific UV discharge image output;
[0045] The third data set is randomly divided into k parts, k-1 parts are used as training sets, and 1 part is used as a validation set. The training set and the validation set are rotated k times in the effectiveness evaluation network. The above cross entropy loss function is used as the loss function of the effectiveness evaluation network to participate in training and verification, and a trained effectiveness evaluation network is obtained.
[0046] Preferably, after inputting the ambient temperature of the target ultraviolet optical sensor under the working state and the image data collected at the corresponding temperature and the corresponding working parameters of the ultraviolet optical sensor into the trained image compensation network, the obtained compensated image of the ultraviolet optical captured image is input into the trained effectiveness evaluation network to obtain the equipment effectiveness evaluation result of the target ultraviolet optical sensor, which specifically includes:
[0047] Preprocessing the ambient temperature of the target ultraviolet optical sensor under the working state, the image data collected by the corresponding temperature, and the corresponding working parameters of the ultraviolet optical sensor to obtain the data to be measured;
[0048] After the data to be tested is input into the trained image compensation network, a compensated image of the image photographed by ultraviolet optics is obtained;
[0049] After the compensation image of the ultraviolet optical image is defogged, a defogged image is obtained;
[0050] The dehazed image is input into the effectiveness evaluation network to obtain the effectiveness evaluation result of the target UV optical sensor.
[0051] A device for evaluating the effectiveness of ultraviolet optical detection equipment in a low-temperature environment, comprising:
[0052] The first data processing module is used to obtain the ambient temperature of the plurality of ultraviolet optical sensors under working conditions and the image data collected at the corresponding temperature, the working parameters of the corresponding ultraviolet optical sensors and the validity data after each operation; and perform data preprocessing to divide the processed data into a first data set and a second data set;
[0053] Network construction module, used to construct image compensation network and construct and effectiveness evaluation network;
[0054] A network training module, used to train the image compensation network using the first data set to obtain a trained image compensation network, and to train the effectiveness evaluation network using the third data set to obtain a trained effectiveness evaluation network;
[0055] A second data processing module is used to input the second data set into the trained image compensation network to obtain a compensated image of the ultraviolet optical image and corresponding sensor device effectiveness data, and synthesize a third data set;
[0056] The target evaluation module is used to input the ambient temperature of the target ultraviolet optical sensor under the working state, the image data collected at the corresponding temperature, and the corresponding working parameters of the ultraviolet optical sensor into the trained image compensation network, and then input the compensated image of the ultraviolet optical image obtained into the trained effectiveness evaluation network to obtain the equipment effectiveness evaluation result of the target ultraviolet optical sensor.
[0057] A computer storage medium, wherein the computer readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the effectiveness evaluation method of the low-temperature environment ultraviolet optical detection equipment as described above are implemented.
[0058] An electronic device comprises a memory and a processor: the memory is used to store computer executable instructions, the processor is used to execute the computer executable instructions, and the computer executable instructions, when executed by the processor, implement the steps of the effectiveness evaluation method of the ultraviolet optical detection equipment in a low-temperature environment as described above.
[0059] The beneficial effects of the present invention are:
[0060] (1) Using pix2pix as the compensation network, when temperature directly or indirectly affects imaging, has better robustness and generalization ability than the compensation formula derived based on the theoretical model, and the neural network can perform real-time dynamic compensation;
[0061] (2) The introduction of residual connections allows the gradient to be propagated back to the shallow layers more smoothly, alleviating the gradient vanishing problem and facilitating the training of deeper networks;
[0062] (3) Building a deep model that can effectively extract image features and thus improve the accuracy of model evaluation;
[0063] (4) Through residual connections, the network can directly pass the features of the previous layer to the next layer, making the network more efficient, reducing the risk of overfitting, and reducing the number of parameters and computational complexity in the network.
[0064] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention and implement it according to the contents of the specification, the following is a detailed description of the preferred embodiments of the present invention in conjunction with the accompanying drawings. The specific implementation of the present invention is given in detail by the following embodiments and their accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0065] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:
[0066] Figure 1 This is a flow chart of a method for evaluating the effectiveness of a low-temperature environment ultraviolet optical detection device in Example 1;
[0067] Figure 2 It is a module diagram of the effectiveness evaluation network based on the ResNet model in Example 1;
[0068] Figure 3This is a module diagram of an effectiveness evaluation device for a low-temperature environment ultraviolet optical detection device in Example 2. DETAILED DESCRIPTION
[0069] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.
[0070] like Figure 1 As shown, the present invention provides a method for evaluating the effectiveness of ultraviolet optical detection equipment in a low-temperature environment, comprising the following steps:
[0071] S1, obtaining the ambient temperature of multiple ultraviolet optical sensors under working conditions and image data collected at corresponding temperatures, the corresponding working parameters of the ultraviolet optical sensors and the sensor effectiveness data after each operation, and performing data preprocessing to divide the processed data into a first data set and a second data set; specifically including:
[0072] S11, obtaining the ambient temperature of multiple ultraviolet optical sensors under working conditions and image data collected at corresponding temperatures, working parameters of the corresponding ultraviolet optical sensors, and effectiveness data after each operation;
[0073] S12. Preprocess the acquired data, grayscale the image, and convert the color image into a grayscale image, so that the amount of calculation for subsequent processing is reduced to one third of the original amount. Specifically, Gray = 0.299*R + 0.578*G + 0.114*B
[0074] Among them: R is the red channel component of the image; G is the green channel component of the image; B is the blue channel component of the image; Gray is the grayscale image;
[0075] Then use the bicubic interpolation method to adjust the image size to match the data input shape of the subsequent neural network model, specifically:
[0076]
[0077] Where: (x, y) represents the coordinates of the pixel to be interpolated, f(x, y) represents the value that should be inserted after calculation of the pixel to be interpolated, (x i ,y j ) represents the pixels in the 4x4 region around the pixel to be interpolated, i.e., i, j = 0, 1, 2, 3, and W(*) is a BiCubic function used to determine the influence of the surrounding pixels on the value to be interpolated at the pixel to be interpolated;
[0078] For example, W(*) can take the following forms:
[0079]
[0080] Where a = -0.5, x is the input distance difference, that is, xx i or yy j .
[0081] S13. Divide the preprocessed data into a first data set and a second data set.
[0082] S2, build an image compensation network based on the pix2pix model;
[0083] The image compensation network includes a generator module and a discriminator module;
[0084] The generator module adopts a U-Net network structure, including: an encoding stage and a decoding stage. In the encoding stage, the size of the input image is gradually reduced, and image features are extracted to form a feature map, so as to provide image features at various levels for learning for image compensation; in the decoding stage, the image size is gradually restored, the feature map of the encoding stage is spliced with the feature map of the decoding stage, the feature map is converted into an image output, and the encoded features are converted into a compensated image and output; in the two stages, image compensation is performed with the iterative optimization of parameters, so that the compensated image finally output is closer to the ultraviolet image of a normal sensor at standard temperature.
[0085] The discriminator module adopts the patchGAN network structure, including: convolution layer, batch normalization layer and sigmoid activation function, which are used to eliminate the noise and blur of the input image. Convolution layer, batch normalization layer and sigmoid activation function, the convolution layer extracts the features of the input image for neural network learning; the features are sent to the normalization layer to eliminate the errors caused by the non-uniform range of image feature values, and finally the activation function introduces nonlinearity so that the neural network can better fit the target image. Through multi-layer superposition, the output image of the neural network can be continuously close to the output image of the standard temperature normal sensor during the iteration of network parameters.
[0086] S3, using the first data set to train the image compensation network to obtain a trained image compensation network, specifically including:
[0087] S31. The joint loss function is used as the training loss function of the image compensation network. The expression is as follows:
[0088]
[0089] Where, L cGAN represents the loss function of the conditional generative adversarial network, L1 represents the L1 distance loss function, and λ is a weight parameter;
[0090] The L Cgan The loss function of the conditional generative adversarial network is expressed as:
[0091] -L cGAN (G, D) = E x，y [logD(x,y)]+E x，z [log(1-D(x,G(x,z)))]
[0092] In the formula, G represents the generator, D represents the discriminator, x is the input UV image, y is the standard UV image, z is the random noise vector, and E x，y represents the expectation of the real data distribution, E x，z represents the expectation of the generated data distribution, G(x, z) represents the fake samples generated by the generator G, and D(x, y) represents the probability estimate of the similarity between the inputs x and y by the discriminator D;
[0093] The L1 is the distance loss function, expressed as:
[0094]
[0095] In the formula, G represents the generator, D represents the discriminator, x is the input ultraviolet image, and y is the standard ultraviolet image.
[0096] S32. Randomly divide the first data set into k parts, k-1 parts as training sets, and 1 part as verification set. In the image compensation network model, the training set and the verification set are rotated k times in sequence. The above-mentioned joint loss function is used as the loss function of the compensation network to participate in training and verification, and a trained image compensation network is obtained.
[0097] S4. Input the second data set into the trained image compensation network to obtain the compensated image and the corresponding sensor device validity data, and integrate them into a third data set; specifically, the compensated image and the corresponding sensor device validity data are named with the same number (the image can be in jpg, png, mat and other formats, and the validity data can be in txt and other formats), and then all the compensated images and sensor device validity data are placed in the same folder as the third data set.
[0098] S5. Build a validity evaluation network based on the ResNet model;
[0099] like Figure 2 As shown, the ResNet module is built based on the BN structure, including IB (Identity Block) units and CB (Conv Block) units; IB units and CB units are connected in different numbers to form different residual blocks, and the stacking of different residual blocks forms the ResNet module.
[0100] In this embodiment, the ResNet-101 network is used. The ResNet-101 deep residual network is composed of conventional neural network modules such as an input layer, a convolutional layer, a residual block, a pooling layer, a fully connected layer, and a softmax layer connected in sequence. Its main feature is the introduction of residual connection, that is, the output of a shallower neural network layer is used as the input of a deeper neural network layer, thereby solving the gradient vanishing problem and allowing the neural network to have more stacked layers.
[0101] The compensated image data is extracted through the multi-layer convolution layer in ResNet to form a feature map. The feature map can be converted into a 1xN value output after the final fully connected layer and other structures, where each value represents the probability that the image belongs to the corresponding category, where the ath value represents the probability that the input image data belongs to the ath category (between 0 and 1). For example, N can be set to 2, that is, 0 represents that the sensor is invalid, and 1 represents that the sensor is valid. Then the output of the evaluation network is a 1x2 matrix, and the number of N needs to be the same as the sensor validity label category in the third training set and have the same meaning. The residual block in the neural network is used to prevent the problem of being unable to train due to gradient explosion or disappearance, and the use of residual blocks allows ResNet to have many layers. During the training process, the compensated image data in the third data set is used to input the evaluation network to obtain a 1xN output. Then, the sensor validity label in the third training set is compared with the category with the highest probability in the output. If they are not the same category, it means that the model prediction is wrong, otherwise it is correct. Finally, the loss function is calculated based on the above correct or incorrect results to update the parameters of the neural network.
[0102] S6. Use the third data set to evaluate the effectiveness of the network, and obtain a trained effectiveness evaluation network, as follows:
[0103] S61. Use the cross entropy loss function (the conventional loss function for classification evaluation tasks) as the cost function to train the ResNet-101 deep residual network model, specifically:
[0104]
[0105] In the formula, i represents the i-th picture, g i Represents the real UV discharge image; s i represents the predicted UV discharge image output of the network, k is the number of classes in the UV discharge image dataset; N represents the spatial resolution of the specific UV discharge image output;
[0106] S62. Randomly divide the third data set into k parts, with k-1 parts as training sets and 1 part as validation sets. In the effectiveness evaluation network, the training set and the validation set are rotated k times in turn. The above cross entropy loss function is used as the loss function of the effectiveness evaluation network to participate in training and validation, and a trained effectiveness evaluation network is obtained.
[0107] S7, after inputting the ambient temperature of the target ultraviolet optical sensor under the working state and the image data collected at the corresponding temperature and the corresponding working parameters of the ultraviolet optical sensor into the trained image compensation network, the compensated image of the ultraviolet optical image obtained is input into the trained effectiveness evaluation network to obtain the equipment effectiveness evaluation result of the target ultraviolet optical sensor, which is as follows:
[0108] S71, pre-processing the ambient temperature of the target ultraviolet optical sensor under the working state, the image data collected by the corresponding temperature, and the corresponding working parameters of the ultraviolet optical sensor to obtain the data to be measured;
[0109] S72, after inputting the data to be measured into the trained image compensation network, a compensated image of the image photographed by ultraviolet optics is obtained;
[0110] S73, performing defogging processing on the compensated image of the ultraviolet optical image to obtain a defogging image;
[0111] S74. The dehazed image is input into the effectiveness evaluation network to obtain the effectiveness evaluation result of the target ultraviolet optical sensor.
[0112] Example 2
[0113] A device for evaluating the effectiveness of ultraviolet optical detection equipment in a low-temperature environment, comprising:
[0114] The first data processing module is used to obtain the ambient temperature of the plurality of ultraviolet optical sensors under working conditions and the image data collected at the corresponding temperature, the working parameters of the corresponding ultraviolet optical sensors and the sensor effectiveness data after each operation; and perform data preprocessing to divide the processed data into a first data set and a second data set;
[0115] Network construction module, used to construct image compensation network and construct and effectiveness evaluation network;
[0116] A network training module, used to train the image compensation network using the first data set to obtain a trained image compensation network, and to train the effectiveness evaluation network using the third data set to obtain a trained effectiveness evaluation network;
[0117] A second data processing module is used to input the second data set into the trained image compensation network to obtain a compensated image of the ultraviolet optical image and corresponding sensor device effectiveness data, and synthesize a third data set;
[0118] The target evaluation module is used to input the ambient temperature of the target ultraviolet optical sensor under the working state, the image data collected at the corresponding temperature, and the corresponding working parameters of the ultraviolet optical sensor into the trained image compensation network, and then input the compensated image of the ultraviolet optical image obtained into the trained effectiveness evaluation network to obtain the equipment effectiveness evaluation result of the target ultraviolet optical sensor.
[0119] Example 3
[0120] A computer storage medium, wherein the computer readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for evaluating the effectiveness of a low-temperature environment ultraviolet optical detection device as described in Example 1 are implemented.
[0121] Example 4
[0122] An electronic device comprises a memory and a processor: the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions. When the computer executable instructions are executed by the processor, the steps of the effectiveness evaluation method of low-temperature environment ultraviolet optical detection equipment as described in Example 1 are implemented.
[0123] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any ordinary technician in the industry can smoothly implement the present invention as shown in the drawings and described above. However, any equivalent changes, modifications and evolutions made by technicians familiar with the profession without departing from the scope of the technical solution of the present invention using the technical content disclosed above are all equivalent embodiments of the present invention. At the same time, any equivalent changes, modifications and evolutions made to the above embodiments based on the essential technology of the present invention are still within the protection scope of the technical solution of the present invention.
[0124] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0125] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0126] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0127] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
Claims
1. A method for evaluating the effectiveness of ultraviolet optical detection equipment in a low-temperature environment, characterized in that: The following steps are involved: Acquire the ambient temperature of multiple ultraviolet optical sensors under working conditions, image data collected at corresponding temperatures, working parameters of corresponding ultraviolet optical sensors, and sensor effectiveness data after each operation, perform data preprocessing, and divide the processed data into a first data set and a second data set; Construct image compensation network; Using the first data set to train the image compensation network to obtain a trained image compensation network; Inputting the second data set into the trained image compensation network to obtain a compensated image of the ultraviolet optical image and corresponding sensor device effectiveness data, and synthesizing a third data set; Constructing effectiveness evaluation network; The effectiveness evaluation network is trained using the third data set to obtain a trained effectiveness evaluation network; After the ambient temperature under the working state of the target ultraviolet optical sensor, the image data collected at the corresponding temperature, and the corresponding working parameters of the ultraviolet optical sensor are input into the trained image compensation network, the compensated image of the ultraviolet optical captured image is input into the trained effectiveness evaluation network to obtain the equipment effectiveness evaluation result of the target ultraviolet optical sensor.
2. The effectiveness evaluation method of a low-temperature environment ultraviolet optical detection device according to claim 1 is characterized in that: The data preprocessing includes image grayscale conversion and image size uniformity processing.
3. The method for evaluating the effectiveness of a low-temperature environment ultraviolet optical detection device according to claim 1, characterized in that: The image compensation network is constructed based on the pix2pix model, including a generator module and a discriminator module; The generator module adopts a U-Net network structure to convert the input image into a feature map and then into a compensated image output; The discriminator module adopts a patchGAN network structure to eliminate noise and blur of the input image.
4. The method for evaluating the effectiveness of a low-temperature environment ultraviolet optical detection device according to claim 3, characterized in that: The first data set is used to train the image compensation network to obtain a trained image compensation network, which specifically includes: The joint loss function is set up as the training loss function of the image compensation network as follows: Where, L cGAN represents the loss function of the conditional generative adversarial network, L1 represents the L1 distance loss function, λ is a weight parameter, G represents the generator module, and D represents the discriminator module; The first data set is randomly divided into k parts, k-1 parts are used as training sets, and 1 part is used as a validation set. The training set and the validation set are rotated k times in the image compensation network. The joint loss function is used as the loss function of the image compensation network to participate in training and validation, and a trained image compensation network is obtained.
5. The method for evaluating the effectiveness of a low-temperature environment ultraviolet optical detection device according to claim 4, characterized in that: The L Cgan The loss function of the conditional generative adversarial network is expressed as: L cGAN (G,D)=E x，y [logD(x,y)]+E x，z [log(1-D(x,G(x,z)))] In the formula, G represents the generator module, D represents the discriminator module, x is the input UV image, y is the standard UV image, z is the random noise vector, and E x,y represents the expectation of the real data distribution, E x,z represents the expectation of the generated data distribution, G(x,z) represents the fake samples generated by the generator G, and D(x,y) represents the probability estimate of the similarity between the inputs x and y by the discriminator D; The L1 is the distance loss function, expressed as: In the formula, G represents the generator module, D represents the discriminator module, x is the input ultraviolet image, and y is the standard ultraviolet image.
6. According to the method for evaluating the effectiveness of a low-temperature environment ultraviolet optical detection device in claim 1, the effectiveness evaluation network is trained using the third data set to obtain a trained effectiveness evaluation network, specifically as follows: Set the cross entropy loss function as the loss function, the expression is as follows: In the formula, i represents the i-th picture, g i Represents the real UV discharge image; s i represents the predicted UV discharge image output of the network, k is the number of classes in the UV discharge image dataset; N represents the spatial resolution of the specific UV discharge image output; The third data set is randomly divided into k parts, k-1 parts are used as training sets, and 1 part is used as a validation set. The training set and the validation set are rotated k times in the effectiveness evaluation network. The cross entropy loss function is used as the loss function of the effectiveness evaluation network to participate in training and verification, and a trained effectiveness evaluation network is obtained.
7. The method for evaluating the effectiveness of a low-temperature environment ultraviolet optical detection device according to claim 6, characterized in that: After inputting the ambient temperature of the target ultraviolet optical sensor under the working state, the image data collected at the corresponding temperature, and the corresponding working parameters of the ultraviolet optical sensor into the trained image compensation network, the compensated image of the ultraviolet optical image obtained is input into the trained effectiveness evaluation network to obtain the equipment effectiveness evaluation result of the target ultraviolet optical sensor, which specifically includes: Preprocessing the ambient temperature of the target ultraviolet optical sensor under the working state, the image data collected by the corresponding temperature, and the corresponding working parameters of the ultraviolet optical sensor to obtain the data to be measured; After the data to be tested is input into the trained image compensation network, a compensated image of the image photographed by ultraviolet optics is obtained; After the compensation image of the ultraviolet optical image is defogged, a defogged image is obtained; The dehazed image is input into the effectiveness evaluation network to obtain the effectiveness evaluation result of the target UV optical sensor.
8. A device for evaluating the effectiveness of ultraviolet optical detection equipment in a low temperature environment, characterized in that: include: A first data processing module is used to obtain the ambient temperature of multiple ultraviolet optical sensors under working conditions and image data collected at corresponding temperatures, working parameters of the corresponding ultraviolet optical sensors, and sensor effectiveness data after each operation; and performing data preprocessing to divide the processed data into a first data set and a second data set; Network construction module, used to construct image compensation network and construct and effectiveness evaluation network; A network training module, used to train the image compensation network using the first data set to obtain a trained image compensation network, and to train the effectiveness evaluation network using the third data set to obtain a trained effectiveness evaluation network; A second data processing module is used to input the second data set into the trained image compensation network to obtain a compensated image of the ultraviolet optical image and corresponding sensor device effectiveness data, and synthesize a third data set; The target evaluation module is used to input the ambient temperature of the target ultraviolet optical sensor under the working state, the image data collected at the corresponding temperature, and the corresponding working parameters of the ultraviolet optical sensor into the trained image compensation network, and then input the compensated image of the ultraviolet optical image obtained into the trained effectiveness evaluation network to obtain the equipment effectiveness evaluation result of the target ultraviolet optical sensor.
9. A computer storage medium, wherein the computer readable storage medium stores a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method for evaluating the effectiveness of a low-temperature environment ultraviolet optical detection device as described in any one of claims 1 to 7 are implemented.
10. An electronic device, characterized in that: It comprises a memory and a processor: the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions. When the computer executable instructions are executed by the processor, the steps of the effectiveness evaluation method of the low-temperature environment ultraviolet optical detection equipment as described in any one of claims 1 to 7 are implemented.