A base station detection model training method and device and electronic equipment

Abstract

Embodiments of the present application provide a base station detection model training method and device and electronic equipment. The method comprises: obtaining a first alarm image output by a base station detection model; removing invalid alarm images from the first alarm image; in the case of at least two valid alarm images, comparing the image similarity of the at least two valid alarm images to obtain a target alarm image; classifying the target alarm image according to a target object framed in the target alarm image to obtain at least one sample data set; and training the base station detection model based on the at least one sample data set. In this way, the image data is de-duplicated, removed and classified before being input into the base station detection model, so that the image data input into the base station detection model is single-object image data, thereby increasing the speed of improving the detection accuracy of the base station detection model for single objects.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer vision, and in particular to a base station detection model training method and device and electronic equipment. BACKGROUND

[0002] A patrol camera is arranged in a machine room of a base station to monitor the base station. The patrol camera sends the captured images to a server, and a base station detection model in the server detects the captured images, so as to determine whether a device change or a plant growth occurs in the machine room.

[0003] Since the patrol device is continuously operated and the captured images are not classified, a staff member usually inputs all images (for example, images of the base station, images of an air conditioner in the machine room, images of a machine room door, and images of plants in the machine room) captured by the patrol device into the base station detection model for detection, thereby causing a problem that the detection accuracy of the base station detection model for a single object (the base station, the air conditioner, the machine room door, or the plants) is improved at a slow speed. SUMMARY

[0004] Embodiments of the present application provide a base station detection model training method, device and electronic equipment to solve the problem that the detection accuracy of the base station detection model for a single object (the base station, the air conditioner, the machine room door, or the plants) is improved at a slow speed in the prior art.

[0005] In a first aspect, the embodiments of the present application provide a base station detection model training method, comprising:

[0006] obtaining a first alarm image output by a base station detection model, the first alarm image being an image with a similarity to a preset image less than a first preset value;

[0007] removing an invalid alarm image in the first alarm image, the first alarm image including a valid alarm image and the invalid alarm image, the valid alarm image representing an image with a real change, and the invalid alarm image representing an image without a real change;

[0008] in a case where there are at least two valid alarm images, performing image similarity comparison on the at least two valid alarm images to obtain a target alarm image, the target alarm image including part of the valid alarm images;

[0009] classifying the target alarm image according to a target object framed in the target alarm image to obtain at least one sample data set, the target object including a base station, an air conditioner and plants, and the sample data set being a collection of the target alarm images with the same target object;

[0010] Train the base station detection model based on the at least one sample data set.

[0011] Optionally, the base station detection model is a Yolov5 network model constructed based on a Yolov5 target detection algorithm.

[0012] Optionally, in the case of having at least two effective alarm images, the at least two effective alarm images are compared in image similarity to obtain a target alarm image, comprising:

[0013] Comparing the similarity of any two images in the at least two effective alarm images to obtain the similarity of each effective alarm image to other effective alarm images;

[0014] According to the similarity of each effective alarm image to other effective alarm images, the target alarm image is determined.

[0015] Optionally, the target alarm image is determined according to the similarity of each effective alarm image to other effective alarm images, comprising:

[0016] Comparing the similarity of each effective alarm image to other effective alarm images with a second preset value;

[0017] Removing other effective alarm images with a similarity greater than the second preset value to determine the target alarm image.

[0018] Optionally, before the target object in the target alarm image is classified according to the target object framed in the target alarm image to obtain at least one sample data set, further comprising:

[0019] Identifying the target object in the target alarm image;

[0020] Framing the target object in the target alarm image.

[0021] In a second aspect, the embodiments of the present application further provide a training device of a base station detection model, comprising:

[0022] An acquisition module is configured to acquire a first alarm image output by a base station detection model, wherein the first alarm image is an image with a similarity to a preset image less than a first preset value;

[0023] A removal module is configured to remove an invalid alarm image in the first alarm image, wherein the first alarm image comprises an effective alarm image and the invalid alarm image, the effective alarm image represents an image with real changes, and the invalid alarm image represents an image without real changes;

[0024] The comparison module is configured to compare image similarities of the at least two effective alarm images to obtain a target alarm image, the target alarm image including part of the effective alarm images.

[0025] The classification module is configured to classify the target alarm image according to a target object framed in the target alarm image to obtain at least one sample data set, the target object including a base station, an air conditioner, and a green plant, and the sample data set being a collection of the target alarm images with the same target object.

[0026] The training module is configured to train the base station detection model based on the at least one sample data set.

[0027] Optionally, the comparison module includes:

[0028] The comparison unit is configured to compare similarities of any two of the at least two effective alarm images to obtain a similarity of each of the effective alarm images to other effective alarm images.

[0029] The determination unit is configured to determine the target alarm image according to the similarity of each of the effective alarm images to other effective alarm images.

[0030] Optionally, the determination unit includes:

[0031] The comparison subunit is configured to compare the similarity of each of the effective alarm images to other effective alarm images with a second preset value.

[0032] The determination subunit is configured to remove other effective alarm images with a similarity greater than the second preset value to determine the target alarm image.

[0033] Optionally, the classification module includes:

[0034] The identification unit is configured to identify the target object in the target alarm image.

[0035] The framing unit is configured to frame the target object in the target alarm image.

[0036] In a third aspect, an electronic device is provided, which includes a processor, a memory, and a program or instructions stored in the memory and executable on the processor, and the program or instructions, when executed by the processor, implement the steps of the training method of the base station detection model according to the first aspect.

[0037] In the embodiment of the present application, a first alarm image output by a base station detection model is acquired; invalid alarm images in the first alarm image are removed; in the case of at least two valid alarm images, the at least two valid alarm images are compared in image similarity to obtain a target alarm image; the target alarm image is classified according to a target object framed in the target alarm image to obtain at least one sample data set; and the base station detection model is trained based on the at least one sample data set. In this way, before image data is input to the base station detection model, the images are de-duplicated, removed and classified, so that the image data input to the base station detection model is image data of a single object, thereby increasing the speed of improving the detection accuracy of the base station detection model for a single object. BRIEF DESCRIPTION OF DRAWINGS

[0038] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiments or prior art description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0039] Figure 1 is a flow diagram of a base station detection model training method provided by an embodiment of the present application;

[0040] Figure 2 is a structural diagram of a base station detection model training device provided by an embodiment of the present application;

[0041] Figure 3 is a structural diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0042] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0043] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the structures used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first", "second", etc. are generally of a kind and do not limit the number of objects, for example, the first object can be one or more. In addition, "and / or" in the specification and claims indicates at least one of the connected objects, and the character " / " generally indicates that the objects before and after are in an "or" relationship.

[0044] The training method, device and electronic equipment of the base station detection model provided by the embodiments of the present application will be described in detail below in combination with the drawings, specific embodiments and application scenarios.

[0045] Referring to Figure 1 , Figure 1 is a flowchart of a training method of a base station detection model provided by the embodiments of the present application, as Figure 1 shown, comprising the following steps:

[0046] Step 101, obtaining a first alarm image output by a base station detection model, the first alarm image being an image with a similarity to a preset image less than a first preset value;

[0047] In the embodiments of the present application, one or more cameras are arranged in the machine room of the base station, which is not limited further here.

[0048] It should be noted that the images (including images of the base station, images of the air conditioner in the machine room, images of the machine room door and images of the green plants in the machine room) captured by the cameras are converted into base64 streams by the designated path image of the inspection server, collected and stored in the file server, and the base64 streams are converted into image data for storage.

[0049] In the embodiments of the present application, the image data stored in the file server is detected by the base station detection model, and then the first alarm image output by the base station detection model is obtained.

[0050] It should be understood that the above-mentioned preset image can be any one of the images of the machine room, the normal operation of the base station, the normal closing of the machine room door, the normal operation of the air conditioner and the absence of green plants, which are artificially captured; or the preset image can be captured by the camera under the condition that the machine room is detected in the field and operates normally.

[0051] It should be noted that the image of the machine room photographed by the camera is input into the base station detection model, and the preset image is input into the base station detection model; the base station detection model compares the image of the machine room photographed by the camera with the preset image, and outputs the first alarm image. The first alarm image is an image of the machine room with a similarity less than a first preset value.

[0052] Further, the first preset value can be set according to experience, for example, 90%, or can be set to 95%, which is not limited further here.

[0053] For example, when the first preset value is 90%, the base station detection model detects the image of the air conditioner in the machine room photographed by the camera and the preset image, and obtains that the similarity of the image of the air conditioner in the machine room photographed by the camera and the preset image is 80%, and the base station detection model outputs the image of the air conditioner in the machine room photographed by the camera as the first alarm image.

[0054] In another example, when the first preset value is 90%, the base station detection model detects the image of the air conditioner in the machine room photographed by the camera and the preset image, and obtains that the similarity of the image of the air conditioner in the machine room photographed by the camera and the preset image is 100%, and the base station detection model determines that the image of the air conditioner in the machine room photographed by the camera is an image without alarm.

[0055] Step 102, removing invalid alarm images in the first alarm image, the first alarm image including valid alarm images and the invalid alarm images, the valid alarm images representing images with real changes, and the invalid alarm images representing images without real changes;

[0056] It should be understood that the valid alarm image is determined by the base station detection model as the first alarm image, and the position corresponding to the machine room of the valid alarm image has a real change relative to the position corresponding to the machine room of the preset image.

[0057] It should be understood that the invalid alarm image is determined by the base station detection model as the first alarm image, and the position corresponding to the machine room of the invalid alarm image has no real change relative to the position corresponding to the machine room of the preset image.

[0058] Further, the real change can be caused by different lightings when the preset image and the invalid alarm image are photographed, so that the base station detection model determines the invalid alarm image as the first alarm image.

[0059] Step 103, in the case of having at least two of the effective alarm images, performing image similarity comparison on the at least two effective alarm images to obtain a target alarm image, the target alarm image including part of the effective alarm images.

[0060] In the embodiments of the present application, the execution subject of the above steps can be a server, or other electronic devices with computing capability.

[0061] For example, in the case of having five of the effective alarm images, performing image similarity comparison on the five effective alarm images to obtain the target alarm image, the number of the target alarm images can be five or four, which is determined according to the specific similarity.

[0062] Optionally, in some embodiments, the target alarm image includes part of the effective alarm images, which can be understood as that the number of the target alarm images is equal to the number of the effective alarm images, or that the number of the target alarm images is less than the number of the effective alarm images.

[0063] Step 104, classifying the target alarm image according to the target object framed in the target alarm image to obtain at least one sample data set, the target object including a base station, an air conditioner and a green plant, and the sample data set being a collection of the target alarm images with the same target object.

[0064] The execution subject of the step 104 can be a server, or other electronic devices with computing capability.

[0065] It should be noted that the target object framed in the target alarm image can be a base station, an air conditioner in a machine room, a machine room door, a green plant in a machine room, etc., which is not limited further herein.

[0066] For example, if the target object framed in the target alarm image is an air conditioner in a machine room, the target alarm image is classified into a sample data set corresponding to the air conditioner; if the target object framed in the target alarm image is a base station, the target alarm image is classified into a sample data set corresponding to the base station; if the target object framed in the target alarm image is a green plant, the target alarm image is classified into a sample data set corresponding to the green plant.

[0067] In another embodiment, if the target object framed in the target alarm image includes both a base station and an air conditioner, the target alarm image can be classified into a sample data set corresponding to the base station, or into a sample data set corresponding to the air conditioner.

[0068] It should be noted that the images in the same sample data set have the same target object, but not only one target object in the image.

[0069] Step 105: Train the base station detection model based on the at least one sample dataset.

[0070] The entity executing step 105 can be a server or other electronic devices with computing capabilities.

[0071] It should be noted that the base station detection model is trained based on the at least one sample dataset. This should be understood as different base station detection models corresponding to different sample datasets.

[0072] For example, if the data is based on a sample dataset corresponding to an air conditioner, then the base station detection model corresponding to the air conditioner is trained; if the data is based on a sample dataset corresponding to a base station, then the base station detection model corresponding to the base station is trained.

[0073] In this embodiment, a first alarm image output by the base station detection model is obtained; invalid alarm images are removed from the first alarm image; if at least two valid alarm images are available, image similarity comparison is performed on the at least two valid alarm images to obtain a target alarm image; the target alarm image is classified according to the target object selected within the target alarm image to obtain at least one sample dataset; and the base station detection model is trained based on the at least one sample dataset. Thus, before inputting image data into the base station detection model, the images are deduplicated, removed, and classified, ensuring that the image data input into the base station detection model consists of images of a single object, thereby accelerating the improvement of the base station detection model's detection accuracy for single objects.

[0074] Optionally, in some embodiments, the base station detection model is a YOLOv5 network model built based on the YOLOv5 target detection algorithm.

[0075] It should be noted that the Yolov5 network model consists of a backbone network and a feature fusion network. The backbone network is CSPDarkNet, and the feature fusion network is PANNet. Its working principle is as follows: The backbone network processes the image by performing operations such as convolution to extract four feature maps of different depths. The feature fusion network receives the four feature maps output by the backbone network and performs convolution, concatenation, and other operations to fuse the feature maps. Finally, it outputs the discrimination criteria.

[0076] The training process of the Yolov5 network model is as follows: First, set the initial detection categories, image size, learning rate, optimizer, and other network parameters.

[0077] Next, each batch of input pictures is subjected to Mosaic data enhancement, that is, random scaling, random cropping, and random arrangement for splicing; and finally, adaptive anchor box calculation is performed, and initial length-width anchor boxes are set for different data sets. In network training, the network outputs a prediction box based on the initial anchor box, and then compares it with the real box, calculates the difference between the two, and then updates the network parameters in reverse.

[0078] The prediction process of the Yolov5 network model is as follows: the picture to be detected is scaled to the image size set during training, and is input into the Yolov5 network model to obtain an initial prediction box. After non-maximum suppression processing of the initial prediction box, a target box is output.

[0079] In the embodiments of the present application, by using the improved Yolov5 network model, the detection accuracy of images captured by cameras in the machine room can be improved.

[0080] Optionally, in some embodiments, the image similarity comparison of the at least two effective alarm images to obtain the target alarm image comprises:

[0081] The similarity of any two images in the at least two effective alarm images is compared to obtain the similarity of each effective alarm image to other effective alarm images.

[0082] For example, if the at least two effective alarm images are A, B, C, D, and E five effective alarm images; first, similarity comparison is performed between A and B, C, D, and E respectively to obtain four similarity values; then, similarity comparison is performed between B and C, D, and E respectively to obtain three similarity values; then, similarity comparison is performed between C and D and E respectively to obtain two similarity values; and finally, similarity comparison is performed between D and E to obtain one similarity value.

[0083] In another example, if the at least two effective alarm images are A, B, C, D, E, and F five effective alarm images; first, similarity comparison is performed between A and B, C, D, E, and F respectively to obtain five similarity values; then, similarity comparison is performed between B and C, D, E, and F respectively to obtain four similarity values; then, similarity comparison is performed between C and D, E, and F respectively to obtain three similarity values; then, similarity comparison is performed between D and E and F to obtain two similarity values; and finally, similarity comparison is performed between E and F to obtain one similarity value.

[0084] In another embodiment, if the effective alarm image B is not determined as the target alarm image, then there is no need to compare the similarity of B with C, D and E respectively.

[0085] The target alarm image is determined according to the similarity of each of the effective alarm images with other effective alarm images.

[0086] For example, if the at least two effective alarm images are A, B, C, D and E, then the similarity of A with B, C, D and E is compared respectively to obtain four similarity values, and then whether B, C, D and E are target alarm images is determined according to the similarity values corresponding to B, C, D and E respectively.

[0087] Optionally, in some embodiments, the determining the target alarm image according to the similarity of each of the effective alarm images with other effective alarm images comprises:

[0088] Comparing the similarity of each of the effective alarm images with other effective alarm images with a second preset value;

[0089] Removing other effective alarm images with similarity greater than the second preset value to determine the target alarm image.

[0090] It should be understood that the second preset value is set according to experience, which can be 75% or 70%, and is not limited further herein.

[0091] In the embodiments of the present application, if the at least two effective alarm images are A, B, C, D and E, then the similarity of A with B, C, D and E is compared respectively to obtain four similarity values, and the four similarities are 60%, 70%, 80% and 90% respectively and correspond to B, C, D and E respectively; if the second preset value is 75%, then D and E are removed, and A, B and C are determined as target alarm images; further, B is taken as a reference to calculate the similarity with C, if the similarity of B with C is 80%, then A and B are determined as target alarm images.

[0092] Optionally, in some embodiments, before the target object in the target alarm image is framed and the target alarm image is classified to obtain at least one sample data set, the method further comprises:

[0093] Identifying the target object in the target alarm image;

[0094] box the target object in the target alarm image.

[0095] In the embodiments of the present application, the execution subject of the above steps can be a server, or other electronic devices with computing power.

[0096] Optionally, in some embodiments, the way of identifying the target object in the target alarm image can be a Two-Stage algorithm represented by Faster R-CNN, or a One-Stage algorithm represented by SSD and YOLO.

[0097] It should be noted that after being identified, the target object can be boxed by using a minimum rectangular frame.

[0098] Referring to Figure 2 , Figure 2 is a structural schematic diagram of a base station detection model training device provided by an embodiment of the present application, as shown in Figure 2 The base station detection model training device includes:

[0099] The acquisition module 201 is configured to acquire a first alarm image output by a base station detection model, the first alarm image being an image with a similarity to a preset image less than a first preset value;

[0100] The removal module 202 is configured to remove invalid alarm images in the first alarm image, the first alarm image including valid alarm images and the invalid alarm images, the valid alarm images representing images with real changes, and the invalid alarm images representing images without real changes;

[0101] The comparison module 203 is configured to, in a case where there are at least two valid alarm images, compare the at least two valid alarm images in terms of image similarity, to obtain a target alarm image, the target alarm image including part of the valid alarm images;

[0102] The classification module 204 is configured to classify the target alarm image according to a target object boxed in the target alarm image, to obtain at least one sample data set, the target object including a base station, an air conditioner, and a green plant, and the sample data set being a collection of the target alarm images with the same target object.

[0103] The training module 205 is configured to train the base station detection model based on the at least one sample data set.

[0104] Optionally, in some embodiments, the comparison module 203 includes:

[0105] The comparison unit is configured to compare the similarity of any two of the at least two effective alarm images to obtain the similarity of each of the effective alarm images to other effective alarm images.

[0106] The determination unit is configured to determine the target alarm image according to the similarity of each of the effective alarm images to other effective alarm images.

[0107] Optionally, in some embodiments, the determination unit comprises:

[0108] The comparison subunit is configured to compare the similarity of each of the effective alarm images to other effective alarm images with a second preset value.

[0109] The determination subunit is configured to remove other effective alarm images with a similarity greater than the second preset value to determine the target alarm image.

[0110] Optionally, the classification module comprises:

[0111] The identification unit is configured to identify the target object in the target alarm image.

[0112] The frame selection unit is configured to frame the target object in the target alarm image.

[0113] The training device of the base station detection model in the embodiments of the present application can be a device, or a component, an integrated circuit, or a chip in a terminal. The device can be a mobile electronic device, or a non-mobile electronic device. Exemplarily, the mobile electronic device can be a mobile phone, a tablet computer, a notebook computer, a palm computer, a vehicle-mounted electronic device, a wearable device, an ultra-mobile personal computer (UMPC), a netbook, or a personal digital assistant (PDA), etc., and the non-mobile electronic device can be a server, a network attached storage (NAS), a personal computer (PC), a television (TV), a teller machine, or a self-service machine, etc., and the embodiments of the present application are not limited in this regard.

[0114] The training device of the base station detection model in the embodiments of the present application can be a device with an operating system. The operating system can be an Android operating system, an ios operating system, or other possible operating systems, and the embodiments of the present application are not limited in this regard.

[0115] The training device of the base station detection model provided in the embodiment of the application can realize Figure 1 The various processes implemented by the training device of the base station detection model in the method embodiment are not repeated here.

[0116] Referring to Figure 3 , Figure 3 is a structural schematic diagram of an electronic device provided in the embodiment of the application, as shown in Figure 3 The electronic device includes a processor 301, a memory 302, a program or instruction 3021 stored in the memory 302 and executable on the processor 301, the program or instruction being executed by the processor 301 to implement the various processes of the base station detection model training method embodiment and achieve the same technical effects, and details are not repeated here.

[0117] It should be noted that the electronic device in the embodiment of the application includes the mobile electronic device and the non-mobile electronic device described above.

[0118] It should be noted that in this document, the terms "comprising", "containing" or any other variant thereof are intended to cover non-exclusive inclusion, so that the processes, methods, articles or devices including a series of elements not only include those elements, but also include other elements not explicitly listed or inherent to such processes, methods, articles or devices. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of another identical element in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the method and device in the present application is not limited to performing functions in the order discussed, but can also include performing functions in a substantially simultaneous manner or in the opposite order, for example, the described method can be performed in an order different from that described, and various steps can also be added, omitted or combined. In addition, the features described with reference to some examples can be combined in other examples.

[0119] The embodiments of the application are described above in combination with the drawings, but the application is not limited to the above specific embodiments, and the above specific embodiments are only illustrative, not restrictive, and those skilled in the art can make many forms under the inspiration of the application without departing from the scope of the application and the scope protected by the claims.

Claims

1. A training method for a base station detection model, characterized in that, The method comprises: obtaining a first alarm image output by a base station detection model, the first alarm image being an image with a similarity to a preset image less than a first preset value; removing an invalid alarm image in the first alarm image, the first alarm image comprising a valid alarm image and the invalid alarm image, the valid alarm image representing an image with a real change, and the invalid alarm image representing an image without a real change; in a case where there are at least two valid alarm images, comparing the at least two valid alarm images in terms of image similarity to obtain a target alarm image, the target alarm image comprising part of the valid alarm images; classifying the target alarm image according to a target object framed in the target alarm image to obtain at least one sample data set, the target object comprising a base station, an air conditioner, and a green plant, and the sample data set being a collection of the target alarm images with the same target object; training the base station detection model based on the at least one sample data set; before the classifying the target alarm image according to the target object framed in the target alarm image to obtain at least one sample data set, the method further comprises: identifying the target object in the target alarm image; and framing the target object in the target alarm image. 2.The method of training a base station detection model according to claim 1, wherein, The base station detection model is a Yolov5 network model constructed based on a Yolov5 target detection algorithm. 3.The method of claim 1, wherein, The comparing the at least two valid alarm images in terms of image similarity to obtain a target alarm image comprises: comparing any two images in the at least two valid alarm images in terms of similarity to obtain the similarity of each valid alarm image to other valid alarm images; and determining the target alarm image according to the similarity of each valid alarm image to other valid alarm images. 4.The method of claim 3, wherein, The determining the target alarm image according to the similarity of each valid alarm image to other valid alarm images comprises: comparing the similarity of each valid alarm image to other valid alarm images with a second preset value; and removing other valid alarm images with a similarity greater than the second preset value to determine the target alarm image.

5. A training device for a base station detection model, characterized in that, The device comprises: an obtaining module configured to obtain a first alarm image output by a base station detection model, the first alarm image being an image with a similarity to a preset image less than a first preset value; a removing module configured to remove an invalid alarm image in the first alarm image, the first alarm image comprising a valid alarm image and the invalid alarm image, the valid alarm image representing an image with a real change, and the invalid alarm image representing an image without a real change; a comparing module configured to, in a case where there are at least two valid alarm images, compare the at least two valid alarm images in terms of image similarity to obtain a target alarm image, the target alarm image comprising part of the valid alarm images; and a classifying module configured to classify the target alarm image according to a target object framed in the target alarm image to obtain at least one sample data set, the target object comprising a base station, an air conditioner, and a green plant, and the sample data set being a collection of the target alarm images with the same target object. The classification module is configured to classify the target alarm image according to a target object framed in the target alarm image, to obtain at least one sample data set, wherein the target object includes a base station, an air conditioner, and a green plant, and the sample data set is a set of the target alarm images with the same target object. The training module is configured to train the base station detection model based on the at least one sample data set. The classification module includes: An identification unit configured to identify the target object in the target alarm image. A framing unit configured to frame the target object in the target alarm image.

6. The training apparatus of a base station detection model according to claim 5, wherein, The comparison module includes: A comparison unit configured to compare the similarity of any two images in the at least two effective alarm images, to obtain the similarity of each effective alarm image with other effective alarm images. A determination unit configured to determine the target alarm image according to the similarity of each effective alarm image with other effective alarm images.

7. The training apparatus of a base station detection model according to claim 6, wherein, The determination unit includes: A comparison subunit configured to compare the similarity of each effective alarm image with other effective alarm images with a second preset value. A determination subunit configured to remove other effective alarm images with a similarity greater than the second preset value, and to determine the target alarm image.

8. An electronic device, comprising: A processor, a memory, and a program or instructions stored on the memory and executable on the processor, wherein the program or instructions are executed by the processor to implement the steps of the base station detection model training method according to any one of claims 1 to 4.

Images