A densely arranged power equipment detection method, system, device and storage medium

Abstract

The present invention discloses a method, system, device and storage medium for detecting densely arranged power equipment, comprising: acquiring visible light images and infrared imaging images of photovoltaic panels; marking single devices in the visible light images by means of horizontal rectangular frames; determining a center point set and an outer contour key point set of the power equipment according to the marked visible light images and the infrared imaging images of the photovoltaic panels; clustering each center point in the center point set of the power equipment, clustering each outer contour key point in the outer contour key point set of the power equipment, and then using the clustering results of each center point in the center point set of the power equipment and the clustering results of each outer contour key point in the outer contour key point set of the power equipment as the detection results of the status of densely arranged power equipment. The method, system, device and storage medium can detect densely arranged power equipment with excellent detection effect.

Description

Technical Field

[0001] The present invention belongs to the technical field of power equipment detection, and relates to a method, system, equipment and storage medium for densely arranged power equipment detection. Background Art

[0002] my country is currently at a critical stage of clean energy transformation. Clean energy power generation equipment represented by solar energy and wind energy has been deployed in large quantities. The large-scale access of new power equipment has put forward new requirements for the operation and maintenance of equipment. At the same time, considering the uneven distribution of new energy resources in my country, new energy equipment is generally distributed in a scattered manner, and a large number of them are built in areas where outdoor operation and maintenance personnel are difficult to reach. Research on new detection and status monitoring technologies for power equipment has important practical application value.

[0003] With the rapid development of artificial intelligence technology, target detection technology based on visible light and infrared images has made great progress. Its accuracy and adaptability have been significantly improved, and a series of applications have been achieved in the field of power equipment detection. Traditional deep learning-based target detection technology includes two categories: single-stage detection network and two-stage detection network. The design idea of ​​the single-stage detection network is simple and direct. A series of preset center points are evenly set on the original image to judge the probability of the device center and the size of the device in the neighboring area of ​​each center point. Represented by algorithms such as SSD, YOLO series, and RetinaNet, the single-stage detection network has fast detection speed and low computational requirements. It has been widely used in edge computing with high speed requirements and relatively scarce computing power. The two-stage network is more complex in structure. After generating dense candidate boxes on the image to be detected, a binary classifier is first used to screen valid candidate boxes, and then the candidate boxes are further fine-tuned and category detected. The two-stage network is oriented to high-computing power scenarios. Through the organic combination of network structure and components, it has obvious advantages in detection accuracy.

[0004] However, with the continuous expansion of detection scenarios and the complexity of detection targets, the single horizontal rectangle detection method has obvious defects, which limits the further improvement of detection effect. Summary of the invention

[0005] The purpose of the present invention is to overcome the shortcomings of the above-mentioned prior art and provide a method, system, device and storage medium for detecting densely arranged power equipment. The method, system, device and storage medium can detect densely arranged power equipment with excellent detection effect.

[0006] In order to achieve the above object, the present invention adopts the following technical scheme:

[0007] In one aspect, the present invention provides a method for detecting the status of densely arranged power equipment, characterized in that it includes:

[0008] Obtain visible light images and infrared imaging images of photovoltaic panels;

[0009] Annotating a single device in the visible light image by a horizontal rectangular frame;

[0010] Determine a center point set and an outer contour key point set of the power equipment according to the annotated visible light image and the infrared imaging image of the photovoltaic panel;

[0011] Cluster each center point in the center point set of the electric equipment, cluster each outer contour key point in the outer contour key point set of the electric equipment, and then use the clustering results of each center point in the center point set of the electric equipment and the clustering results of each outer contour key point in the outer contour key point set of the electric equipment as the detection results of the densely arranged electric equipment status.

[0012] The densely arranged power equipment status detection method of the present invention is further improved in that:

[0013] The specific process of obtaining the visible light image and infrared imaging image of the photovoltaic panel is as follows:

[0014] Determine the flight trajectory of the drone in the area to be photographed, determine the photographing order of each photovoltaic panel, and then use the drone to photograph each photovoltaic panel to obtain visible light images and infrared images of the photovoltaic panels.

[0015] The specific process of labeling the single device in the visible light image with the horizontal rectangular frame is: using the horizontal rectangular frame to label the single device in the visible light image using the free rectangular frame four-point labeling method.

[0016] The specific process of determining the center point set and the outer contour key point set of the power equipment according to the annotated visible light image and the infrared imaging image of the photovoltaic panel is as follows:

[0017] Get single device t i The key point set of visible light images is i Each key point in the visible light image key point set is divided into the center point pc 2,i And the key points of the outer contour po 2,i ;

[0018] Get single device t i The infrared imaging image key point set of a single device t i Each key point in the infrared imaging image key point set is divided into the center point pc of the device 3,i With the outer contour key point po 3,i ;

[0019] According to the center point pc 2,i , key points of outer contour po2,i 、Center point pc 3,i And the key points of the outer contour po 3,i Determine the center point set and the key point aggregation of the outer contour of the power equipment.

[0020] The key point set of the visible light image of a single device is obtained, and each key point in the key point set of the visible light image of the single device is divided into a center point pc 2,i And the key points of the outer contour po 2,i The specific operations are:

[0021] Intercept single device i The corresponding visible light image v i and feature layer f 1,i , after scaling, it is input into the trained key point detection network for feature extraction, and the sub-feature layer is obtained

[0022] Sub-feature layer With feature layer f 1,i The fusion result is then used to extract the key points of a single device to obtain the single device t i The key point set p of the visible light image i ;

[0023] The key point set p of the visible light image i Each key point in is divided into the center point pc 2,i And the key points of the outer contour po 2,i .

[0024] The infrared imaging image key point set of a single device is obtained, and each key point in the infrared imaging image key point set of a single device is divided into a center point pc of the single device. 3,i With the outer contour key point po 3,i The specific process is:

[0025] The single device area in the infrared imaging image is intercepted and input into the trained infrared key point detection model to obtain the sub-feature layer.

[0026] Sub-feature layer With feature layer f 1,i The fusion is performed, and the key points of a single device are extracted from the fusion result to obtain the single device t i The key point set of infrared imaging images;

[0027] Single device t i Each key point in the infrared imaging image key point set is divided into a single device t i The center point pc 3,i With the outer contour key point po 3,i .

[0028] According to the center point pc 2,i , key points of outer contour po 2,i 、Center point pc 3,i And the key points of the outer contour po 3,i The specific process of determining the center point set and the key point aggregation of the outer contour of the power equipment is as follows:

[0029] Sub-feature layer With sub-feature layer Splice and restore to the original size using 1x1 convolution to obtain the fused sub-feature layer

[0030] Fusion sub-feature layer Extract key points of a single device and divide the extracted key points into single device t i The center point pc 4,i And the key points of the outer contour po 4,i ;

[0031] Using the central point pc 2,i 、Center point pc 3,i and center point pc 4,i Construct the center point set of the power equipment; use the key points of the outer contour po 2,i , key points of outer contour po 3,i And the key points of the outer contour po 4,i Construct a set of key points of the outer contour of the power equipment.

[0032] In a second aspect of the present invention, the present invention provides a densely arranged power equipment state detection system, comprising:

[0033] An acquisition module, used to acquire visible light images and infrared imaging images of photovoltaic panels;

[0034] A labeling module, used for labeling a single device in the visible light image by means of a horizontal rectangular frame;

[0035] A determination module, used to determine a set of center points and a set of key points of an outer contour of the power equipment according to the annotated visible light image and the infrared imaging image of the photovoltaic panel;

[0036] A clustering module is used to cluster each center point in the center point set of the power equipment, cluster each outer contour key point in the outer contour key point set of the power equipment, and then use the clustering results of each center point in the center point set of the power equipment and the clustering results of each outer contour key point in the outer contour key point set of the power equipment as the detection results of the densely arranged power equipment status.

[0037] In a third aspect of the present invention, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method for detecting the status of densely arranged power equipment when executing the computer program.

[0038] In a fourth aspect of the present invention, 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 steps of the densely arranged power equipment status detection method are implemented.

[0039] The present invention has the following beneficial effects:

[0040] During specific operation, the densely arranged power equipment detection method, system, equipment and storage medium described in the present invention mark single equipment in the visible light image through a horizontal rectangular frame to adapt to densely arranged, irregularly shaped power equipment with obvious key feature points. The center point set and the outer contour key point set of the power equipment are determined based on the marked visible light image and the infrared imaging image of the photovoltaic panel. That is, the visible light and infrared image fusion method is used to solve the detection missing problem caused by the complexity of the application scenarios of the power equipment. In addition, the detection results of the key points are finally clustered to improve the detection effect and reduce the missed detection phenomenon. The operation is simple and convenient. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:

[0042] Figure 1 is a flow chart of the present invention;

[0043] Figure 2 It is a schematic diagram of the structure of the present invention. DETAILED DESCRIPTION

[0044] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only 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 should fall within the scope of protection of the present invention.

[0045] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0046] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0047] Embodiment 1

[0048] refer to Figure 1 The densely arranged power equipment state detection method of the present invention comprises the following steps:

[0049] 1) Use drones to capture visible light and infrared images of photovoltaic panels;

[0050] Specifically, the flight trajectory of the drone in the area to be photographed is first determined, the photographing order of each photovoltaic panel is determined, and then the drone is used to photograph each photovoltaic panel to obtain visible light images and infrared images of the photovoltaic panels.

[0051] 2) Use a horizontal rectangular frame to mark the target device in the visible light image using a free rectangular frame four-point marking method, where the four vertices of a photovoltaic panel are marked in the clockwise order of upper left, upper right, lower right, and lower left in the image to form a rectangular marking;

[0052] 3) Input the visible light image into the trained Cascade R-CNN network for power equipment detection to obtain the prediction regression frame b1, where:

[0053] f1=F1(v) (1)

[0054] b1=G1(f1) (2)

[0055] Among them, F1 is the feature extraction network in the Cascade R-CNN network, G1 is the regression prediction network in the Cascade R-CNN network, and f1 is the feature map extracted by the feature extraction network.

[0056] 4) Obtain the key point set p of the visible light image of a single device i

[0057] Specifically, for a single device t in the detection area marked by the horizontal rectangular box i , the single device t i The corresponding visible light image v i and feature layer f 1,i The image is intercepted and scaled to a fixed size. The trained key point detection network is then used to extract features from the scaled visible light image to obtain the sub-feature layer. Then the sub-feature layer With feature layer f 1,i The fusion is then performed, and then the key points of a single device are extracted from the fused result to obtain the key point set p of the visible light image of the single device i ,Right now

[0058] f 2,i =F2(v i ) (3)

[0059] p 2,i =G2(f 1,i ,f 2,i ) (4)

[0060] 5) Set the key point set p of the visible light image i Each key point in is divided into the center point pc 2,i And the key points of the outer contour po 2,i ;

[0061] Specifically, according to the equipment arrangement and the corresponding power equipment structural characteristics, the visible light image key point set p i Each key point in is divided into the center point pc 2,i And the key points of the outer contour po 2,i Among them, the center point pc is the center point of the power equipment in the current visible light image, and the key points of the outer contour are the key marking points on the equipment shell determined according to the structural characteristics of the power equipment, that is,

[0062] pc 2,i ,po 2,i =G2(f 1,i ,f 2,i ) (5)

[0063] 6) Capture the single device area in the infrared imaging image and then feed it into the trained infrared key point detection model to obtain the sub-feature layer Then the sub-feature layer With feature layer f 1,i The fusion is performed, and then the key points of a single device are extracted from the fusion result to obtain the key point set of the infrared imaging image of a single device. Then, each key point in the key point set of the infrared imaging image of a single device is divided into the center point pc of the device. 3,i With the outer contour key point po3,i ,Right now

[0064] f 3,i =F3(v i ) (6)

[0065] pc 3,i ,po 3,i =G3(f 1,i ,f 3,i ) (7)

[0066] 7) Sub-feature layer With sub-feature layer Concatenate them and then use 1x1 convolution to restore them to their original size to get the fused sub-feature layer Finally, the fusion sub-feature layer Extract key points of a single device, and then divide the extracted key points of the single device into the center point pc of the single device 4,i And the key points of the outer contour po 4,i ,Right now

[0067] pc 4,i ,po 4,i =G4(f 1,i ,f 2,i ,f 3,i ,f 4,i ) (8)

[0068] 8) Get the center point set PC = pc2∩pc3∩pc4 and the outer contour key point set PO = po2∩po3∩po4, set the neighborhood range ∈ and the core point threshold The DBSCAN density clustering method is used based on the neighborhood range ∈ and the core point threshold Traverse the set PC and the set PO, calculate the number of associated points within the neighborhood range ∈ of each key point, and classify them to determine the core points of the set PC and the set PO;

[0069] Taking the center point set as an example, for the center point x∈PC, set N ∈ (x) is the neighborhood range ∈ of x, ρ(x) is the density of x, then

[0070] N ∈ (x)={y∈PC:d(y,x)<∈} (9)

[0071] ρ(x)=|N ∈ (x)| (10)

[0072] Among them, ρ(x) is an integer. Then the center point x is a core point, denoted as PC c is the set of all core points in PC.nc =PC\PC c It is the set of all non-core points in PC.

[0073] Taking the center point set as an example, when x∈PC nc ,and satisfy

[0074] y∈N ∈ (x)∩PC c (11)

[0075] Then there is a core point within the neighborhood range ∈ of x, x is a boundary point of the set PC, and all core points that contain each other or can be connected through boundary points and their boundary points constitute the same set.

[0076] 9) According to the core points of the set PC and the set PO, the key points of a single device are adaptively clustered, and the clustering results are used as densely arranged power equipment status detection.

[0077] Specifically, for power equipment J, considering the appearance and structural characteristics of power equipment J, the reference neighborhood range ∈ J And reference core point threshold According to the actual width w and actual height h of the image, calculate its actual neighborhood range Right now

[0078]

[0079] For a set of core points {x k}∈PC c , when the core points {x k Each point in the set is within the neighborhood of every other point in the set. When {x k} is a group of homogeneous points, and the homogeneous centroid c and centroid threshold ∈ c , if satisfied

[0080]

[0081]

[0082] Then {x k} is an alternative homogeneous point, and the homogeneous centroid c is used instead of {x k} as the prediction result.

[0083] For example, according to the arrangement characteristics of photovoltaic panels, the key points of photovoltaic panels are divided into: photovoltaic panel center point, photovoltaic panel edge point and photovoltaic panel splicing point. Among them, the photovoltaic panel splicing point is the common splicing point of four closely arranged photovoltaic panels, and the remaining photovoltaic panel vertices are all photovoltaic panel edge points.

[0084] Taking the photovoltaic panel splicing point set as an example, the key point adaptive clustering method is applied. First, for the photovoltaic panel equipment, set the reference neighborhood range ∈ J =0.02, reference core point threshold Then, based on the actual width w = 512 and the actual height h = 512 of the image, calculate its actual neighborhood range For all core points, calculate the homogeneous centroid c and cluster them, and the centroid threshold ∈ c =0.01. The final output result of the photovoltaic panel splicing point is obtained.

[0085] Embodiment 3

[0086] The densely arranged power equipment state detection system of the present invention comprises:

[0087] Acquisition module 1, used to acquire visible light images and infrared imaging images of photovoltaic panels;

[0088] A labeling module 2, used for labeling a single device in the visible light image by means of a horizontal rectangular frame;

[0089] A determination module 3 is used to determine a set of center points and a set of key points of an outer contour of the power equipment according to the annotated visible light image and the infrared imaging image of the photovoltaic panel;

[0090] The clustering module 4 is used to cluster the center points in the center point set of the power equipment, cluster the outer contour key points in the outer contour key point set of the power equipment, and then use the clustering results of the center points in the center point set of the power equipment and the clustering results of the outer contour key points in the outer contour key point set of the power equipment as the detection results of the densely arranged power equipment status.

[0091] Embodiment 3

[0092] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the densely arranged power equipment state detection method are implemented. The memory may include a memory, such as a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk memory, etc. The processor, the network interface, and the memory are interconnected through an internal bus. The internal bus may be an industrial standard architecture bus, a peripheral component interconnection standard bus, an extended industrial standard structure bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The memory is used to store programs. Specifically, the program may include a program code, and the program code includes computer operation instructions. The memory may include a memory and a non-volatile memory, and provide instructions and data to the processor.

[0093] Embodiment 4

[0094] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the densely arranged power device state detection method are implemented. Specifically, the computer-readable storage medium includes, but is not limited to, for example, a volatile memory and / or a non-volatile memory. The volatile memory may include a random access memory (RAM) and / or a cache memory (cache), etc. The non-volatile memory may include a read-only memory (ROM), a hard disk, a flash memory, an optical disk, a magnetic disk, etc.

[0095] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0096] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0097] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0098] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0099] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for detecting the status of densely arranged power equipment, characterized in that: include: Obtain visible light images and infrared imaging images of photovoltaic panels; Annotating a single device in the visible light image by a horizontal rectangular frame; Determine a center point set and an outer contour key point set of the power equipment according to the annotated visible light image and the infrared imaging image of the photovoltaic panel; Clustering each center point in the center point set of the electric equipment, clustering each outer contour key point in the outer contour key point set of the electric equipment, and then using the clustering results of each center point in the center point set of the electric equipment and the clustering results of each outer contour key point in the outer contour key point set of the electric equipment as the detection result of the densely arranged electric equipment state; The specific process of determining the center point set and the outer contour key point set of the power equipment according to the annotated visible light image and the infrared imaging image of the photovoltaic panel is as follows: Get single device t i The key point set of visible light images is i Each key point in the visible light image key point set is divided into the center point pc 2，i And the key points of the outer contour po 2，i ; Get single device t i The infrared imaging image key point set of a single device t i Each key point in the infrared imaging image key point set is divided into the center point pc of the device 3，i With the outer contour key point po 3，i ; According to the center point pc 2，i , key points of outer contour po 2，i 、Center point pc 3，i And the key points of the outer contour po 3，i Determine the center point set and the key point aggregation of the outer contour of the power equipment; The key point set of the visible light image of a single device is obtained, and each key point in the key point set of the visible light image of the single device is divided into a center point pc 2，i And the key points of the outer contour po 2，i The specific operations are: Intercept single device i The corresponding visible light image v i and feature layer f 1，i , after scaling, it is input into the trained key point detection network for feature extraction, and the sub-feature layer is obtained Sub-feature layer With feature layer f 1，i The fusion result is then used to extract the key points of a single device to obtain the single device t i The key point set p of the visible light image i ; The key point set p of the visible light image i Each key point in is divided into a center point pC 2，i And the key points of the outer contour po 2，i ; The infrared imaging image key point set of a single device is obtained, and each key point in the infrared imaging image key point set of a single device is divided into a center point pc of the single device. 3，i With the outer contour key point po 3，i The specific process is: The single device area in the infrared imaging image is intercepted and input into the trained infrared key point detection model to obtain the sub-feature layer. Sub-feature layer With feature layer f 1，i The fusion result is then used to extract the key points of a single device to obtain the single device t i The key point set of infrared imaging images; Set a single device i Each key point in the infrared imaging image key point set is divided into a single device t i The center point pc 3，i With the outer contour key point po 3，i ; According to the center point pc 2，i , key points of outer contour po 2，i 、Center point pc 3，i And the key points of the outer contour po 3，i The specific process of determining the center point set and the key point aggregation of the outer contour of the power equipment is as follows: Sub-feature layer With sub-feature layer Splice and restore to the original size using 1xl convolution to get the fusion sub-feature layer Fusion sub-feature layer Extract key points of a single device and divide the extracted key points into single device t i The center point pc 4，i And the key points of the outer contour po 4，i ; Using the central point pc 2，i 、Center point pc 3，i and center point pc 4，i Construct a collection of central points for power equipment; Using the outer contour key points po 2，i , key points of outer contour po 3，i And the key points of the outer contour po 4，i Construct a set of key points of the outer contour of the power equipment.

2. The method for detecting the status of densely arranged power equipment according to claim 1, characterized in that: The specific process of obtaining the visible light image and infrared imaging image of the photovoltaic panel is as follows: Determine the flight trajectory of the drone in the area to be photographed, determine the photographing order of each photovoltaic panel, and then use the drone to photograph each photovoltaic panel to obtain visible light images and infrared images of the photovoltaic panels.

3. The method for detecting the status of densely arranged power equipment according to claim 1, characterized in that: The specific process of labeling the single device in the visible light image with the horizontal rectangular frame is: using the horizontal rectangular frame to label the single device in the visible light image using the free rectangular frame four-point labeling method.

4. A densely arranged power equipment status detection system, characterized in that: include: An acquisition module (1) is used to acquire visible light images and infrared imaging images of the photovoltaic panel; A labeling module (2), used for labeling a single device in the visible light image by means of a horizontal rectangular frame; A determination module (3) is used to determine a set of center points and a set of key points of an outer contour of the power equipment based on the annotated visible light image and the infrared imaging image of the photovoltaic panel; A clustering module (4) is used to cluster each center point in the center point set of the electric equipment, cluster each outer contour key point in the outer contour key point set of the electric equipment, and then use the clustering results of each center point in the center point set of the electric equipment and the clustering results of each outer contour key point in the outer contour key point set of the electric equipment as the detection result of the densely arranged electric equipment state; The specific process of determining the center point set and the outer contour key point set of the power equipment according to the annotated visible light image and the infrared imaging image of the photovoltaic panel is as follows: Get single device t i The key point set of visible light images is i Each key point in the visible light image key point set is divided into the center point pc 2，i And the key points of the outer contour po 2，i ; Get single device t i The infrared imaging image key point set of a single device t i Each key point in the infrared imaging image key point set is divided into the center point pc of the device 3，i With the outer contour key point po 3，i ; According to the center point pc 2，i , key points of outer contour po 2，i 、Center point pc 3，i And the key points of the outer contour po 3，i Determine the center point set and the key point aggregation of the outer contour of the power equipment; The key point set of the visible light image of a single device is obtained, and each key point in the key point set of the visible light image of the single device is divided into a center point pc 2，i And the key points of the outer contour po 2，i The specific operations are: Intercept single device i The corresponding visible light image v i and feature layer f 1，i , after scaling, it is input into the trained key point detection network for feature extraction, and the sub-feature layer is obtained Sub-feature layer With feature layer f 1，i The fusion result is then used to extract the key points of a single device to obtain the single device t i The key point set p of the visible light image i ; The key point set p of the visible light image i Each key point in is divided into the center point pc 2，i And the key points of the outer contour po 2.i ; The infrared imaging image key point set of a single device is obtained, and each key point in the infrared imaging image key point set of a single device is divided into a center point pc of the single device. 3，i With the outer contour key point po 3，i The specific process is: The single device area in the infrared imaging image is intercepted and input into the trained infrared key point detection model to obtain the sub-feature layer. Sub-feature layer With feature layer f 1，i The fusion result is then used to extract the key points of a single device to obtain the single device t i The key point set of infrared imaging images; Set a single device i Each key point in the infrared imaging image key point set is divided into a single device t i The center point pc 3，i With the outer contour key point po 3，i ; According to the center point pc 2，i , key points of outer contour po 2，i 、Center point pc 3，i And the key points of the outer contour po 3，i The specific process of determining the center point set and the key point aggregation of the outer contour of the power equipment is as follows: Sub-feature layer With sub-feature layer Splice and restore to the original size using 1x1 convolution to get the fused sub-feature layer Fusion sub-feature layer Extract key points of a single device and divide the extracted key points into single device t i The center point pc 4，i And the key points of the outer contour po 4，i ; Using the central point pc 2，i 、Center point pc 3，i and center point pc 4，i Construct a collection of central points for power equipment; Using the outer contour key points po 2，i , key points of outer contour po 3，i And the key points of the outer contour po 4，i Construct a set of key points of the outer contour of the power equipment.

5. 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 steps of the densely arranged power equipment status detection method according to any one of claims 1 to 3 are implemented.

6. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the densely arranged power equipment status detection method as claimed in any one of claims 1 to 3 are implemented.

Images