Bird intrusion monitoring method and device, computer equipment and storage medium
Through image detection and tracking technology, bird targets around the substation are identified and tracked, and alarms are triggered when the target enters the anti-invasion area, which solves the problem that the existing technology cannot effectively monitor bird invasion, realizes real-time monitoring and early warning of bird invasion, and improves the safety management level of the substation.
Patent Information
- Application Number
- CN202510063237.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-15
- Publication Date
- 2025-04-29
AI Technical Summary
The existing technology cannot effectively monitor and early warning of bird invasions around substations, resulting in equipment damage and safety accidents.
Image detection technology and image tracking technology are used to identify and track bird targets through nuclear-related filtering algorithms and pre-training models, and alarms are triggered when the target enters the anti-invasion area.
Real-time monitoring and early warning of bird invasions has been achieved, the safety management level of substations has been improved, and equipment damage and safety accidents have been avoided.
Smart Images
Figure CN119904813A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of image processing technology, and in particular to a bird invasion monitoring method, device, computer equipment and storage medium. Background Art
[0002] In recent years, with the rapid development of the power system, substations, as key nodes for power transmission, are crucial to the reliability of the entire power grid for their safe and stable operation. However, bird invasion of substations may not only cause physical damage to equipment, but also cause safety accidents such as short circuits, seriously affecting the normal operation of substations. Therefore, the safety management of substations has become one of the focuses of the power industry. With the advancement of technology, the management of substations is gradually transforming from traditional manual inspections to intelligent and automated monitoring systems to improve efficiency and safety.
[0003] Existing substation monitoring technologies mainly focus on monitoring the equipment and environment within the substation, and are mainly implemented in the form of monitoring systems. These monitoring systems include unit modules such as data acquisition, transmission, processing and control. Through the coordinated work of these units, the overall operation of the substation can be monitored, such as monitoring the equipment temperature and environmental parameters within the substation. When the monitoring system detects abnormal data, an alarm signal can be issued in a timely manner.
[0004] However, these monitoring systems mainly focus on the station environment and equipment status, and lack effective monitoring methods for the surrounding environment of the substation, especially the situation of bird invasion. When birds enter the substation area, the existing technology often cannot detect and warn in time, which may cause equipment damage and safety accidents, causing serious economic losses and safety hazards to the power system. It can be seen that the current technical research and application of bird invasion monitoring in substations is relatively insufficient. Summary of the invention
[0005] The purpose of the present invention is to provide a bird invasion monitoring method, device, computer equipment and storage medium, aiming to solve the problem that the research and application of bird invasion monitoring technology in the prior art are relatively insufficient.
[0006] In a first aspect, an embodiment of the present invention provides a bird invasion monitoring method, comprising:
[0007] Confirming the scene to be monitored and the anti-intrusion area in the scene to be monitored;
[0008] Get the current frame image of the scene to be monitored;
[0009] Perform target detection processing on the current frame image to confirm the bird target in the current frame image;
[0010] Extracting and fusing the features of the bird target and the surrounding area of the bird target by using a kernelized correlation filtering algorithm to obtain a current feature vector;
[0011] The displacement direction of the current feature vector is evaluated through the pre-trained model to obtain the target tracking area of the next frame of the image;
[0012] When the target tracking area appears in the anti-intrusion area, a bird intrusion alarm is triggered.
[0013] In a second aspect, an embodiment of the present invention provides a bird intrusion monitoring device, comprising:
[0014] An area confirmation unit, used to confirm the scene to be monitored and the anti-intrusion area in the scene to be monitored;
[0015] A camera unit, used to obtain a current frame image of the scene to be monitored;
[0016] The target detection unit is used to perform target detection processing on the current frame image to confirm the bird target in the current frame image;
[0017] A feature extraction unit, used for performing feature extraction and fusion processing on the bird target and the surrounding area of the bird target by using a kernelized correlation filtering algorithm to obtain a current feature vector;
[0018] The target tracking unit is used to evaluate the displacement direction of the current feature vector through the pre-trained model to obtain the target tracking area of the next frame image;
[0019] An alarm unit is used to trigger a bird intrusion alarm when the target tracking area appears in the anti-intrusion area.
[0020] In a third aspect, an embodiment of the present invention provides a computer device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the bird intrusion monitoring method described in the first aspect when executing the computer program.
[0021] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor executes the bird intrusion monitoring method described in the first aspect above.
[0022] The beneficial effects of the embodiments of the present invention are: using image detection technology to identify and locate bird targets appearing in monitoring scenes and anti-intrusion areas, and using image tracking technology to track the identified bird targets to determine their real-time positions, and triggering an alarm signal when the bird targets enter the anti-intrusion area, thereby realizing real-time monitoring and early warning of bird invasions, and effectively improving the safety management level of substations. BRIEF DESCRIPTION OF THE DRAWINGS
[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0024] Figure 1 A schematic diagram of a process for monitoring bird invasion provided by an embodiment of the present invention;
[0025] Figure 2 A schematic diagram of a sub-process of a bird invasion monitoring method provided by an embodiment of the present invention;
[0026] Figure 3 A schematic diagram of another sub-process of the bird invasion monitoring method provided by an embodiment of the present invention;
[0027] Figure 4 A schematic diagram of another sub-process of the bird invasion monitoring method provided by an embodiment of the present invention;
[0028] Figure 5 A schematic diagram of another sub-process of the bird invasion monitoring method provided by an embodiment of the present invention;
[0029] Figure 6 A schematic diagram of another sub-process of the bird invasion monitoring method provided by an embodiment of the present invention;
[0030] Figure 7 A schematic block diagram of a bird intrusion monitoring device provided by an embodiment of the present invention;
[0031] Figure 8 A schematic block diagram of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0032] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0033] It should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.
[0034] It should also be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, the singular forms "a", "an" and "the" are intended to include plural forms unless the context clearly indicates otherwise.
[0035] It should be further understood that the term "and / or" used in the present description and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.
[0036] See also Figure 1 , Figure 1 A schematic diagram of a process for monitoring bird invasion provided by an embodiment of the present invention;
[0037] like Figure 1 As shown, the method includes steps S101 to S106.
[0038] S101, confirming the scene to be monitored and the anti-intrusion area in the scene to be monitored;
[0039] In this step, such as some substation monitoring scenarios, anti-intrusion areas can be delineated according to the actual needs of the substation, usually the substation entrances and exits and surrounding key facility areas. One or more areas can be delineated as anti-intrusion areas using polygons of any shape. The anti-intrusion areas are used to trigger subsequent alarm signals.
[0040] S102, obtaining a current frame image of the scene to be monitored;
[0041] In this step, the image of the scene to be monitored can be acquired in real time through the camera module deployed in the scene to be monitored.
[0042] S103, performing target detection processing on the current frame image to confirm the bird target in the current frame image;
[0043] In this step, a deep learning algorithm such as the Yolo_v3 target detection network can be used to perform target detection processing on the current frame image, so as to accurately output the bird target in the current frame image.
[0044] S104, extracting and fusing features of the bird target and the surrounding area of the bird target by using a kernelized correlation filtering algorithm to obtain a current feature vector;
[0045] In this step, the kernelized correlation filtering algorithm can be used to perform detailed feature extraction and fusion processing on the bird target and its surrounding area, thereby effectively capturing key information containing the unique attributes of the target itself, and integrating this information into a comprehensive current feature vector, providing a solid foundation for subsequent target tracking tasks.
[0046] S105, evaluating the displacement direction of the current feature vector through the pre-trained model to obtain the target tracking area of the next frame image;
[0047] In this step, the motion trajectory of the target object in continuous frames can be effectively identified and located, thereby achieving precise target tracking. This not only improves the accuracy of target tracking, but also enhances the robustness and adaptability of the entire system, enabling it to maintain efficient and stable performance in various complex scenarios.
[0048] S106. When the target tracking area appears in the anti-intrusion area, a bird intrusion alarm is triggered.
[0049] In this embodiment, based on the process of steps S101 to S106, image detection technology is used to identify and locate bird targets appearing in the monitoring scene and the anti-intrusion area, and image tracking technology is used to track the identified bird targets to determine their real-time positions, and an alarm signal is triggered when the bird targets enter the anti-intrusion area, thereby realizing real-time monitoring and early warning of bird invasions, and effectively improving the safety management level of the substation.
[0050] In one embodiment, if Figure 2 As shown, step S103 includes:
[0051] S201, performing multiple downsampling processes on the current frame image to obtain multiple downsampling feature maps of different scales;
[0052] S202, fusing multiple downsampled feature maps of different scales according to a required scale to obtain a fused feature map of the required scale;
[0053] S203, extracting features from the fused feature map to confirm the bird target in the fused feature map;
[0054] S204, using a perceptual generative adversarial network to generate super-resolution features for the bird target, and superimposing the super-resolution features on the bird target on the fusion feature map to obtain a feature enhancement map;
[0055] S205: extract features from the feature enhancement image to confirm the bird target in the current frame image.
[0056] This example is mainly implemented by the Yolo_v3 target detection network. Yolo_v3 uses darknet_53 as its basic network. The network is composed of 53 convolutional layers stacked together to extract key features from the image. These convolutional layers are organized into multiple residual structures (res), each of which contains two convolutional layers followed by batch normalization (BN) and ReLU activation layers.
[0057] The specific process is as follows: first, in step S201, a plurality of down-sampled feature maps of different scales are obtained. This process involves gradually reducing the current frame image. Each reduction will extract the key features in the image, thereby forming down-sampled feature maps of different scales. Specifically, a 32-fold down-sampled feature map suitable for detecting large-size targets can be obtained by down-sampling 5 times, a 16-fold down-sampled feature map suitable for detecting medium-size targets can be obtained by down-sampling 4 times, and an 8-fold down-sampled feature map suitable for detecting small-size targets can be obtained by down-sampling 3 times.
[0058] Next, in step S202, feature fusion technology is used to further improve the detection effect. The sampled feature maps of different scales are combined to form a comprehensive fused feature map. This fused feature map contains information at each scale and can better reflect the details and structure in the current frame image.
[0059] Next, step S203 performs feature extraction on the fused feature map to confirm the bird target in the current frame image.
[0060] Next, step S204 introduces a perceptual generative adversarial network (Perceptual GAN), which includes a generator network and a discriminator network; the generator network recreates the bird target data, and the discriminator network determines whether the bird target data is real or created by the generator network. The two are trained adversarially to generate super-resolution features of the bird target, which can improve the details and clarity of the bird target. These super-resolution features are then superimposed on the bird target on the fusion feature map, thereby obtaining a feature enhancement map that enhances the feature expression of the bird target and improves the performance of detecting the bird target;
[0061] Finally, step S205 performs feature extraction on the feature enhancement image to confirm the bird target in the current frame image and locate it.
[0062] More specifically, Figure 3 As shown, step S202 includes:
[0063] S301, selecting the required scale of the fused feature map;
[0064] S302, down-sampling the down-sampled feature graphs larger than the required scale, and up-sampling the down-sampled feature graphs smaller than the required scale, to obtain multiple sampled feature graphs of the same size as the required scale;
[0065] S303: Fusing multiple sampling feature maps of the same size as the required scale to obtain a fused feature map of the required scale.
[0066] In this embodiment, for ease of understanding, the scale required for the 8-fold down-sampling feature map is used for introduction. The 32-fold down-sampling feature map needs to be up-sampled once and then fused with the 16-fold down-sampling feature map. Then, the fused 16-fold down-sampling feature map is up-sampled once and then fused with the 8-fold down-sampling feature map to obtain the required 8-fold down-sampling feature map. It should be noted that the 8-fold down-sampling feature map obtained after fusion here can better capture the detailed information of the target, thereby improving the detection effect, compared with the 8-fold down-sampling feature map directly obtained in the aforementioned step S201.
[0067] Similarly. If the scale required by the 16x down-sampled feature map is used, the 32x down-sampled feature map needs to be up-sampled once and fused with the 16x down-sampled feature map first, and then the 8x down-sampled feature map needs to be down-sampled once and fused with the 16x down-sampled feature map to obtain the fused 16x down-sampled feature map. If the scale required by the 32x down-sampled feature map is used, the 8x down-sampled feature map needs to be down-sampled once and fused with the 16x down-sampled feature map first, and then the fused 16x down-sampled feature map needs to be down-sampled once and fused with the 32x down-sampled feature map to obtain the fused 32x down-sampled feature map.
[0068] In one embodiment, if Figure 4 As shown, step S104 includes:
[0069] S401, extracting the edge direction and intensity distribution of the bird target to obtain HOG features;
[0070] S402, extracting pixel information of the surrounding area of the bird target to obtain grayscale features;
[0071] S403: Fusing the HOG feature and the grayscale feature according to a preset weight to obtain a current feature vector.
[0072] In the process of extracting the bird target, this embodiment first focuses on obtaining the edge direction and intensity distribution of the bird target. This method can obtain the HOG feature, i.e., the directional gradient histogram feature, for subsequent processing. After extracting the edge direction and intensity distribution of the bird target, further attention is paid to obtaining the pixel information of the surrounding area of the bird target, and the grayscale feature can be obtained by using the cluster analysis algorithm.
[0073] After extracting HOG features and grayscale features, these two features are fused according to the preset weights to obtain a comprehensive current feature vector, which will contain information such as the edge, shape, texture, etc. of the target, providing strong support for subsequent classification and recognition tasks.
[0074] Specifically, the preset weights in step S403 may be fused with equal weights, that is, each feature is given an equal weight.
[0075] More specifically, in order to improve the characteristic information of the current feature vector, dynamic weight fusion can be used. The faster the moving speed of the bird target is, the higher the fusion weight of the HOG feature is.
[0076] In some other embodiments, in order to further improve the characteristic information of the current feature vector, the fused feature map or the feature enhancement map is also convolved to obtain the corresponding CNN features, and the CNN features are added to S403 to participate in the fusion, that is, the CNN features, HOG features and grayscale features are fused according to preset weights to obtain the current feature vector.
[0077] Furthermore, the preset weights can be fused with equal weights, that is, the three features of CNN features, HOG features and grayscale features are given equal weights.
[0078] In a specific weight distribution method, the preset weights can also give CNN features a weight of 0.5, and give HOG features and grayscale features a total weight of 0.5. HOG features and grayscale features can be fused using dynamic weights. The faster the bird target moves, the higher the fusion weight of the HOG feature. This distribution method pays more attention to CNN features and is suitable for situations where local features are relatively obvious.
[0079] In one embodiment, if Figure 5 As shown, step S105 includes:
[0080] S501, input the current feature vector into the pre-trained model to perform operations in different displacement directions, and calculate the Gaussian correlation value after each displacement;
[0081] S502, confirming the predicted position of the bird target in the next frame by the displacement position with the maximum Gaussian correlation value;
[0082] S503: Use the surrounding area of the predicted position of the bird target in the next frame as the target tracking area.
[0083] In this embodiment, the pre-trained model is a model obtained by pre-training, by preparing a sample picture with a bird target in advance, selecting a target tracking area in the sample picture, and performing a cyclic displacement operation on the bird target in the target tracking area to generate samples at different positions, wherein the cyclic displacement operation is to shift the bird target to the left or right in each row of the picture grid, and the displacement of each row is 1 greater than that of the previous row, thereby obtaining samples of the target at different positions. These samples are used for sample training classifiers, and the model is saved.
[0084] Training and updating of pre-trained models: Use the above samples to train the classifier, and then in a new frame of the picture, continue to perform cyclic displacement in the target tracking area to obtain new samples to be detected. Use the trained samples to train the classifier and calculate the Gaussian correlation value of the samples to be detected. The sample with the largest Gaussian correlation value is the new position of the bird target. Calculate the offset of the bird target based on the relative offset of the new position, and use the new features to update the model, and repeat this cycle to track the position of the target in real time.
[0085] Based on this, steps S501 to S503 can be implemented using the pre-trained model obtained by the above-mentioned pre-training. Specifically, first, the current feature vector is input into the pre-trained model for operations in different displacement directions, and the Gaussian correlation value after each displacement is calculated for subsequent analysis and processing. Next, the largest one is found from all the calculated Gaussian correlation values. The displacement position corresponding to the maximum value is the predicted position of the bird target in the next frame of the picture. This step is crucial for accurately tracking the target. Finally, the predicted position of the bird target in the next frame of the picture is taken as the center, and the surrounding area is selected as the target tracking area. The size of this target tracking area may need to be adjusted according to actual conditions to ensure that the target can be effectively covered and background interference is reduced. The target tracking area can be a rectangle with the center of the detection frame as the center point, and the width and height are 1.5 to 2.5 times of the detection frame respectively.
[0086] In one embodiment, if Figure 6 As shown, step S106 includes:
[0087] S601, when the center position of the target tracking area enters the anti-intrusion area, it is determined that a bird has abnormally entered the anti-intrusion area and an alarm message is triggered;
[0088] S602: A specific boundary box is flashed on the monitoring interface according to the alarm information, and the alarm information is pushed to the operation and maintenance terminal.
[0089] In this embodiment, when the center of the target tracking area enters the preset anti-intrusion area, the system automatically determines that a bird has abnormally entered the anti-intrusion area and immediately triggers a corresponding alarm message to alert the monitoring personnel. After receiving the alarm message, the system will flash a specific boundary box on the monitoring interface to attract the attention of the monitoring personnel. At the same time, the system will also push the alarm information to the operation and maintenance terminal in real time to ensure that the operation and maintenance personnel can obtain the alarm information in time so as to quickly take corresponding treatment measures.
[0090] The embodiment of the present invention further provides a bird intrusion monitoring device, which is used to execute any embodiment of the above-mentioned bird intrusion monitoring method. Figure 7 , Figure 7 It is a schematic block diagram of a bird intrusion monitoring device provided in an embodiment of the present invention.
[0091] like Figure 7 As shown, the bird intrusion monitoring device 700 includes: an area confirmation unit 701, a camera unit 702, a target detection unit 703, a feature extraction unit 704, a target tracking unit 705 and an alarm unit 706.
[0092] The area confirmation unit 701 is used to confirm the scene to be monitored and the anti-intrusion area in the scene to be monitored;
[0093] The camera unit 702 is used to obtain a current frame image of the scene to be monitored;
[0094] The target detection unit 703 is used to perform target detection processing on the current frame image to confirm the bird target in the current frame image;
[0095] The feature extraction unit 704 is used to extract and fuse the features of the bird target and the surrounding area of the bird target by using a kernelized correlation filter algorithm to obtain a current feature vector;
[0096] The target tracking unit 705 is used to evaluate the displacement direction of the current feature vector through the pre-trained model to obtain the target tracking area of the next frame image;
[0097] The alarm unit 706 is used to trigger a bird intrusion alarm when the target tracking area appears in the anti-intrusion area.
[0098] The device uses image detection technology to identify and locate bird targets that appear in the monitoring scene and anti-intrusion area, and uses image tracking technology to track the identified bird targets to determine their real-time location, and triggers an alarm signal when the bird targets enter the anti-intrusion area, thereby realizing real-time monitoring and early warning of bird invasions, effectively improving the safety management level of substations.
[0099] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described devices and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.
[0100] The bird invasion monitoring device can be implemented in the form of a computer program. Figure 8 Runs on the computer device shown.
[0101] See also Figure 8 , Figure 8 800 is a schematic block diagram of a computer device provided by an embodiment of the present invention. The computer device 800 is a server, which can be an independent server or a server cluster composed of multiple servers.
[0102] See also Figure 8 The computer device 800 includes a processor 802 , a memory and a network interface 805 connected via a system bus 801 , wherein the memory may include a non-volatile storage medium 803 and an internal memory 804 .
[0103] The non-volatile storage medium 803 may store an operating system 8031 and a computer program 8032. When the computer program 8032 is executed, the processor 802 may execute a bird intrusion monitoring method.
[0104] The processor 802 is used to provide computing and control capabilities and support the operation of the entire computer device 800.
[0105] The internal memory 804 provides an environment for the operation of the computer program 8032 in the non-volatile storage medium 803. When the computer program 8032 is executed by the processor 802, the processor 802 can execute the bird intrusion monitoring method.
[0106] The network interface 805 is used for network communication, such as providing data information transmission, etc. Those skilled in the art will understand that Figure 8 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device 800 to which the solution of the present invention is applied. The specific computer device 800 may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.
[0107] Those skilled in the art will understand that Figure 8The embodiments of the computer device shown in the figure do not constitute a limitation on the specific composition of the computer device. In other embodiments, the computer device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently. For example, in some embodiments, the computer device may only include a memory and a processor. In such embodiments, the structure and function of the memory and the processor are the same as those of the embodiment of the present invention. Figure 8 The embodiments shown are consistent and will not be described again here.
[0108] It should be understood that in the embodiment of the present invention, the processor 802 may be a central processing unit (CPU), and the processor 802 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.
[0109] In another embodiment of the present invention, a computer-readable storage medium is provided. The computer-readable storage medium may be a non-volatile computer-readable storage medium. The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, the bird intrusion monitoring method of the embodiment of the present invention is implemented.
[0110] The storage medium is a physical, non-transient storage medium, for example, it can be a USB flash drive, a mobile hard disk, a read-only memory (ROM), a magnetic disk or an optical disk, etc., which can store program codes.
[0111] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described equipment, devices and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.
[0112] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.
Claims
1. A method for monitoring bird invasion, characterized in that: include: Confirming the scene to be monitored and the anti-intrusion area in the scene to be monitored; Get the current frame image of the scene to be monitored; Perform target detection processing on the current frame image to confirm the bird target in the current frame image; Extracting and fusing the bird target and the surrounding area of the bird target by using a kernelized correlation filtering algorithm to obtain a current feature vector; The displacement direction of the current feature vector is evaluated through the pre-trained model to obtain the target tracking area of the next frame of the image; When the target tracking area appears in the anti-intrusion area, a bird intrusion alarm is triggered.
2. The bird invasion monitoring method according to claim 1, characterized in that: The performing target detection processing on the current frame image to confirm the bird target in the current frame image includes: Perform multiple downsampling processes on the current frame image to obtain multiple downsampled feature maps of different scales; The downsampled feature maps of multiple different scales are fused at the required scale to obtain a fused feature map of the required scale; Performing feature extraction on the fused feature map to confirm the bird target in the fused feature map; Generate super-resolution features for the bird target using a perceptual generative adversarial network, and superimpose the super-resolution features on the bird target on the fusion feature map to obtain a feature enhancement map; Feature extraction is performed on the feature enhancement image to confirm the bird target in the current frame image.
3. The bird invasion monitoring method according to claim 2, characterized in that: The step of fusing the multiple downsampled feature maps of different scales at a desired scale to obtain a fused feature map of the desired scale includes: Selecting a desired scale of the fused feature map; Down-sampling the down-sampled feature maps larger than the required scale, and up-sampling the down-sampled feature maps smaller than the required scale, to obtain multiple sampled feature maps of the same size as the required scale; Multiple sampling feature maps of the same size as the required scale are fused to obtain a fused feature map of the required scale.
4. The bird invasion monitoring method according to claim 1, characterized in that: The kernelized correlation filtering algorithm is used to extract and fuse the features of the bird target and the surrounding area of the bird target to obtain a current feature vector, including: Extracting the edge direction and intensity distribution of the bird target to obtain HOG features; Extracting pixel information of the surrounding area of the bird target to obtain grayscale features; The HOG feature and the grayscale feature are fused according to a preset weight to obtain a current feature vector.
5. The bird invasion monitoring method according to claim 4, characterized in that: The faster the moving speed of the bird target is, the higher the fusion weight of the HOG feature is.
6. The bird invasion monitoring method according to claim 1, characterized in that: The method of evaluating the displacement direction of the current feature vector by using the pre-trained model to obtain the target tracking area of the next frame of the image includes: Input the current feature vector into the pre-trained model to perform operations in different displacement directions, and calculate the Gaussian correlation value after each displacement; The displacement position with the maximum Gaussian correlation value is used to confirm the predicted position of the bird target in the next frame; The surrounding area of the predicted position of the bird target in the next frame is used as the target tracking area.
7. The bird invasion monitoring method according to claim 1, characterized in that: The triggering of a bird intrusion alarm when the target tracking area appears in the anti-intrusion area includes: When the center position of the target tracking area enters the anti-intrusion area, it is determined that a bird has abnormally entered the anti-intrusion area and an alarm message is triggered; A specific boundary box is used as a flashing prompt on the monitoring interface according to the alarm information, and the alarm information is pushed to the operation and maintenance terminal.
8. A bird invasion monitoring device, characterized in that: include: An area confirmation unit, used to confirm the scene to be monitored and the anti-intrusion area in the scene to be monitored; A camera unit, used to obtain a current frame image of the scene to be monitored; The target detection unit is used to perform target detection processing on the current frame image to confirm the bird target in the current frame image; A feature extraction unit, used for performing feature extraction and fusion processing on the bird target and the surrounding area of the bird target by using a kernelized correlation filtering algorithm to obtain a current feature vector; The target tracking unit is used to evaluate the displacement direction of the current feature vector through the pre-trained model to obtain the target tracking area of the next frame image; An alarm unit is used to trigger a bird intrusion alarm when the target tracking area appears in the anti-intrusion area.
9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the bird intrusion monitoring method according to any one of claims 1 to 7 is implemented.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, enables the processor to perform the bird invasion monitoring method according to any one of claims 1 to 7.