River pollution identification and early warning method based on computer vision

Through the river pollution identification and early warning method based on computer vision, the improved ResNet-18 model and high-definition camera are used to achieve efficient, economical and real-time monitoring of river water quality, and solve the high cost and poor timeliness of river pollution monitoring in the existing technology. The problem can detect pollution incidents as soon as possible.

CN119091290AInactive Publication Date: 2024-12-06CHINA YANGTZE POWER
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202411024779.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-29
Publication Date
2024-12-06
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing river pollution monitoring methods rely on manual sampling and sensor real-time monitoring, which is costly, poor timeliness, and it is difficult to fully cover water bodies and effectively identify complex pollutants.

Method used

Using a river pollution identification and early warning method based on computer vision, the river water quality image data set is used to build an improved ResNet-18 model based on attention mechanism, and automatically identify and monitor river water quality images, and use a high-definition camera to take real-time shots. Combining image processing and machine learning algorithms, automatic identification and early warning of pollution conditions is achieved.

Benefits of technology

It realizes efficient, economical and real-time monitoring of river water quality, can effectively identify and early warning of pollution incidents, provides strong technical support for river water quality management, and has the advantages of non-contact, monitoring the entire river channel, and is low-cost.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119091290A_ABST
    Figure CN119091290A_ABST
Patent Text Reader

Abstract

The invention discloses a river pollution identification and early warning method based on computer vision. The method comprises the following steps: step 1, making a river water quality image data set; step 2, constructing an improved ResNet-18 model based on an attention mechanism; 3, constructing a training set, a verification set and a test set; 4, training the constructed improved ResNet-18 model by using the training set, adjusting hyper-parameters of the model by using the verification set, and inputting the test set into the trained deep learning network for performance evaluation; 5, the model is stored and deployed, the trained improved ResNet-18 model is used for river water quality pollution detection, and when the read river image is judged as polluted water quality, an alarm signal is triggered; according to the invention, real-time monitoring and early warning of river pollution are realized; the system has the advantages of non-contact, capability of monitoring the whole riverway, low cost and the like, and can discover illegal drainage in the first time.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of river pollution identification and early warning, and in particular to a river pollution identification and early warning method based on computer vision. Background Art

[0002] Currently, river pollution monitoring mainly relies on two methods: manual sampling and real-time monitoring by sensors. The manual sampling method requires professionals to regularly collect water samples on site and then bring them back to the laboratory for analysis. Although this method can obtain accurate data, it is costly and time-sensitive. It is also limited by the number and distribution of sampling points and cannot fully cover the entire water body, which may miss important pollution sources. Although the real-time monitoring method by sensors can reflect changes in river water quality in real time, the deployment and maintenance costs of sensors are high, and it is difficult to effectively identify some complex pollutants.

[0003] With the rapid development of computer vision technology, it has shown strong capabilities in image processing, target recognition, pattern recognition and other fields. Computer vision technology can extract useful information through the analysis of image or video data, and realize automatic recognition and monitoring of targets. Therefore, the application of computer vision technology in the field of river pollution monitoring has great potential and advantages, and is a technical problem that needs to be solved urgently. Summary of the invention

[0004] The purpose of the present invention is to overcome the above-mentioned shortcomings and provide a river pollution identification and early warning method based on computer vision to solve the problems raised in the background technology.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is: a river pollution identification and early warning method based on computer vision, comprising the following steps:

[0006] Step 1, create a river water quality image dataset;

[0007] Step 2: Build an improved ResNet-18 model based on the attention mechanism;

[0008] Step 3: Construct a training set, a validation set, and a test set; the training set is used to train the neural network, the validation set is used to adjust the model hyperparameters, and the test set is used to test the model performance;

[0009] Step 4: Use the training set to train the constructed improved ResNet-18 model, use the validation set to adjust the model's hyperparameters, and then input the test set into the trained deep learning network for performance evaluation;

[0010] Step 5: Save the model and deploy it. Use the trained improved ResNet-18 model to detect river water pollution. When the read river image is identified as polluted water, an alarm signal is triggered.

[0011] Furthermore, the step 1 specifically includes the following steps:

[0012] Step 1.1, collecting video surveillance data of the observation point: installing a black light camera on the river bank of the observation point; then, selecting multiple video clips in four scenes of sunny day, rainy day, foggy day and snowy day in the database of the monitoring server under two scenarios of day and night;

[0013] Step 1.2, video frame processing: use the VideoCapture function in the OpenCV extension library cv2 to perform frame processing on the collected river monitoring video clips of the observation point to obtain the river water quality image of the observation point;

[0014] Step 1.3, collect river pollution images: Since there are few water quality images of river pollution at the observation point, publicly available river pollution water quality images are collected through online channels;

[0015] Step 1.4, image preprocessing: For river pollution water quality images: use the image cropping tool to crop out the polluted area in the image, and use the Padding function to adjust the cropped image to a square size, and then scale it proportionally to a 224 pixel × 224 pixel image; For river water quality images at observation points: use the function to automatically crop out a 224 pixel × 224 pixel image in the center of the image.

[0016] Furthermore, the step 2 specifically includes the following steps:

[0017] Step 2.1, build the classic ResNet-18 model, including the initial convolutional layer, residual block, transition layer, and fully connected layer;

[0018] Step 2.2, embed the hybrid attention mechanism in the residual block of ResNet-18.

[0019] Furthermore, the classic ResNet-18 model built in step 2.1 is as follows:

[0020] After the input layer is the initial convolution layer, which is used to extract the initial features of the image; after the initial layer is the stacked residual blocks, each residual block contains two 3×3 convolution layers, and each convolution layer is followed by a batch normalization layer and a ReLU activation function; secondly, the residual connection enables the network to learn feature representations other than identity mapping by adding the input directly to the output of the residual block; in order to change the size and number of channels of the feature map, a transition layer is inserted between the stacked residual blocks. The transition layer contains a 1×1 convolution layer to increase the number of channels, and a 3×3 convolution layer or average pooling layer with a stride of 2 to reduce the size of the feature map; at the end of the model there is a fully connected layer to map the features extracted by the convolution layer to the output space.

[0021] Furthermore, the step 2.2 is specifically as follows:

[0022] The hybrid attention mechanism consists of a channel attention module M C and the spatial attention module M S Composition, focusing on important information from the two dimensions of features and space; by C and the spatial attention module M S The hybrid attention mechanism is embedded in the residual block of ResNet-18 by replacing the two convolutional layers in the residual block of ResNet-18 respectively.

[0023] Furthermore, the structure after embedding the hybrid attention mechanism in the residual block of ResNet-18 is as follows:

[0024] use Represents the input of the structure, where H represents the height of the input feature map, W represents the width of the input feature map, and C represents the number of channels of the input feature map;

[0025] Use F as the channel attention module M C The input of the channel attention module is obtained, and then the output M C (F), and then multiply it by the input F to get the channel attention map F'; F' then serves as the spatial attention module M S The input of the spatial attention module is then used to obtain the output M S (F′), and then multiply it by F' to get the output F″ of the hybrid attention mechanism; the function is as follows:

[0026]

[0027] in, Indicates the corresponding element multiplication operation;

[0028]

[0029] Where σ represents the sigmoid activation function, MLP(·) represents the multi-layer perceptron network, AvgPool(·) represents the average pooling operation, MaxPool(·) represents the maximum pooling operation, W0 and W1 represent the weights from the input layer to the hidden layer and from the hidden layer to the output layer of the multi-layer perceptron, respectively. represents the output of average pooling, represents the output of the maximum pooling;

[0030]

[0031] Among them, f 7×7 (·) represents a convolution operation with a kernel size of 7×7. represents the output of average pooling, represents the output after maximum pooling, It means that the feature map obtained by dimensionally concatenating the outputs of average pooling and maximum pooling is used as the input of the convolution.

[0032] Furthermore, the construction of the training set, validation set, and test set described in step 3 is as follows:

[0033] For the collected river images of M observation points and N water pollution images, the two types of image samples are randomly divided into training set, validation set and test set in a ratio of 7:2:1;

[0034] Then the number of training set samples is: 0.7×M+0.7×N;

[0035] The number of samples in the validation set is: 0.2×M+0.2×N;

[0036] The number of test set samples is: 0.1×M+0.1×N.

[0037] Furthermore, the improved ResNet-18 model constructed by training in step 4 includes the following steps:

[0038] Step 4.1, set the hyperparameters required for training, including epoch, batch_size, learning rate, training optimizer, and loss function; the loss function of the improved ResNet-18 model constructed uses the cross entropy loss function;

[0039] Step 4.2, set the model evaluation index: the evaluation index uses four evaluation indexes used for classification tasks, namely accuracy, precision, recall and F1-score; among them, the polluted water quality image is regarded as a positive sample, and the unpolluted water quality image is regarded as a negative sample, then:

[0040]

[0041] Among them, TP means that the positive sample is successfully predicted as a positive sample, TN means that the negative sample is successfully predicted as a negative sample, FP means that the negative sample is incorrectly predicted as a positive sample, and FN means that the positive sample is incorrectly predicted as a negative sample;

[0042] Step 4.3, initialize the network parameters, and input the training set into the constructed improved ResNet-18 model to start the training process. According to the calculated loss, update the model parameters through the gradient descent algorithm until the training ends. Then freeze the model parameters, input the validation set into the trained improved ResNet-18 model, optimize and adjust the model through the evaluation indicators, and finally, input the test set into the optimized improved ResNet-18 model to evaluate the performance of the model.

[0043] Furthermore, the step 5 comprises the following steps:

[0044] Step 5.1, save the model with the best evaluation index during the verification process, the model includes the model architecture and parameters;

[0045] Step 5.2, in the river monitoring system, write a real-time video data reading script to achieve uninterrupted reading of video data and pre-process the video data;

[0046] Step 5.3, call the saved improved ResNet-18 model to monitor river water pollution and return the current water quality status.

[0047] The beneficial effects of the present invention are as follows:

[0048] The method of the present invention utilizes a high-definition camera to take real-time photos of the river, and analyzes and processes the collected images through image processing technology and machine learning algorithms to achieve automatic identification and early warning of river pollution conditions; compared with traditional methods, the method of the present invention can achieve efficient, economical, and real-time monitoring of river water quality, effectively identify and warn of pollution incidents, and provide strong technical support for river water quality management and protection.

[0049] 2. The present invention designs an improved ResNet-18 model based on the attention mechanism to realize water pollution detection.

[0050] 3. The method of the present invention can realize real-time monitoring and early warning of river pollution with the help of river monitoring system and computer vision technology; it has the advantages of non-contact, monitoring the entire river, low cost, etc., and can detect illegal discharge in the first time. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a flow chart of a river pollution identification and early warning method based on computer vision;

[0052] Figure 2 Schematic diagram of the improved residual module. DETAILED DESCRIPTION

[0053] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0054] like Figure 1 As shown, a river pollution identification and early warning method based on computer vision includes the following steps:

[0055] Step 1: Create a river water quality image dataset, including the following steps:

[0056] Step 1.1, collect video surveillance data of the observation point: install a black light camera on the river bank of the observation point. Then, select multiple video clips in four scenes of sunny, rainy, foggy and snowy days in both day and night scenarios from the database of the monitoring server.

[0057] Step 1.2, video frame processing: Use the VideoCapture function in the OpenCV extension library cv2 to perform frame processing on the collected observation point river monitoring video clips to obtain the observation point river water quality image.

[0058] Step 1.3, collect river pollution images: Given that there are few water quality images of river pollution at the observation point, publicly available river pollution water quality images are collected through online channels.

[0059] Step 1.4, image preprocessing: For river pollution water quality images: use the image cropping tool to crop out the polluted area in the image, and use the Padding function to adjust the cropped image to a square size, and then scale it proportionally to a 224 pixel × 224 pixel image; For river water quality images at observation points: use the function to automatically crop out a 224 pixel × 224 pixel image in the center of the image.

[0060] Step 2: Build an improved ResNet-18 model based on the attention mechanism (such as Figure 2 As shown), comprising the following steps:

[0061] Step 2.1, build the classic ResNet-18 model, including the initial convolutional layer, residual block, transition layer, and fully connected layer.

[0062] Step 2.2, embed the hybrid attention mechanism in the residual block of ResNet-18.

[0063] Preferably, the classic ResNet-18 model is constructed as described in step 2.1, specifically as follows: after the input layer is an initial convolution layer for extracting the initial features of the image; after the initial layer is a stacked residual block, each residual block contains two 3×3 convolution layers, and each convolution layer is followed by a batch normalization layer and a ReLU activation function, and the residual connection adds the input directly to the output of the residual block, so that the network can learn feature representations other than the identity mapping; in order to change the size and number of channels of the feature map, a transition layer is inserted between the stacked residual blocks (the transition layer contains a 1×1 convolution layer for reducing the number of channels, and a 3×3 convolution layer or an average pooling layer with a stride of 2 for reducing the size of the feature map); at the end of the model there is a fully connected layer for mapping the features extracted by the convolution layer to the output space.

[0064] Preferably, the hybrid attention mechanism is embedded in the residual block of ResNet-18 in step 2.2, as follows:

[0065] The hybrid attention mechanism consists of a channel attention module M C and the spatial attention module M S The channel attention module M C and the spatial attention module M S The hybrid attention mechanism is embedded in the residual block of ResNet-18 by replacing the two convolutional layers in the residual block of ResNet-18 respectively.

[0066] The structure after embedding the hybrid attention mechanism in the residual block of ResNet-18 is as follows:

[0067] use represents the input of the residual block embedded in the hybrid attention mechanism, where H represents the height of the input feature map, W represents the width of the input feature map, and C represents the number of channels of the input feature map;

[0068] Use F as the channel attention module M C The input of the channel attention module is obtained, and then the output M C (F), and then multiply it by the input F to get the channel attention map F'; F' then serves as the spatial attention module M S The input of the spatial attention module is then used to obtain the output M S (F′), and then multiply it by F' to get the output F″ of the hybrid attention mechanism; the function is as follows:

[0069]

[0070] in, Indicates the corresponding element multiplication operation;

[0071]

[0072] Where σ represents the sigmoid activation function, MLP(·) represents the multi-layer perceptron network, AvgPool(·) represents the average pooling operation, MaxPool(·) represents the maximum pooling operation, W0 and W1 represent the weights from the input layer to the hidden layer and from the hidden layer to the output layer of the multi-layer perceptron, respectively. represents the output of average pooling, represents the output of the maximum pooling;

[0073]

[0074] Among them, f 7×7 (·) represents a convolution operation with a kernel size of 7×7. represents the output of average pooling, represents the output after maximum pooling, Indicates that the feature map obtained by dimensional concatenation of the outputs of average pooling and maximum pooling is used as the input of the convolution;

[0075] Step 3: Construct a training set, a validation set, and a test set; the training set is used to train the neural network, the validation set is used to adjust the model hyperparameters, and the test set is used to test the model performance;

[0076] Preferably, the training set, validation set and test set are constructed in step 3 as follows:

[0077] For the collected river images of M observation points and N water pollution images, the two types of image samples are randomly divided into training set, validation set and test set in a ratio of 7:2:1;

[0078] Then the number of training set samples is: 0.7×M+0.7×N;

[0079] The number of samples in the validation set is: 0.2×M+0.2×N;

[0080] The number of test set samples is: 0.1×M+0.1×N.

[0081] Step 4: Use the training set to train the constructed improved ResNet-18 model, use the validation set to adjust the model's hyperparameters, and then input the test set into the trained deep learning network for performance evaluation;

[0082] Preferably, the improved ResNet-18 model constructed by training in step 4 comprises the following steps:

[0083] Step 4.1, set the hyperparameters required for training, including epoch, batch_size, learning rate, training optimizer, and loss function; the loss function of the improved ResNet-18 model constructed uses the cross entropy loss function;

[0084] Step 4.2, set the model evaluation index: the evaluation index uses four widely used evaluation indexes for classification tasks, namely accuracy, precision, recall and F1-score. Among them, the polluted water quality image is regarded as a positive sample (Positive), and the unpolluted water quality image is regarded as a negative sample (Negative), then:

[0085]

[0086]

[0087] Among them, TP means that the positive sample is successfully predicted as a positive sample, TN means that the negative sample is successfully predicted as a negative sample, FP means that the negative sample is incorrectly predicted as a positive sample, and FN means that the positive sample is incorrectly predicted as a negative sample;

[0088] Step 4.3, initialize the network parameters, and input the training set into the constructed improved ResNet-18 model to start the training process. According to the calculated loss, update the model parameters through the gradient descent algorithm until the training ends (the loss value is stable or the iteration ends); then freeze the model parameters, input the validation set into the trained improved ResNet-18 model, optimize and adjust the model through the evaluation indicators, and finally, input the test set into the optimized improved ResNet-18 model to evaluate the performance of the model.

[0089] Step 5: Save the model and deploy it. Use the trained improved ResNet-18 model to detect river water pollution. When the read river image is identified as polluted water, an alarm signal is triggered.

[0090] Preferably, step 5 includes the following steps:

[0091] Step 5.1, save the model (including model architecture and parameters) when the evaluation index is optimal during the verification process;

[0092] Step 5.2, in the river monitoring system, write a real-time video data reading script to achieve uninterrupted reading of video data and pre-process the video data;

[0093] Step 5.3, call the saved improved ResNet-18 model to monitor river water pollution and return the current water quality status.

[0094] The above embodiments are only preferred technical solutions of the present invention and should not be regarded as limiting the present invention. The protection scope of the present invention shall be the technical solutions recorded in the claims, including equivalent replacement solutions of the technical features in the technical solutions recorded in the claims. That is, equivalent replacement improvements within this scope are also within the protection scope of the present invention.

Claims

1. A river pollution identification and early warning method based on computer vision, characterized by: The following steps are involved: Step 1, create a river water quality image dataset; Step 2: Build an improved ResNet-18 model based on the attention mechanism; Step 3: construct training set, validation set and test set; The training set is used to train the neural network, the validation set is used to adjust the model hyperparameters, and the test set is used to test the model performance; Step 4: Use the training set to train the constructed improved ResNet-18 model, use the validation set to adjust the model's hyperparameters, and then input the test set into the trained deep learning network for performance evaluation; Step 5: Save the model and deploy it. Use the trained improved ResNet-18 model to detect river water pollution. When the read river image is identified as polluted water, an alarm signal is triggered.

2. The computer vision-based river pollution identification and early warning method according to claim 1, characterized in that: The step 1 specifically comprises the following steps: Step 1.1, collecting video surveillance data of the observation point: installing a black light camera on the river bank of the observation point; then, selecting multiple video clips in four scenes of sunny day, rainy day, foggy day and snowy day in the database of the monitoring server under two scenarios of day and night; Step 1.2, video frame processing: use the VideoCapture function in the OpenCV extension library cv2 to perform frame processing on the collected river monitoring video clips of the observation point to obtain the river water quality image of the observation point; Step 1.3, collect river pollution images: Since there are few water quality images of river pollution at the observation point, publicly available river pollution water quality images are collected through online channels; Step 1.4, image preprocessing: For river pollution water quality images: use the image cropping tool to crop out the polluted area in the image, and use the Padding function to adjust the cropped image to a square size, and then scale it proportionally to a 224 pixel × 224 pixel image; For river water quality images at observation points: use the function to automatically crop out a 224 pixel × 224 pixel image in the center of the image.

3. The computer vision-based river pollution identification and early warning method according to claim 1, characterized in that: The step 2 specifically includes the following steps: Step 2.1, build the classic ResNet-18 model, including the initial convolutional layer, residual block, transition layer, and fully connected layer; Step 2.2, embed the hybrid attention mechanism in the residual block of ResNet-18.

4. The computer vision-based river pollution identification and early warning method according to claim 3 is characterized by: The classic ResNet-18 model built in step 2.1 is as follows: After the input layer is the initial convolution layer, which is used to extract the initial features of the image; after the initial layer is the stacked residual blocks, each residual block contains two 3×3 convolution layers, and each convolution layer is followed by a batch normalization layer and a ReLU activation function; secondly, the residual connection enables the network to learn feature representations other than identity mapping by adding the input directly to the output of the residual block; in order to change the size and number of channels of the feature map, a transition layer is inserted between the stacked residual blocks. The transition layer contains a 1×1 convolution layer to increase the number of channels, and a 3×3 convolution layer or average pooling layer with a stride of 2 to reduce the size of the feature map; at the end of the model there is a fully connected layer to map the features extracted by the convolution layer to the output space.

5. The computer vision-based river pollution identification and early warning method according to claim 4 is characterized by: The step 2.2 is specifically as follows: The hybrid attention mechanism consists of a channel attention module M C and the spatial attention module M S Composition, focusing on important information from two dimensions: feature and space; By adding the channel attention module M C and the spatial attention module M S The hybrid attention mechanism is embedded in the residual block of ResNet-18 by replacing the two convolutional layers in the residual block of ResNet-18 respectively.

6. The computer vision-based river pollution identification and early warning method according to claim 5 is characterized by: The structure after embedding the hybrid attention mechanism in the residual block of ResNet-18 is as follows: use Represents the input of the structure, where H represents the height of the input feature map, W represents the width of the input feature map, and C represents the number of channels of the input feature map; Use F as the channel attention module M C The input of the channel attention module is obtained, and then the output M C (F), and then multiply it by the input F to get the channel attention map F'; F' then serves as the spatial attention module M S The input of the spatial attention module is then used to obtain the output M S (F′), and then multiply it by F' to get the output F″ of the hybrid attention mechanism; the function is as follows: in, Indicates the corresponding element multiplication operation; Where σ represents the sigmoid activation function, MLP(·) represents the multi-layer perceptron network, AvgPool(·) represents the average pooling operation, MaxPool(·) represents the maximum pooling operation, W0 and W1 represent the weights from the input layer to the hidden layer and from the hidden layer to the output layer of the multi-layer perceptron, respectively. represents the output of average pooling, represents the output of the maximum pooling; Among them, f 7×7 (·) represents a convolution operation with a kernel size of 7×7. represents the output of average pooling, represents the output after maximum pooling, It means that the feature map obtained by dimensionally concatenating the outputs of average pooling and maximum pooling is used as the input of the convolution.

7. The computer vision-based river pollution identification and early warning method according to claim 1 is characterized by: The construction of training set, validation set and test set described in step 3 is as follows: For the collected river images of M observation points and N water pollution images, the two types of image samples are randomly divided into training set, validation set and test set in a ratio of 7:2:1; Then the number of training set samples is: 0.7×M+0.7×N; The number of samples in the validation set is: 0.2×M+0.2×N; The number of test set samples is: 0.1×M+0.1×N.

8. The computer vision-based river pollution identification and early warning method according to claim 1 is characterized by: The training of the improved ResNet-18 model constructed in step 4 includes the following steps: Step 4.1, set the hyperparameters required for training, including epoch, batch_size, learning rate, training optimizer, and loss function; the loss function of the improved ResNet-18 model constructed uses the cross entropy loss function; Step 4.2, set the model evaluation index: the evaluation index uses four evaluation indexes used for classification tasks, namely accuracy, precision, recall and F1-score; among them, the polluted water quality image is regarded as a positive sample, and the unpolluted water quality image is regarded as a negative sample, then: Among them, TP means that the positive sample is successfully predicted as a positive sample, TN means that the negative sample is successfully predicted as a negative sample, FP means that the negative sample is incorrectly predicted as a positive sample, and FN means that the positive sample is incorrectly predicted as a negative sample; Step 4.3, initialize the network parameters, and input the training set into the constructed improved ResNet-18 model to start the training process. According to the calculated loss, update the model parameters through the gradient descent algorithm until the training ends. Then freeze the model parameters, input the validation set into the trained improved ResNet-18 model, optimize and adjust the model through the evaluation indicators, and finally, input the test set into the optimized improved ResNet-18 model to evaluate the performance of the model.

9. The computer vision-based river pollution identification and early warning method according to claim 1, characterized in that: The step 5 comprises the following steps: Step 5.1, save the model with the best evaluation index during the verification process, the model includes the model architecture and parameters; Step 5.2, in the river monitoring system, write a real-time video data reading script to achieve uninterrupted reading of video data and pre-process the video data; Step 5.3, call the saved improved ResNet-18 model to monitor river water pollution and return the current water quality status.

Citation Information

Patent Citations

  • SAR (Synthetic Aperture Radar) image target identification method combining transfer learning and attention mechanism

    CN114863283A

  • River floating pollutant identification method based on convolutional neural network

    CN115497015A

  • Steel structure corrosion identification method and device, medium and equipment

    CN117934901A

  • Improved YOLOv8n-based garbage identification method for trash rack of hydraulic power plant

    CN118155135A