Damaged historic building prototype digital imaging system based on deep learning

Abstract

The invention relates to the technical field of historic building restoration, in particular to a damaged historic building prototype digital imaging system based on deep learning, which comprises a construction layer, an expansion layer and a reduction layer, structural parameters of a damaged building are uploaded in the construction layer, the construction layer constructs a damaged building model based on the structural parameters of the damaged building, the expansion layer inputs building structural parameters and feeds back the input building structural parameters to the construction layer, and the construction layer constructs a building model based on the building structural parameters input in the expansion layer. The method comprises the steps that a damaged building model is constructed and transmitted to an expansion layer to be stored, the expansion layer collects a damaged building facade image on the damaged building model, the range of the damaged building facade image is determined, and the building facade image is further extracted from the stored building model. And a mode of collecting the external vertical surface of the damaged building model in the damaged building model brings digital restoration processing for the damaged building and assists the development of the restoration work of the damaged building.

Description

Technical Field

[0001] The present invention relates to the technical field of ancient building restoration, and particularly relates to a digital imaging system for damaged ancient building prototypes based on deep learning. Background Art

[0002] An ancient building refers to an old building with historical, cultural, artistic and scientific value;

[0003] Ancient building restoration refers to the restoration of damaged ancient buildings, aiming to protect and restore the original appearance and value of historical buildings.

[0004] Currently, for the restoration work of damaged ancient buildings, full-time personnel often carry out a long-term restoration work on the damaged ancient buildings to be restored based on historical documents, on-site investigation, data analysis and speculation. The restored ancient buildings are directly affected by the integrity of the relevant historical documents of the ancient buildings, the accuracy, effectiveness of on-site investigation, data analysis and speculation, and the restoration technology of the staff. Moreover, during the restoration process of damaged ancient buildings, there is no complete reference for the restoration prototype of ancient buildings, resulting in a difference between the finally restored damaged ancient buildings and the expected restoration effect, and a difference from the original appearance of the damaged ancient buildings. Summary of the Invention

[0005] In view of the above-mentioned drawbacks of the prior art, the present invention provides a digital imaging system for damaged ancient building prototypes based on deep learning, which solves the technical problems raised in the above background art.

[0006] To achieve the above object, the present invention is realized through the following technical solutions:

[0007] A digital imaging system for damaged ancient building prototypes based on deep learning includes: a construction layer, an expansion layer and a restoration layer;

[0008] The structural parameters of the damaged building are uploaded in the construction layer. The construction layer constructs a damaged building model based on the structural parameters of the damaged building. The expansion layer inputs the structural parameters of the building and feeds back the input structural parameters of the building to the construction layer. The construction layer constructs a building model based on the structural parameters of the building input in the expansion layer and transmits it to the expansion layer for storage. The expansion layer collects the external facade images of the damaged building on the damaged building model, determines the range of the external facade images of the damaged building, and further extracts the external facade images on the stored building model, enlarges the external facade images of the damaged building based on the external facade images of the building, and the expansion layer further transmits the enlarged images to the restoration layer. The restoration layer restores the damaged building model by applying the enlarged images and the external facade images of the damaged building.

[0009] The expansion layer includes a storage module, an extraction module, and an image expansion module. The storage module is used to input building structure parameters and transmit the building structure parameters to the construction layer. The extraction module is used to traverse the building models stored in the storage module and extract building facade images from the building models. The image expansion module is used to receive the damaged building model constructed in the construction layer, obtain the damaged building facade image on the damaged building model, use the damaged building facade image as the expansion target, and perform image expansion based on the building facade image extracted by the extraction module.

[0010] The storage module is wirelessly interconnected with the extraction module and the image expansion module. The storage module is wirelessly interconnected with a transmission module. The transmission module is wirelessly interconnected with a construction module and an upload module. The image expansion module is wirelessly interconnected with a restoration module. The restoration module is wirelessly interconnected with an evaluation module and a selection module.

[0011] Furthermore, the construction layer includes an upload module, a construction module, and a transmission module. The upload module is used to upload damaged building structure parameters. The construction module is used to receive the damaged building structure parameters uploaded by the upload module and construct a damaged building model based on the damaged building structure parameters. The transmission module is used to obtain the damaged building model constructed by the construction module and feedback the damaged building model to the expansion layer.

[0012] Among them, the damaged building structure parameters uploaded by the upload module include: the occupied area size and location information of the damaged building, the specification size and location information of the damaged building structure, and the spatial size and location information of the complete outer contour of the damaged building. The spatial size and location information of the complete outer contour of the damaged building are manually set by the system-end user. The damaged surface specification size of the damaged structure on the damaged building is included in the damaged building structure specification size. When the construction module constructs the damaged building model, it determines the damaged building model construction area in combination with the occupied area size and location information of the damaged building and the spatial size and location information of the complete outer contour of the damaged building.

[0013] Furthermore, when the storage module transmits the building structure parameters to the construction layer, it uses the transmission module as the transmission target. The transmission module completes the forwarding of the building structure parameters to the construction module. The construction module constructs a building model based on the building structure parameters. The building model constructed by the construction module is synchronously returned to the storage module via the transmission module and stored. After the extraction module extracts the building facade image on the building model, the extracted building facade image is synchronously sent to the storage module and stored separately from the building model.

[0014] Furthermore, the building structure parameters input in the storage module include damaged building structure parameters. When the extraction module extracts the building facade images of the building model, the extraction perspectives are: directly in front, directly behind, directly to the left, directly to the right, and directly above. When the image enlargement module enlarges the damaged building facade image, it retrieves the building facade image from the storage module for enlargement. The logic for retrieving the building facade image is as follows:

[0015]

[0016] where: SSIM(a, b) is the similarity between the building facade image and the damaged building facade image; σ a and σ b are the means of the building facade image and the damaged building facade image; μ a and μ b are the standard deviations of the building facade image and the damaged building facade image; m is the number of retrieved building facade images; γ is a constraint constant; λ is the integrity of the damaged building;

[0017] wherein, the number of retrieved building facade images is rounded up.

[0018] Furthermore, when the image enlargement module retrieves the building facade image, it calculates the similarity between the building facade images of each group of building models in the storage module and the facade image of the damaged building model based on Equation (1), further calculates the average similarity of the two sets of image collections, and then records the calculation result of the average similarity as the similarity between the building model and the damaged building model. Finally, it sorts the building models in descending order according to the similarity between the building model and the damaged building model;

[0019] After obtaining the number m of retrieved building facade images, the image enlargement module sequentially retrieves the building facade images of the corresponding number of similar building models from the storage module based on the descending order queue;

[0020] Among them, in the logic of retrieving the building facade image, the constraint constant γ > 0, and the constraint constant γ makes the number of retrieved building facade images by the image enlargement module obey: the higher the integrity λ of the damaged building, the fewer the number m of retrieved building facade images, and the lower the integrity λ of the damaged building, the more the number m of retrieved building facade images.

[0021] Furthermore, the integrity λ of the damaged building is obtained by the following formula:

[0022]

[0023] where: v rid$V_{d}$ is the volume of the damaged building corresponding to the damaged building model; $V$ is the original volume of the damaged building determined based on the occupied size of the damaged building and the complete outer contour of the damaged building; $k$ is the ratio of the load-bearing structure to the overall model in the damaged building model.

[0024] Further, during the operation stage of the enlargement module, all the damaged building facade images obtained on the damaged building model are taken as a set, denoted as Set One, and the retrieved building facade images are taken as a set, denoted as Set Two. Taking each damaged building facade image in Set One as the enlargement target, and applying Set Two in combination with any enlargement software or program to perform enlargement processing on the enlargement target;

[0025] Among them, when performing enlargement processing on the enlargement target, the enlargement area of the enlargement target is determined based on the complete outer contour of the damaged building, and the enlargement processing range of the enlargement target is limited based on the determined enlargement area. When applying the enlargement software or program to perform enlargement processing on the enlargement target, Set Two is used as the data support for the enlargement processing of the enlargement software or program.

[0026] Further, the restoration layer includes a restoration module, an evaluation module, and a selection module. The restoration module is used to receive the enlargement processing result of the enlargement target of the enlargement module in the expansion layer, identify the position of the enlargement target corresponding to the enlargement processing result on the damaged building model, and cover the enlarged image corresponding to the enlargement processing result at the corresponding position on the damaged building model. The evaluation module is used to receive the damaged building model after all the enlarged images are covered at the corresponding positions on the damaged building model based on the processing of the restoration module, and evaluate the rationality of the surfaces of the damaged building model covered with enlarged images. The selection module is used to obtain the rationality evaluation results of the surfaces of the damaged building model covered with enlarged images in the evaluation module, select the surfaces of the damaged building model covered with enlarged images based on the rationality evaluation results, restore the selected surfaces to the damaged building facade images, and transmit them to the enlargement module in the expansion layer for reprocessing by the enlargement module, and drive the restoration layer to run again until the surfaces of the damaged building model covered with enlarged images are all reasonable, and then output the damaged building model covered with enlarged images obtained by the final restoration module processing.

[0027] Further, the rationality evaluation logic of the surfaces of the damaged building model covered with enlarged images in the evaluation module is expressed as:

[0028]

[0029] Where: F is the rationality performance value of each surface of the damaged building model covered with the enlarged image; I(i,j) is the pixel value of the enlarged image at the position (i, j); I(x,y) is the pixel value within the neighborhood centered on (i, j) in the enlarged image; n is the number of pixels in the neighborhood; N is the neighborhood range; θ is the similarity between the two sets of image blocks obtained by symmetrically dividing the enlarged image along the central axis; ε is the adjustment factor;

[0030] Among them, the adjustment factor ε ∈ (0,1]. The larger θ is, the larger the value of the adjustment factor ε. Conversely, the smaller the value of the adjustment factor ε. There is a rationality determination threshold set in the evaluation module. The evaluation module compares the rationality determination threshold with F obtained from the above formula to determine whether the surfaces of the damaged building model covered with the enlarged images are reasonable. The damaged building surfaces covered with the enlarged images with unreasonable evaluation results by the evaluation module correspond to the damaged building facades, that is, the selection module selects the target surface.

[0031] Furthermore, when the restoration module runs to cover the enlarged image corresponding to the enlarged processing result on the corresponding position of the damaged building model, the enlarged image is stretched so that the enlarged image completely fits the surface of the damaged building model and falls within the area defined by the occupied area size of the damaged building and the complete outer contour of the damaged building.

[0032] Adopting the technical solution provided by the present invention, compared with the known public technology, it has the following beneficial effects:

[0033] The present invention provides a digital imaging system for damaged ancient building prototypes based on deep learning. During the operation of the system, by constructing a damaged building model and then collecting the facade of the damaged building model in the damaged building model, digital restoration processing is carried out for the damaged building to assist in the development of the damaged building repair work. And during the digital restoration processing of the damaged building, based on a large number of ancient building models, data support is provided for the AI image enlargement technology used in the digital restoration processing of the damaged building, ensuring that the damaged building digitally restored based on AI image enlargement calculation has a better restoration effect and is more fitting with the original appearance of the damaged building, effectively improving the work efficiency of damaged building repair and avoiding a large difference between the repaired damaged building and the original appearance of the damaged building. Description of the Drawings

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0035] Figure 1It is a schematic structural diagram of a digital imaging system for damaged ancient building prototypes based on deep learning. Specific implementation manners

[0036] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0037] The present invention will be further described below with reference to the embodiments.

[0038] Embodiment 1:

[0039] A digital imaging system for damaged ancient building prototypes based on deep learning in this embodiment, as Figure 1 shown, includes: a construction layer, an expansion layer, and a restoration layer;

[0040] The structural parameters of the damaged building are uploaded in the construction layer. The construction layer constructs a damaged building model based on the structural parameters of the damaged building. The expansion layer inputs the structural parameters of the building, feeds back the input structural parameters of the building to the construction layer. The construction layer constructs a building model based on the structural parameters of the building input in the expansion layer and transmits it to the expansion layer for storage. The expansion layer collects the external facade image of the damaged building on the damaged building model, determines the range of the external facade image of the damaged building, and further extracts the external facade image on the stored building model, enlarges the external facade image of the damaged building based on the external facade image of the building. The expansion layer further transmits the enlarged image to the restoration layer, and the restoration layer restores the damaged building model by applying the enlarged image and the external facade image of the damaged building;

[0041] The construction layer includes an upload module, a construction module, and a transmission module. The upload module is used to upload the structural parameters of the damaged building. The construction module is used to receive the structural parameters of the damaged building uploaded in the upload module and construct a damaged building model based on the structural parameters of the damaged building. The transmission module is used to obtain the damaged building model constructed in the construction module and feed back the damaged building model to the expansion layer;

[0042] Among them, the damaged building structure parameters uploaded by the upload module include: the occupied area size and location information of the damaged building, the specification size and location information of the damaged building structure, and the spatial size and location information of the space where the complete outer contour of the damaged building is located. The spatial size and location information of the space where the complete outer contour of the damaged building is located is manually set by the system-end user. The damaged surface specification size of the damaged structure on the damaged building is included in the damaged building structure specification size. When the construction module constructs the damaged building model, it determines the damaged building model construction area in combination with the occupied area size and location information of the damaged building and the spatial size and location information of the space where the complete outer contour of the damaged building is located;

[0043] The expansion layer includes a storage module, an extraction module, and an image expansion module. The storage module is used to input building structure parameters and transmit the building structure parameters to the construction layer. The extraction module is used to traverse the building models stored in the storage module and extract the building facade images from the building models. The image expansion module is used to receive the damaged building model constructed in the construction layer, obtain the damaged building facade image on the damaged building model, use the damaged building facade image as the image expansion target, and perform image expansion based on the building facade image extracted by the extraction module;

[0044] The restoration layer includes a restoration module, an evaluation module, and a selection module. The restoration module is used to receive the image expansion processing result of the image expansion target in the expansion layer, identify the position of the image expansion target corresponding to the image expansion processing result on the damaged building model, and cover the corresponding image of the image expansion processing result on the corresponding position on the damaged building model. The evaluation module is used to receive the damaged building model after all the expanded images are covered on the corresponding positions on the damaged building model based on the processing of the restoration module, and evaluate the rationality of the surfaces covered with the expanded images on the damaged building model. The selection module is used to obtain the rationality evaluation results of the surfaces covered with the expanded images on the damaged building model in the evaluation module, select the surfaces on the damaged building model covered with the expanded images based on the rationality evaluation results, restore the selected surfaces to the damaged building facade image, and transmit them to the image expansion module in the expansion layer. After being processed again by the image expansion module, it drives the restoration layer to run again until the surfaces covered with the expanded images on the damaged building model are all reasonable, and then outputs the damaged building model covered with the expanded images finally processed by the restoration module;

[0045] The storage module is wirelessly interconnected with the extraction module and the image expansion module. The storage module is wirelessly interconnected with the transmission module. The transmission module is wirelessly interconnected with the construction module and the upload module. The image expansion module is wirelessly interconnected with the restoration module. The restoration module is wirelessly interconnected with the evaluation module and the selection module.

[0046] In this embodiment, the upload module runs to upload the damaged building structure parameters. The construction module synchronously receives the damaged building structure parameters uploaded in the upload module and constructs a damaged building model based on the damaged building structure parameters. The transmission module obtains in real time the damaged building model constructed in the construction module and feeds back the damaged building model to the expansion layer. The storage module further inputs the building structure parameters and transmits the building structure parameters to the construction layer. The extraction module runs later to traverse the building models stored in the storage module and extracts the building facade images from the building models. Then, the image expansion module receives the damaged building model constructed in the construction layer, obtains the damaged building facade image on the damaged building model, uses the damaged building facade image as the image expansion target, and expands the image based on the building facade image extracted by the extraction module. Then, the restoration module receives the image expansion processing result of the image expansion target in the expansion layer, identifies the position of the image expansion processing result source image expansion target on the damaged building model, and covers the image corresponding to the image expansion processing result on the corresponding position on the damaged building model. Finally, the evaluation module receives the damaged building model after all the image expansion images are covered on the corresponding positions on the damaged building model based on the processing of the restoration module, evaluates the rationality of the surfaces covered with the image expansion images on the damaged building model, the selection module obtains the rationality evaluation results of the surfaces covered with the image expansion images on the damaged building model in the evaluation module, selects the surfaces on the damaged building model covered with the image expansion images based on the rationality evaluation results, restores the selected surfaces to the damaged building facade image, and transmits them to the image expansion module in the expansion layer. After being processed again by the image expansion module, it drives the restoration layer to run again until the surfaces covered with the image expansion images on the damaged building model are all reasonable, and then outputs the damaged building model covered with the image expansion images finally obtained by the processing of the restoration module;

[0047] Through the operation of the system in the above embodiment, a digital appearance restoration method is brought to the damaged ancient buildings, providing a better reference for the restoration operation of the damaged ancient buildings, and enabling the restoration work of the damaged ancient buildings to be carried out more efficiently and accurately.

[0048] Embodiment 2:

[0049] At the specific implementation level, on the basis of Embodiment 1, this embodiment further specifically describes a digital imaging system for damaged ancient building prototypes based on deep learning in Embodiment 1 with reference to Figure 1 :

[0050] When the storage module transfers the building structure parameters to the construction layer, the transfer module is used as the transfer target. The transfer module forwards the building structure parameters to the construction module. Based on the building structure parameters, the construction module constructs a building model. The building model constructed by the construction module is synchronously returned to the storage module via the transfer module and stored. After the extraction module extracts the building facade image on the building model, the extracted building facade image is synchronously sent to the storage module and stored separately from the building model.

[0051] Through the above settings, further operational logic support is provided for the operation of the storage module, ensuring more stable operation of each operation layer in the system.

[0052] Such as Figure 1 As shown, the building structure parameters input into the storage module include damaged building structure parameters. When the extraction module extracts the building facade image of the building model, the extraction perspectives are: directly in front, directly behind, directly to the left, directly to the right, and directly above. When the zooming module performs zooming processing on the damaged building facade image, it retrieves the building facade image from the storage module for zooming. The logic for retrieving the building facade image is:

[0053]

[0054] In the formula: SSIM(a,b) is the similarity between the building facade image and the damaged building facade image; σ a 、σ b are the means of the building facade image and the damaged building facade image; μ a 、μ b are the standard deviations of the building facade image and the damaged building facade image; m is the number of retrieved building facade images; γ is the constraint constant; λ is the integrity of the damaged building;

[0055] Among them, the number of retrieved building facade images m is rounded up;

[0056] When the zooming module retrieves the building facade image, it calculates the similarity between the building facade images belonging to each building model in the storage module and the facade image of the damaged building model based on formula (1), further calculates the average similarity of the two sets of image collections, then records the calculation result of the average similarity as the similarity between the building model and the damaged building model, and finally sorts the building models in descending order according to the similarity between the building model and the damaged building model;

[0057] After obtaining the number of retrieved building facade images m, the zooming module sequentially retrieves the corresponding number of building model facade images belonging to similar building models from the storage module based on the descending order queue;

[0058] Among them, in the building facade image retrieval logic, the constraint constant γ > 0. The constraint constant γ makes the number of building facade images retrieved by the image enlargement module obey the following rule: the higher the damage to the building integrity λ, the fewer the number of building facade images m retrieved; the lower the damage to the building integrity λ, the more the number of building facade images m retrieved.

[0059] Through the above building facade image retrieval logic, it is set to provide further operation data support for the operation of the extraction module and the image enlargement module in the expansion layer, so that the image enlargement module finally enlarges the damaged building facade images with a specified group of building facade images.

[0060] Embodiment 3:

[0061] At the specific implementation level, on the basis of Embodiment 1, this embodiment further specifically describes a digital imaging system for damaged ancient building prototypes based on deep learning in Embodiment 1 with reference to Figure 1 :

[0062] The damage to the building integrity λ is obtained by the following formula:

[0063]

[0064] In the formula: v rid is the volume of the damaged building corresponding to the damaged building model; V is the original volume of the damaged building determined based on the occupied area size of the damaged building and the complete outer contour of the damaged building; k is the ratio of the load-bearing structure in the damaged building model to the whole model.

[0065] The above-defined calculation formula for the damage to the building integrity λ provides further operation data support for the building facade image retrieval logic.

[0066] As Figure 1 shown, during the operation stage of the image enlargement module, all the damaged building facade images obtained on the damaged building model are used as a set, denoted as Set One, and the retrieved building facade images are used as a set, denoted as Set Two. Taking each group of damaged building facade images in Set One as the enlargement target, and applying Set Two in combination with any image enlargement software or program, the enlargement target is processed for enlargement.

[0067] Among them, when processing the enlargement target for enlargement, the enlargement area of the enlargement target is determined based on the complete outer contour of the damaged building, and the enlargement processing range of the enlargement target is limited based on the determined enlargement area. When applying the image enlargement software or program to process the enlargement target for enlargement, Set Two is used as the data support for the enlargement processing of the image enlargement software or program.

[0068] Based on the above description, a further detailed introduction is made to the operation process and principle of the image enlargement module.

[0069] AsFigure 1 As shown in the figure, the rationality evaluation logic for each surface of the damaged building model covered with enlarged images in the evaluation module is expressed as follows:

[0070]

[0071] In the formula: F is the rationality performance value of each surface of the damaged building model covered with enlarged images; I(i,j) is the pixel value of the enlarged image at the position (i, j); I(x,y) is the pixel value within the neighborhood centered on (i, j) in the enlarged image; n is the number of pixels in the neighborhood; N is the neighborhood range; θ is the similarity between two groups of image blocks obtained by symmetrically dividing the enlarged image along the central axis; ε is the adjustment factor;

[0072] Among them, the adjustment factor ε ∈ (0, 1]. The larger θ is, the larger the value of the adjustment factor ε. Conversely, the smaller the value of the adjustment factor ε. A rationality determination threshold is set in the evaluation module. The evaluation module compares the rationality determination threshold with F obtained from the above formula to determine whether each surface of the damaged building model covered with enlarged images is reasonable. The evaluation result of the evaluation module for the damaged building surface covered with enlarged images that is unreasonable corresponds to the damaged building facade, that is, the selection module selects the target surface.

[0073] Through the above settings, the rationality evaluation logic for each surface of the damaged building model covered with enlarged images is further defined, ensuring the stable output of the evaluation results of the evaluation module.

[0074] As Figure 1 shown in the figure, when the reduction module runs and covers the enlarged image corresponding to the enlarged processing result at the corresponding position on the damaged building model, the enlarged image is stretched so that the enlarged image completely fits the surface of the damaged building model and falls within the area defined by the occupied area size of the damaged building and the complete outer contour of the damaged building.

[0075] Based on the above settings, the enlarged image is finally combined with the damaged building model to obtain the complete damaged building model corresponding to the ancient building appearance model.

[0076] In summary, during the operation of the system in the above embodiments, by constructing a damaged building model and then collecting the damaged building facade in the damaged building model, digital restoration processing is carried out for the damaged building to assist in the development of the damaged building repair work. And during the digital restoration processing of the damaged building, based on a large number of ancient building models, data support is provided for the AI enlarged image technology used in the digital restoration processing of the damaged building, ensuring that the digitally restored damaged building completed based on AI enlarged image calculation has a better restoration effect, is more in line with the original appearance of the damaged building, effectively improving the efficiency of the damaged building repair work and avoiding a large difference between the repaired damaged building and the original appearance of the damaged building.

[0077] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements will not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A digital imaging system for damaged ancient building prototypes based on deep learning, characterized in that: include: Construction layer, expansion layer and restoration layer; The structural parameters of the damaged building are uploaded in the construction layer, and the construction layer constructs a damaged building model based on the structural parameters of the damaged building. The extension layer inputs the building structural parameters, and feeds back the input building structural parameters to the construction layer. The construction layer constructs a building model based on the building structural parameters input in the extension layer, and transmits it to the extension layer for storage. The extension layer collects the damaged building facade image on the damaged building model, determines the range of the damaged building facade image, and further extracts the building facade image on the stored building model, expands the damaged building facade image based on the building facade image, and the extension layer further transmits the expanded image to the restoration layer, and the restoration layer uses the expanded image and the damaged building facade image to restore the damaged building model; The expansion layer includes a storage module, an extraction module and an expansion module. The storage module is used to input building structure parameters and transmit the building structure parameters to the construction layer. The extraction module is used to traverse the building model stored in the storage module and extract the building facade image on the building model. The expansion module is used to receive the damaged building model constructed in the construction layer, obtain the damaged building facade image on the damaged building model, take the damaged building facade image as the expansion target, and expand the image based on the building facade image extracted by the extraction module. The storage module is interactively connected to the extraction module and the expansion module through a wireless network, the storage module is interactively connected to the transmission module through a wireless network, the transmission module is interactively connected to the construction module and the upload module through a wireless network, the expansion module is interactively connected to the restoration module through a wireless network, and the restoration module is interactively connected to the evaluation module and the selection module through a wireless network.

2. According to the deep learning-based digital imaging system for damaged ancient building prototypes according to claim 1, it is characterized in that: The construction layer includes an upload module, a construction module and a transmission module. The upload module is used to upload the damaged building structure parameters. The construction module is used to receive the damaged building structure parameters uploaded in the upload module and construct a damaged building model based on the damaged building structure parameters. The transmission module is used to obtain the damaged building model constructed in the construction module and feed the damaged building model back to the expansion layer. Among them, the damaged building structure parameters uploaded in the upload module include: the size and location information of the damaged building, the specifications and location information of the damaged building structure, and the space size and location information of the complete outer contour of the damaged building. The space size and location information of the complete outer contour of the damaged building are manually set by the system end user. The specifications and dimensions of the damaged building structure include the specifications and dimensions of the damaged surface of the damaged structure on the damaged building. When the construction module constructs the damaged building model, the construction area of ​​the damaged building model is determined in combination with the size and location information of the damaged building, and the space size and location information of the complete outer contour of the damaged building.

3. According to the deep learning-based digital imaging system for damaged ancient building prototypes according to claim 1, it is characterized in that: When the storage module transmits the building structure parameters to the construction layer, the transmission module is used as the transmission target, and the transmission module completes the forwarding of the building structure parameters to the construction module. The building model is constructed based on the building structure parameters by the construction module, and the building model constructed by the construction module is synchronously returned to the storage module via the transmission module and stored. After the extraction module executes the building facade image extraction on the building model, the extracted building facade image is synchronously sent to the storage module and stored separately from the building model.

4. According to the deep learning-based digital imaging system for damaged ancient building prototypes according to claim 1, it is characterized in that: The building structure parameters inputted into the storage module include damaged building structure parameters. When the extraction module extracts the building facade image of the building model, the extraction perspectives are: front, back, left, right, and top. When the expansion module expands the damaged building facade image, the building facade image is retrieved from the storage module for expansion. The logic for retrieving the building facade image is: Where: SSIM(a,b) is the similarity between the building facade image and the damaged building facade image; σ a , σb is the mean of the building facade image and the damaged building facade image; μ a , μ b is the standard deviation of the building facade image and the damaged building facade image; m is the number of building facade images retrieved; γ is the constraint constant; λ is the damage to building integrity; Among them, the number of building facade images retrieved is rounded up.

5. According to claim 4, a digital imaging system for damaged ancient building prototypes based on deep learning is characterized in that: When the image expansion module retrieves the building facade image, the building facade image of each group of building models in the storage module and the facade image of the damaged building model are similarly calculated based on formula (1), and the similarity mean of the two groups of image sets is further calculated, and the similarity mean calculation result is recorded as the similarity between the building model and the damaged building model, and finally the building models are arranged in descending order according to the similarity between the building model and the damaged building model; After the number m of building facade images retrieved is obtained, the image expansion module sequentially retrieves the building model facade images of the corresponding number of similar building models from the storage module based on the descending order queue; Among them, in the logic of retrieving building facade images, the constraint constant γ>0, and the constraint constant γ makes the number of building facade images retrieved by the expansion module obey: the higher the damage to the building integrity λ, the smaller the number of building facade images retrieved m; the lower the damage to the building integrity λ, the more the number of building facade images retrieved m.

6. The deep learning-based digital imaging system for damaged ancient building prototypes according to claim 4 is characterized in that: The damaged building integrity λ is obtained by the following formula: Where: v rid is the volume of the damaged building corresponding to the damaged building model; V is the original volume of the damaged building determined based on the size of the damaged building and the complete outer contour of the damaged building; k is the ratio of the load-bearing structure in the damaged building model to the entire model.

7. The deep learning-based digital imaging system for damaged ancient building prototypes according to claim 1 is characterized in that: In the operation phase of the image expansion module, all damaged building facade images acquired on the damaged building model are taken as a set, recorded as set one, and the retrieved building facade images are taken as a set, recorded as set two. Each set of damaged building facade images in set one is taken as an image expansion target, and set two is used in combination with any image expansion software or program to perform image expansion processing on the image expansion target; Among them, when expanding the target, the expansion area of ​​the target is determined based on the complete outer contour of the damaged building, and the expansion processing range of the target is limited based on the determined expansion area. When expanding the target using expansion software or program, set 2 is used as the expansion processing data support for the expansion software or program.

8. The deep learning-based digital imaging system for damaged ancient building prototypes according to claim 1 is characterized in that: The restoration layer includes a restoration module, an evaluation module and a selection module. The restoration module is used to receive the expansion target expansion processing result of the expansion module in the extension layer, identify the position of the expansion target of the expansion processing result source on the damaged building model, and cover the corresponding position of the expanded image on the damaged building model with the expanded image corresponding to the expansion processing result. The evaluation module is used to receive the damaged building model after all the expanded images are covered on the corresponding positions of the damaged building model based on the restoration module processing, and evaluate the rationality of each surface covered with the expanded image on the damaged building model. The selection module is used to obtain the rationality evaluation result of each surface covered with the expanded image on the damaged building model in the evaluation module, select the surface on the damaged building model covered with the expanded image based on the rationality evaluation result, restore the selected surface to the damaged building facade image, and transmit it to the expansion module in the extension layer, and process it again by the expansion module, and drive the restoration layer to run again, until all the surfaces covered with the expanded image on the damaged building model are reasonable, and then the damaged building model covered with the expanded image obtained by the final restoration module processing is output.

9. The deep learning-based digital imaging system for damaged ancient building prototypes according to claim 8, characterized in that: The rationality evaluation logic of each surface covered with the expanded image on the damaged building model in the evaluation module is expressed as: Where: F is the rationality performance value of each surface covered with the expanded image on the damaged building model; I(i, j) is the pixel value of the expanded image at the position (i, j); I(x,y) is the pixel value in the neighborhood centered at (i,j) in the expanded image; n is the number of pixels in the neighborhood; N is the range of the neighborhood; θ is the similarity of the two groups of image blocks obtained by symmetrical segmentation along the axis of the expanded image; ε is the adjustment factor; Among them, the adjustment factor ε∈(0,1], the larger the θ is, the larger the value of the adjustment factor ε is, and vice versa, the smaller the value of the adjustment factor ε is. A rationality judgment threshold is set in the evaluation module. The evaluation module compares the rationality judgment threshold with F obtained by the above formula to determine whether the surfaces covered with the expanded image on the damaged building model are reasonable. The evaluation result of the evaluation module is that the damaged building surface covered with the expanded image is unreasonable, which corresponds to the damaged building facade, that is, the selection module selects the target surface.

10. The deep learning-based digital imaging system for damaged ancient building prototypes according to claim 8, characterized in that: When the restoration module runs to cover the expanded image corresponding to the expansion processing result at the corresponding position on the damaged building model, the expanded image is stretched so that the expanded image completely fits the surface of the damaged building model and falls within the area defined by the size of the damaged building and the complete outer contour of the damaged building.

Images