Metal roof intelligent installation control system based on machine intelligence

Abstract

The invention relates to the technical field of hybrid intelligence, in particular to a metal roof intelligent installation control system based on machine intelligence, which comprises an intelligent installation data acquisition module for acquiring intelligent installation data, an installation environment sensing module for eliminating reflective interference, and a control module for controlling the intelligent installation of a metal roof. The intelligent risk prediction module obtains the boundary distance of a safety area, the edge intelligent cooperation module measures and calculates computing resources, and the intelligent manipulator control module allocates initial tasks. Environment self-adaption, risk controllability and efficient resource utilization in the installation process are intelligently achieved through a machine, and the safety, accuracy and resource utilization rate of metal roof installation are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of hybrid intelligent technology, and in particular to a metal roof intelligent installation control system based on machine intelligence. Background Art

[0002] The reflection of metal roofing materials causes some areas of the image captured by the camera to be too bright, making it difficult to clearly present key information such as the edges of the metal plates and installation marks. It is difficult for installers to accurately judge the installation position and angle of the plates, which in turn affects the installation accuracy and increases the risk of rework. In rainy and foggy weather, the scattering and absorption of light by raindrops will reduce visibility, affecting the imaging quality of machine intelligent monitoring equipment, resulting in blurred images and loss of details, making it difficult for the monitoring system to accurately obtain real-time information from the installation site. The current metal roof installation process lacks effective risk prediction methods. As the intelligence level of metal roof installation increases, the amount of data and computing tasks that the installation control system needs to process are becoming increasingly complex. However, existing machine intelligence usually adopts a fixed computing resource allocation model and cannot dynamically adjust computing resources according to the needs of actual tasks.

[0003] Chinese Patent Publication No. CN101614058B discloses a rooftop photovoltaic installation system, including upper, lower, left, and right frames of a photovoltaic module, as well as keels arranged under the longitudinal gaps between the photovoltaic modules. This shows that this solution still has the problem of being unable to use machine intelligence to accurately install the metal roof in reflective conditions and rainy and foggy weather. The lack of effective risk prediction methods leads to dangers in the machine intelligence during the installation process, and the inability to flexibly allocate computing power resources leads to low utilization of machine intelligence resources. Summary of the Invention

[0004] To this end, the present invention provides a metal roof intelligent installation control system based on machine intelligence to overcome the problems in the prior art, such as the inability of machine intelligence to accurately install metal roofs under reflective conditions and in rainy and foggy weather, the lack of effective risk prediction methods resulting in dangers of machine intelligence during the metal roof installation process, and the inability to flexibly allocate computing power resources resulting in low utilization of machine intelligence resources.

[0005] To achieve the above objectives, the present invention provides a metal roof intelligent installation control system based on machine intelligence, comprising:

[0006] Smart installation data acquisition module, used to acquire smart installation data;

[0007] An environmental perception module is installed to judge the reflective area based on the intelligent installation data to obtain a reflective judgment result, eliminate reflective interference based on the reflective judgment result to obtain target environmental data, build an environmental perception model based on the target environmental data, and obtain reflective elimination data based on the environmental perception model;

[0008] An intelligent risk prediction module is used to obtain the safe zone boundary distance based on the reflection elimination data and adjust the environmental perception model based on the safe zone boundary distance;

[0009] The edge intelligent collaboration module is used to measure computing resources within the boundary distance of the safe area based on the intelligent twin model to obtain computing resource measurement results, and is also used to perform edge intelligent collaboration on the computing resource measurement results based on the edge intelligent collaboration method, and is also used to adjust the penalty of the edge intelligent collaboration method;

[0010] The intelligent manipulator control module is used to perform initial task allocation based on the calculation resource measurement results and to adjust the allocation process of the initial task allocation.

[0011] Furthermore, the installation environment perception module judges the reflective area according to the intelligent installation data using a reflective judgment method, and the reflective interference elimination method includes:

[0012] Step A01: According to the first polarized light intensity I0 and the second polarized light intensity I 45 , the third polarized light intensity I 90 and the fourth polarized light intensity I 135 Calculate the total light intensity S0 and get the total light intensity S0, set S0 = I0 + I 45 +I 90 +I 135 ;

[0013] Step A02: according to the first polarized light intensity I0 and the third polarized light intensity I 90 Calculate the first light intensity difference S1 to obtain the first light intensity difference S1, and set S1 = I0-I 90 ;

[0014] Step A03, according to the second polarized light intensity I 45 and the fourth polarized light intensity I 135 Calculate the second light intensity difference S2 to obtain the second light intensity difference S2, and set S2 = I 45 -I 135 ;

[0015] Step A04, calculate the polarization degree P according to the total light intensity S0, the first light intensity difference S1 and the second light intensity difference S2, and obtain the polarization degree P, and set

[0016] Step A05: Compare the polarization degree P with the preset polarization degree P0, set 0.6≤P0≤0.8, judge the reflective state of the roof area based on the comparison result, and output the reflective judgment result based on the judgment result, wherein:

[0017] When P>P0, the installation environment perception module determines that the reflective state of the roof area is reflective, and outputs the reflective area as the reflective judgment result;

[0018] When P≤P0, the installation environment perception module determines that the reflective state of the roof area is non-reflective, and outputs the non-reflective area as the reflective judgment result.

[0019] Furthermore, the installation environment perception module performs a de-reflection process on the reflective area according to a de-reflection process method, and the de-reflection process method includes:

[0020] Step B01: Based on the dark channel I of the image dark ={I r , I g , I b}, transmittance control constant ω, set 0.9≤ω≤0.95, and the background light estimation value A to calculate the initial transmittance τ0, and get the initial transmittance τ0, set

[0021] Step B02, calculating the optimized transmittance τq based on the first optimization coefficient ak, the second optimization coefficient bk and the reflective point pixel Iq, obtaining the optimized transmittance τq, and setting τq = ak × Iq ​​+ bk;

[0022] Step B03, calculate the real texture channel Jc according to the optimized transmittance τq, the first channel intensity Ic and the second channel intensity Ac, and obtain the real texture channel Jc, and set

[0023] Step B04: output the real texture channel Jc and the optimized transmittance τ0 as target environment data.

[0024] Furthermore, the installation environment perception module constructs an environment perception model according to the target environment data through an environment perception model construction method, and the environment perception model construction method includes:

[0025] Step S01, initialize the parameters of the convolutional neural network model according to the historical environment database: set the generator weight coefficient λ1 and the discriminator weight coefficient λ2, setting λ1 = 0.5, λ2 = 1-λ1 = 0.5;

[0026] Step S02, calculating the total loss L according to the binary cross entropy loss function LG, the reconstruction loss Lr, the generator weight coefficient λ1 and the discriminator weight coefficient λ2, obtaining the total loss L, and setting L = λ1 × LG + λ2 × Lr;

[0027] Step S03: Compare the total loss L with the preset total loss L0, set 0.22≤L0≤0.31, judge the accuracy of the convolutional neural network model based on the comparison result, and adjust the binary cross entropy loss function LG based on the judgment result, where:

[0028] When L≤L0, the installed environment perception module determines that the accuracy of the convolutional neural network model is accurate, does not perform loss adjustment on the binary cross entropy loss function LG, and outputs the convolutional neural network model as the environment perception model;

[0029] When L>L0, the installation environment perception module determines that the accuracy of the convolutional neural network model is inaccurate, and adjusts the binary cross entropy loss function LG. According to the adjustment coefficient θ1, set Perform loss adjustment on the binary cross entropy loss function LG to obtain the adjusted binary cross entropy loss function LG', set LG'=LG×θ1, replace the binary cross entropy loss function LG with the adjusted binary cross entropy loss function LG', and recalculate the total loss L;

[0030] Step S04: Acquire the current geographical environment, compare the current geographical environment with the historical geographical environment, determine the consistency between the current geographical environment and the historical geographical environment based on the comparison result, and perform a secondary loss adjustment on the preset total loss L0 based on the determination result, wherein:

[0031] When the current geographical environment is consistent with the historical geographical environment, no secondary loss adjustment is made to the preset total loss L0;

[0032] When the current geographical environment is inconsistent with the historical geographical environment, a secondary loss adjustment is performed on the preset total loss L0. According to the environmental coefficient γ, 1.1≤γ≤1.5 is set, and the preset total loss L0 is adjusted to obtain the adjusted preset total loss L0`. L0`=γ×L0 is set, and the preset total loss L0 is replaced by the adjusted preset total loss L0`. The total loss L is then re-compared with the adjusted preset total loss L0`.

[0033] Step S05 , calculating the environmental change factor Hy according to the temperature difference Wc and the light intensity difference Gq to obtain the environmental change factor Hy, and setting Hy=0.58×Wc+0.42×Gq;

[0034] Step S06: Compare the environmental change factor Hy with the preset environmental change factor Hy0, set 0.16≤Hy0≤0.39, judge the environmental change state according to the comparison result, and perform a third loss adjustment on the adjustment process of the second loss adjustment according to the judgment result, wherein:

[0035] When Hy≤Hy0, the installation environment perception module determines that the environmental change state is a small-scale change, and does not perform the third loss adjustment in the adjustment process of the second loss adjustment;

[0036] When Hy>Hy0, the installed environment perception module determines that the environmental change state is a large-scale change, and performs a third loss adjustment on the adjustment process of the secondary loss adjustment. According to the environmental adjustment coefficient θ2, θ2=1+(Hy-Hy0) / Hy0 is set, and the adjustment process of the secondary loss adjustment is performed three times to obtain the preset total loss L0`` after the secondary adjustment. L0``=L0×θ2 is set, and the preset total loss L0 is replaced with the preset total loss L0`` after the secondary adjustment, and the total loss L is re-compared with the preset total loss L0`` after the secondary adjustment.

[0037] Furthermore, when the installed environmental perception module acquires the reflection elimination data according to the environmental perception model, the fog point cloud data is input into the environmental perception model to obtain clear point cloud data, the optimized transmittance τq and the clear point cloud data are input into the reflection elimination information model to obtain the target category probability, target position and target posture information, and the target category probability, target position and target posture information are output as reflection elimination data.

[0038] Furthermore, the intelligent risk prediction module obtains the safe area boundary distance according to the reflection elimination data using a safe area boundary distance acquisition method, and the safe area boundary distance acquisition method includes:

[0039] Step C01, obtaining the target category according to the target category probability in the reflection removal data;

[0040] Step C03: constructing an intelligent twin model based on target category, target location, and target posture information;

[0041] Step C04: construct a safe area in the intelligent twin model, set the center coordinates of the safe area to (x0, y0, z0) and the radius to rq;

[0042] Step C04, constructing a risk prediction model and outputting the target location coordinates from the risk prediction model;

[0043] Step C05, calculate the safety area boundary distance d based on the target position coordinate point CRM (xm, ym, zm), the safety area center coordinates CRR (x0, y0, z0) and the safety area radius rq, and obtain the safety area boundary distance d, set

[0044] Furthermore, when the intelligent risk prediction module adjusts the environmental perception model according to the safe area boundary distance, the intelligent risk prediction module compares the safe area boundary distance d with the first preset boundary distance d1 and the second preset boundary distance d2, sets d1=10 cm and d2=15 cm, judges the risk level of the target location according to the comparison result, and adjusts the environmental perception model according to the judgment result, wherein:

[0045] When d≤d1, the intelligent risk prediction module determines that the risk level of the target location is low risk and does not adjust the environment perception model;

[0046] When d1<d≤d2, the intelligent risk prediction module determines that the risk level of the target location is medium risk, and adjusts the environmental perception model. According to the first model adjustment coefficient α1, α1=1+(d-d1) / d2 is set, and the environmental perception model is adjusted to obtain the first adjusted generator weight coefficient λ1`, and λ1`=λ1×α1 is set. The generator weight coefficient λ1 is replaced with the first adjusted generator weight coefficient λ1`, and the total loss L in the environmental perception model is recalculated according to the first adjusted generator weight coefficient λ1`;

[0047] When d>d2, the intelligent risk prediction module determines that the risk level of the target location is high risk, and adjusts the environmental perception model. According to the second model adjustment coefficient α2, α2=1.5+(d-d2) / d2 is set, and the environmental perception model is adjusted to obtain the second adjusted generator weight coefficient λ1``, and λ1``=λ1×α2 is set. The generator weight coefficient λ1 is replaced with the second adjusted generator weight coefficient λ1``, and the total loss L in the environmental perception model is recalculated according to the second adjusted generator weight coefficient λ1``.

[0048] Furthermore, when the intelligent risk prediction module performs distance adjustment on the model adjustment process according to the target movement speed, the intelligent risk prediction module compares the target movement speed Mv with the preset movement speed Mv0, sets 0.5m / s≤Mv0≤0.7m / s, judges the target movement speed state according to the comparison result, and adjusts the safety zone boundary distance d according to the judgment result, wherein:

[0049] When Mv≤Mv0, the intelligent risk prediction module determines that the target moving speed is slow and does not adjust the distance d of the safety zone boundary;

[0050] When Mv>Mv0, the intelligent risk prediction module determines that the target moving speed state is fast, adjusts the distance d of the safety area boundary, sets d>d2, determines that the target position risk level is high risk, and adjusts the environmental perception model.

[0051] Furthermore, when the intelligent risk prediction module adjusts the speed of the distance adjustment process according to the metal roof slope, the intelligent risk prediction module compares the metal roof slope Jp with the preset metal roof slope Jp0, sets 15°≤Jp0≤30°, judges the metal roof slope state according to the comparison result, and adjusts the preset moving speed Mv0 according to the judgment result, wherein:

[0052] When Jp≤Jp0, the intelligent risk prediction module determines that the slope of the metal roof is gentle and does not adjust the preset moving speed Mv0;

[0053] When Jp>Jp0, the intelligent risk prediction module determines that the slope of the metal roof is not gentle, adjusts the preset moving speed Mv0, and sets the slope coefficient δ. The preset moving speed Mv0 is adjusted to obtain the adjusted preset moving speed Mv0`, Mv0` is set to Mv0×δ, the target moving speed Mv is replaced with the adjusted preset moving speed Mv0`, and the target moving speed Mv is re-compared with the adjusted preset moving speed Mv0`.

[0054] Furthermore, the edge intelligent collaboration module measures computing resources within the safe area boundary distance using a computing resource measurement method based on the intelligent twin model, and the computing resource measurement method includes:

[0055] Step E01: define the state parameter Si for the edge node Ni, and set Si = {Ci, Mi, Bi}, where Ci is the computing capability value, Mi is the memory evaluation value, and Bi is the network bandwidth evaluation value;

[0056] Step E02 , calculating the computing capacity value Ci according to the current available CPU cycle number Cia and the total CPU cycle number Cit to obtain the computing capacity value Ci, and setting Ci=Cia / Cit;

[0057] Step E03, calculating the memory evaluation value Mi according to the current available memory Mia and the total memory Mit to obtain the memory evaluation value Mi, and setting Mi=Mia / Mit;

[0058] Step E04: Calculate the network bandwidth evaluation value Bi based on the current available network bandwidth Bia and the total network bandwidth Bit to obtain the network bandwidth evaluation value Bi, and set Bi=Bia / Bit;

[0059] In step E05 , the computing capability value Ci, the memory evaluation value Mi, the network bandwidth evaluation value Bi, and the load evaluation value Li are used as computing resource measurement results.

[0060] Compared with the existing technology, the beneficial effect of the present invention lies in that the various modules work together to eliminate environmental interference, dynamically regulate risks, intelligently allocate resources and precisely control the manipulator during the metal roof installation process, so as to cope with the complex working conditions of machine intelligence in reflective conditions and rainy and foggy weather. At the same time, it accurately predicts installation risks and dynamically allocates computing resources, thereby improving the accuracy, safety and resource utilization of machine intelligence when applied to metal roof installation. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 This is a schematic diagram of the structure of the metal roof intelligent installation control system based on machine intelligence in this embodiment;

[0062] Figure 2 This is a schematic diagram of the structure of the intelligent installation device for metal roofing in this embodiment. DETAILED DESCRIPTION

[0063] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention.

[0064] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0065] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside", and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0066] Furthermore, it should be noted that, in the description of the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0067] See also Figure 1 As shown in FIG, which is a structural diagram of the metal roof intelligent installation control system based on machine intelligence in this embodiment, the system includes:

[0068] Smart installation data acquisition module, used to acquire smart installation data;

[0069] An environment perception module is installed to judge the reflective area according to the intelligent installation data to obtain a reflection judgment result, eliminate the reflection interference according to the reflection judgment result, obtain target environment data, build an environment perception model according to the target environment data, and obtain reflection elimination data according to the environment perception model. The environment perception module is installed and connected to the intelligent installation data acquisition module;

[0070] An intelligent risk prediction module is used to obtain the safe area boundary distance based on the reflection elimination data and adjust the environment perception model based on the safe area boundary distance. The intelligent risk prediction module is connected to the installation environment perception module;

[0071] The edge intelligent collaboration module is used to measure computing resources within the boundary distance of the safe area according to the intelligent twin model to obtain computing resource measurement results. It is also used to perform edge intelligent collaboration on the computing resource measurement results according to the edge intelligent collaboration method and to adjust the penalty of the edge intelligent collaboration method. The edge intelligent collaboration module is connected to the intelligent risk prediction module;

[0072] The intelligent manipulator control module is used to perform initial task allocation on the computing resource measurement results and to adjust the allocation process of the initial task allocation. The intelligent manipulator control module is connected to the edge intelligent collaboration module.

[0073] Specifically, the machine intelligence-based intelligent installation control system for metal roofs is applied to intelligent installation devices for metal roofs. The machine intelligence-based intelligent installation control system for metal roofs eliminates environmental interference, dynamically regulates risks, intelligently allocates resources, and precisely controls manipulators during the metal roof installation process through the collaboration of various modules, so as to cope with complex working conditions under reflective conditions and rainy and foggy weather, and dynamically allocates computing resources, thereby improving the safety, accuracy, and resource utilization of metal roof installation. Among them, the machine intelligence-based intelligent installation control system for metal roofs acquires intelligent installation data through an intelligent installation data acquisition module, so as to facilitate subsequent installation control based on the intelligent installation data and improve the accuracy of the installation process. The machine intelligence-based intelligent installation control system for metal roofs senses the metal roof installation environment through an installation environment perception module. Perception, so as to use machine intelligence to dynamically adjust the installation method according to the installation environmental conditions, thereby improving the safety and accuracy of the installation. The metal roof intelligent installation control system based on machine intelligence uses an intelligent risk prediction module to predict the environmental risks of metal roof operations in real time, so as to optimize the operation method of machine intelligence in real time, thereby improving the safety of metal roof installation. The metal roof intelligent installation control system based on machine intelligence decomposes the computing tasks through the edge intelligent collaboration module and selects the best edge node to perform the computing tasks, thereby ensuring the efficient use of computing resources and improving resource utilization. The metal roof intelligent installation control system based on machine intelligence controls the manipulator through the intelligent manipulator control module to increase the accuracy of the manipulator in metal roof installation, thereby improving the operational safety and accuracy of machine intelligence.

[0074] Specifically, the intelligent installation data acquisition module acquires the intelligent installation data, which includes polarization data, cloud point cloud data, ambient temperature, light intensity, target moving speed and metal roof slope. The polarization data includes the first polarization intensity I0, the second polarization intensity I 45 , the third polarized light intensity I 90 and the fourth polarized light intensity I 135 The intelligent installation data acquisition module acquires polarization data through a polarization camera, acquires cloud point cloud data through a lidar, acquires ambient temperature through a temperature sensor, acquires light intensity through a light intensity sensor, acquires target moving speed through a speed sensor, and acquires the slope of the metal roof through an angle sensor.

[0075] Specifically, the first polarized light intensity I0 refers to the light intensity value collected when the polarization direction is 0°, and the second polarized light intensity I45 Refers to the light intensity value collected when the polarization direction is 45°, the third polarized light intensity I 90 Refers to the light intensity value collected when the polarization direction is 90°, the four polarized light intensity I 135 Refers to the light intensity value collected when the polarization direction is 135°.

[0076] Specifically, the intelligent installation data acquisition module acquires the intelligent installation data so as to perform installation control according to the intelligent installation data and improve the accuracy of the installation process.

[0077] Specifically, the installation environment perception module judges the reflective area according to the intelligent installation data using a reflective judgment method, and the reflective interference elimination method includes:

[0078] Step A01: According to the first polarized light intensity I0 and the second polarized light intensity I 45 , the third polarized light intensity I 90 and the fourth polarized light intensity I 135 Calculate the total light intensity S0 and get the total light intensity S0, set S0 = I0 + I 45 +I 90 +I 135 ;

[0079] Step A02: according to the first polarized light intensity I0 and the third polarized light intensity I 90 Calculate the first light intensity difference S1 to obtain the first light intensity difference S1, and set S1 = I0-I 90 ;

[0080] Step A03, according to the second polarized light intensity I 45 and the fourth polarized light intensity I 135 Calculate the second light intensity difference S2 to obtain the second light intensity difference S2, and set S2 = I 45 -I 135 ;

[0081] Step A04, calculate the polarization degree P according to the total light intensity S0, the first light intensity difference S1 and the second light intensity difference S2, and obtain the polarization degree P, and set

[0082] Step A05: Compare the polarization degree P with the preset polarization degree P0, set 0.6≤P0≤0.8, judge the reflective state of the roof area based on the comparison result, and output the reflective judgment result based on the judgment result, wherein:

[0083] When P>P0, the installation environment perception module determines that the reflective state of the roof area is reflective, and outputs the reflective area as the reflective judgment result;

[0084] When P≤P0, the installation environment perception module determines that the reflective state of the roof area is non-reflective, and outputs the non-reflective area as the reflective judgment result.

[0085] Specifically, the total light intensity refers to the total intensity of the optical signal calculated based on the first polarized light intensity, the second polarized light intensity, the third polarized light intensity and the fourth polarized light intensity, which is used to characterize the brightness information of the roof area. The first light intensity difference refers to the value calculated based on the first polarized light intensity and the third polarized light intensity to characterize the polarization degree in the horizontal and vertical directions of the roof area. The second light intensity difference refers to the value calculated based on the second polarized light intensity and the fourth polarized light intensity to characterize the polarization degree in the diagonal direction of the roof area. The polarization degree refers to the physical property that describes the vibration direction and spatial distribution law of light in the roof area. The polarization degree refers to a value calculated based on the total light intensity, the first light intensity difference, and the second light intensity difference to reflect the overall polarization degree of the roof area. The preset polarization degree refers to a preset value used to judge the reflective state of the roof area. The reflective state of the roof area refers to the situation in which the roof area reflects light under illumination. The reflective state of the roof area includes the reflective state of the roof area being reflective and the reflective state of the roof area being non-reflective. The reflection judgment result refers to the reflective state of the roof area output according to the reflective state of the roof area, and the reflection judgment result includes a reflective area and a non-reflective area.

[0086] Specifically, the installation environment perception module judges the reflective area to accurately select the reflective area of ​​the roof, facilitates the subsequent de-reflection processing of the reflective area of ​​the roof, removes reflective interference, and thus improves the accuracy of installation.

[0087] Specifically, the installation environment perception module performs a de-reflection process on the reflective area according to a de-reflection process method, and the de-reflection process method includes:

[0088] Step B01: Based on the dark channel I of the image dark ={I r , I g , I b}, transmittance control constant ω, set 0.9≤ω≤0.95, and the background light estimation value A to calculate the initial transmittance τ0, and get the initial transmittance τ0, set

[0089] Step B02, calculating the optimized transmittance τq based on the first optimization coefficient ak, the second optimization coefficient bk and the reflective point pixel Iq, obtaining the optimized transmittance τq, and setting τq = ak × Iq ​​+ bk;

[0090] Step B03, calculate the real texture channel Jc according to the optimized transmittance τq, the first channel intensity I c and the second channel intensity Ac, and obtain the real texture channel Jc, and set

[0091] Step B04: output the real texture channel Jc and the optimized transmittance τ0 as target environment data.

[0092] Specifically, the dark channel of the image refers to the minimum value of the pixel point in the reflective area in the RGB channel, including the minimum value I corresponding to the red channel r , the green channel corresponds to the minimum value I g , the blue channel corresponds to the minimum value I b The RGB channel refers to a method for describing the color of pixels in the reflective area, and all visible colors are represented by different intensities of the three primary colors of red, green, and blue, including a red channel, a green channel, and a blue channel. The transmittance control constant refers to a preset value that controls the degree of influence of the image dark channel on the transmittance. The background light estimation value refers to the 95% quantile of the image dark channel, which is used to separate the background light value of the pixels in the reflective area. The background light refers to the light intensity in the reflective area. For example, the image dark channels of all pixels in the reflective area are arranged in ascending order to obtain an image dark channel sequence. It is assumed that there are m image dark channels in the image dark channel sequence, and the background light estimation value is s. Setting s = m × 95% obtains the background light estimation value s. The first optimization coefficient refers to a coefficient that measures the importance of the reflective point pixel when calculating the optimized transmittance τq. The second optimization coefficient refers to a coefficient that affects the optimized transmittance when calculating the optimized transmittance τq. This embodiment does not limit the method for obtaining the first optimization coefficient and the second optimization coefficient. Relevant technicians in the field can freely choose according to actual needs, such as obtaining the first optimization coefficient and the second optimization coefficient through the least squares method. The reflective point pixel refers to the pixel value corresponding to the pixel point in the reflective area. This embodiment does not limit the method for obtaining the reflective point pixel. Relevant technicians in this field can freely choose according to actual needs, such as obtaining the reflective point pixel through software measurement. The first channel intensity refers to the pixel value size of the pixel point in the reflective area in the RGB channel, which is used to reflect the brightness of the pixel color in the reflective area in the image. The second channel intensity refers to the pixel value size of the background light estimation value in the RGB channel, which is used to reflect the brightness of the background light color in the image. This embodiment does not limit the method for obtaining the first channel intensity and the second channel intensity. Relevant technicians in this field can freely choose according to actual needs, such as obtaining through software measurement. The real texture channel refers to the actual pixel value of the pixel point in the reflective area when it is not reflective, calculated based on the optimized transmittance, the first channel intensity, and the second channel intensity.

[0093] Specifically, the installation environment perception module restores the true physical shape of the roof by de-reflecting the reflective area, so as to avoid inaccurate judgment of the true shape of the roof caused by reflected light during the installation of the metal roof, thereby improving the accuracy of the metal roof installation.

[0094] Specifically, the installation environment perception module constructs an environment perception model according to the target environment data through an environment perception model construction method, and the environment perception model construction method includes:

[0095] Step S01, initialize the parameters of the convolutional neural network model according to the historical environment database: set the generator weight coefficient λ1 and the discriminator weight coefficient λ2, setting λ1 = 0.5, λ2 = 1-λ1 = 0.5;

[0096] Step S02, calculating the total loss L according to the binary cross entropy loss function LG, the reconstruction loss Lr, the generator weight coefficient λ1 and the discriminator weight coefficient λ2, obtaining the total loss L, and setting L = λ1 × LG + λ2 × Lr;

[0097] Step S03: Compare the total loss L with the preset total loss L0, set 0.22≤L0≤0.31, judge the accuracy of the convolutional neural network model based on the comparison result, and adjust the binary cross entropy loss function LG based on the judgment result, where:

[0098] When L≤L0, the installed environment perception module determines that the accuracy of the convolutional neural network model is accurate, does not perform loss adjustment on the binary cross entropy loss function LG, and outputs the convolutional neural network model as the environment perception model;

[0099] When L>L0, the installation environment perception module determines that the accuracy of the convolutional neural network model is inaccurate, and adjusts the binary cross entropy loss function LG. According to the adjustment coefficient θ1, set Perform loss adjustment on the binary cross entropy loss function LG to obtain the adjusted binary cross entropy loss function LG', set LG'=LG×θ1, replace the binary cross entropy loss function LG with the adjusted binary cross entropy loss function LG', and recalculate the total loss L;

[0100] Step S04: Acquire the current geographical environment, compare the current geographical environment with the historical geographical environment, determine the consistency between the current geographical environment and the historical geographical environment based on the comparison result, and perform a secondary loss adjustment on the preset total loss L0 based on the determination result, wherein:

[0101] When the current geographical environment is consistent with the historical geographical environment, no secondary loss adjustment is made to the preset total loss L0;

[0102] When the current geographical environment is inconsistent with the historical geographical environment, a secondary loss adjustment is performed on the preset total loss L0. According to the environmental coefficient γ, 1.1≤γ≤1.5 is set, and the preset total loss L0 is adjusted to obtain the adjusted preset total loss L0`. L0`=γ×L0 is set, and the preset total loss L0 is replaced by the adjusted preset total loss L0`. The total loss L is then re-compared with the adjusted preset total loss L0`.

[0103] Step S05 , calculating the environmental change factor Hy according to the temperature difference Wc and the light intensity difference Gq to obtain the environmental change factor Hy, and setting Hy=0.58×Wc+0.42×Gq;

[0104] Step S06: Compare the environmental change factor Hy with the preset environmental change factor Hy0, set 0.16≤Hy0≤0.39, judge the environmental change state according to the comparison result, and perform a third loss adjustment on the adjustment process of the second loss adjustment according to the judgment result, wherein:

[0105] When Hy≤Hy0, the installation environment perception module determines that the environmental change state is a small-scale change, and does not perform the third loss adjustment in the adjustment process of the second loss adjustment;

[0106] When Hy>Hy0, the installed environment perception module determines that the environmental change state is a large-scale change, and performs a third loss adjustment on the adjustment process of the secondary loss adjustment. According to the environmental adjustment coefficient θ2, θ2=1+(Hy-Hy0) / Hy0 is set, and the adjustment process of the secondary loss adjustment is performed three times to obtain the preset total loss L0`` after the secondary adjustment. L0``=L0×θ2 is set, and the preset total loss L0 is replaced with the preset total loss L0`` after the secondary adjustment, and the total loss L is re-compared with the preset total loss L0`` after the secondary adjustment.

[0107] Specifically, the historical environment database refers to the historical rain and fog point cloud data and the clear point cloud data corresponding to the historical rain and fog point cloud data. The historical rain and fog point cloud data is used as the input data of the environment perception model, and the clear point cloud data corresponding to the historical rain and fog point cloud data is used as the output data of the environment perception model. The historical rain and fog point cloud data refers to the three-dimensional point cloud data of the metal roof collected under rainy and foggy weather conditions in history. The clear point cloud data refers to the three-dimensional point cloud data generated in rainless and foggy weather in history. The three-dimensional point cloud data refers to a collection of a large number of discrete points in the metal roof obtained by lidar, which is used to reflect the shape contour and surface texture of the metal roof. The convolutional neural network model refers to a neural network used to describe the historical rain and fog point cloud data. According to the statistical model of the linear relationship between the clear point cloud data corresponding to the historical rain and fog point cloud data, the parameter initialization refers to the process of setting the generator weight coefficient and the discriminator weight coefficient of the convolutional neural network model. This embodiment does not limit the specific setting method of parameter initialization. Relevant technical personnel in this field can freely choose according to actual needs, such as manual setting. The generator weight coefficient refers to the coefficient used to control the strength of the connection between neurons in the convolutional neural network model, thereby determining the mapping relationship between the historical rain and fog point cloud data and the clear point cloud data corresponding to the historical rain and fog point cloud data. The discriminator weight coefficient refers to the coefficient used to control the convolutional neural network model's ability to discriminate historical rain and fog point cloud data. The two The meta-cross entropy loss function refers to a function that measures the difference between the output data of the convolutional neural network model and the real data, and is used to optimize the convolutional neural network model. The reconstruction loss refers to a function that measures the physical morphology difference between the output data of the optimized convolutional neural network model and the real clear point cloud, and is used to constrain the physical authenticity of the output data. The preset total loss L0 refers to a preset value for judging the accuracy of the convolutional neural network model. The accuracy of the convolutional neural network model refers to the judgment based on the total loss and the preset total loss to reflect the matching accuracy between the output data of the convolutional neural network model and the real data. The accuracy of the convolutional neural network model includes the accuracy of the convolutional neural network model as accurate and the convolution neural network as accurate. The accuracy of the network model is inaccurate. The current geographical environment refers to the natural geographical conditions at the time point of the current metal roof installation, such as wind, sand and humidity. The historical geographical environment refers to the natural geographical conditions when the metal roof was installed last time. This embodiment does not limit the method of obtaining the current geographical environment and the historical geographical environment. Relevant technicians in this field can freely choose according to actual needs, such as obtaining through the Internet. The consistency between the current geographical environment and the historical geographical environment refers to the degree of matching between the current geographical environment and the historical geographical environment. The consistency between the current geographical environment and the historical geographical environment includes the consistency between the current geographical environment and the historical geographical environment and the inconsistency between the current geographical environment and the historical geographical environment.The temperature difference Wc refers to the difference between the ambient temperature wt at the current metal roof installation time point and the ambient temperature wk at the previous moment, with Wc set to wt-wk. The ambient temperature refers to the ambient temperature at the current metal roof installation time point. The light intensity difference Gq refers to the difference between the light intensity gt at the current metal roof installation time point and the ambient temperature gk at the previous moment, with Gq set to gt-gk. The light intensity is a numerical value that measures the light intensity in the environment at the current metal roof installation time point. The environmental change factor Hy is a numerical value calculated based on the temperature difference Wc and the light intensity difference Gq to reflect changes in environmental conditions. The preset environmental change factor Hy0 is a preset value for determining the environmental change state. The environmental change state refers to the scale of the environmental change range determined based on the environmental change factor Hy and the preset environmental change factor Hy0. The environmental change state includes a small-scale change and a large-scale change.

[0108] Specifically, the installation environment perception module improves the ability to perceive the environment during installation by constructing an environmental perception model, thereby flexibly adjusting the installation process and improving installation accuracy and safety. When the accuracy of the convolutional neural network model is inaccurate, the binary cross entropy loss function is reduced to improve the authenticity of the output data of the environmental perception model and improve installation accuracy. When the current geographical environment is inconsistent with the historical geographical environment, the preset total loss is increased to avoid inaccurate results of the environmental perception model due to changes in environmental conditions, thereby improving the adaptability of the environmental perception model. When the environmental change state is a large-scale change, the preset total loss is increased to avoid the impact of large changes in environmental conditions on the installation process and improve the safety and accuracy of the installation.

[0109] Specifically, when the installed environmental perception module obtains the reflection elimination data according to the environmental perception model, the cloud point cloud data is input into the environmental perception model to obtain clear point cloud data, and the optimized transmittance τq and clear point cloud data are input into the reflection elimination information model to obtain the target category probability, target position and target posture information, and the target category probability, target position and target posture information are output as reflection elimination data.

[0110] Specifically, the cloud point cloud data refers to the three-dimensional point cloud data of the metal roof collected by the lidar in foggy weather, the clear point cloud data refers to the data corresponding to the cloud point cloud data obtained by the environmental perception model, which is used to reflect the real physical form of the metal roof, and the reflection elimination information model refers to a convolutional neural network model with the optimized transmittance and clear point cloud data as input data, and the target category probability, target position and target posture information corresponding to the optimized transmittance and clear point cloud data as output data. The installed environmental perception module constructs the reflection elimination information model through the reflection elimination information model construction method. This embodiment does not use the reflection elimination information model. The specific method of model construction is limited, and relevant technical personnel in this field can freely choose according to actual needs, such as using historical optimized transmittance, historical clear point cloud data, historical target category probability, historical target position and historical target posture information as data sets for training the reflection elimination information model, and training the reflection elimination information model. The target category probability refers to the probability of classifying the metal roof component, such as the probability of determining that the target category is a metal roof fixing bracket is 97%, the target position refers to the specific location of the metal roof component, and the target posture information refers to the physical form of the metal roof component, such as the placement direction of the metal roof fixing bracket.

[0111] Specifically, the installation environment perception module acquires the reflection elimination data through the reflection elimination information model, so as to facilitate subsequent accurate installation according to the reflection elimination data, thereby improving installation accuracy.

[0112] Specifically, the intelligent risk prediction module obtains the safe area boundary distance according to the reflection elimination data using a safe area boundary distance acquisition method, and the safe area boundary distance acquisition method includes:

[0113] Step C01, obtaining the target category according to the target category probability in the reflection removal data;

[0114] Step C03: constructing an intelligent twin model based on target category, target location, and target posture information;

[0115] Step C04: construct a safe area in the intelligent twin model, set the center coordinates of the safe area to (x0, y0, z0) and the radius to rq;

[0116] Step C04, constructing a risk prediction model and outputting the target location coordinates from the risk prediction model;

[0117] Step C05, calculate the safety area boundary distance d based on the target position coordinate point CRM (xm, ym, zm), the safety area center coordinates CRR (x0, y0, z0) and the safety area radius rq, and obtain the safety area boundary distance d, set

[0118] Specifically, the target category refers to the specific component category of the metal roof output according to the target category probability in the reflection elimination data. This embodiment does not limit the specific method of obtaining the target category according to the target category probability in the reflection elimination data. Relevant technical personnel in this field can freely choose according to actual needs, such as setting the category with a target category probability of 90% as the target category. The intelligent twin model refers to the visual model obtained by modeling according to the target category. This embodiment does not limit the specific construction method of the intelligent twin model. Relevant technical personnel in this field can freely choose according to actual needs, such as through software modeling. The safe area refers to the spherical safe operating range pre-set in the intelligent twin model. This embodiment There is no limitation on the specific setting method of the safety area. Relevant technicians in this field can freely choose according to actual needs, such as software setting. The center coordinates of the safety area refer to the center positioning coordinates set in advance in the safety area. This embodiment does not limit the specific positioning method of the center coordinates of the safety area. Relevant technicians in this field can freely choose according to actual needs, such as software setting for positioning. The x0 refers to the value of the center of the safety area on the X axis in the three-dimensional coordinate system, the y0 refers to the value of the center of the safety area on the Y axis in the three-dimensional coordinate system, and the z0 refers to the value of the center of the safety area on the Z axis in the three-dimensional coordinate system. The three-dimensional coordinate system refers to the coordinates of the center of the safety area established in the safety area for positioning the center coordinates of the safety area. The positioning coordinate system, this embodiment does not limit the specific establishment method of the three-dimensional coordinate system. Relevant technical personnel in this field can freely choose according to actual needs, such as setting the current installation operation position as the origin position of the three-dimensional coordinate system. The safety area radius refers to the radius length of the safety area. This embodiment does not limit the safety area radius. Relevant technical personnel in this field can freely choose according to actual needs, such as setting rq = 50cm. The risk prediction model refers to a convolutional neural network model that uses target category, target position and target posture information as input data and the target position coordinates corresponding to the target category, target position and target posture information as output data. The target position coordinates are used to represent the target position at the future ty time point. Among them, the xm refers to the value of the target position on the X-axis in the three-dimensional coordinate system, the ym refers to the value of the target position on the Y-axis in the three-dimensional coordinate system, and the zm refers to the value of the target position on the Z-axis in the three-dimensional coordinate system. The installation environment perception module constructs the risk prediction model through the risk prediction model construction method. This embodiment does not limit the specific method of the risk prediction model construction method. Relevant technical personnel in this field can freely choose according to actual needs, such as using historical target categories, historical target positions, historical target posture information, and target position coordinates as data sets for training the risk prediction model to train the risk prediction model. The safe area boundary distance refers to the distance between the target position coordinates and the safe area boundary.

[0119] Specifically, the intelligent risk prediction module obtains the distance to the boundary of the safe area so as to perform risk prediction based on the distance to the boundary of the safe area, thereby improving the safety during the installation process.

[0120] Specifically, when the intelligent risk prediction module adjusts the environmental perception model according to the safe area boundary distance, the intelligent risk prediction module compares the safe area boundary distance d with the first preset boundary distance d1 and the second preset boundary distance d2, sets d1 = 10 cm and d2 = 15 cm, judges the risk level of the target location based on the comparison result, and adjusts the environmental perception model based on the judgment result, wherein:

[0121] When d≤d1, the intelligent risk prediction module determines that the risk level of the target location is low risk and does not adjust the environment perception model;

[0122] When d1<d≤d2, the intelligent risk prediction module determines that the risk level of the target location is medium risk, and adjusts the environmental perception model. According to the first model adjustment coefficient α1, α1=1+(d-d1) / d2 is set, and the environmental perception model is adjusted to obtain the first adjusted generator weight coefficient λ1`, and λ1`=λ1×α1 is set. The generator weight coefficient λ1 is replaced with the first adjusted generator weight coefficient λ1`, and the total loss L in the environmental perception model is recalculated according to the first adjusted generator weight coefficient λ1`;

[0123] When d>d2, the intelligent risk prediction module determines that the risk level of the target location is high risk, and adjusts the environmental perception model. According to the second model adjustment coefficient α2, α2=1.5+(d-d2) / d2 is set, and the environmental perception model is adjusted to obtain the second adjusted generator weight coefficient λ1``, and λ1``=λ1×α2 is set. The generator weight coefficient λ1 is replaced with the second adjusted generator weight coefficient λ1``, and the total loss L in the environmental perception model is recalculated according to the second adjusted generator weight coefficient λ1``.

[0124] Specifically, the first preset boundary distance d1 refers to the lower limit of the preset value for judging the risk level of the target position, and the second preset boundary distance d2 refers to the upper limit of the preset value for judging the risk level of the target position. The target position risk level refers to the operational risk level of the target position coordinates judged based on the safety area boundary distance d, the first preset boundary distance d1 and the second preset boundary distance d2. The target position risk level includes a low risk target position risk level, a medium risk target position risk level and a high risk target position risk level.

[0125] Specifically, the installation environment perception module judges the risk level of the target location so that the generator weight coefficient in the environment perception model is adjusted as the distance d from the boundary of the safe area changes. When the risk level of the target location is medium, the generator weight coefficient is increased to increase the strength of the connection between neurons in the convolutional neural network model and improve the accuracy of the environment perception model. When the risk level of the target location is high, the generator weight coefficient is greatly increased to greatly enhance the analysis of the environment perception model, thereby improving the accuracy and efficiency of the installation operation.

[0126] Specifically, when the intelligent risk prediction module performs distance adjustment on the model adjustment process according to the target movement speed, the intelligent risk prediction module compares the target movement speed Mv with the preset movement speed Mv0, sets 0.5m / s≤Mv0≤0.7m / s, judges the target movement speed state according to the comparison result, and adjusts the safety zone boundary distance d according to the judgment result, wherein:

[0127] When Mv≤Mv0, the intelligent risk prediction module determines that the target moving speed is slow and does not adjust the distance d of the safety zone boundary;

[0128] When Mv>Mv0, the intelligent risk prediction module determines that the target moving speed state is fast, adjusts the distance d of the safety area boundary, sets d>d2, determines that the target position risk level is high risk, and adjusts the environmental perception model.

[0129] Specifically, the target moving speed refers to the moving speed of the installation component when installing the metal roof, the preset moving speed Mv0 refers to the preset value for judging the target moving speed state, and the target moving speed state refers to the moving speed of the installation component judged based on the target moving speed Mv and the preset moving speed Mv0. The target moving speed state includes a target moving speed state of slow and a target moving speed state of fast.

[0130] Specifically, the installation environment perception module judges the target movement speed state. When the target movement speed is too fast, it directly defines the target position risk level as high risk and adjusts the environment perception model to avoid the impact of the target movement speed on the installation accuracy, thereby improving the accuracy and efficiency of the installation.

[0131] Specifically, when the intelligent risk prediction module adjusts the speed of the distance adjustment process according to the metal roof slope, the intelligent risk prediction module compares the metal roof slope Jp with the preset metal roof slope Jp0, sets 15°≤Jp0≤30°, judges the metal roof slope state according to the comparison result, and adjusts the preset moving speed Mv0 according to the judgment result, wherein:

[0132] When Jp≤Jp0, the intelligent risk prediction module determines that the slope of the metal roof is gentle and does not adjust the preset moving speed Mv0;

[0133] When Jp>Jp0, the intelligent risk prediction module determines that the slope of the metal roof is not gentle, adjusts the preset moving speed Mv0, and sets the slope coefficient δ. The preset moving speed Mv0 is adjusted to obtain the adjusted preset moving speed Mv0`, Mv0` is set to Mv0×δ, the target moving speed Mv is replaced with the adjusted preset moving speed Mv0`, and the target moving speed Mv is re-compared with the adjusted preset moving speed Mv0`.

[0134] Specifically, the metal roof slope refers to the inclination of the metal roof, the preset metal roof slope refers to a preset value for judging the slope state of the metal roof, the metal roof slope state refers to the flatness of the metal roof slope judged based on the metal roof slope and the preset metal roof slope, and the metal roof slope state includes the metal roof slope state being flat and the metal roof slope state being uneven.

[0135] Specifically, the intelligent risk prediction module judges the slope status of the metal roof. When the slope status of the metal roof is large, it reduces the preset moving speed to reduce the risk of manipulator movement caused by roof inclination, avoid installation deviation, and thus improve the safety and efficiency of installation.

[0136] Specifically, the edge intelligent collaboration module calculates computing resources within the safe area boundary distance using a computing resource calculation method based on the intelligent twin model. The computing resource calculation method includes:

[0137] Step E01: define the state parameter Si for the edge node Ni, and set Si = {Ci, Mi, Bi}, where Ci is the computing capability value, Mi is the memory evaluation value, and Bi is the network bandwidth evaluation value;

[0138] Step E02 , calculating the computing capacity value Ci according to the current available CPU cycle number Cia and the total CPU cycle number Cit to obtain the computing capacity value Ci, and setting Ci=Cia / Cit;

[0139] Step E03, calculating the memory evaluation value Mi according to the current available memory Mia and the total memory Mit to obtain the memory evaluation value Mi, and setting Mi=Mia / Mit;

[0140] Step E04: Calculate the network bandwidth evaluation value Bi based on the current available network bandwidth Bia and the total network bandwidth Bit to obtain the network bandwidth evaluation value Bi, and set Bi=Bia / Bit;

[0141] In step E05 , the computing capability value Ci, the memory evaluation value Mi, the network bandwidth evaluation value Bi, and the load evaluation value Li are used as computing resource measurement results.

[0142] Specifically, the edge node refers to the decomposition of the running computing tasks of the intelligent installation data acquisition module, the installation environment perception module, the intelligent risk prediction module, the edge intelligent collaboration module and the intelligent manipulator control module to obtain multiple subtasks, each subtask corresponds to an edge node, which is used to reflect the computing resources required for a single running computing task of the system. The state parameter refers to a set of values ​​for overall evaluation of the computing resources required for the running computing tasks of the edge node. The state parameters include computing power value, memory evaluation value, and network bandwidth evaluation value. This embodiment does not limit the way of defining state parameters, such as software definition. The computing power value refers to the central processing computing value required when the edge node performs the running computing task. The memory evaluation value refers to the memory space required when the edge node performs the running computing task. The network bandwidth evaluation value refers to the network data transmission capacity required when the edge node performs the running computing task. The currently available number of CPU cycles refers to the remaining central processing unit of the data processing system. Computing processing capacity, the total number of CPU cycles refers to the total central processing capacity of the data processing system. This embodiment does not limit the method for obtaining the currently available number of CPU cycles and the total number of CPU cycles. Relevant technical personnel in this field can freely choose according to actual needs, such as a graphical interface tool. The currently available memory refers to the remaining available memory space in the data processing system, and the total memory refers to the total memory space in the data processing system. This embodiment does not limit the method for obtaining the currently available memory and the total memory. Relevant technical personnel in this field can freely choose according to actual needs, such as obtaining through a computer window. The currently available network bandwidth refers to the remaining available network transmission capacity in the data processing system, and the total network bandwidth refers to the total available network transmission capacity in the data processing system. This embodiment does not limit the method for obtaining the currently available network bandwidth and the total network bandwidth. Relevant technical personnel in this field can freely choose according to actual needs, such as accessing the router management interface through a browser.

[0143] Specifically, the edge intelligent collaboration module breaks down the running computing tasks of each module by defining status parameters for the edge nodes, and obtains the computing resources required by the edge nodes to perform task calculations, so as to facilitate subsequent edge intelligent collaboration, thereby improving the resource allocation capabilities and efficiency of installation operations.

[0144] Specifically, the edge intelligent collaboration module performs edge intelligent collaboration on the computing resource measurement results according to the edge intelligent collaboration method, and the edge intelligent collaboration method includes:

[0145] Step F01, calculate the task value Vic according to the preset load penalty factor §, the preset reward value Rc, the computing power value Ci, the memory evaluation value Mi, the network bandwidth evaluation value Bi, the load evaluation value Li and the preset load penalty factor §, set 0<§<1, obtain the task value Vic, set Vic=Rc×0.6×Ci×0.2×Mi×0.2×Bi-§×Li;

[0146] Step F02: Compare the task result evaluation value AC with the preset evaluation value AC0, set 90≤AC0≤100, judge the task completion status based on the comparison result, and adjust the preset reward value Rc based on the judgment result, where:

[0147] When AC≥AC0, the edge intelligent collaboration module determines that the task completion is of high quality, and adjusts the preset reward value Rc according to the reward coefficient sny, setting 0.7≤sny≤0.9 to obtain the first adjusted preset reward value Rc1, setting Rc1=(1+sny)×Rc, replacing the preset reward value Rc with the first adjusted preset reward value Rc1, and recalculating the task value Vic;

[0148] When AC<AC0, the edge intelligent collaboration module determines that the task completion is of low quality, and adjusts the preset reward value Rc. The preset reward value Rc is adjusted according to the reward coefficient sny, and 0.7≤sny≤0.9 is set to obtain the second adjusted preset reward value Rc2. Rc2=sny×Rc is set, and the preset reward value Rc is replaced with the second adjusted preset reward value Rc2, and the task value Vic is recalculated.

[0149] Specifically, the preset reward value refers to a pre-set positive feedback coefficient for calculating the task value Vic, the load evaluation value refers to the actual total load value of the edge node when executing the computing task, the present embodiment does not limit the method for obtaining the load evaluation value, and relevant technical personnel in this field can freely choose according to actual needs, such as obtaining it through software calculation, the preset load penalty factor refers to a coefficient for measuring the importance of the load evaluation value in the task value, the task result evaluation value refers to a specific value reflecting the quality of completion of a single running computing task by the winning edge node, the winning edge node refers to the edge node with the highest task value among all edge nodes of each module, the present embodiment does not limit the specific method for obtaining the edge node with the highest task value, and relevant technical personnel in this field can freely choose according to actual needs, such as obtaining it through software calculation, the preset load penalty factor refers to a coefficient for measuring the importance of the load evaluation value in the task value, the task result evaluation value refers to a specific value reflecting the quality of completion of a single running computing task by the winning edge node, and the winning edge node refers to the edge node with the highest task value among all edge nodes of each module. Technical personnel can make free choices based on actual needs, such as arranging the task values ​​corresponding to all edge nodes of each module in order from large to small, and taking the edge node with the largest task value as the winning edge node. The single running computing task refers to a single computing task obtained after decomposing the running computing tasks of each module. This embodiment does not limit the specific method of decomposing the running computing tasks of each module. Relevant technical personnel in this field can make free choices based on actual needs, such as decomposing through software. The preset evaluation value refers to the preset value for judging the task completion status. The task completion status refers to the completion quality of the single running computing task performed by the winning edge node. The task completion status includes high-quality task completion and low-quality task completion.

[0150] Specifically, the edge intelligent collaboration module adjusts the task value based on the task completion status of the winning edge node, so as to select the edge node that needs computing resources most for calculation in real time, thereby intelligently collaborating computing resources and improving the efficiency of computing resource utilization.

[0151] Specifically, the edge intelligent collaboration module performs a penalty adjustment on the edge intelligent collaboration method according to the load evaluation value Li, compares the load evaluation value Li with the preset load evaluation value Li0, sets 30%≤Li0≤70%, judges the load status of the edge node according to the comparison result, and adjusts the preset load penalty factor § according to the judgment result, where:

[0152] When Li≤Li0, the edge intelligent collaboration module determines that the load condition of the edge node is normal and does not make any penalty adjustment to the preset load penalty factor §;

[0153] When Li>Li0, the edge intelligent collaboration module determines that the load condition of the edge node is an abnormal load, and performs penalty adjustment on the preset load penalty factor §. According to the penalty adjustment coefficient α2, α2=Li0 / Li is set, and the preset load penalty factor § is penalty adjusted to obtain the adjusted preset load penalty factor §`, and §`=§×α2 is set. The preset load penalty factor § is replaced with the adjusted preset load penalty factor §`, and the task value Vic is recalculated.

[0154] Specifically, the preset load assessment value Li0 refers to a preset value for judging the load condition of the edge node, the load condition of the edge node refers to the normality of the load of the edge node judged based on the load assessment value Li and the preset load assessment value Li0, the load condition of the edge node refers to the normality of the load of the edge node judged based on the load assessment value and the preset load assessment value, and the load condition of the edge node includes the load condition of the edge node being a normal load and the load condition of the edge node being an abnormal load.

[0155] Specifically, the edge intelligent collaboration module judges the load condition of the edge node. When the load condition of the edge node is abnormal, it reduces the value of the preset load penalty factor § to avoid the instability of the intelligent collaboration process caused by excessive load on the edge node, thereby improving system stability and accuracy.

[0156] Specifically, when the edge intelligent collaboration module performs throughput adjustment on the penalty adjustment process according to the system throughput, the edge intelligent collaboration module compares the system throughput Th with the first preset throughput Th1 and the second preset throughput Th2, sets Th1=165 units / second and Th2=330 units / second, judges the load condition of the system throughput according to the comparison result, and adjusts the throughput of the preset load evaluation value Li0 according to the judgment result, wherein:

[0157] When Th>Th2, the edge intelligent collaboration module determines that the load condition of the system throughput is high load, performs throughput adjustment on the preset load evaluation value Li0, and adjusts the throughput of the preset load evaluation value Li0 according to the first load adjustment coefficient gcx1, sets gcx1=1+(Th-Th2) / Th2, obtains the first adjusted preset load evaluation value Li0`, sets Li0`=gcx1×Li0, replaces the preset load evaluation value Li0 with the first adjusted preset load evaluation value Li0`, and re-compares the load evaluation value Li with the first adjusted preset load evaluation value Li0`;

[0158] When Th1<Th≤Th2, the edge intelligent collaboration module determines that the load condition of the system throughput is medium load and does not adjust the throughput of the preset load evaluation value Li0;

[0159] When Th≤Th1, the edge intelligent collaboration module determines that the load condition of the system throughput is low load, and adjusts the throughput of the preset load assessment value Li0. The throughput of the preset load assessment value Li0 is adjusted according to the second load adjustment coefficient gcx2, and gcx2=1-(Th1-Th) / Th1 is set to obtain the second adjusted preset load assessment value Li0``, and Li0``=gcx2×Li0 is set to replace the preset load assessment value Li0 with the second adjusted preset load assessment value Li0``, and the load assessment value Li is re-compared with the second adjusted preset load assessment value Li0``.

[0160] Specifically, the system throughput refers to a measure of the number of tasks processed by a data processing system within a preset time. This embodiment does not limit the preset time, and relevant technical personnel in this field can freely select it according to actual needs. For example, the preset time is set to 1 hour. The first preset throughput refers to the lower limit of the preset value for judging the load condition of the system throughput. The second preset throughput refers to the upper limit of the preset value for judging the load condition of the system throughput. The load condition of the system throughput refers to the load degree of the system throughput judged based on the system throughput, the first preset throughput, and the second preset throughput. The load condition of the system throughput includes the load condition of the system throughput being low load, the load condition of the system throughput being medium load, and the load condition of the system throughput being high load.

[0161] Specifically, the edge intelligent collaboration module judges the load condition of the system throughput. When the load condition of the system throughput is high, the preset load assessment value is increased to ensure that multiple tasks can be completed on time, thereby improving the processing efficiency of the installation tasks. When the load condition of the system throughput is low, the preset load assessment value is reduced to save computing resources, thereby improving the resource utilization of the data processing system.

[0162] Specifically, the intelligent manipulator control module divides the computing resource measurement results into the first allocation ratio ZK, the second allocation ratio ZI and the third allocation ratio ZU respectively to obtain the first task, the second task and the third task, and assigns the first task to the manipulator motion control task, the second task to the manipulator installation quality inspection task, and the third task to the manipulator communication task, setting ZK=40%, ZI=30%, and ZU=30%.

[0163] Specifically, the manipulator motion control task refers to the computing power required for the computing task of controlling the motion of the manipulator, the manipulator installation quality inspection task refers to the computing power required for the computing task of inspecting the installation quality of the manipulator, and the manipulator communication task refers to the computing power required for the computing task of communicating between the manipulator and the manipulator control management system. This embodiment does not limit the specific allocation method of allocating the first task to the manipulator motion control task, the second task to the manipulator installation quality inspection task, and the third task to the manipulator communication task. Relevant technical personnel in this field can freely choose according to actual needs, such as software allocation.

[0164] Specifically, the intelligent manipulator control module disassembles the computing tasks for controlling the manipulator and sets reasonable initial computing resources, so as to adjust the initial task allocation process during the subsequent installation process and flexibly mobilize computing resources, thereby improving resource utilization and installation accuracy.

[0165] Specifically, the intelligent manipulator control module compares the manipulator motion accuracy index Jj with the preset manipulator motion accuracy index Jj0, sets 0.77≤Ji0≤0.89, judges the state of the manipulator motion accuracy based on the comparison result, and allocates and adjusts the manipulator motion control task ZK based on the judgment result, where:

[0166] When Jj≥Jj0, the intelligent manipulator control module determines that the manipulator motion accuracy is normal, and does not allocate or adjust the manipulator motion control task ZK;

[0167] When Jj<Jj0, the intelligent manipulator control module determines that the manipulator motion accuracy is abnormal, and allocates and adjusts the manipulator motion control task ZK. The manipulator motion control task ZK is allocated and adjusted according to the resource improvement coefficient sck, and sck is set to 0.5×(Jj0 / Jj). The adjusted manipulator motion control task ZK` is obtained, and ZK` is set to ZK×sck. The manipulator motion control task ZK is replaced with the adjusted manipulator motion control task ZK`, and the manipulator is re-controlled.

[0168] Specifically, the manipulator motion accuracy index refers to a specific value for evaluating the control accuracy of the manipulator performing manipulator motion control tasks. This embodiment does not limit the method of obtaining the manipulator motion accuracy index. Relevant technical personnel in this field can freely choose according to actual conditions, such as manual setting. The preset manipulator motion accuracy index refers to a preset value for judging the state of the manipulator motion accuracy. The state of the manipulator motion accuracy refers to the normal degree of control accuracy of the manipulator performing manipulator motion control tasks. The state of the manipulator motion accuracy includes normal manipulator motion accuracy and abnormal motion accuracy.

[0169] Specifically, the intelligent manipulator control module judges the state of the manipulator's motion accuracy. When the manipulator's motion accuracy is abnormal, it increases the proportion of computing resource measurement results of the manipulator motion control task, thereby improving the manipulator's operating accuracy.

[0170] Specifically, the intelligent manipulator control module compares the installation time Ta with the preset installation time Ta0, sets 10h≤Ta0≤15h, judges the status of the installation time according to the comparison result, and adjusts the time according to the manipulator motion accuracy index Jj according to the judgment result, where:

[0171] When Ta≤Ta0, the intelligent manipulator control module determines that the installation time is short and does not adjust the manipulator motion accuracy index Jj;

[0172] When Ta>Ta0, the intelligent manipulator control module determines that the installation time is long, and adjusts the manipulator motion precision index Jj according to the time adjustment coefficient sdf. Among them, e is the natural logarithm, and the adjusted manipulator motion accuracy index Jj` is obtained. Set Jj`=Jj×sdf, compare the adjusted manipulator motion accuracy index Jj` with the preset manipulator motion accuracy index Jj0, and re-judge the state of the manipulator motion accuracy. At the same time, adjust the time of the manipulator installation quality inspection task ZI according to the quality adjustment coefficient sdd, and set Where e is the natural logarithm, and the adjusted manipulator installation quality inspection task ZI' is obtained. Set ZI'=ZI×sdd, replace the manipulator installation quality inspection task ZI with the adjusted manipulator installation quality inspection task ZI', re-control the manipulator, and adjust the manipulator communication task ZU according to the communication adjustment coefficient sdg. Set Wherein, e is the natural logarithm, and the adjusted manipulator communication task ZU` is obtained, and ZU`=ZU×sdg is set, and the manipulator communication task ZU is replaced with the adjusted manipulator communication task ZU`, and the manipulator is controlled again.

[0173] Specifically, the installation progress time refers to the length of time it takes for the robot to complete the installation task. This embodiment does not limit the method of obtaining the installation progress time. Relevant technical personnel in this field can freely choose according to actual needs, such as software timing. The preset installation time refers to the preset value for judging the status of the installation progress time. The status of the installation progress time refers to the length of time required for the installation progress time judged based on the installation progress time and the preset installation time. The status of the installation progress time includes the status of the installation progress time being short and the status of the installation progress time being long.

[0174] Specifically, the intelligent manipulator control module judges the status of the installation time. When the installation time is too long, it reduces the manipulator motion accuracy index, increases the manipulator installation quality inspection task and reduces the manipulator communication task, so as to concentrate computing power resources on the manipulator motion control and quality inspection, thereby improving the accuracy and efficiency of the installation operation.

[0175] See also Figure 2 As shown in FIG, which is a schematic diagram of the structure of the metal roof intelligent installation device of this embodiment, the device includes:

[0176] A laser radar 1 connected to a fixed base 8, wherein the laser radar 1 is connected to a metal roof intelligent installation control system 2 based on machine intelligence via an internal connecting line, for acquiring cloud point cloud data;

[0177] A machine intelligence-based intelligent metal roof installation control system 2, which is connected to a temperature sensor 3, a light intensity sensor 4, an angle sensor 6, a speed sensor 7, and a balance board 10. The machine intelligence-based intelligent metal roof installation control system 2 is connected to a laser radar 1, a polarization camera 11, and a manipulator 5 via internal connecting lines to control the metal roof intelligent installation device;

[0178] A temperature sensor 3, which is connected to the metal roof intelligent installation control system 2 based on machine intelligence and is used to obtain the ambient temperature;

[0179] A light intensity sensor 4, which is connected to the metal roof intelligent installation control system 2 based on machine intelligence, and is used to obtain light intensity;

[0180] A manipulator 5 connected to the machine intelligence-based metal roof intelligent installation control system 2, a movable axis group and a polarization camera 11 for installing the metal roof;

[0181] A movable axis group, comprising a first movable axis 501, a second movable axis 502, a third movable axis 503 and a fourth movable axis 504. The first movable axis 501 is connected to the polarization camera 11 and the manipulator 5, the second movable axis 502 is connected to the manipulator 5, the third movable axis 503 is connected to the manipulator 5, and the fourth movable axis 504 is connected to the manipulator 5. The movable axis group is used to fix the manipulator;

[0182] An angle sensor 6, which is connected to the metal roof intelligent installation control system 2 based on machine intelligence and is used to obtain the slope of the metal roof;

[0183] A speed sensor 7, which is connected to the metal roof intelligent installation control system 2 based on machine intelligence and is used to obtain the target moving speed;

[0184] A fixed base 8, which is connected to the balance board 10, the dust shield assembly, the polarization camera 11 and the laser radar 1, and is used to fix the manipulator;

[0185] A dustproof baffle assembly, comprising a first dustproof baffle 901 and a second dustproof baffle 902, connected to the fixed base 8 and the balance plate 10 for dust prevention;

[0186] A balancing board 10, which is connected to the machine intelligence-based metal roof intelligent installation control system 2, the fixed base 8 and the dust baffle assembly to maintain balance;

[0187] The polarization camera 11 is connected to the fixed base 8, the first movable axis 501 and the manipulator 5, and is used to obtain polarization data, which includes the first polarization intensity I0, the second polarization intensity I 45 , the third polarized light intensity I 90 and the fourth polarized light intensity I 135 ;

[0188] The operating end 12 is connected to the manipulator 5 and is used to install the metal roof.

[0189] Specifically, the metal roof intelligent installation device is used for metal roof installation. The metal roof intelligent installation device performs precise operation and control on the metal roof installation process so as to accurately deal with complex working conditions such as reflection, rain, fog, and slope, thereby improving the safety, accuracy and efficiency of the installation operation. Among them, the metal roof intelligent installation device collects cloud and fog point cloud data through a lidar to eliminate the impact of meteorological interference on installation positioning, thereby improving the roof perception accuracy in complex environments. The metal roof intelligent installation device obtains polarized light intensity data in four directions through a polarization camera to accurately identify the reflective area of ​​the roof, thereby improving the reliability of the positioning of the installation components. The metal roof intelligent installation device monitors the ambient temperature and light parameters in real time through temperature and light intensity sensors, thereby improving the system's ability to recognize environmental changes. The metal roof intelligent installation device obtains the roof slope and the movement speed of the manipulator in real time through angle and speed sensors, thereby improving the stability and safety of the installation process.

[0190] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

Claims

1. Intelligent installation control system for metal roof based on machine intelligence, characterized by: The system comprises: Smart installation data acquisition module, used to acquire smart installation data; An environmental perception module is installed to judge the reflective area based on the intelligent installation data to obtain a reflective judgment result, eliminate reflective interference based on the reflective judgment result to obtain target environmental data, build an environmental perception model based on the target environmental data, and obtain reflective elimination data based on the environmental perception model; An intelligent risk prediction module is used to obtain the safe zone boundary distance based on the reflection elimination data and adjust the environmental perception model based on the safe zone boundary distance; The edge intelligent collaboration module is used to measure computing resources within the boundary distance of the safe area based on the intelligent twin model to obtain computing resource measurement results, and is also used to perform edge intelligent collaboration on the computing resource measurement results based on the edge intelligent collaboration method, and is also used to adjust the penalty of the edge intelligent collaboration method; The intelligent manipulator control module is used to perform initial task allocation based on the calculation resource measurement results and to adjust the allocation process of the initial task allocation.

2. The metal roof intelligent installation control system based on machine intelligence according to claim 1 is characterized in that: The installation environment perception module judges the reflective area according to the intelligent installation data using a reflective judgment method, and the reflective interference elimination method includes: Step A01: According to the first polarized light intensity I0 and the second polarized light intensity I 45 , the third polarized light intensity I 90 and the fourth polarized light intensity I 135 Calculate the total light intensity S0 and get the total light intensity S0, set S0 = I0 + I 45 +I 90 +I 135 ; Step A02: according to the first polarized light intensity I0 and the third polarized light intensity I 90 Calculate the first light intensity difference S1 to obtain the first light intensity difference S1, and set S1 = I0-I 90 ; Step A03, according to the second polarized light intensity I 45 and the fourth polarized light intensity I 135 Calculate the second light intensity difference S2 to obtain the second light intensity difference S2, and set S2 = I 45 -I 135 ; Step A04, calculating the polarization degree P according to the total light intensity S0, the first light intensity difference S1 and the second light intensity difference S2 to obtain the polarization degree P; Step A05: Compare the polarization degree P with the preset polarization degree P0, set 0.6≤P0≤0.8, judge the reflective state of the roof area based on the comparison result, and output the reflective judgment result based on the judgment result, wherein: When P>P0, the installation environment perception module determines that the reflective state of the roof area is reflective, and outputs the reflective area as the reflective judgment result; When P≤P0, the installation environment perception module determines that the reflective state of the roof area is non-reflective, and outputs the non-reflective area as the reflective judgment result.

3. The metal roof intelligent installation control system based on machine intelligence according to claim 2 is characterized in that: The installation environment perception module performs a de-reflection process on the reflective area according to a de-reflection process method, wherein the de-reflection process method includes: Step B01: Based on the dark channel I of the image dark ={I r , I g , I b }, transmittance control constant ω, set to 0.9≤ω≤0.95, and the background light estimation value A to calculate the initial transmittance τ0 to obtain the initial transmittance τ0; Step B02, calculating the optimized transmittance τq based on the first optimization coefficient ak, the second optimization coefficient bk and the reflective point pixel Iq, obtaining the optimized transmittance τq, and setting τq = ak × Iq ​​+ bk; Step B03, calculating the real texture channel Jc according to the optimized transmittance τq, the first channel intensity Ic, and the second channel intensity Ac to obtain the real texture channel Jc; Step B04: output the real texture channel Jc and the optimized transmittance τ0 as target environment data.

4. The metal roof intelligent installation control system based on machine intelligence according to claim 3 is characterized in that: The installation environment perception module constructs an environment perception model according to the target environment data through an environment perception model construction method, and the environment perception model construction method includes: Step S01, initialize the parameters of the convolutional neural network model according to the historical environment database: set the generator weight coefficient λ1 and the discriminator weight coefficient λ2, setting λ1 = 0.5, λ2 = 1-λ1 = 0.5; Step S02, calculating the total loss L according to the binary cross entropy loss function LG, the reconstruction loss Lr, the generator weight coefficient λ1 and the discriminator weight coefficient λ2, obtaining the total loss L, and setting L = λ1 × LG + λ2 × Lr; Step S03: Compare the total loss L with the preset total loss L0, set 0.22≤L0≤0.31, judge the accuracy of the convolutional neural network model based on the comparison result, and adjust the binary cross entropy loss function LG based on the judgment result, where: When L≤L0, the installed environment perception module determines that the accuracy of the convolutional neural network model is accurate, does not perform loss adjustment on the binary cross entropy loss function LG, and outputs the convolutional neural network model as the environment perception model; When L>L0, the installation environment perception module determines that the accuracy of the convolutional neural network model is inaccurate, and performs loss adjustment on the binary cross entropy loss function LG. The binary cross entropy loss function LG is adjusted according to the adjustment coefficient θ1 to obtain the adjusted binary cross entropy loss function LG`, and LG`=LG×θ1 is set. The binary cross entropy loss function LG is replaced with the adjusted binary cross entropy loss function LG`, and the total loss L is recalculated; Step S04: Acquire the current geographical environment, compare the current geographical environment with the historical geographical environment, determine the consistency between the current geographical environment and the historical geographical environment based on the comparison result, and perform a secondary loss adjustment on the preset total loss L0 based on the determination result, wherein: When the current geographical environment is consistent with the historical geographical environment, no secondary loss adjustment is made to the preset total loss L0; When the current geographical environment is inconsistent with the historical geographical environment, a secondary loss adjustment is performed on the preset total loss L0. According to the environmental coefficient γ, 1.1≤γ≤1.5 is set, and the preset total loss L0 is adjusted to obtain the adjusted preset total loss L0`. L0`=γ×L0 is set, and the preset total loss L0 is replaced by the adjusted preset total loss L0`. The total loss L is then re-compared with the adjusted preset total loss L0`. Step S05 , calculating the environmental change factor Hy according to the temperature difference Wc and the light intensity difference Gq to obtain the environmental change factor Hy, and setting Hy=0.58×Wc+0.42×Gq; Step S06: Compare the environmental change factor Hy with the preset environmental change factor Hy0, set 0.16≤Hy0≤0.39, judge the environmental change state according to the comparison result, and perform a third loss adjustment on the adjustment process of the second loss adjustment according to the judgment result, wherein: When Hy≤Hy0, the installation environment perception module determines that the environmental change state is a small-scale change, and does not perform the third loss adjustment in the adjustment process of the second loss adjustment; When Hy>Hy0, the installed environment perception module determines that the environmental change state is a large-scale change, and performs a third loss adjustment on the adjustment process of the secondary loss adjustment. According to the environmental adjustment coefficient θ2, θ2=1+(Hy-Hy0) / Hy0 is set, and the adjustment process of the secondary loss adjustment is performed three times to obtain the preset total loss L0`` after the secondary adjustment. L0``=L0×θ2 is set, and the preset total loss L0 is replaced with the preset total loss L0`` after the secondary adjustment, and the total loss L is re-compared with the preset total loss L0`` after the secondary adjustment.

5. The machine intelligence-based metal roof intelligent installation control system according to claim 4 is characterized in that: When the installed environmental perception module acquires the reflection elimination data according to the environmental perception model, the fog point cloud data is input into the environmental perception model to obtain clear point cloud data, the optimized transmittance τq and the clear point cloud data are input into the reflection elimination information model to obtain the target category probability, target position and target posture information, and the target category probability, target position and target posture information are output as the reflection elimination data.

6. The machine intelligence-based metal roof intelligent installation control system according to claim 5 is characterized in that: The intelligent risk prediction module obtains the safety area boundary distance according to the reflection elimination data using a safety area boundary distance acquisition method, and the safety area boundary distance acquisition method includes: Step C01, obtaining the target category according to the target category probability in the reflection removal data; Step C03: constructing an intelligent twin model based on target category, target location, and target posture information; Step C04: construct a safe area in the intelligent twin model, set the center coordinates of the safe area to (x0, y0, z0) and the radius to rq; Step C04, constructing a risk prediction model and outputting the target location coordinates from the risk prediction model; Step C05 , calculating the safety area boundary distance d based on the target position coordinate point CRM (xm, ym, zm), the center coordinates of the safety area CRR (x0, y0, z0) and the safety area radius rq to obtain the safety area boundary distance d.

7. The metal roof intelligent installation control system based on machine intelligence according to claim 6 is characterized in that: When the intelligent risk prediction module adjusts the environmental perception model according to the safe area boundary distance, the intelligent risk prediction module compares the safe area boundary distance d with the first preset boundary distance d1 and the second preset boundary distance d2, sets d1=10cm and d2=15cm, judges the risk level of the target location according to the comparison result, and adjusts the environmental perception model according to the judgment result, wherein: When d≤d1, the intelligent risk prediction module determines that the risk level of the target location is low risk and does not adjust the environment perception model; When d1<d≤d2, the intelligent risk prediction module determines that the risk level of the target location is medium risk, and adjusts the environmental perception model. According to the first model adjustment coefficient α1, α1=1+(d-d1) / d2 is set, and the environmental perception model is adjusted to obtain the first adjusted generator weight coefficient λ1`, and λ1`=λ1×α1 is set. The generator weight coefficient λ1 is replaced with the first adjusted generator weight coefficient λ1`, and the total loss L in the environmental perception model is recalculated according to the first adjusted generator weight coefficient λ1`; When d>d2, the intelligent risk prediction module determines that the risk level of the target location is high risk, and adjusts the environmental perception model. According to the second model adjustment coefficient α2, α2=1.5+(d-d2) / d2 is set, and the environmental perception model is adjusted to obtain the second adjusted generator weight coefficient λ1``, and λ1``=λ1×α2 is set. The generator weight coefficient λ1 is replaced with the second adjusted generator weight coefficient λ1``, and the total loss L in the environmental perception model is recalculated according to the second adjusted generator weight coefficient λ1``.

8. The machine intelligence-based metal roof intelligent installation control system according to claim 7 is characterized in that: When the intelligent risk prediction module performs distance adjustment on the model adjustment process according to the target movement speed, the intelligent risk prediction module compares the target movement speed Mv with the preset movement speed Mv0, sets 0.5m / s≤Mv0≤0.7m / s, judges the target movement speed state according to the comparison result, and adjusts the safety zone boundary distance d according to the judgment result, wherein: When Mv≤Mv0, the intelligent risk prediction module determines that the target moving speed is slow and does not adjust the distance d of the safety zone boundary; When Mv>Mv0, the intelligent risk prediction module determines that the target moving speed state is fast, adjusts the distance d of the safety area boundary, sets d>d2, determines that the target position risk level is high risk, and adjusts the environmental perception model.

9. The machine intelligence-based metal roof intelligent installation control system according to claim 8, characterized in that: When the intelligent risk prediction module adjusts the speed of the distance adjustment process according to the metal roof slope, the intelligent risk prediction module compares the metal roof slope Jp with the preset metal roof slope Jp0, sets 15°≤Jp0≤30°, judges the metal roof slope state according to the comparison result, and adjusts the preset moving speed Mv0 according to the judgment result, wherein: When Jp≤Jp0, the intelligent risk prediction module determines that the slope of the metal roof is gentle and does not adjust the preset moving speed Mv0; When Jp>Jp0, the intelligent risk prediction module determines that the slope of the metal roof is uneven, adjusts the preset moving speed Mv0, and adjusts the preset moving speed Mv0 according to the slope coefficient δ to obtain the adjusted preset moving speed Mv0`, sets Mv0`=Mv0×δ, replaces the target moving speed Mv with the adjusted preset moving speed Mv0`, and re-compares the target moving speed Mv with the adjusted preset moving speed Mv0`.

10. The metal roof intelligent installation control system based on machine intelligence according to claim 9 is characterized in that: The edge intelligent collaboration module calculates computing resources within the safe area boundary distance using a computing resource calculation method based on the intelligent twin model. The computing resource calculation method includes: Step E01: define the state parameter Si for the edge node Ni, and set Si = {Ci, Mi, Bi}, where Ci is the computing capability value, Mi is the memory evaluation value, and Bi is the network bandwidth evaluation value; Step E02 , calculating the computing capacity value Ci according to the current available CPU cycle number Cia and the total CPU cycle number Cit to obtain the computing capacity value Ci, and setting Ci=Cia / Cit; Step E03, calculating the memory evaluation value Mi according to the current available memory Mia and the total memory Mit to obtain the memory evaluation value Mi, and setting Mi=Mia / Mit; Step E04: Calculate the network bandwidth evaluation value Bi based on the current available network bandwidth Bia and the total network bandwidth Bit to obtain the network bandwidth evaluation value Bi, and set Bi=Bia / Bit; In step E05 , the computing capability value Ci, the memory evaluation value Mi, the network bandwidth evaluation value Bi, and the load evaluation value Li are used as computing resource measurement results.

Images