Steel tower detection method and system based on meteorological spectrum

Abstract

The present application relates to a steel tower detection method and system based on meteorological spectroscopy. A steel tower type library is established through a three-dimensional data model library. Each steel tower type has at least one detection feature area. Each steel tower type in the feature library forms a mapping relationship with at least one detection feature area. The drone carries a multi-spectral camera to obtain images of the steel tower, focusing on the detection feature area. Therefore, when formulating the flight route of the drone, the present application formulates the flight route according to the distribution of multiple detection feature areas in the feature library that form a mapping relationship with the steel tower type. This is conducive to the thermal vortex equipment to stimulate the steel tower to generate electric heat, and the meteorological spectrum generated by the electric heat of the steel tower itself is used to obtain the stress damage inside the steel tower. More specifically, the drone flies based on the route, which saves time and effort and obtains high saturation of information. The spectrum can efficiently feedback the damage information inside the steel tower. The thermal neural network based on the spectrum can efficiently and quickly determine whether the steel tower components are damaged.

Description

Technical Field

[0001] The present application relates to the technical field of steel tower detection, and in particular to a steel tower detection method and system based on meteorological spectroscopy. Background Art

[0002] Steel towers are widely used in communications and power transmission. Since steel towers are generally high and are load-bearing components, the monitoring and detection of steel towers is the focus of daily maintenance projects for steel towers. The load-bearing capacity of steel towers is mainly determined by the following factors: the vertical height of the tower body, the geometric dimensions of the steel components within the design index, the surface hardness and fatigue strength of the steel structure, the anti-corrosion coating of the steel structure, and the structural completeness of the connectors or connections.

[0003] Traditional monitoring and detection methods are mainly based on visual image analysis. This type of monitoring and detection method cannot determine the internal conditions of steel tower components, such as: powdering, cracking, peeling, failure and shedding of anti-corrosion coatings, internal cracks in steel components, deformation and rust of non-visual surfaces, breakage and loosening of connectors or connections, fatigue strength of steel, etc.

[0004] Therefore, it is necessary to propose a steel tower detection method and system based on meteorological spectroscopy, as traditional image-based steel tower monitoring and detection methods cannot determine the defects of the internal conditions of steel tower components. Summary of the invention

[0005] Based on this, it is necessary to propose a steel tower detection method and system based on meteorological spectroscopy as traditional image-based steel tower monitoring and detection methods cannot determine the defects of the internal conditions of steel tower components.

[0006] The present application provides a steel tower detection method based on meteorological spectroscopy, comprising:

[0007] Establish a steel tower type library, extract detection feature areas from the steel tower image of each steel tower type, and include each detection feature area into the feature library; form a mapping relationship between each steel tower type and at least one detection feature area in the feature library;

[0008] Turn on the thermal vortex device to generate electric heat on the steel tower to be investigated, and generate a meteorological spectrum based on the electric heat;

[0009] The tower type of the steel tower to be detected is received, and the flight route of the UAV is formulated according to the distribution of multiple detection feature areas in the feature library that form a mapping relationship with the tower type of the steel tower; the trajectory of the UAV during flight is spirally ascending or spirally descending with the central axis of the steel tower, where the formula of the cylindrical spiral line is:

[0010]

[0011] Where: X, Y, and Z are the three-dimensional spatial coordinates of the UAV, respectively; a is the projection radius of the helix on the XOY plane; b is the pitch of the helix; θ is the rotation angle of the UAV around the central axis of the helix, and the unit is radians. The left-handed helix takes a positive sign, and the right-handed helix takes a negative sign.

[0012] The power of the thermal vortex equipment is adjusted, and based on the flight route of the UAV, the UAV is controlled to perform multiple navigation tasks to photograph the steel tower to be explored, so as to obtain multiple groups of meteorological spectra of the steel tower to be explored; each group of meteorological spectra corresponds to a power of the thermal vortex equipment;

[0013] According to the detection feature area corresponding to the steel tower to be detected, each meteorological spectrum is segmented to form an initial operation graph of each meteorological spectrum; one meteorological spectrum corresponds to at least one initial operation graph;

[0014] Establishing and training a thermal neural network to generate a trained thermal neural network;

[0015] Select an initial operation graph;

[0016] Incorporating the initial operation graph into the trained thermal neural network, starting the trained thermal neural network, and judging whether the steel tower entity corresponding to the initial operation graph is qualified according to the operation result of the trained thermal neural network; the steel tower entity includes one or more of the steel components, connectors and anti-corrosion coatings of the steel tower to be inspected;

[0017] If the steel tower entity corresponding to the initial operation diagram is unqualified, the information that the steel tower entity corresponding to the initial operation diagram is unqualified is fed back, and the process returns to selecting an initial operation diagram until all initial operation diagrams are selected.

[0018] The present application also provides a steel tower detection system based on meteorological spectroscopy, comprising:

[0019] A host computer, used to execute the above steel tower detection method based on meteorological spectrum;

[0020] A thermal vortex device, which is communicatively connected with the host computer;

[0021] An unmanned aerial vehicle system, the unmanned aerial vehicle system comprises an unmanned aerial vehicle and a multispectral camera, the multispectral camera is mounted on the unmanned aerial vehicle, the unmanned aerial vehicle is communicatively connected to the host computer, and the multispectral camera is communicatively connected to the host computer.

[0022] The present application relates to a steel tower detection method and system based on meteorological spectroscopy. A steel tower type library is established through a three-dimensional data model library. Each steel tower type has at least one detection feature area. Each steel tower type in the feature library forms a mapping relationship with at least one detection feature area. The drone carries a multi-spectral camera to obtain images of the steel tower, focusing on the detection feature area. Therefore, when formulating the flight route of the drone, the present application formulates the flight route according to the distribution of multiple detection feature areas in the feature library that form a mapping relationship with the steel tower type. This is conducive to the thermal vortex equipment to stimulate the steel tower to generate electric heat, and the meteorological spectrum generated by the electric heat of the steel tower itself is used to obtain the stress damage inside the steel tower. More specifically, the drone flies based on the route, which saves time and effort and obtains high saturation of information. At the same time, the spectrum can efficiently feedback the damage information inside the steel tower. The thermal neural network based on the spectrum can efficiently and quickly determine whether the steel tower components are damaged. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A schematic flow chart of a steel tower detection method based on meteorological spectroscopy provided in one embodiment of the present application.

[0024] Figure 2 A schematic diagram of the operation of a steel tower detection system based on meteorological spectroscopy provided in one embodiment of the present application.

[0025] Figure 3 A transmission principle diagram of a steel tower material according to a steel tower detection method based on meteorological spectroscopy provided in one embodiment of the present application.

[0026] Figure 4 A connection diagram of a steel tower detection system based on meteorological spectroscopy provided in one embodiment of the present application.

[0027] Reference numerals:

[0028] 100-host computer; 200-thermal vortex equipment; 300-UAV system; 310-UAV; 320-multispectral camera. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0030] The present application provides a steel tower detection method based on meteorological spectroscopy.

[0031] like Figure 1 As shown, in one embodiment of the present application, a steel tower detection method based on meteorological spectroscopy includes:

[0032] S000, establish a steel tower type library.

[0033] S100, extracting detection feature areas from the steel tower image of each steel tower type, and adding each detection feature area into a feature library, in which a mapping relationship is formed between each steel tower type and at least one detection feature area.

[0034] Specifically, a mapping relationship is formed between each steel tower type and at least one detection feature area. A steel tower type library is established through a three-dimensional data model library, each steel tower type has at least one detection feature area, and each steel tower type in the feature library forms a mapping relationship with at least one detection feature area.

[0035] More specifically, in the process of industrial flaw detection, this application focuses on the load-bearing objects of the steel tower, such as communication antennas and transmission cables. The transmission cables themselves may have magnetic saturation and other conditions. When dealing with the alternating magnetic field generated by the thermal vortex equipment, the ability to resist magnetic field shock is relatively strong. However, the communication antenna itself has a weak ability to resist magnetic field shock, so the magnetic field frequency and intensity of the thermal vortex equipment will be limited. The upper limit of this limit lies in the heat resistance value of the solder strength of the communication antenna and the inductor saturation magnetic field intensity. Therefore, the acquisition of spectral images of the detection feature area needs to be closely coordinated with the flight route of the drone, so as to achieve high safety and low maintenance risk of the load-bearing objects of the steel tower.

[0036] S200, turning on the thermal vortex equipment to generate electric heat at the steel tower to be inspected, and generating a meteorological spectrum based on the electric heat.

[0037] Specifically, the flight route is designed to facilitate the thermal vortex equipment to stimulate the steel tower to generate electric heat, and the meteorological spectrum generated by the steel tower's own electric heat can be used to obtain the stress damage inside the steel tower. The drone flies based on the route, saving time and effort and obtaining a high degree of information.

[0038] More specifically, the vortex generator is mainly used to generate an alternating magnetic field. Traditional vortex generators need to efficiently utilize the alternating magnetic field, so the vortex generator is close to the vortex generator, which is the steel tower. However, the vortex generator in the present application can be far away from the steel tower, which improves the operating efficiency.

[0039] It is worth mentioning that the thermal vortex equipment mainly uses medium-frequency electromagnetic waves to generate alternating magnetic fields, such as 7000 Hz electromagnetic waves, which continuously bombard the tower for 3 minutes.

[0040] S300, receiving the steel tower type of the steel tower to be detected, and formulating a flight route for the UAV according to the distribution of multiple detection feature areas in the feature library that form a mapping relationship with the steel tower type.

[0041] S400, adjusting the power of the thermal vortex device, and based on the flight route of the drone, controlling the drone to perform multiple navigation tasks, photographing the steel tower to be explored, so as to obtain multiple groups of meteorological spectra of the steel tower to be explored. Each group of meteorological spectra corresponds to a power of the thermal vortex device.

[0042] Specifically, there are two main purposes for adjusting the power of the thermal vortex equipment:

[0043] The first purpose is that different structures of the steel tower have different materials, and different materials produce different degrees of distinct meteorological spectral characteristics at different temperatures. This application can maximize the extraction of the most distinctive features of the steel tower through different thermal vortex equipment powers, and judge whether the structural components of the steel tower are damaged based on the features.

[0044] The second purpose is that the temperature of the steel tower itself is different in different working environments and seasons, such as the temperature of the steel tower itself in winter and the temperature of the steel tower itself in summer. Different thermal vortex equipment power gradients can make the spectral image of the steel tower have a wider coupling range.

[0045] S500, segmenting each meteorological spectrum according to the detection feature area corresponding to the steel tower to be detected, so as to form an initial operation graph of each meteorological spectrum. A meteorological spectrum corresponds to at least one initial operation graph.

[0046] Specifically, the role of extracting the exploration feature area is that: in a panoramic image of the meteorological spectrum, a large number of exploration feature areas will be included. Subdividing the panoramic image of the meteorological spectrum into multiple initial operation images of the meteorological spectrum is conducive to improving the operation efficiency of the thermal neural network.

[0047] More specifically, multiple adjacent extracted detection feature areas can be segmented progressively. Figure 2 As shown, in the panoramic view of the meteorological spectrum composed of two extracted exploration feature areas A and B, it can be divided into exploration feature area A and exploration feature area B, or it can be divided into exploration feature area A-1, exploration feature area A-2 and exploration feature area B, and the exploration feature area A-1 and the exploration feature area A-2 together constitute the exploration feature area A. This can ensure that the information loss rate of the initial operation graph is low.

[0048] S600, establishing and training a thermal neural network to generate a trained thermal neural network.

[0049] S700: Select an initial operation graph.

[0050] S800, incorporating the initial operation diagram into the trained thermal neural network, starting the trained thermal neural network, and judging whether the steel tower entity corresponding to the initial operation diagram is qualified according to the operation results of the trained thermal neural network.

[0051] The steel tower entity includes one or more of steel components, connectors and anti-corrosion coatings of the steel tower to be inspected.

[0052] S900, if the steel tower entity corresponding to the initial operation diagram is unqualified, then the information that the steel tower entity corresponding to the initial operation diagram is unqualified is fed back, and the process of selecting an initial operation diagram is returned until all initial operation diagrams are selected.

[0053] The present embodiment relates to an image-based steel tower monitoring method, in which a steel tower type library is established through a three-dimensional data model library, each steel tower type has at least one detection feature area, and each steel tower type in the feature library forms a mapping relationship with at least one detection feature area. The drone carries a multi-spectral camera to obtain images of the steel tower, focusing on the detection feature area. Therefore, when formulating the flight route of the drone, the present application formulates the flight route according to the distribution of multiple detection feature areas in the feature library that form a mapping relationship with the steel tower type, which is conducive to the thermal vortex equipment to stimulate the steel tower to generate electric heat, and obtain the internal stress damage of the steel tower through the meteorological spectrum generated by the electric heat of the steel tower itself. More specifically, the drone flies based on the route, which saves time and effort and obtains high saturation of information. At the same time, the spectrogram can efficiently feedback the damage information inside the steel tower. The thermal neural network based on the spectrogram can efficiently and quickly determine whether the steel tower components are damaged.

[0054] In one embodiment of the present application, S100 includes:

[0055] S110, receiving steel tower types, and adding each steel tower type into the steel tower type library.

[0056] Specifically, the tower type includes one or more of a guyed tower, a single-tube tower, a three-tube tower and an angle steel tower.

[0057] Different tower types can affect the flight trajectory of the drone, and thus affect the information saturation of the spectral panorama. It is worth mentioning that the spectral panorama is the spectrum of the entire steel tower.

[0058] The completeness of the tower type is conducive to the improvement of the UAV flight route, thereby improving the information saturation of the spectral panorama.

[0059] For S120, choose a steel tower type.

[0060] S130, determine one or more of the shape, thickness and material of the steel components of the steel tower, determine one or more of the shape, thickness and material of the connecting parts or connecting parts of the steel tower, and determine one or more of the thickness and material of the anti-corrosion coating of the steel tower.

[0061] Specifically, steel components, connectors or connecting parts, and anti-corrosion coatings constitute all types of the detection feature area. In fact, connectors or connecting parts are all connecting structures between steel components.

[0062] When the UAV is formulating its flight trajectory, the increase in the density of the exploration feature area will increase the difficulty of the UAV's shooting.

[0063] More specifically, in order to simplify the difficulty of drone shooting, based on the tower shape of the steel tower and the assemblability of the detection feature area, the steel tower is mostly a symmetrical structure. Therefore, the trajectory of the drone during flight is mostly based on the central axis of the steel tower, spiraling up or down. The host computer can determine the direction of the drone's operation according to the density of the detection feature area. At the same time, the drone's multispectral camera is on the gimbal, which can provide the camera with pitch and swing positions while keeping the camera stable.

[0064] This greatly reduces the difficulty of taking spectral images with the camera and improves the information saturation and shooting efficiency of the spectral panorama.

[0065] S140, based on the three-dimensional data model of the steel tower, determine the physical assembly order of the steel components, connectors, and anti-corrosion coatings in the steel tower.

[0066] Specifically, the connecting parts may be bolts and nuts, and the connecting parts may be welds. When a bolt or nut is missing, a gap will appear in the spectrum. At this time, the thermal neural network will quickly find the missing point.

[0067] Similarly, cracks in welds are different from metal fatigue in steel components. However, once cracks and metal fatigue occur, textures will appear on the spectrum and the spectrum will no longer be smooth. At this time, the thermal neural network will quickly find the damage point.

[0068] S150, returning to the step of selecting a steel tower type until all steel tower types are selected.

[0069] S160, based on the physical assembly order, the steel tower image of each steel tower type is divided into detection feature areas. The present embodiment relates to a method for establishing a steel tower type library and a feature library. Based on the steel tower type library and the feature library, the present application obtains the spectrum of the steel tower and analyzes the spectrum. This monitoring-detection method can effectively determine the internal conditions of the components of the steel tower, such as: powdering, cracking, peeling, failure and shedding of the anti-corrosion coating, internal cracks of steel components, deformation and rust of non-visual surfaces, fracture and loosening of connectors or connecting parts, fatigue strength of steel, etc.

[0070] like Figure 3 As shown, in one embodiment of the present application, S130 includes:

[0071] S131, obtaining the chemical element composition ratio of each steel component of the steel tower type.

[0072] S132, obtaining the material of each steel component based on the chemical element composition ratio of each steel component.

[0073] S133, select a steel member.

[0074] S134, select a thermal vortex device power.

[0075] S135, obtaining a plurality of sample meteorological spectra of the steel structure under the power of the thermal vortex equipment.

[0076] Specifically, a meteorological spectrum is obtained based on a multi-spectral camera. One shooting angle of the steel component entity corresponds to a meteorological spectrum.

[0077] Since the steel structure may be subjected to shear force and other forces at different angles, each part of the steel structure may be damaged by bending, metal fatigue, etc. As the initial sample of the thermal neural network sample library, a structurally intact and undamaged sample needs to record spectra at various angles and temperatures.

[0078] S136, returning to the step of selecting a thermal turbine device power until all thermal turbine device powers are selected.

[0079] S137, returning to the selection of a steel component, until all steel components are selected, and each sample meteorological spectrum is incorporated into the sample library of the thermal neural network.

[0080] Faced with different chemical compositions, such as Figure 3 , the transmission bands of different materials are different, based on this, meteorological spectroscopy can perform layered processing.

[0081] In fact, the method for acquiring the sample meteorological spectrum of the connecting member or the connecting portion is the same as S131 to S137, so S130 also includes:

[0082] S138, obtaining the chemical element composition ratio of the connecting parts or connecting parts of the tower-shaped steel tower.

[0083] Specifically, similarly, the connecting piece or connecting portion also has the same working state as the steel component entity.

[0084] S139, based on the chemical element composition ratio of the connector or the connector part, all meteorological spectra of the connector or the connector part are obtained, and each meteorological spectrum is included in the sample library of the thermal neural network.

[0085] This embodiment relates to a method for acquiring a thermal neural network sample library. Acquisition of the sample library is conducive to the thermal vortex device stimulating the steel tower to generate electric heat, and obtaining the internal stress damage of the steel tower through the meteorological spectrum generated by the electric heat of the steel tower itself. The judgment basis of the thermal neural network lies in the referenceability of sample data with intact structure. The sample library can greatly improve the judgment accuracy of the thermal neural network.

[0086] In one embodiment of the present application, S130 further includes:

[0087] S139a, select a material for the anti-corrosion coating.

[0088] S139b, select a thickness of the anti-corrosion coating.

[0089] S139c, applying the anti-corrosion coating to the steel structure and applying the anti-corrosion coating to the connecting parts to obtain a sample meteorological spectrum of the steel structure coated with the anti-corrosion coating and a sample meteorological spectrum of the connecting parts coated with the anti-corrosion coating.

[0090] S139d, returning to the step of selecting a thickness of an anti-corrosion coating until the thickness of all anti-corrosion coatings are selected.

[0091] S139e, returning to the process of selecting a material for the anti-corrosion coating, until all materials for the anti-corrosion coating are selected, and incorporating each meteorological spectrum into the sample library of the thermal neural network.

[0092] Specifically, under actual use conditions, steel components, connectors or joints are often coated with anti-corrosion coatings. Powdering, cracking, peeling, failure and shedding of anti-corrosion coatings are also daily tasks for monitoring and testing steel towers.

[0093] The acquisition of the sample library is conducive to the thermal vortex equipment to stimulate the steel tower to generate electric heat, and the meteorological spectrum generated by the electric heat of the steel tower itself can be used to obtain the adverse conditions of the anti-corrosion coating, such as powdering, cracking, peeling, failure and shedding. The judgment basis of the thermal neural network lies in the reference of the sample data with intact structure. The sample library can greatly improve the judgment accuracy of the thermal neural network.

[0094] It is worth mentioning that at different temperatures, the spectra excited by heat in the steel structure itself, the connection or connector itself, and the coating itself are different. The thermal neural network can separate different sub-spectra under the same spectrum through the grayscale feature matrix.

[0095] In one embodiment of the present application, S300 includes:

[0096] S310, according to the steel tower type of the steel tower to be inspected, call the three-dimensional data model of the steel tower type, and parse the three-dimensional data model of the steel tower type to determine the type of steel components, the type of connectors, and the type of anti-corrosion coating of the steel tower type.

[0097] Specifically, to determine the type of steel components of the steel tower, the type of connectors or connecting parts, and the type of anti-corrosion coating.

[0098] S320, according to the type of steel components, the type of connecting parts, and the type of anti-corrosion coating of the steel tower, retrieve the physical assembly order of the steel components, connecting parts, and anti-corrosion coating in the steel tower.

[0099] S330, based on the physical assembly order of steel components, connectors, and anti-corrosion coatings in the steel tower type, with steel components and connectors as target points, calculates the flight route of the UAV according to the optimal algorithm.

[0100] This embodiment relates to a target point collection method for a steel tower. When a UAV is making a flight trajectory, an increase in the density of the detection feature area will increase the difficulty of the UAV's shooting.

[0101] More specifically, in order to simplify the difficulty of drone shooting, based on the tower shape of the steel tower and the assemblability of the detection feature area, the steel tower is mostly a symmetrical structure. Therefore, the trajectory of the drone during flight is mostly based on the central axis of the steel tower, spiraling up or down. The host computer can determine the direction of the drone's operation according to the density of the detection feature area. At the same time, the drone's multispectral camera is on the gimbal, which can provide the camera with pitch and swing positions while keeping the camera stable.

[0102] This greatly reduces the difficulty of taking spectral images with the camera and improves the information saturation and shooting efficiency of the spectral panorama.

[0103] like Figure 2 As shown, in one embodiment of the present application, the optimal algorithm in S300 includes:

[0104] S341, calling the three-dimensional data model of the steel tower, and simplifying the three-dimensional data model of the steel tower into a target point wireframe diagram based on the target point.

[0105] S342, based on the density of the target points in the target point wireframe diagram, define the UAV flight navigation direction as a line from a low-density target point collection area to a high-density target point collection area, and generate the UAV flight route according to the UAV flight navigation direction.

[0106] This embodiment involves an optimal algorithm. The trajectory of the drone during flight is mostly about the central axis of the steel tower, spiraling up or spiraling down, where the formula of the cylindrical spiral line is:

[0107]

[0108] Where: X, Y, and Z are the three-dimensional spatial coordinates of the UAV, respectively; a is the projection radius of the helix on the XOY plane; b is the pitch of the helix; θ is the rotation angle of the UAV around the central axis of the helix; and the unit is radians. The left-handed helix takes a positive sign, and the right-handed helix takes a negative sign.

[0109] Based on the UAV's flight trajectory being a cylindrical spiral, the remaining parameters to be determined are the spatial orientation of the cylindrical spiral caused by the density of target points:

[0110] Define the weight of a target point as Ai.

[0111] Then there is a surface body with weights about the target point set in the three-dimensional data space:

[0112] L(A i X, A i Y, A i Z),A i∈[A] Formula 2

[0113] Among them, [A] is the weight matrix of the target point set.

[0114] For L(A i X, A i Y, A i Z) Take the first-order spatial gradient and get the Nabla operator:

[0115]

[0116] in, is the Nabla operator of target point i, is the partial differential derivative of the target point i in the X direction, is the partial differential derivative of the target point i in the Y direction, is the partial differential derivative of the target point i in the Z direction, k is the basis vector in the X direction, n is the basis vector in the Y direction, and m is the basis vector in the Z direction.

[0117] The Nabla operator and the UAV flight navigation direction are used as a line pointing from a low-density target point collection area to a high-density target point collection area. The rotation direction of the spiral line of the route and the direction of the starting point and the end point of the route can be known.

[0118] In one embodiment of the present application, S600 includes:

[0119] S611, select a sample meteorological spectrum in the sample library of the thermal neural network.

[0120] S612, stratifying the sample meteorological spectrum based on the grayscale feature matrix to obtain a plurality of sub-spectral graphs of the sample meteorological spectra.

[0121] S613, selecting a sub-spectrum of a sample meteorological spectrum.

[0122] S614, extracting texture features from the sub-spectral graphs of the sample meteorological spectrum.

[0123] S615, returning to the process of selecting a sub-spectrum of the sample meteorological spectrum until all sub-spectrums of the sample meteorological spectra are selected.

[0124] S616, returning a sample meteorological spectrum in the sample library of the selected thermal neural network until all sample meteorological spectra are selected.

[0125] S617, establish a screening layer.

[0126] S618, each texture feature is incorporated into the initial part of the screening layer, so that the screening layer obtains a feature judgment value of each texture feature.

[0127] Specifically, a grayscale matrix [K] is defined, and the grayscale is equally divided from 0% to 100% and incorporated into the one-dimensional grayscale matrix [K].

[0128] The sub-spectral image of the selected meteorological spectrum is divided into a two-dimensional pixel matrix [N] with pixels as units, where [N] is the two-dimensional spatial position matrix of the image.

[0129] Perform cross product on the grayscale matrix [K] and the pixel matrix [N] to obtain the joint matrix [M].

[0130] Each element matrix in the joint matrix [M] is normalized to obtain the texture feature matrix [Q].

[0131] The texture feature matrix [Q] is incorporated into the initial part of the filtering layer.

[0132] This embodiment involves the establishment of a screening layer. The samples in the screening layer in the initial thermal neural network only have sample gas phase spectra of intact steel components themselves, connections or connectors themselves, and coatings themselves. These meteorological spectra correspond to texture feature matrices [Q].

[0133] During the training process of the thermal neural network, the screening layer will include the gas phase spectra that the thermal neural network has misjudged. These meteorological spectra correspond to texture feature matrices [Q] that can be used for the later learning of the thermal neural network, which is conducive to improving the judgment accuracy of the thermal neural network.

[0134] In one embodiment of the present application, S600 further includes:

[0135] S621, based on the grayscale feature matrix, each sample meteorological spectrum in the sample library of the thermal neural network is stratified to establish a judgment layer.

[0136] S622, receiving a feature judgment value of each texture feature so that the judgment layer has a judgment criterion.

[0137] S623, incorporate the screening layer and the judgment layer into the thermal neural network.

[0138] Specifically, the judgment layer will pre-execute steps S612 and S614 on the received meteorological spectrum, and extract the texture feature matrix [Q] from the received meteorological spectrum.

[0139] The subsequent execution steps of the judgment layer are to compare the texture feature matrix [Q] of the received meteorological spectrum with the texture feature matrix [Q] library in the screening layer to determine the matching degree between the texture feature matrix [Q] of the received meteorological spectrum and the texture feature matrix [Q] library in the screening layer, and finally output the matching result.

[0140] It is worth mentioning that the texture feature matrix [Q] library in the screening layer is the judgment criterion.

[0141] In one embodiment of the present application, S600 further includes:

[0142] S631, receiving a target meteorological spectrum; the target meteorological spectrum is a meteorological spectrum of a steel component that has been determined to be damaged and coated with an anti-corrosion coating, and a meteorological spectrum of a connector that has been determined to be damaged and coated with an anti-corrosion coating.

[0143] S632, mixing the target meteorological spectrum with the sample meteorological spectrum in the sample library to generate training samples. The mixing method is random mixing.

[0144] S633, incorporating the training sample into the thermal neural network to train the thermal neural network and obtain the judgment result of the thermal neural network. Determine the success rate of the thermal neural network judgment according to the judgment result of the thermal neural network.

[0145] S634, based on the success rate of the thermal neural network judgment, determine whether the success rate of the thermal neural network judgment is greater than or equal to a success rate threshold.

[0146] S635, if the success rate determined by the thermal neural network is greater than or equal to the success rate threshold, a trained thermal neural network is generated.

[0147] S636, if the success rate of the thermal neural network judgment is less than the success rate threshold, the target meteorological spectrum or sample meteorological spectrum that is judged incorrectly in the training sample is marked.

[0148] Specifically, each time the judgment layer makes a judgment, it will feed back the judgment result and the texture feature matrix [Q] of the received meteorological spectrum and the texture feature matrix [Q] matching object of the texture feature matrix [Q] library in the screening layer.

[0149] When there is no matching object in the texture feature matrix [Q] library, or the matching degree is too low, the thermal neural network will mark the received meteorological spectrum, incorporate the received meteorological spectrum into the screening layer, and determine through physical experiments whether the entity corresponding to the received meteorological spectrum is damaged, and map the result with the texture feature matrix [Q] of the received meteorological spectrum in the screening layer.

[0150] S637, including the marked target meteorological spectrum or sample meteorological spectrum into the screening layer to obtain a new feature judgment value.

[0151] S638, modifying the judgment layer based on the new feature judgment value, and returning to the mixing of the target meteorological spectrum with the meteorological spectrum in the sample library to generate training samples, until the success rate of the thermal neural network judgment is greater than or equal to the success rate threshold.

[0152] The present application provides a steel tower detection system based on meteorological spectroscopy.

[0153] like Figure 4 As shown, in one embodiment of the present application, a steel tower detection system based on meteorological spectroscopy includes a host computer 100, a thermal vortex device 200 and a drone system 300.

[0154] The host computer 100 is used to execute the above-mentioned image-based steel tower monitoring method.

[0155] The vortex device 200 is in communication connection with the host computer 100 .

[0156] Specifically, the thermal vortex device 200 is an electromagnetic wave transmitting device.

[0157] The drone system 300 includes a drone 310 and a multispectral camera 320 . The multispectral camera 320 is mounted on the drone 310 . The drone 310 is communicatively connected to the host computer 100 . The multispectral camera 320 is communicatively connected to the host computer 100 .

[0158] The present embodiment relates to a steel tower detection system based on meteorological spectrum. The host computer 100 establishes a steel tower type library through a three-dimensional data model library. Each steel tower type has at least one detection feature area. Each steel tower type in the feature library forms a mapping relationship with at least one detection feature area. The drone 310 carries a multi-spectral camera 320 to obtain images of the steel tower, focusing on the detection feature area. Therefore, when formulating the flight route of the drone 310, the present application formulates the flight route according to the distribution of multiple detection feature areas in the feature library that form a mapping relationship with the steel tower type, which is conducive to the thermal vortex device 200 to stimulate the steel tower to generate electric heat, and obtain the internal stress damage of the steel tower through the meteorological spectrum generated by the electric heat of the steel tower itself. More specifically, the drone 310 flies based on the route, which saves time and effort and obtains high fullness of information. At the same time, the spectrum can efficiently feedback the damage information inside the steel tower. The thermal neural network based on the spectrum can efficiently and quickly determine whether the steel tower components are damaged.

[0159] The technical features of the above-described embodiments may be arbitrarily combined, and the execution order of the method steps is not limited. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0160] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A steel tower detection method based on meteorological spectroscopy, characterized in that: include: Extracting detection feature areas from the steel tower image of each steel tower type, and adding each detection feature area into a feature library, in which each steel tower type is mapped to at least one detection feature area; Turn on the thermal vortex device to generate electric heat on the steel tower to be investigated, and generate a meteorological spectrum based on the electric heat; The tower type of the steel tower to be detected is received, and the flight route of the UAV is formulated according to the distribution of multiple detection feature areas in the feature library that form a mapping relationship with the tower type of the steel tower; the trajectory of the UAV during flight is spirally ascending or spirally descending with the central axis of the steel tower, where the formula of the cylindrical spiral line is: Where: X, Y, and Z are the three-dimensional spatial coordinates of the UAV, respectively; a is the projection radius of the helix on the XOY plane; b is the pitch of the helix; θ is the rotation angle of the UAV around the central axis of the helix, and the unit of θ is radians. The left-handed helix takes a positive sign, and the right-handed helix takes a negative sign. The power of the thermal vortex equipment is adjusted, and based on the flight route of the UAV, the UAV is controlled to perform multiple navigation tasks to photograph the steel tower to be explored, so as to obtain multiple groups of meteorological spectra of the steel tower to be explored; each group of meteorological spectra corresponds to a power of the thermal vortex equipment; According to the detection feature area corresponding to the steel tower to be detected, each meteorological spectrum is segmented to form an initial operation graph of each meteorological spectrum; one meteorological spectrum corresponds to at least one initial operation graph; Establishing and training a thermal neural network to generate a trained thermal neural network; Select an initial operation graph; Incorporating the initial operation graph into the trained thermal neural network, starting the trained thermal neural network, and judging whether the steel tower entity corresponding to the initial operation graph is qualified according to the operation result of the trained thermal neural network; the steel tower entity includes one or more of the steel components, connectors and anti-corrosion coatings of the steel tower to be inspected; If the steel tower entity corresponding to the initial operation diagram is unqualified, the information that the steel tower entity corresponding to the initial operation diagram is unqualified is fed back, and the initial operation diagram is returned until all initial operation diagrams are selected.

2. The steel tower detection method based on meteorological spectrum according to claim 1 is characterized in that: The extraction of the detection feature area from the steel tower image of each steel tower type and the inclusion of each detection feature area into the feature library include: receiving steel tower types, and adding each steel tower type into a steel tower type library; the steel tower type includes one or more of a guyed tower, a single-tube tower, a three-tube tower, and an angle steel tower; Choose a steel tower type; Determine one or more of the shape, thickness and material of the steel components of the steel tower, determine one or more of the shape, thickness and material of the connecting parts or connecting parts of the steel tower, and determine one or more of the thickness and material of the anti-corrosion coating of the steel tower; Based on the three-dimensional data model of the steel tower, determine the physical assembly order of the steel components, connectors, and anti-corrosion coatings in the steel tower; Return to the steel tower type until all steel tower types have been selected; The steel tower image of each type of steel tower is divided into exploration feature areas based on the entity assembly order.

3. The steel tower detection method based on meteorological spectrum according to claim 2 is characterized in that: The step of determining one or more of the shape, thickness and material of the steel structure of the steel tower, determining one or more of the shape, thickness and material of the connecting piece or connecting part of the steel tower, and determining one or more of the thickness and material of the anti-corrosion coating of the steel tower includes: Obtaining the chemical element composition ratio of each steel component of the steel tower type; Based on the chemical element composition ratio of each steel component, the material of each steel component is obtained; Select a steel member; Select a thermal vortex equipment power; Acquire a plurality of sample meteorological spectra of the steel structure under the power of the vortex device; Return the power of the vortex device until all the vortex device powers are selected; Return to the steel component of this type until all steel components are selected, and include each sample meteorological spectrum into the sample library of the thermal neural network.

4. The steel tower detection method based on meteorological spectroscopy according to claim 3 is characterized in that: The step of determining one or more of the shape, thickness and material of the steel structure of the steel tower, determining one or more of the shape, thickness and material of the connecting piece or connecting part of the steel tower, and determining one or more of the thickness and material of the anti-corrosion coating of the steel tower also includes: Choose a material for the anti-corrosion coating; Choose an anti-corrosion coating thickness; Applying the anti-corrosion coating to the steel member and applying the anti-corrosion coating to the connecting member to obtain a sample meteorological spectrum of the steel member coated with the anti-corrosion coating and a sample meteorological spectrum of the connecting member coated with the anti-corrosion coating; Return the thickness of the anti-corrosion coating until the thickness of all anti-corrosion coatings are selected; The material of the anti-corrosion coating is returned until all the materials of the anti-corrosion coating are selected, and each meteorological spectrum is included in the sample library of the thermal neural network.

5. The steel tower detection method based on meteorological spectrum according to claim 4 is characterized in that: The receiving of the steel tower type of the steel tower to be detected and formulating a flight route of the UAV according to the distribution of a plurality of detection feature areas in the feature library that form a mapping relationship with the steel tower type include: According to the steel tower type of the steel tower to be inspected, the three-dimensional data model of the steel tower type is called, and the three-dimensional data model of the steel tower type is analyzed to determine the type of steel components, the type of connectors, and the type of anti-corrosion coating of the steel tower type; According to the type of steel components, the type of connectors, and the type of anti-corrosion coating of the steel tower, the physical assembly order of the steel components, the connectors, and the anti-corrosion coating in the steel tower is retrieved; Based on the physical assembly order of steel components, connectors, and anti-corrosion coatings in the steel tower type, the flight route of the UAV is calculated according to the optimal algorithm, taking the steel components and connectors as target points.

6. The steel tower detection method based on meteorological spectrum according to claim 5 is characterized in that: The receiving of the steel tower type of the steel tower to be detected and formulating the flight route of the UAV according to the distribution of multiple detection feature areas in the feature library that form a mapping relationship with the steel tower type also includes: Define the weight of a target point as A i ; In the three-dimensional data space, there exists a surface body with weights about the target point set L(A i X,A i Y,A i Z),A i ∈[A] Where [A] is the weight matrix of the target point set; For L(A i X, A i Y, A i Z) Take the first-order spatial gradient and get the Nabla operator in, is the Nabla operator of target point i, is the partial differential derivative of the target point i in the X direction, is the partial differential derivative of the target point i in the Y direction, is the partial differential derivative of the target point i in the Z direction, k is the basis vector in the X direction, n is the basis vector in the Y direction, and m is the basis vector in the Z direction; Taking the Nabla operator and the UAV flight direction as a line, pointing from the low-density target point collection area to the high-density target point collection area, we can know the rotation direction of the spiral line of the route and the direction of the starting point and the end point of the route.

7. The steel tower detection method based on meteorological spectrum according to claim 6 is characterized in that: The step of establishing and training a thermal neural network to generate a trained thermal neural network includes: Select a sample meteorological spectrum from the sample library of the thermal neural network; The sample meteorological spectrum is layered based on the grayscale feature matrix to obtain a plurality of sub-spectral graphs of the sample meteorological spectra; Select a sample meteorological spectrum; Extracting texture features from the sub-spectrum of the sample meteorological spectrum; Return the sub-spectrum of the sample meteorological spectrum until all sub-spectrums of the sample meteorological spectrum are selected; Return the selected sample meteorological spectra until all sample meteorological spectra have been selected; Establishing the screening layer; Each texture feature is incorporated into the initial part of the screening layer so that the screening layer obtains a feature judgment value of each texture feature.

8. The steel tower detection method based on meteorological spectrum according to claim 7 is characterized in that: The step of establishing and training the thermal neural network to generate a trained thermal neural network also includes: Define the grayscale matrix [K]; Divide the grayscale into equal parts from 0% to 100%; The grayscale segmentation results are incorporated into the one-dimensional grayscale matrix [K]; The sub-spectrum of the selected meteorological spectrum is divided into a two-dimensional pixel matrix [N] with pixels as units; [N] is the two-dimensional spatial position matrix of the image; Perform cross multiplication on the grayscale matrix [K] and the pixel matrix [N] to obtain the joint matrix [M]; Normalize each element matrix in the joint matrix [M] to obtain the texture feature matrix [Q]; The texture feature matrix [Q] is incorporated into the initial part of the filtering layer.

9. The steel tower detection method based on meteorological spectrum according to claim 8 is characterized in that: The step of establishing and training the thermal neural network to generate a trained thermal neural network also includes: Extracting texture feature matrix [Q] from the received meteorological spectrum; Compare the texture feature matrix [Q] of the received meteorological spectrum with the texture feature matrix [Q] library in the screening layer; Determine the matching degree between the texture feature matrix [Q] of the received meteorological spectrum and the texture feature matrix [Q] of the texture feature matrix [Q] library in the screening layer; Finally output the matching results; When there is no matching object in the texture feature matrix [Q] library, or the matching degree is too low, the thermal neural network will mark the received meteorological spectrum and include the received meteorological spectrum in the screening layer; determine through physical experiments whether the entity corresponding to the received meteorological spectrum is damaged, and map the result with the texture feature matrix [Q] of the received meteorological spectrum in the screening layer.

10. A steel tower detection system based on meteorological spectroscopy, characterized in that: include: A host computer, used to execute the steel tower detection method based on meteorological spectrum as described in any one of claims 1 to 9; A thermal vortex device, which is communicatively connected with the host computer; An unmanned aerial vehicle system, the unmanned aerial vehicle system comprises an unmanned aerial vehicle and a multispectral camera, the multispectral camera is mounted on the unmanned aerial vehicle, the unmanned aerial vehicle is communicatively connected to the host computer, and the multispectral camera is communicatively connected to the host computer.

Images