A method for extracting building change features based on real-life 3D models

Abstract

The present invention provides a method for extracting building change features based on a real-life three-dimensional model. Step 1: Establish and update a real-life three-dimensional model of a target building; Step 2: Extract the initial change features of the target building in the real-life three-dimensional model according to a preset extraction rule; Step 3: Add detail features, optimize multi-feature difference images, and then obtain the final change features of the target building and output them. The building change features can be extracted accurately and efficiently, avoiding the subjective influence of manual extraction, saving time and effort.

Description

Technical Field

[0001] The invention relates to the technical field of architecture, and in particular to a method for extracting building change features based on a real-scene three-dimensional model. Background Art

[0002] At present, for buildings, one of the important parts of the city, the development of the country has driven the rapid urbanization, and at the same time it has also brought about constant changes in urban landform information. Only by having an accurate understanding of these constantly changing building information can we have a macro grasp of local landform information.

[0003] The traditional pixel-based building extraction method has great drawbacks, and the interference of non-buildings with highly similar spectral information is difficult to deal with. Therefore, it is of great significance to find an automatic, accurate and efficient building extraction method. Summary of the invention

[0004] The present invention provides a building change feature extraction method based on a real-scene three-dimensional model, which is used to accurately obtain the change features of a target building.

[0005] A method for extracting building change features based on real-life 3D models:

[0006] Step 1: Create and update a real-life 3D model of the target building;

[0007] Step 2: Extract the initial change features of the target building in the real-life 3D model according to the pre-set extraction rules;

[0008] Step 3: Add detail features, optimize the multi-feature difference image, and then obtain the final change features of the target building and output them.

[0009] Preferably, the method for extracting building change features based on a real-scene three-dimensional model, step 1: establishing and updating a real-scene three-dimensional model of a target building, comprises:

[0010] Step 1.1: Obtain location data of a target building, and plan a collection path for the target building according to the location data;

[0011] Step 1.2: photographing the target building according to the acquisition path, and obtaining the positional relationship between different shooting points during the shooting process, and obtaining the pixel characteristics of the images of the corresponding shooting points;

[0012] Step 1.3: discretizing the corresponding image according to the pixel characteristics to obtain a three-dimensional point cloud of the target building;

[0013] Step 1.4: According to the positional relationship between different shooting points, a standard coordinate system of the target building is established, and the three-dimensional point cloud of the target building is inserted into the standard coordinate system to establish a real-scene three-dimensional model of the target building.

[0014] Preferably, the method for extracting building change features based on a real-scene 3D model, step 1: establishing and updating a real-scene 3D model of a target building, comprises: updating the real-scene 3D model of the target building, the steps of which include:

[0015] Acquire a plurality of current images of the target building, and perform first-stage training on the current images to obtain a plurality of first training images;

[0016] Segmenting the first training image in a first manner to obtain first data, performing a first preprocessing on the first data to obtain first pixel information, segmenting the first training image in a second manner to obtain second data, and performing a second preprocessing on the second data to obtain second pixel information;

[0017] Performing a weighted average of the first pixel information and the second pixel information to obtain third pixel information;

[0018] When the third pixel information reaches a preset standard, the first training image is determined to be qualified and stored in an image information library;

[0019] Otherwise, performing a second stage of training on the first training image to obtain a second training image, removing duplicate images in the image information library, and establishing a temporary real-scene three-dimensional model based on the remaining images in the image information library and the second training image;

[0020] Put the temporary real-scene 3D model into a standard coordinate system and compare it with the real-scene 3D model to obtain a difference data set;

[0021] Performing target difference classification on the difference data set, and obtaining a short data set of the same target according to the classification result, performing feature extraction on the short data set, and obtaining the morphological difference of the target location of the target building;

[0022] The real-scene three-dimensional model is updated according to the morphological differences of the target positions.

[0023] Preferably, the method for extracting building change features based on a real-scene three-dimensional model, step 2: extracting initial change features of a target building in the real-scene three-dimensional model according to a preset extraction rule, comprises:

[0024] Step 2.1: Establish a three-dimensional model database, and store the standard real-scene three-dimensional model of the target building in the three-dimensional model database;

[0025] Step 2.2: Obtain the current real-scene 3D model of the target building, divide the current real-scene 3D model into N small windows, and obtain the gray-level co-occurrence matrix of each small window, obtain the texture feature value of the corresponding small window, construct a texture feature matrix according to all the texture feature values, and obtain the first feature of the target building;

[0026] Step 2.3: According to the 3D point cloud of the current real-scene 3D model, each vertex of the target building is obtained, a vertex graph is generated, and a second feature of the target building is obtained;

[0027] Step 2.4: acquiring a standard first feature and a standard second feature of the target building according to the standard real-scene three-dimensional model of the target building;

[0028] Step 2.5: According to the difference between the first feature and the standard first feature, and the difference between the second feature and the standard second feature, the initial change feature of the target building is obtained.

[0029] Preferably, the method for extracting building change features based on a real-life three-dimensional model, step 3: adding detail features, optimizing the multi-feature difference image, and then obtaining the final change features of the target building, and outputting, includes:

[0030] Step 3.1: Divide the target building into regions based on the first feature to obtain different regional objects, obtain color point clouds of the regional objects, and obtain color information;

[0031] Determine the spectral information of each point cloud in the region object according to the color information, establish a spectral matrix of the region object, and obtain the eigenvalue of the spectral matrix;

[0032] Based on the characteristic value, a spectral characteristic of the target building is obtained as a third characteristic;

[0033] Step 3.2: Obtain the second feature, and process the second feature information based on the morphological information of the target building to obtain the fourth feature of the target building;

[0034] Step 3.3: Obtain detail features: The third feature and the fourth feature are compared with the corresponding standard values ​​respectively to obtain feature differences, and the multi-feature difference image constructed by the initial change features is optimized based on the feature differences to obtain the final change features of the target building.

[0035] Preferably, the method for extracting building change features based on a real-scene three-dimensional model, step 1.1: obtaining location information of a target building, and planning an image acquisition path of the target building according to the location information, comprises:

[0036] Step 1.1.1: Obtain location information of a target building, and determine the construction scope of the target building according to the location information;

[0037] Step 1.1.2: Based on the construction range of the target building, infer the characteristic landmarks that need to be passed through to collect the image of the target building, obtain the initial path, and calculate the correlation between each characteristic landmark to determine the optimal path for collecting the image of the target building;

[0038] Obtain the weight value of each characteristic landmark on the initial path, and use formula (1) to calculate the correlation between adjacent characteristic landmarks;

[0039]

[0040] Among them, ξ represents the correlation degree of the adjacent feature landmarks, p j represents the weight value of the current feature landmark j, j ≥ 2; p j-1 represents the weight value of the previous feature landmark j-1; p j+1 represents the weight value of the next feature landmark j+1; n represents the total number of possible paths of the current feature landmark in all directions, n≥1;

[0041] When the correlation degree ξ of the adjacent feature landmarks is within the prediction range, the current feature landmark j is retained, and the correlation degree calculation of the adjacent nodes of the next feature landmark j+1 is continued;

[0042] The combining unit is used to combine all the reserved characteristic landmarks to construct the best path for collecting the image of the target building, and output it for display.

[0043] Preferably, the method for extracting building change features based on a real-life three-dimensional model, step 3.3: obtaining detail features: comparing the third feature and the fourth feature with corresponding standard values ​​respectively to obtain feature differences, optimizing a multi-feature difference image constructed from initial change features based on the feature differences to obtain the final change features of the target building, including:

[0044] Step 3.3.1: obtaining a multi-feature difference image before optimization, and performing fusion optimization on the multi-feature difference image according to different fusion rules to obtain a plurality of initial multi-feature difference images;

[0045] Step 3.3.2: performing a first evaluation on the initial multi-feature difference image based on an image evaluation index, and obtaining a first difference between the initial multi-feature difference image and a standard multi-feature difference image;

[0046] If the first difference is within a preset difference range, it indicates that the initial multi-feature difference image meets the basic requirements, and a second evaluation is performed;

[0047] Otherwise, it indicates that the initial multi-feature difference image does not meet the basic requirements, the fusion rule corresponding to the initial multi-feature difference image is abandoned, and the proportion of each feature corresponding to the fusion rule is adjusted according to the first difference and the image evaluation index to generate a new fusion rule;

[0048] Step 3.3.3: Based on the standard multi-feature difference image, obtain a second evaluation of the initial multi-feature difference image that meets the basic requirements, evaluate the degree of morphological loss of the target building, and obtain the morphological loss rate;

[0049] Step 3.3.4: judging whether the loss rate of the initial multi-feature difference image is greater than a preset threshold value according to the morphological loss rate; if the loss rate is greater than the preset threshold value, judging that the initial multi-feature difference image is unqualified and removing it;

[0050] Otherwise, the initial multi-feature difference image is determined to be qualified, the initial multi-feature difference image with the smallest loss rate is selected as the final multi-feature difference image, and all the qualified initial multi-feature difference images are temporarily stored;

[0051] Step 3.3.5: When the user needs to select the current multi-feature difference image, all qualified initial multi-feature difference images are obtained, and the user selects from the qualified initial multi-feature difference images according to his / her own needs;

[0052] The user selection result is obtained, the user demand is analyzed, and the standard for prioritizing the recommendation of the initial multi-feature difference image is updated according to the user demand.

[0053] Preferably, the method for extracting building change features based on a real-scene three-dimensional model comprises: denoising a multi-feature difference image, comprising:

[0054] Step 4.1: Obtain an optimized multi-feature difference image, and obtain the noise signal variance of the optimized multi-feature difference image;

[0055] Step 4.2: De-noise the optimized multi-feature difference image based on the dual complex wave signal, and calculate the estimated value of the signal wave coefficient of the optimized multi-feature difference image without noise according to formula (2):

[0056]

[0057] Among them, δ 2 represents the estimated value of the signal wave coefficient of the optimized multi-feature difference image without noise; M represents the number of pixels in the pixel area N of the optimized multi-feature difference image with noise; C represents the signal wave coefficient of the optimized multi-feature difference image with noise at the current moment; represents the noise signal variance of the optimized multi-feature difference image;

[0058] Based on the estimated value of the signal wave coefficient without noise of the multi-feature difference image, a denoising model is established:

[0059]

[0060] Wherein, H represents the denoising model of the multi-feature difference image; A represents the signal wave coefficient of the multi-feature difference image without noise; C represents the signal wave coefficient of the multi-feature difference image with noise; represents C=A+B, B represents the wave coefficient of the noise signal of the multi-feature difference image; λ represents the regularization number; inf{} represents the lower bound function; Ω represents the integral range, which is determined according to the size of the noise signal; represents the gradient operator;

[0061] Step 4.3: performing denoising processing on the multi-feature difference image according to the denoising model, and when the clarity of the multi-feature difference image reaches a preset requirement, determining that the denoising processing of the multi-feature difference image is completed;

[0062] Otherwise, the regularization number is adjusted and the multi-feature difference image is denoised again.

[0063] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings.

[0064] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0066] Figure 1 It is a flow chart of a method for extracting building change features based on a real-scene three-dimensional model in an embodiment of the present invention;

[0067] Figure 2 This is a flowchart of step 1 of a method for extracting building change features based on a real-scene three-dimensional model in an embodiment of the present invention;

[0068] Figure 3Flow chart of step 2 of a method for extracting building change features based on a real-scene three-dimensional model in an embodiment of the present invention;

[0069] Figure 4 This is a flowchart of step 3 of a method for extracting building change features based on a real-scene three-dimensional model in an embodiment of the present invention. DETAILED DESCRIPTION

[0070] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0071] Embodiment 1:

[0072] The present invention provides a method for extracting building change features based on a real-scene three-dimensional model. Figure 1 As shown, including:

[0073] Step 1: Create and update a real-life 3D model of the target building;

[0074] Step 2: Extract the initial change features of the target building in the real-life 3D model according to the pre-set extraction rules;

[0075] Step 3: Add detail features, optimize the multi-feature difference image, and then obtain the final change features of the target building and output them.

[0076] In this embodiment, the pre-set extraction rule refers to extracting the architectural features of the target building, and the features include, for example, any one of the length, width, height, shape, color, etc. of the wall.

[0077] In this embodiment, the initial change feature refers to the difference between the first feature and the standard first feature, and the difference between the second feature and the standard second feature, and the multi-feature difference image is obtained from the initial change feature, thereby obtaining the change feature of the target building; the final change feature refers to the image after the initial feature is adjusted and optimized according to the difference between the third feature and the fourth feature and the standard value, thereby obtaining the change feature of the target building; the first feature refers to the texture feature of the target building; the second feature refers to the morphological characteristics of the target building; the third feature refers to the spectral feature of the target building; the fourth feature refers to the morphological building index feature of the target building.

[0078] In this embodiment, the detail features refer to the spectral features and morphological building index features of the target building.

[0079] In this embodiment, the multi-feature difference image refers to an image constructed by fusing multiple features of the obtained real-scene three-dimensional model.

[0080] The beneficial effect of the above technical solution is that the present invention accurately captures the characteristics of the target building by establishing a real-scene three-dimensional model, and extracts the initial change characteristics of the target building in the real-scene three-dimensional model according to the preset extraction rules, adds detail features according to the classification results, optimizes the multi-feature difference image, and then obtains the final change characteristics of the target building and outputs them. Accurately and efficiently obtain the change characteristics of the objects in the target building.

[0081] Embodiment 2:

[0082] Based on Example 1, step 1: establish and update the real-scene 3D model of the target building, such as Figure 2 As shown, including:

[0083] Step 1.1: Obtain location data of a target building, and plan a collection path for the target building according to the location data;

[0084] Step 1.2: photographing the target building according to the acquisition path, and obtaining the positional relationship between different shooting points during the shooting process, and obtaining the pixel characteristics of the images of the corresponding shooting points;

[0085] Step 1.3: discretizing the corresponding image according to the pixel characteristics to obtain a three-dimensional point cloud of the target building;

[0086] Step 1.4: According to the positional relationship between different shooting points, a standard coordinate system of the target building is established, and the three-dimensional point cloud of the target building is inserted into the standard coordinate system to establish a real-scene three-dimensional model of the target building.

[0087] In this embodiment, the collection path refers to the moving path of the collection device (for example, a drone), and the collection path includes multiple shooting points, which is convenient for obtaining images of buildings in different locations and ensuring the comprehensiveness of the collection.

[0088] In this embodiment, the pixel characteristic refers to the grayscale change of the pixel.

[0089] In this embodiment, the standard coordinate system refers to a coordinate system constructed according to the architectural characteristics of the target building, and the purpose is to facilitate the comparison of real-scene three-dimensional models of different periods.

[0090] The beneficial effects of the above technical solution are: obtaining the location data of the target building, planning the acquisition path of the target building according to the location data, avoiding the problem of incomplete image acquisition caused by the irregular acquisition range, processing the acquired images and establishing a real-life three-dimensional model in a standard coordinate system, so as to facilitate the comparison of the real-life three-dimensional models in different periods and roughly observe the changes in the buildings.

[0091] Embodiment 3:

[0092] Based on Example 1, step 1: establishing and updating a real-scene 3D model of a target building, including: updating the real-scene 3D model of the target building, the steps of which include:

[0093] Acquire a plurality of current images of the target building, and perform first-stage training on the current images to obtain a plurality of first training images;

[0094] Segmenting the first training image in a first manner to obtain first data, performing a first preprocessing on the first data to obtain first pixel information, segmenting the first training image in a second manner to obtain second data, and performing a second preprocessing on the second data to obtain second pixel information;

[0095] Performing a weighted average of the first pixel information and the second pixel information to obtain third pixel information;

[0096] When the third pixel information reaches a preset standard, the first training image is determined to be qualified and stored in an image information library;

[0097] Otherwise, performing a second stage of training on the first training image to obtain a second training image, removing duplicate images in the image information library, and establishing a temporary real-scene three-dimensional model based on the remaining images in the image information library and the second training image;

[0098] Put the temporary real-scene 3D model into a standard coordinate system and compare it with the real-scene 3D model to obtain a difference data set;

[0099] Performing target difference classification on the difference data set, and obtaining a short data set of the same target according to the classification result, performing feature extraction on the short data set, and obtaining the morphological difference of the target location of the target building;

[0100] The real-scene three-dimensional model is updated according to the morphological differences of the target positions.

[0101] In this embodiment, the first stage of training refers to training the images acquired by the acquisition device; the second stage of training refers to training the images trained in the first stage.

[0102] In this embodiment, the first training image refers to the image acquired by the acquisition device after the first stage of training; the second training image refers to the first training image of the second stage of training; the current image, the first training image, and the second training image are all multiple images.

[0103] In this embodiment, the first method is to determine the adaptive threshold of the image, and use different thresholds to segment the image according to the local features of the image; the second method refers to using edge detection to detect the places where the gray level or structure has a sudden change, so as to divide the image into regions, and the image is segmented according to the region division;

[0104] In this embodiment, the first pixel information refers to the pixels of the multiple image blocks obtained by segmentation according to the threshold value; the second pixel information refers to the pixels of the multiple image blocks obtained by region segmentation.

[0105] In this embodiment, the image information library is used to store the first training image and the image used for initially establishing the real-scene three-dimensional model.

[0106] In this embodiment, the difference data set refers to a set of difference values ​​at different positions obtained by comparing the temporary real-scene three-dimensional model with the real-scene three-dimensional model by placing the model in a standard coordinate system.

[0107] In this embodiment, the classification result refers to the result obtained by classifying the data in the difference data set according to the location of the corresponding target building.

[0108] In this embodiment, the short data set refers to a set of data corresponding to the same position of the target building according to the classification result.

[0109] In this embodiment, the target position refers to the position on the target building corresponding to the short data set.

[0110] The beneficial effect of the above technical solution is: the present invention performs a first training on the current image to obtain a first training image that meets the requirements, and stores it in an image information library to eliminate duplicate images in the image information library, establishes a temporary real-scene three-dimensional model based on the remaining images in the image information library and the second training image, and compares the temporary real-scene three-dimensional model with the real-scene three-dimensional model to obtain a difference data set.

[0111] The target difference is classified for the difference data set, and the classification result is used to obtain a short data set of the same target. The feature extraction of the short data set is performed to obtain the morphological difference of the target building and the target position; according to the morphological difference of the target position, the real-life 3D model is updated. The difference between the real-life 3D model and the current temporary 3D model is accurately found, and the exact position is found on the target building. The real-life 3D model is accurately updated according to the position so that accurate change features can be obtained during the next detection.

[0112] Embodiment 4:

[0113] Based on Example 1, step 2: extract the initial change features of the target building in the real-scene 3D model according to the preset extraction rules, such as Figure 3 As shown, including:

[0114] Step 2.1: Establish a three-dimensional model database, and store the standard real-scene three-dimensional model of the target building in the three-dimensional model database;

[0115] Step 2.2: Obtain the current real-scene 3D model of the target building, divide the current real-scene 3D model into N small windows, and obtain the gray-level co-occurrence matrix of each small window, obtain the texture feature value of the corresponding small window, construct a texture feature matrix according to all the texture feature values, and obtain the first feature of the target building;

[0116] Step 2.3: According to the 3D point cloud of the current real-scene 3D model, each vertex of the target building is obtained, a vertex graph is generated, and a second feature of the target building is obtained;

[0117] Step 2.4: acquiring a standard first feature and a standard second feature of the target building according to the standard real-scene three-dimensional model of the target building;

[0118] Step 2.5: According to the difference between the first feature and the standard first feature, and the difference between the second feature and the standard second feature, the initial change feature of the target building is obtained.

[0119] In this embodiment, the 3D model database refers to a database storing different versions of real-scene 3D models due to model updates and before updates, including standard real-scene 3D models.

[0120] In this embodiment, the first feature refers to the texture feature of the target building; the second feature refers to the morphological characteristic of the target building.

[0121] The beneficial effect of the above technical solution is that the present invention obtains the first feature and the second feature of the current real-scene 3D model of the target building, and obtains the initial change feature of the target building according to the difference between the first feature and the standard first feature, and the second feature and the standard second feature. The change feature of the target building has a preliminary analysis, which can be used to speculate the change direction of the target building.

[0122] Embodiment 5:

[0123] Based on Example 1, step 3: add detail features, optimize the multi-feature difference image, and then obtain the final change features of the target building, and output, such as Figure 4 As shown, including:

[0124] Step 3.1: Divide the target building into regions based on the first feature to obtain different regional objects, obtain color point clouds of the regional objects, and obtain color information;

[0125] Determine the spectral information of each point cloud in the region object according to the color information, establish a spectral matrix of the region object, and obtain the eigenvalue of the spectral matrix;

[0126] Based on the characteristic value, a spectral characteristic of the target building is obtained as a third characteristic;

[0127] Step 3.2: Obtain the second feature, and process the second feature information based on the morphological information of the target building to obtain the fourth feature of the target building;

[0128] Step 3.3: Obtain detail features: The third feature and the fourth feature are compared with the corresponding standard values ​​respectively to obtain feature differences, and the multi-feature difference image constructed by the initial change features is optimized based on the feature differences to obtain the final change features of the target building.

[0129] In this embodiment, the color information refers to the color saturation and color expression of the color point cloud.

[0130] In this embodiment, the spectral information refers to the reflection spectrum of the point cloud determined according to the color information; the spectral matrix refers to the matrix composed of the reflection spectra of the point clouds in the same area.

[0131] In this embodiment, the third feature refers to the spectral feature of the target building; and the fourth feature refers to the morphological building index feature of the target building.

[0132] The beneficial effect of the above technical solution is: the present invention processes the first feature to obtain the third feature, processes the second feature to obtain the fourth feature, obtains the standard values ​​corresponding to the third feature and the fourth feature, and respectively calculates the differences between the third feature and the fourth feature and the corresponding standard values. At the same time, based on the difference optimization, the multi-feature difference image is obtained to obtain the final change feature of the target building, thereby compensating for the problem of incomplete image information caused by the fusion of the first feature and the second feature.

[0133] Embodiment 6:

[0134] Based on Example 2, step 1.1: obtaining the location information of the target building, and planning the image acquisition path of the target building according to the location information, includes:

[0135] Step 1.1.1: Obtain location information of a target building, and determine the construction scope of the target building according to the location information;

[0136] Step 1.1.2: Based on the construction range of the target building, infer the characteristic landmarks that need to be passed through to collect the image of the target building, obtain the initial path, and calculate the correlation between each characteristic landmark to determine the optimal path for collecting the image of the target building;

[0137] Obtain the weight value of each characteristic landmark on the initial path, and use formula (1) to calculate the correlation between adjacent characteristic landmarks;

[0138]

[0139] Among them, ξ represents the correlation degree of the adjacent feature landmarks, p j represents the weight value of the current feature landmark j, j ≥ 2; p j-1 represents the weight value of the previous feature landmark j-1; p j+1 represents the weight value of the next feature landmark j+1; n represents the total number of possible paths of the current feature landmark in all directions, n≥1;

[0140] When the correlation degree ξ of the adjacent feature landmarks is within the prediction range, the current feature landmark j is retained, and the correlation degree calculation of the adjacent nodes of the next feature landmark j+1 is continued;

[0141] The combining unit is used to combine all the reserved characteristic landmarks to construct the best path for collecting the image of the target building, and output it for display.

[0142] In this embodiment, the characteristic landmark refers to a landmark object in the image acquisition moving path.

[0143] In this embodiment, the initial path refers to the image acquisition moving path preliminarily divided according to the construction scope of the target building; the optimal path refers to the image acquisition moving path obtained by combining the retained characteristic landmarks.

[0144] In this embodiment, the weight value refers to the importance of the landmark object in the moving path of the image acquisition, and is expressed as a probability value.

[0145] The beneficial effect of the above technical solution is: the present invention determines the construction scope of the target building according to the location data of the target building, and based on the construction scope of the target building, infers the characteristic landmarks that need to be passed to collect the image of the target building, obtains the initial path, and calculates the correlation between each characteristic landmark, determines the best path for collecting the image of the target building, eliminates unimportant landmarks, and obtains the best collection path, which greatly reduces the image collection time. At the same time, it also avoids the problem of incomplete image collection due to unclear collection scope.

[0146] Embodiment 7:

[0147] Based on Example 5, step 3.3: Step 3.3: Obtain detail features: The third feature and the fourth feature are compared with the corresponding standard values ​​respectively to obtain feature differences, and the multi-feature difference image constructed by the initial change features is optimized based on the feature differences to obtain the final change features of the target building, including:

[0148] Step 3.3.1: obtaining a multi-feature difference image before optimization, and performing fusion optimization on the multi-feature difference image according to different fusion rules to obtain a plurality of initial multi-feature difference images;

[0149] Step 3.3.2: performing a first evaluation on the initial multi-feature difference image based on an image evaluation index, and obtaining a first difference between the initial multi-feature difference image and a standard multi-feature difference image;

[0150] If the first difference is within a preset difference range, it indicates that the initial multi-feature difference image meets the basic requirements, and a second evaluation is performed;

[0151] Otherwise, it indicates that the initial multi-feature difference image does not meet the basic requirements, the fusion rule corresponding to the initial multi-feature difference image is abandoned, and the proportion of each feature corresponding to the fusion rule is adjusted according to the first difference and the image evaluation index to generate a new fusion rule;

[0152] Step 3.3.3: Based on the standard multi-feature difference image, obtain a second evaluation of the initial multi-feature difference image that meets the basic requirements, evaluate the degree of morphological loss of the target building, and obtain the morphological loss rate;

[0153] Step 3.3.4: judging whether the loss rate of the initial multi-feature difference image is greater than a preset threshold value according to the morphological loss rate; if the loss rate is greater than the preset threshold value, judging that the initial multi-feature difference image is unqualified and removing it;

[0154] Otherwise, the initial multi-feature difference image is determined to be qualified, the initial multi-feature difference image with the smallest loss rate is selected as the final multi-feature difference image, and all the qualified initial multi-feature difference images are temporarily stored;

[0155] Step 3.3.5: When the user needs to select the current multi-feature difference image, all qualified initial multi-feature difference images are obtained, and the user selects from the qualified initial multi-feature difference images according to his / her own needs;

[0156] The user selection result is obtained, the user demand is analyzed, and the standard for prioritizing the recommendation of the initial multi-feature difference image is updated according to the user demand.

[0157] In this embodiment, the multi-feature difference image before optimization refers to the multi-feature difference image without adding the third feature and the fourth feature.

[0158] In this embodiment, the fusion rule refers to a rule for performing image fusion with different proportions of various features.

[0159] In this embodiment, the initial multi-feature difference image refers to a multi-feature difference image obtained by fusion according to different fusion rules.

[0160] In this embodiment, the image evaluation indexes include color, clarity, noise, etc.

[0161] In this embodiment, the first evaluation refers to evaluating the image evaluation index of the initial multi-feature difference image; the first difference refers to the difference between the initial multi-feature difference image and the standard multi-feature difference image with respect to the image evaluation index.

[0162] In this embodiment, the second evaluation is performed based on the difference in morphology between the building morphology described by the initial multi-feature difference image and the target building itself; the morphology loss rate refers to the change in morphology of the initial multi-feature difference image relative to the target building itself.

[0163] In this embodiment, the preferential recommendation of the initial multi-feature difference image refers to the initial multi-feature difference image pre-screened for the user according to the user's needs, and the standard is determined according to the user's needs.

[0164] The beneficial effect of the above technical solution is that the present invention performs fusion optimization on the multi-feature difference image according to different fusion rules, obtains multiple initial multi-feature difference images, evaluates the initial multi-feature difference images, selects the best result according to the evaluation result, and thus screens out the best fusion rule. In addition, the user can select the fusion rule according to his own needs, realizing human-computer interaction, and at the same time, analyzes the user needs, and updates the standard for the priority recommendation of the initial multi-feature difference image according to the needs.

[0165] Embodiment 8:

[0166] Based on the fourth embodiment, the method further includes: denoising the multi-feature difference image, including:

[0167] Step 4.1: Obtain an optimized multi-feature difference image, and obtain the noise signal variance of the optimized multi-feature difference image;

[0168] Step 4.2: De-noise the optimized multi-feature difference image based on the dual complex wave signal, and calculate the estimated value of the signal wave coefficient of the optimized multi-feature difference image without noise according to formula (2):

[0169]

[0170] Among them, δ 2 represents the estimated value of the signal wave coefficient of the optimized multi-feature difference image without noise; M represents the number of pixels in the pixel area N of the optimized multi-feature difference image with noise; C represents the signal wave coefficient of the optimized multi-feature difference image with noise at the current moment; represents the noise signal variance of the optimized multi-feature difference image;

[0171] Based on the estimated value of the signal wave coefficient without noise of the multi-feature difference image, a denoising model is established:

[0172]

[0173] Wherein, H represents the denoising model of the multi-feature difference image; A represents the signal wave coefficient of the multi-feature difference image without noise; C represents the signal wave coefficient of the multi-feature difference image with noise; represents C=A+B, B represents the wave coefficient of the noise signal of the multi-feature difference image; λ represents the regularization number; inf{} represents the lower bound function; Ω represents the integral range, which is determined according to the size of the noise signal; represents the gradient operator;

[0174] Step 4.3: performing denoising processing on the multi-feature difference image according to the denoising model, and when the clarity of the multi-feature difference image reaches a preset requirement, determining that the denoising processing of the multi-feature difference image is completed;

[0175] Otherwise, the regularization number is adjusted and the multi-feature difference image is denoised again.

[0176] The beneficial effect of the above technical solution is: the present invention obtains an optimized multi-feature difference image, obtains the noise signal variance of the multi-feature difference image, denoises the multi-feature difference image based on a dual complex wave signal, and establishes a denoising model, thereby avoiding the distortion of the multi-feature difference image caused by noise, which leads to errors in the extraction of building change features.

[0177] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for extracting building change features based on a real-scene three-dimensional model, characterized in that: Step 1: Create and update a real-life 3D model of the target building; Step 2: Extract the initial change features of the target building in the real-life 3D model according to the pre-set extraction rules; Step 3: Add detail features, optimize the multi-feature difference image, and then obtain the final change features of the target building and output them; Among them, step 1: establishing and updating the real-life 3D model of the target building, including: Step 1.1: Obtain location data of a target building, and plan a collection path for the target building according to the location data; Step 1.2: photographing the target building according to the acquisition path, and obtaining the positional relationship between different shooting points during the shooting process, and obtaining the pixel characteristics of the images of the corresponding shooting points; Step 1.3: discretizing the corresponding image according to the pixel characteristics to obtain a three-dimensional point cloud of the target building; Step 1.4: According to the positional relationship between different shooting points, a standard coordinate system of the target building is established, and the three-dimensional point cloud of the target building is inserted into the standard coordinate system to establish a real-scene three-dimensional model of the target building; The steps of updating the real-scene three-dimensional model of the target building include: Acquire a plurality of current images of the target building, and perform first-stage training on the current images to obtain a plurality of first training images; Segmenting the first training image in a first manner to obtain first data, performing a first preprocessing on the first data to obtain first pixel information, segmenting the first training image in a second manner to obtain second data, and performing a second preprocessing on the second data to obtain second pixel information; Performing a weighted average of the first pixel information and the second pixel information to obtain third pixel information; When the third pixel information reaches a preset standard, the first training image is determined to be qualified and stored in an image information library; Otherwise, performing a second stage of training on the first training image to obtain a second training image, removing duplicate images in the image information library, and establishing a temporary real-scene three-dimensional model based on the remaining images in the image information library and the second training image; Put the temporary real-scene 3D model into a standard coordinate system and compare it with the real-scene 3D model to obtain a difference data set; Performing target difference classification on the difference data set, and obtaining a short data set of the same target according to the classification result, performing feature extraction on the short data set, and obtaining the morphological difference of the target location of the target building; According to the morphological differences of the target positions, the real-scene three-dimensional model is updated; Wherein, step 2: extracting the initial change features of the target building in the real-scene 3D model according to the preset extraction rules, including: Step 2.1: Establish a three-dimensional model database, and store the standard real-scene three-dimensional model of the target building in the three-dimensional model database; Step 2.2: Obtain the current real-scene 3D model of the target building, divide the current real-scene 3D model into N small windows, and obtain the gray-level co-occurrence matrix of each small window, obtain the texture feature value of the corresponding small window, construct a texture feature matrix according to all the texture feature values, and obtain the first feature of the target building; Step 2.3: According to the 3D point cloud of the current real-scene 3D model, each vertex of the target building is obtained, a vertex graph is generated, and a second feature of the target building is obtained; Step 2.4: acquiring a standard first feature and a standard second feature of the target building according to the standard real-scene three-dimensional model of the target building; Step 2.5: Obtaining the initial change feature of the target building according to the difference between the first feature and the standard first feature, and the difference between the second feature and the standard second feature; Among them, step 3: adding detail features, optimizing the multi-feature difference image, and then obtaining the final change features of the target building and outputting them, including: Step 3.1: Divide the target building into regions based on the first feature to obtain different regional objects, obtain color point clouds of the regional objects, and obtain color information; Determine the spectral information of each point cloud in the region object according to the color information, establish a spectral matrix of the region object, and obtain the eigenvalue of the spectral matrix; Based on the characteristic value, a spectral characteristic of the target building is obtained as a third characteristic; Step 3.2: Obtain the second feature, and process the second feature information based on the morphological information of the target building to obtain the fourth feature of the target building; Step 3.3: Obtain detail features: The third feature and the fourth feature are compared with the corresponding standard values ​​respectively to obtain feature differences, and the multi-feature difference image constructed by the initial change features is optimized based on the feature differences to obtain the final change features of the target building.

2. The method for extracting building change features based on a real-scene three-dimensional model according to claim 1, characterized in that: Step 1.1: Obtaining the location data of the target building, and planning the image acquisition path of the target building according to the location data, including: Step 1.1.1: Obtain location information of a target building, and determine the construction scope of the target building according to the location information; Step 1.1.2: Based on the construction range of the target building, infer the characteristic landmarks that need to be passed through to collect the image of the target building, obtain the initial path, and calculate the correlation between each characteristic landmark to determine the optimal path for collecting the image of the target building; Obtain the weight value of each characteristic landmark on the initial path, and use formula (1) to calculate the correlation between adjacent characteristic landmarks; Among them, ξ represents the correlation degree of the adjacent feature landmarks, p j represents the weight value of the current feature landmark j, j ≥ 2; p j-1 represents the weight value of the previous feature landmark j-1; p j+1 represents the weight value of the next feature landmark j+1; n represents the total number of possible paths of the current feature landmark in all directions, n≥1; When the correlation degree ξ of the adjacent feature landmarks is within the prediction range, the current feature landmark j is retained, and the correlation degree calculation of the adjacent nodes of the next feature landmark j+1 is continued; The combining unit is used to combine all the reserved characteristic landmarks to construct the best path for collecting the image of the target building, and output it for display.

3. The method for extracting building change features based on a real-scene three-dimensional model according to claim 1, characterized in that: Step 3.3: Get detailed features: The third and fourth characteristics are compared with the corresponding standard values. Obtaining feature differences, and optimizing a multi-feature difference image constructed by initial change features based on the feature differences to obtain final change features of the target building, including: Step 3.3.1: obtaining a multi-feature difference image before optimization, and performing fusion optimization on the multi-feature difference image according to different fusion rules to obtain a plurality of initial multi-feature difference images; Step 3.3.2: performing a first evaluation on the initial multi-feature difference image based on an image evaluation index, and obtaining a first difference between the initial multi-feature difference image and a standard multi-feature difference image; If the first difference is within a preset difference range, it indicates that the initial multi-feature difference image meets the basic requirements, and a second evaluation is performed; Otherwise, it indicates that the initial multi-feature difference image does not meet the basic requirements, the fusion rule corresponding to the initial multi-feature difference image is abandoned, and the proportion of each feature corresponding to the fusion rule is adjusted according to the first difference and the image evaluation index to generate a new fusion rule; Step 3.3.3: Based on the standard multi-feature difference image, obtain a second evaluation of the initial multi-feature difference image that meets the basic requirements, evaluate the degree of morphological loss of the target building, and obtain the morphological loss rate; Step 3.3.4: judging whether the loss rate of the initial multi-feature difference image is greater than a preset threshold value according to the morphological loss rate; if the loss rate is greater than the preset threshold value, judging that the initial multi-feature difference image is unqualified and removing it; Otherwise, the initial multi-feature difference image is determined to be qualified, the initial multi-feature difference image with the smallest loss rate is selected as the final multi-feature difference image, and all the qualified initial multi-feature difference images are temporarily stored; Step 3.3.5: When the user needs to select the current multi-feature difference image, all qualified initial multi-feature difference images are obtained, and the user selects from the qualified initial multi-feature difference images according to his / her own needs; The user selection result is obtained, the user demand is analyzed, and the standard for prioritizing the recommendation of the initial multi-feature difference image is updated according to the user demand.

4. The method for extracting building change features based on a real-scene three-dimensional model according to claim 1, characterized in that: include: Denoise multi-feature difference images, including: Step 4.1: Obtain an optimized multi-feature difference image, and obtain the noise signal variance of the optimized multi-feature difference image; Step 4.2: De-noise the optimized multi-feature difference image based on the dual complex wave signal, and calculate the estimated value of the signal wave coefficient of the optimized multi-feature difference image without noise according to formula (2): Among them, δ 2 represents the estimated value of the signal wave coefficient of the optimized multi-feature difference image without noise; M represents the number of pixels in the pixel area N of the optimized multi-feature difference image with noise; C represents the signal wave coefficient of the optimized multi-feature difference image with noise at the current moment; represents the noise signal variance of the optimized multi-feature difference image; Based on the estimated value of the signal wave coefficient without noise of the multi-feature difference image, a denoising model is established: Wherein, H represents the denoising model of the multi-feature difference image; A represents the signal wave coefficient of the multi-feature difference image without noise; C represents the signal wave coefficient of the multi-feature difference image with noise; represents C=A+B, B represents the wave coefficient of the noise signal of the multi-feature difference image; λ represents the regularization number; inf{} represents the lower bound function; Ω represents the integral range, which is determined according to the size of the noise signal; ▽ represents the gradient operator; Step 4.3: performing denoising processing on the multi-feature difference image according to the denoising model, and when the clarity of the multi-feature difference image reaches a preset requirement, determining that the denoising processing of the multi-feature difference image is completed; Otherwise, the regularization number is adjusted and the multi-feature difference image is denoised again.

Images