A method, monitoring system, medium, and product for intelligent monitoring and recognition of birds

Abstract

This application relates to the field of intelligent monitoring technology, and particularly to a method for intelligent monitoring and identification of birds, a monitoring system, a medium, and a product. The method includes obtaining a target image, where the target image is an image corresponding to the area to be monitored; importing the target image into a trained target recognition model to obtain corresponding bird recognition information; obtaining all bird recognition information corresponding to a preset time period, and integrating all bird recognition information within the preset time period to obtain a bird recognition data set; obtaining a standard data set corresponding to the preset time period, matching the standard data set with the bird recognition data set to obtain a data set matching value, and when the data set matching value is lower than a preset standard value, generating a prompt feedback information based on the bird recognition data set. This application facilitates improving the accuracy of monitoring and identifying birds.

Description

Technical Field

[0001] This application relates to the field of intelligent monitoring technology, and particularly to a method for intelligent monitoring and identification of birds, a monitoring system, a medium, and a product. Background Art

[0002] Birds are an important part of biodiversity and have key identities and roles in the global ecosystem. For example, birds can spread seeds, control pests, etc. By monitoring birds, it is possible to analyze their species, quantity, and distribution changes, which is convenient for evaluating the status of biodiversity. Additionally, it can promptly discover and protect endangered species to prevent the loss of biodiversity. Moreover, since birds are sensitive to environmental changes, the population dynamics and distribution patterns of birds can often reflect the current environmental health status. Therefore, monitoring birds is particularly crucial.

[0003] In related technologies, generally, the method of manually extracting features is used to identify features such as the beak, feet, wings, or neck of birds in the target image, and then the recognition of birds in the target image is completed by extracting details such as the corresponding contours, colors, and textures of the features. However, when using this method to identify birds, it has high requirements for the clarity and quality of the target image. That is, any slight noise, changes in lighting conditions, or differences in shooting angles in the target image may affect the accuracy of feature extraction, thereby affecting the reliability of subsequent recognition. Summary of the Invention

[0004] In order to improve the accuracy of monitoring and identifying birds, this application provides a method for intelligent monitoring and identification of birds, a monitoring system, a medium, and a product.

[0005] In a first aspect, this application provides a method for intelligent monitoring and identification of birds, adopting the following technical solution:

[0006] A method for intelligent monitoring and identification of birds includes:

[0007] Obtain a target image, where the target image is an image corresponding to the area to be monitored;

[0008] Import the target image into a trained target recognition model to obtain corresponding bird recognition information;

[0009] Obtain all bird recognition information corresponding to a preset time period, and integrate all bird recognition information within the preset time period to obtain a bird recognition data set;

[0010] Obtain the standard data set corresponding to the preset time period, match the standard data set with the bird recognition data set to obtain a data set matching value, and when the data set matching value is lower than the preset standard value, generate a prompt feedback information based on the bird recognition data set.

[0011] By adopting the above technical solution, by importing the target image into the trained target recognition model, it is convenient to quickly and accurately identify the bird recognition information contained in the target image. When determining the bird recognition information through the trained target recognition model, the influence of noise, changes in lighting conditions, or shooting angles in the target image on the recognition result is relatively small, and it is also convenient to reduce the influence of human misjudgment or bias on the accuracy of the recognition result. Finally, by matching the standard data set with the bird recognition data set, it is convenient to timely monitor possible abnormal situations in the area to be monitored, and by generating prompt feedback information, it is convenient to remind relevant staff to pay attention to and handle the abnormalities in a timely manner.

[0012] In a possible implementation manner, the training process of the target recognition model includes:

[0013] Obtain an initial recognition model and training sample data. The initial recognition model includes an initial teacher recognition model and an initial student recognition model, and the training sample data includes unlabeled sample pictures and labeled sample pictures;

[0014] Execute the following loop steps until the loss output value corresponding to the initial recognition model is lower than a preset threshold, then stop the loop, and determine the initial recognition model with the loss output value lower than the preset threshold as the target recognition model;

[0015] Among them, the loop steps include:

[0016] Import the unlabeled sample pictures into the initial teacher recognition model to obtain the pseudo-labels corresponding to the unlabeled sample pictures;

[0017] Train the initial student recognition model according to the unlabeled sample pictures and the corresponding pseudo-labels and the labeled sample pictures to obtain an updated initial student recognition model;

[0018] Optimize the updated initial student recognition model based on a preset exponential moving average algorithm to obtain an updated teacher recognition model;

[0019] Determine the loss output value of the initial recognition model based on a preset loss function.

[0020] By adopting the above technical solution, the initial recognition model is trained jointly with unlabeled sample images and labeled sample images, which is convenient for reducing the demand for training sample data during the model training process, thereby improving the generalization ability of the initial recognition model. In addition, through the loop steps, the student model is continuously trained with pseudo-labels, and the initial teacher recognition model is updated with the optimized initial student recognition model. This iterative training method is convenient for enabling the initial teacher recognition model and the initial student recognition model to promote each other and make common progress, thereby continuously improving the recognition performance of the initial recognition model, that is, to ensure that the finally obtained target recognition model has sufficient recognition performance.

[0021] In a possible implementation manner, optimizing the updated initial student recognition model based on the preset exponential moving average algorithm to obtain an updated teacher recognition model includes:

[0022] Obtain the student model parameters corresponding to the updated initial student model and the teacher model parameters corresponding to the updated teacher recognition model;

[0023] Obtain a preset update suppression parameter, and import the update suppression parameter, the student model parameters, and the teacher model parameters into a preset parameter optimization formula to obtain optimized model parameters;

[0024] Update the teacher model parameters based on the optimized model parameters to update the teacher recognition model.

[0025] By adopting the above technical solution, by combining the student model parameters and the teacher model parameters and performing comprehensive calculations using a preset parameter optimization formula after introducing the update suppression parameter, more comprehensive and optimized teacher model parameters can be obtained. By making full use of the complementarity between different models, it is convenient to improve the accuracy of the teacher model parameters.

[0026] In a possible implementation manner, when the dataset matching value is lower than the preset standard value, the method further includes:

[0027] Identify the environmental parameter information and vegetation parameter information corresponding to each bird recognition information in the bird recognition dataset. The environmental parameter information includes water area parameters, water quality parameters, and soil parameters, and the vegetation parameter information includes vegetation types and vegetation areas;

[0028] Identify the weather parameter information corresponding to each bird recognition information in the bird recognition dataset from the collection log data. The weather parameter information includes temperature, precipitation, and light intensity;

[0029] Obtain the simulation area model corresponding to the area to be monitored, and respectively drive the simulation area model for simulation based on the environmental parameter information, vegetation parameter information, and weather parameter information to obtain an environmental impact simulation data set, a vegetation impact simulation data set, and a weather impact simulation data set respectively;

[0030] Respectively match the environmental impact simulation data set, the vegetation impact simulation data set, and the weather impact simulation data set with the standard data set to determine the concerned simulation data set, and the matching difference between the concerned simulation data set and the standard data set is the largest.

[0031] By adopting the above technical solution, when the data set matching value is lower than the preset standard value, the simulation area model can be driven respectively according to the environmental parameter information, vegetation parameter information, and weather parameter information of the area to be monitored, so as to analyze the impacts of environmental, vegetation, and weather factors on bird activities or migrations respectively. Finally, the concerned simulation data set with the greatest impact on bird activities or migrations is fed back, so that relevant staff can adjust the possible anomalies in the area to be monitored according to the concerned simulation data set, thereby reducing the impact of the concerned parameter information on birds and maintaining the ecological balance in the area to be monitored.

[0032] In a possible implementation manner, the method further includes:

[0033] Drive the simulation area model for simulation based on the environmental parameter information, vegetation parameter information, and weather parameter information to obtain an environmental impact layer, a vegetation impact layer, and a weather impact layer respectively;

[0034] When a layer overlay instruction is detected, identify the layer overlay instruction to determine the layer to be overlaid, and overlay the layer to be overlaid to obtain overlay layer feedback information.

[0035] By adopting the above technical solution, the simulation results corresponding to the environmental parameter information, vegetation parameter information, and weather parameter information are displayed in the form of impact layers, which is convenient for intuitively displaying the impacts of different parameter information on birds. In addition, different impact layers can be combined and overlaid according to the viewing requirements, so as to analyze the impacts of the combinations of different parameter information on bird activities, thereby facilitating in-depth understanding and analysis of the laws and connections between parameter information.

[0036] In a possible implementation manner, the method further includes:

[0037] When the bird recognition information corresponding to the target image contains a preset abnormal feature, locate the abnormal position based on the target image;

[0038] Obtain the inspection positions of the preset inspection devices, generate inspection instructions based on the inspection positions and the abnormal positions, control the preset inspection devices to perform inspections according to the inspection instructions, and take close-up photos of the abnormal positions to obtain close-up images;

[0039] Generate abnormal feedback information based on the preset abnormal features and the close-up images.

[0040] By adopting the above technical solutions, through feature recognition and monitoring of the target images, it is convenient to timely discover possible abnormal situations in the target images. When there are abnormal situations in the target images, inspection instructions can be planned in a timely manner according to the inspection positions and the abnormal positions, so as to remotely control the inspection devices to take photos of the abnormal positions, thereby facilitating a more intuitive and clear understanding of the abnormal situations, and further facilitating timely response to and elimination of the abnormal situations.

[0041] In a second aspect, the present application provides a monitoring system, adopting the following technical solutions:

[0042] A monitoring system, the monitoring system includes:

[0043] At least one processor;

[0044] A memory;

[0045] At least one application program, wherein the at least one application program is stored in the memory and is configured to be executed by at least one processor, and the at least one application program is configured to: execute the above-mentioned bird intelligent monitoring and recognition method.

[0046] In a third aspect, the present application provides a computer-readable storage medium, adopting the following technical solutions:

[0047] A computer-readable storage medium, including: a computer program stored therein that can be loaded and executed by a processor to execute the above-mentioned bird intelligent monitoring and recognition method.

[0048] In a fourth aspect, the present application provides a computer program product, adopting the following technical solutions:

[0049] A computer program product, including a computer program, and when the computer program is executed by a processor, it implements the above-mentioned bird intelligent monitoring and recognition method.

[0050] In summary, the present application includes at least one of the following beneficial technical effects:

[0051] By importing the target image into the trained target recognition model, it is convenient to quickly and accurately identify the bird recognition information contained in the target image. When determining the bird recognition information through the trained target recognition model, the influence of noise, changes in lighting conditions, or shooting angles in the target image on the recognition result is relatively small, and it is also convenient to reduce the influence of manual misjudgment or bias on the accuracy of the recognition result. Finally, by matching the standard dataset with the bird recognition dataset, it is convenient to timely monitor possible abnormal situations in the area to be monitored, and by generating prompt feedback information, it is convenient to remind relevant staff to pay attention to and handle the abnormalities in a timely manner.

[0052] When the dataset matching value is lower than the preset standard value, the simulation area models can be driven respectively according to the environmental parameter information, vegetation parameter information, and weather parameter information of the area to be monitored, so as to analyze the influence of factors such as environment, vegetation, and weather on bird activities or migrations. Finally, by feeding back the attention simulation dataset that has the greatest impact on bird activities or migrations, it is convenient for relevant staff to correspondingly adjust the possible abnormalities in the area to be monitored according to the attention simulation dataset, thereby reducing the impact of the attention parameter information on birds and maintaining the ecological balance in the area to be monitored. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 is a schematic flowchart of a method for intelligent monitoring and recognition of birds in an embodiment of the present application;

[0054] Figure 2 is a schematic flowchart of a method for determining an attention simulation dataset in an embodiment of the present application;

[0055] Figure 3 is a schematic structural diagram of a monitoring system in an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0056] The following will further describe the present application in detail Figures 1 to 3 with reference to the accompanying drawings.

[0057] Those skilled in the art can make modifications to this embodiment without creative contributions according to their needs after reading this specification, but as long as they are within the scope of the claims of the present application, they are protected by the Patent Law.

[0058] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without making creative efforts belong to the scope of protection of the present application.

[0059] It should be noted that in the alternative embodiments of the present application, for relevant data such as object information, when the embodiments in the present application are applied to specific products or technologies, object permission or consent needs to be obtained, and the collection, use, and processing of relevant data need to comply with relevant laws, regulations, and standards of relevant countries and regions. That is to say, if the embodiments in the present application involve data related to an object, it needs to be obtained under the authorization and consent of the object, the authorization and consent of relevant departments, and compliance with relevant laws, regulations, and standards of the country and region. In the embodiments, if personal information is involved, the consent of the individual needs to be obtained for all personal information. If sensitive information is involved, the separate consent of the information subject needs to be obtained, and the embodiments also need to be implemented under the authorization and consent of the object.

[0060] Specifically, the embodiments of the present application provide a method for intelligent monitoring and identification of birds, which is executed by a monitoring system. The monitoring system can be a server or a terminal device. Among them, the server can be an independent physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. The terminal device can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc., but is not limited thereto. The terminal device and the server can be directly or indirectly connected through wired or wireless communication methods, and the embodiments of the present application do not limit this.

[0061] Reference Figure 1 , Figure 1 is a schematic flow chart of a method for intelligent monitoring and identification of birds in the embodiments of the present application. The method includes steps S110 - S140, where:

[0062] Step S110: Obtain a target image, where the target image is an image corresponding to the area to be monitored.

[0063] Specifically, the area to be monitored is an area where bird monitoring and identification are required, which can be a wetland, a forest mountain area, an urban park or green space, etc. The specific area is not specifically limited in the embodiments of the present application. The target image can be collected by an image acquisition device set in the area to be monitored and then uploaded to the monitoring system. The target image needs to contain a bird to be identified.

[0064] Step S120: Import the target image into a trained target recognition model to obtain corresponding bird recognition information.

[0065] Specifically, the trained target recognition model can accurately identify the bird information contained in the target image. For example, features such as the beak, feet, wings, or neck of the bird, as well as details such as the contour, color, and texture corresponding to each extracted feature. The target recognition model is a semi-supervised fine-grained training method, and at the same time, a double-threshold loss calculation method is used to verify whether the training result is qualified. Therefore, the accuracy of determining bird recognition information using the trained target recognition model is relatively high. The embodiments of the present application provide a training process for the target recognition model, which may specifically include:

[0066] Obtain an initial recognition model and training sample data. The initial recognition model includes an initial teacher recognition model and an initial student recognition model. The training sample data includes unlabeled sample pictures and labeled sample pictures; perform the following loop steps until the loss output value corresponding to the initial recognition model is lower than a preset threshold, then stop the loop, and determine the initial recognition model with the loss output value lower than the preset threshold as the target recognition model;

[0067] Among them, the loop steps include:

[0068] Import the unlabeled sample pictures into the initial teacher recognition model to obtain the pseudo-labels corresponding to the unlabeled sample pictures; train the initial student recognition model according to the unlabeled sample pictures and the corresponding pseudo-labels and the labeled sample pictures to obtain an updated initial student recognition model; optimize the updated initial student recognition model based on a preset exponential moving average algorithm to obtain an updated teacher recognition model; determine the loss output value of the initial recognition model based on a preset loss function.

[0069] Among them, the training sample data includes unlabeled sample images and labeled sample images. In the embodiments of the present application, a method of mixing unlabeled sample images and labeled sample images is adopted to train the initial recognition model, reducing the high demand for training sample data during the training process. The initial recognition model consists of an initial teacher recognition model and an initial student recognition model. The process of training the initial recognition model with the mixed training sample data can be regarded as a semi-supervised learning process. Among them, it is necessary to predict the unlabeled sample images according to the initial teacher recognition model to generate pseudo-labels corresponding to the unlabeled fake sample images, and bind the obtained pseudo-labels to the corresponding unlabeled sample images to obtain pseudo-labeled sample images. Then, based on the pseudo-labeled sample images and the labeled sample images, the initial student recognition model is jointly trained. Among them, in order to improve the training rate and effect of the initial student recognition model, before training the initial student recognition model based on the pseudo-labeled sample images, preprocessing operations are performed on the pseudo-labeled sample images and the labeled sample images. Among them, the preprocessing operations on the pseudo-labeled sample images are different from those on the labeled sample images. When preprocessing the pseudo-labeled sample images, threshold filtering needs to be performed on the pseudo-labels of the pseudo-labeled sample images, that is, the pseudo-labels exceeding the preset standard threshold are removed to improve the quality of the pseudo-labels, thereby improving the quality of the pseudo-labeled sample images. When preprocessing the labeled sample images, the label edge features in the labeled sample images can be recognized to determine whether the label corresponding to the labeled sample image is a weak label, and weak label enhancement processing is performed on the existing weak labels. Among them, the weak label enhancement processing includes but is not limited to HSV adjustment, rotation, translation, flipping, and Mosaic. The Mosaic processing includes operations such as random cropping, scaling, and image merging. The specific method of preprocessing the labeled sample images is not specifically limited in the embodiments of the present application.

[0070] After preprocessing the pseudo-labeled sample images and the labeled sample images, the initial student recognition model is trained based on the results obtained from the preprocessing until the student loss function value of the initial student recognition model is lower than the preset student threshold. The specific student threshold is not specifically limited in the embodiments of the present application and can be set by relevant technical personnel. Among them, the student loss function value is the weighted sum of the supervised loss function value and the unsupervised loss function value. Among them, the supervised loss function value is obtained after training according to the labeled sample images, and the unsupervised loss function value is obtained after training according to the pseudo-labeled sample images. The calculation formula of the student loss function value is: L = Ls + WuLu, where L is used to represent the student loss function value, Ls is the supervised loss function value, Lu is the unsupervised loss function value, and Wu is the weight of the unsupervised loss. The specific value of Wu is not specifically limited in the embodiments of the present application and can be set by relevant technical personnel according to actual needs.

[0071] When the student loss function value of the initial student recognition model is lower than the preset student threshold, the initial student recognition model is optimized according to the preset exponential moving average algorithm to obtain an updated teacher recognition model, so as to optimize the teacher recognition model. An embodiment of the present application provides a method for optimizing and updating an initial student recognition model based on a preset exponential moving average algorithm, which may specifically include:

[0072] Obtain the student model parameters corresponding to the updated initial student model and the teacher model parameters corresponding to the updated teacher recognition model; obtain a preset update suppression parameter, and import the update suppression parameter, the student model parameters, and the teacher model parameters into a preset parameter optimization formula to obtain optimized model parameters; update the teacher model parameters based on the optimized model parameters to update the teacher recognition model.

[0073] Specifically, the update suppression parameter, the student model parameters, and the teacher model parameters can be imported into , where is used to represent the update suppression parameter. In the embodiment of the present application, the update suppression parameter is 0.5. The specific value is not specifically limited in the embodiment of the present application and can be set by relevant staff according to actual needs; θt is used to represent the teacher model parameters, θs is used to represent the student model parameters, and θxt is used to represent the optimized model parameters, that is, the optimized teacher model parameters. The teacher model parameters are the weights, biases, activation functions, etc. of the initial teacher recognition model. The student model parameters correspond to the teacher model parameters, but the quantity and complexity are lower than the teacher model parameters. The teacher model parameters or the student model parameters can be obtained during the semi-supervised training process. Updating or adjusting the teacher model parameters through the determined optimized model parameters facilitates the update and optimization of the initial teacher recognition model.

[0074] The update strategy of the initial teacher recognition model is the integration of the initial student recognition model in the t-th iteration and the initial teacher recognition model parameters in the previous (t-1) iterations. The initial teacher recognition model can be regarded as an integrated model of the initial student recognition model at different training stages. That is, a part of the weights of the initial teacher recognition model come from the weights of the historical initial teacher recognition model, and the other part comes from the weights of the initial student recognition model at the current stage. Therefore, with the interaction between the initial student recognition model and the initial teacher recognition model, the two types of models can co-evolve and continuously improve the prediction accuracy.

[0075] After optimizing the teacher recognition model, the updated teacher recognition model is used again to predict the pseudo-labels corresponding to other unlabeled sample images to obtain pseudo-labeled sample images. The initial student model is trained again using the pseudo-labeled sample images and other labeled sample images, and this is executed iteratively. At the end of each iteration, the loss output value corresponding to the initial recognition model is calculated according to a preset loss function until the loss output value corresponding to the initial recognition model is lower than a preset threshold. Herein, the preset loss function can be a loss calculation function with double thresholds. The specific loss calculation function and the preset threshold are not specifically limited in the embodiments of this application, as long as they can verify whether the initial recognition model has completed training.

[0076] Step S130: Obtain all bird recognition information corresponding to a preset time period, and integrate all the bird recognition information within the preset time period to obtain a bird recognition data set.

[0077] Specifically, the preset time period can be a period of time before the current moment. The duration corresponding to the preset time period can be 10 days or 20 days. The specific duration is not specifically limited in the embodiments of this application. After sorting all the bird recognition information according to the acquisition time of each bird recognition information, a bird recognition data set can be obtained, which facilitates the overall analysis of the activities or migrations of birds in the area to be monitored.

[0078] Step S140: Obtain a standard data set corresponding to the preset time period, match the standard data set with the bird recognition data set to obtain a data set matching value, and when the data set matching value is lower than a preset standard value, generate a prompt feedback message based on the bird recognition data set.

[0079] Specifically, the standard dataset can be a mean dataset determined by relevant staff based on historical monitoring and identification data. For example, if the preset time period is from October 1st to October 31st, 2023, the standard dataset is the mean of the monitoring and identification data from October 1st to October 30th during the period from 2013 to 2022. The specific content is not specifically limited in the embodiments of this application. By matching the bird identification dataset corresponding to the preset time period with the corresponding standard dataset, it is convenient to analyze and predict whether there are abnormalities in the activities or migrations of birds within the preset time period. Specifically, it can be determined through the dataset matching value. When the dataset matching value is not lower than the preset standard value, it indicates that the migration situation of the bird activities included in the area to be monitored within the preset time period is in a normal state. For example, xx migratory birds generally stay briefly in the area to be monitored from October 1st to October 30th to prepare for the subsequent long-distance flight, and the average number of staying birds is 200. If it is determined from the bird identification dataset that only 100 xx migratory birds stay in the area to be monitored within the preset time period, at this time, the corresponding dataset matching value may be lower than the preset standard value, that is, a prompt feedback message may be generated at this time to remind relevant staff to promptly handle the possible ecological environment problems faced by the area to be monitored.

[0080] For the embodiments of this application, by importing the target image into the trained target recognition model, it is convenient to quickly and accurately identify the bird identification information contained in the target image. When determining the bird identification information through the trained target recognition model, the influence of noise, changes in lighting conditions, or shooting angles in the target image on the recognition result is relatively small, and it is also convenient to reduce the influence of manual misjudgment or bias on the accuracy of the recognition result. Finally, by matching the standard dataset with the bird identification dataset, it is convenient to promptly monitor the possible abnormal situations in the area to be monitored, and by generating prompt feedback information, it is convenient to remind relevant staff to pay attention to and handle the abnormalities in a timely manner.

[0081] Further, when the dataset matching value is lower than the preset standard value, the method provided in the embodiments of this application further includes steps S1 - S4, as Figure 2 shown, where:

[0082] Step S1: Identify the environmental parameter information and vegetation parameter information corresponding to each bird identification information in the bird identification dataset. The environmental parameter information includes water area parameters, water quality parameters, and soil parameters, and the vegetation parameter information includes vegetation types and vegetation areas.

[0083] Specifically, the bird recognition dataset contains bird recognition information corresponding to each acquisition moment within a preset time period. Among them, target images can be collected within the preset time period according to a preset acquisition frequency, and based on this, bird recognition information can be obtained. For example, the target images are collected once every 3 days, and the preset time period is 30 days. That is, the bird recognition dataset corresponding to the preset time period contains 10 pieces of bird recognition information. Since the bird recognition information is obtained from the target images, by analyzing the bird recognition information, it is convenient to determine the environmental parameter information and vegetation parameter information contained in the corresponding target images.

[0084] The water area parameter, that is, the water area, can be recognized from the corresponding target images through feature recognition. The specific feature recognition algorithm is not specifically limited in the embodiments of the present application. For example, an edge detection algorithm can be used. The water quality parameters and soil parameters can be collected by sensors arranged in the area to be monitored and then uploaded to the monitoring system. Since the food sources of birds may come from organisms in the water quality or soil, such as fish or insects, once the water quality or soil in the area to be monitored deteriorates, it is very likely to affect the number of fish or insects, thereby affecting the food supply of birds. Therefore, when it is determined through analyzing the bird recognition dataset in the area to be monitored that the dataset matching value is lower than the preset standard value, it may indicate that there are abnormal conditions in the water quality or soil in the area to be monitored. Therefore, it is necessary to analyze the water quality or soil in the area to be monitored.

[0085] Similarly, since vegetation can provide places for birds to build nests, forage, rest, and hide, and different types of vegetation can support the survival of different types of birds, when the vegetation in the area to be monitored is polluted, damaged, or degraded, it may affect the habitat environment and food sources of birds. Therefore, when it is determined that the dataset matching value is lower than the preset standard value, it may indicate that there are abnormal conditions in the vegetation in the area to be monitored. Therefore, it is necessary to analyze the vegetation in the area to be monitored. Among them, the vegetation parameter information can be recognized from the target images corresponding to the bird recognition information through feature recognition. The specific feature recognition algorithm is not specifically limited in the embodiments of the present application.

[0086] Step S2: Identify the weather parameter information corresponding to each piece of bird recognition information in the bird recognition dataset from the collected log data. The weather parameter information includes temperature, precipitation, and light intensity.

[0087] Specifically, the weather parameter information corresponding to each bird recognition information, that is, the weather parameter information corresponding to each acquisition moment, can be collected by weather sensors installed in the area to be monitored and uploaded to the monitoring system to be converted into acquisition log data, and can be retrieved from the acquisition log data based on the acquisition moment. Since excessively high or low temperatures can affect the physiological functions, foraging, and migration of birds, and precipitation and light intensity can affect the food supply and habitat environment of birds in the area to be monitored, when it is determined that the dataset matching value is lower than the preset standard value, it is also necessary to analyze the weather parameters in the area to be monitored.

[0088] Based on the above method, the environmental parameter information, vegetation parameter information, and weather parameter information corresponding to each acquisition moment within a preset time period can be determined. The environmental parameter information, vegetation parameter information, and weather parameter information corresponding to each acquisition moment are integrated to obtain the integrated environmental parameter information, integrated vegetation parameter information, and integrated weather parameter information corresponding to the preset time period.

[0089] Step S3: Obtain the simulation area model corresponding to the area to be monitored, and drive the simulation area model to perform simulations based on the environmental parameter information, vegetation parameter information, and weather parameter information respectively, to obtain the environmental impact simulation dataset, vegetation impact simulation dataset, and weather impact simulation dataset respectively.

[0090] Specifically, the simulation area model corresponding to the area to be monitored has the same ecological environment as the area to be monitored and can simulate the activities of birds in the area to be monitored. This simulation area model can be determined by relevant staff based on historical experimental data and then uploaded to the monitoring system.

[0091] Regarding the environmental parameter information, the simulation area model is driven based on the environmental parameter information, that is, the simulation area model is driven according to the integrated environmental parameter information. That is to say, the simulation area model is only simulated according to the actual environmental parameter information within the preset time period, and the activities of birds in the area to be monitored within the preset time period are simulated and reviewed. Among them, when driving the simulation area model according to the environmental parameter information, the default vegetation parameter information and the default weather parameter information need to be taken together to drive the simulation area model. The default vegetation parameter information and the default weather parameter information can be determined by relevant staff from the historical collection data, which are the average values corresponding to the same historical stage. For example, the preset time period is from October 1, 2023 to October 30, 2023. One of the acquisition times is October 1. The default vegetation type and default vegetation area corresponding to October 1 are the average values of the vegetation types and the average values of the vegetation areas corresponding to all October 1s during the period from 2010 to 2020 in the historical collection data. After simulating the simulation area model according to the integrated environmental parameter information, the obtained environmental impact simulation data set is the simulated bird recognition data set corresponding to the area to be monitored under the influence of the integrated environmental parameter information.

[0092] Based on the above content, a vegetation impact simulation data set and a weather impact simulation data set can be obtained. When determining the vegetation impact data set, the simulation area model needs to be driven according to the integrated vegetation parameter information, the default environmental parameter information, and the default weather parameter information; when determining the weather impact simulation data set, the simulation area model needs to be driven according to the integrated weather parameter information, the default environmental parameter information, and the simulated vegetation parameter information.

[0093] Step S4: Respectively match the environmental impact simulation data set, the vegetation impact simulation data set, and the weather impact simulation data set with the standard data set to determine the concerned simulation data set, and the matching difference between the concerned simulation data set and the standard data set is the largest.

[0094] Specifically, after determining the environmental impact simulation dataset, the vegetation impact simulation dataset, and the weather impact simulation dataset, the environmental impact simulation dataset, the vegetation impact simulation dataset, and the weather impact simulation dataset can be respectively matched with the standard dataset to obtain the environmental matching difference, the vegetation matching difference, and the weather matching difference. Finally, the environmental matching difference, the vegetation matching difference, and the weather matching difference are compared to determine the concerned simulation dataset with the largest matching difference. For example, when the concerned simulation dataset is the vegetation impact simulation dataset, it can be determined that the main reason for the dataset matching value in the area to be monitored being lower than the preset standard value is that the vegetation in the area to be monitored has abnormal conditions. By feeding back the concerned simulation dataset that has the greatest impact on bird activities or migrations, it is convenient for relevant staff to correspondingly adjust the possible abnormalities in the area to be monitored according to the concerned simulation dataset, thereby reducing the impact of the concerned parameter information on birds and maintaining the ecological balance in the area to be monitored.

[0095] Further, to facilitate the intuitive display of the impacts of different parameter information on birds, the method provided in the embodiment of the present application further includes:

[0096] Based on the environmental parameter information, the vegetation parameter information, and the weather parameter information, drive the simulation area model to perform simulations to respectively obtain an environmental impact layer, a vegetation impact layer, and a weather impact layer; when a layer superposition instruction is detected, identify the layer superposition instruction to determine the layer to be superposed, and superpose the layers to be superposed to obtain the superposed layer feedback information.

[0097] Specifically, driving the simulation area model to perform simulations based on the environmental parameter information, that is, driving the area simulation model to perform simulations according to the integrated environmental parameter information, the default vegetation parameter information, and the default weather parameter information. The corresponding environmental impact layer is an intuitive display method of the environmental impact simulation dataset, that is, the environmental impact layer is a simulation image of the area to be monitored within a preset time period, mainly used to display the impact of the integrated environmental parameter information on bird activities or migrations. When determining the environmental impact layer, only the output type of the simulation area model needs to be adjusted, that is, the original output of the environmental impact simulation dataset is adjusted to output an environmental image layer. Based on the above content, the vegetation impact layer and the weather impact layer can be obtained.

[0098] The layer superposition instruction can be sent by relevant access personnel to the monitoring system through the access terminal. The layer superposition instruction contains the superposition requirement, that is, the layers that need to be superposed. For example, the superposition requirement can be to superpose the environmental impact layer and the vegetation impact layer to facilitate the analysis of the impact of the combination of different parameter information on bird activities. The superposition requirement included in the layer superposition instruction can be determined by means of semantic recognition.

[0099] Further, in order to respond to and eliminate the abnormal situations existing in the target image in a timely manner, the method provided by the embodiment of the present application further includes:

[0100] When the bird recognition information corresponding to the target image contains a preset abnormal feature, locate the abnormal position based on the target image; obtain the inspection position of the preset inspection device, generate an inspection instruction based on the inspection position and the abnormal position, and control the preset inspection device to perform an inspection according to the inspection instruction, and take a close-up photo of the abnormal position to obtain a close-up image; generate abnormal feedback information based on the preset abnormal feature and the close-up image.

[0101] Specifically, during the process of obtaining the target image in real time, it is also necessary to identify and analyze the possible abnormal situations in the target image. Since most of the image acquisition devices set in the area to be monitored are fixed devices, when a preset abnormal feature appears in the target image, it may be unclear. At this time, an inspection route can be formulated based on the abnormal position of the preset abnormal feature and the inspection position of the inspection device, so as to control the inspection device to move from the inspection position to the abnormal position along the inspection route for close-up shooting, without moving the image acquisition device or adjusting the shooting parameters of the image acquisition device, that is, by controlling the inspection device to take a close-up photo of the abnormal position, it is convenient to directly view the abnormal situation, and at the same time, it will not affect the normal target image acquisition process. Among them, the inspection device can be a drone image acquisition device. The specific inspection device is not specifically limited in the embodiment of the present application, as long as it can take a close-up photo of the specified position according to the inspection instruction.

[0102] Among them, the preset abnormal feature can be that the bird's posture is distorted, the legs or wings are missing, the feathers are dull, the feathers contain blood, etc. When the target image contains a preset abnormal feature, it may indicate that the bird corresponding to the target image is injured or dead. The specific preset abnormal feature is not specifically limited in the embodiment of the present application and can be set by relevant staff. By performing feature recognition and monitoring on the target image, it is convenient to timely discover the possible abnormal situations in the target image. When there is an abnormal situation in the target image, the inspection instruction can be planned in a timely manner according to the inspection position and the abnormal position, so as to realize remotely controlling the inspection device to take a photo of the abnormal position, thereby facilitating a more intuitive and clearer understanding of the abnormal situation, and further facilitating timely response to and elimination of the abnormal situation.

[0103] An embodiment of the present application provides a monitoring system, as Figure 3 shown Figure 3The monitoring system 300 shown includes: a processor 301 and a memory 303. Among them, the processor 301 and the memory 303 are connected, such as through a bus 302. Optionally, the monitoring system 300 may further include a transceiver 304. It should be noted that in practical applications, the number of transceivers 304 is not limited to one, and the structure of the monitoring system 300 does not constitute a limitation on the embodiments of the present application.

[0104] The processor 301 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute various exemplary logical blocks, modules, and circuits described in connection with the disclosure of the present application. The processor 301 may also be a combination that implements computing functions, such as a combination including one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0105] The bus 302 may include a path for transmitting information between the above components. The bus 302 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The bus 302 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 only one line is shown in the figure, but it does not mean that there is only one bus or one type of bus.

[0106] The memory 303 can be a ROM (Read Only Memory), or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory), or other types of dynamic storage devices that can store information and instructions. It can also be an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory), or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media, or other magnetic storage devices, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto.

[0107] The memory 303 is used to store the application program code for executing the solution of this application, and is controlled by the processor 301 for execution. The processor 301 is used to execute the application program code stored in the memory 303 to implement the content shown in the foregoing method embodiments.

[0108] Among them, the monitoring system includes but is not limited to: mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Tablet Computers), PMPs (Portable Multimedia Players), in-vehicle terminals (such as in-vehicle navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. It can also be a server, etc. Figure 3 The shown monitoring system is only an example and should not impose any limitations on the functions and usage scope of the embodiments of this application.

[0109] The embodiments of this application provide a computer-readable storage medium, on which a computer program is stored. When it runs on a computer, it enables the computer to execute the corresponding content in the foregoing method embodiments.

[0110] An embodiment of the present application provides a computer program product, which includes a computer program that, when executed by a processor, implements the method in any of the above embodiments. Compared with the related art, in the embodiment of the present application, by importing the target image into the trained target recognition model, it is convenient to quickly and accurately identify the bird recognition information contained in the target image. When determining the bird recognition information through the trained target recognition model, the influence of noise, changes in lighting conditions, or shooting angles in the target image on the recognition result is relatively small, and it is also convenient to reduce the influence of human misjudgment or bias on the accuracy of the recognition result. Finally, by matching the standard data set with the bird recognition data set, it is convenient to timely monitor possible abnormal situations in the area to be monitored, and by generating prompt feedback information, it is convenient to remind relevant staff to pay attention to and handle the abnormalities in a timely manner.

[0111] It should be understood that although the steps in the flowchart of the accompanying drawings are shown sequentially as indicated by the arrows, these steps are not necessarily executed sequentially in the order indicated by the arrows. Unless otherwise clearly stated in this article, the execution of these steps is not strictly limited in order, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed alternately or alternately with at least a part of other steps or sub-steps or stages of other steps.

[0112] The above are only some embodiments of the present application. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present application, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present application.

Claims

1. A method for intelligent monitoring and identification of birds, characterized in that, Including: Obtain a target image, where the target image is an image corresponding to the area to be monitored; Import the target image into a trained target recognition model to obtain corresponding bird recognition information; Obtain all bird recognition information corresponding to a preset time period, and integrate all bird recognition information within the preset time period to obtain a bird recognition data set; Obtain the standard data set corresponding to the preset time period, match the standard data set with the bird recognition data set to obtain a data set matching value, and when the data set matching value is lower than the preset standard value, generate a prompt feedback message based on the bird recognition data set; Among them, when the data set matching value is lower than the preset standard value, it further includes: Identify the environmental parameter information and vegetation parameter information corresponding to each bird recognition information in the bird recognition data set. The environmental parameter information includes water area parameters, water quality parameters, and soil parameters, and the vegetation parameter information includes vegetation types and vegetation areas; Identify the weather parameter information corresponding to each bird recognition information in the bird recognition data set from the collected log data. The weather parameter information includes temperature, precipitation, and light intensity; Obtain the simulation area model corresponding to the area to be monitored, and drive the simulation area model to perform simulations based on the environmental parameter information, vegetation parameter information, and weather parameter information respectively to obtain an environmental impact simulation data set, a vegetation impact simulation data set, and a weather impact simulation data set. Among them, when driving the simulation area model based on the environmental parameter information, adopt the default vegetation parameter information and the default weather parameter information to jointly drive the simulation area model with the environmental parameter information. The default vegetation parameter information is the average value of the vegetation types and the average value of the vegetation areas corresponding to the historical collection data; Match the environmental impact simulation data set, the vegetation impact simulation data set, and the weather impact simulation data set with the standard data set respectively to determine the concerned simulation data set, and the matching difference between the concerned simulation data set and the standard data set is the largest; Among them, it further includes: driving the simulation area model to perform simulations based on the environmental parameter information, vegetation parameter information, and weather parameter information respectively to obtain an environmental impact layer, a vegetation impact layer, and a weather impact layer; When a layer overlay instruction is detected, identify the layer overlay instruction to determine the layer to be overlaid, and overlay the layer to be overlaid to obtain an overlay layer feedback message.

2. The intelligent monitoring and recognition method for birds according to claim 1, characterized in that, The training process of the target recognition model includes: Obtain an initial recognition model and training sample data. The initial recognition model includes an initial teacher recognition model and an initial student recognition model, and the training sample data includes unlabeled sample pictures and labeled sample pictures; Execute the following loop steps until the loss output value corresponding to the initial recognition model is lower than the preset threshold, then stop the loop, and determine the initial recognition model with the loss output value lower than the preset threshold as the target recognition model; Among them, the loop steps include: Import the unlabeled sample pictures into the initial teacher recognition model to obtain the pseudo-labels corresponding to the unlabeled sample pictures; Training the initial student recognition model according to the unlabeled sample images, the corresponding pseudo-labels, and the labeled sample images to obtain an updated initial student recognition model; Optimizing the updated initial student recognition model based on a preset exponential moving average algorithm to obtain an updated teacher recognition model; Determining the loss output value of the initial recognition model based on a preset loss function.

3. The method for intelligent monitoring and recognition of birds according to claim 2, characterized in that, The optimizing the updated initial student recognition model based on a preset exponential moving average algorithm to obtain an updated teacher recognition model includes: Obtaining the student model parameters corresponding to the updated initial student model and the teacher model parameters corresponding to the updated teacher recognition model; Obtaining a preset update suppression parameter, and importing the update suppression parameter, the student model parameters, and the teacher model parameters into a preset parameter optimization formula to obtain optimized model parameters; Updating the teacher model parameters based on the optimized model parameters to update the teacher recognition model.

4. A method for intelligent monitoring and recognition of birds according to claim 1, characterized in that, Further includes: When the bird recognition information corresponding to the target image contains preset abnormal features, locating the abnormal position based on the target image; Obtaining the inspection position of a preset inspection device, generating an inspection instruction based on the inspection position and the abnormal position, controlling the preset inspection device to perform inspections according to the inspection instruction, and taking a close-up image of the abnormal position to obtain a close-up image; Generating abnormal feedback information based on the preset abnormal features and the close-up image.

5. A monitoring system, characterized in that, The monitoring system includes: At least one processor; A memory; At least one application program, where the at least one application program is stored in the memory and configured to be executed by the at least one processor, and the at least one application program is configured to: execute a bird intelligent monitoring and recognition method according to any one of claims 1-4.

6. A computer-readable storage medium, characterized in that, Includes: A computer program stored that can be loaded and executed by a processor to execute a bird intelligent monitoring and recognition method according to any one of claims 1-4.

7. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the steps of a bird intelligent monitoring and recognition method according to any one of claims 1-4.

Images