Real scene three-dimensional modeling method

Abstract

The invention relates to three-dimensional modeling, in particular to a live-action three-dimensional modeling method, which comprises the following steps: acquiring remote sensing images with different shooting angles, screening out contour pixel points of each remote sensing image according to texture factors, and constructing a contour map; detecting matching point pairs among the remote sensing images, and registering the remote sensing images according to the matching points to obtain registered images; optimizing the registration image, and performing three-dimensional curved surface reconstruction on the optimized registration image to obtain a first live-action three-dimensional model; generating an initial live-action three-dimensional model according to the inclined image data, and generating a low-modulus three-dimensional model according to the initial live-action three-dimensional model; performing texture mapping on the low-modulus three-dimensional model according to the mapping data of the initial live-action three-dimensional model, and performing texture repair on the mapped low-modulus three-dimensional model to obtain a second live-action three-dimensional model; according to the technical scheme provided by the invention, the defect that accurate and efficient three-dimensional modeling is difficult to carry out on a complex real scene in the prior art can be effectively overcome.

Description

Technical Field

[0001] The invention relates to three-dimensional modeling, and in particular to a real-scene three-dimensional modeling method. Background Art

[0002] With the acceleration of urbanization and the increasing shortage of land resources, land planning has become increasingly important. Traditional land planning methods mainly rely on two-dimensional maps and data, which are difficult to intuitively display the three-dimensional characteristics and spatial relationships of land, bringing great limitations to planning decisions.

[0003] As an emerging surveying and mapping technology, real-scene 3D modeling can generate realistic 3D models by modeling real scenes in the real world with high precision. This technology can not only provide rich 3D spatial information, but also realize multi-angle and all-round browsing and interaction, providing a new perspective and means for land planning.

[0004] In the field of land planning, the application of real-life 3D modeling mainly involves the collection, processing, and modeling of 3D data. However, the traditional real-life 3D modeling method still has some shortcomings. For example, in the modeling process, a lot of manual operation and intervention are often required, resulting in low modeling efficiency and easy introduction of human errors. In addition, for scenes such as complex terrain and buildings, the traditional real-life 3D modeling method often cannot guarantee the accuracy and integrity of the modeling. Summary of the invention

[0005] 1. Technical issues to be resolved

[0006] In view of the above-mentioned shortcomings of the prior art, the present invention provides a real-scene three-dimensional modeling method, which can effectively overcome the defect of the prior art that it is difficult to accurately and efficiently perform three-dimensional modeling of complex real scenes.

[0007] (II) Technical solution

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0009] A real-scene three-dimensional modeling method comprises the following steps:

[0010] S1, obtaining remote sensing images with different shooting angles, screening out contour pixels of each remote sensing image according to texture factors, and constructing a contour map;

[0011] S2, detecting the matching point pairs between the remote sensing images, and registering the remote sensing images according to the matching points to obtain the registered images;

[0012] S3, optimizing the registered image, and reconstructing the three-dimensional curved surface of the optimized registered image to obtain a first real-scene three-dimensional model;

[0013] S4, generating an initial real-scene 3D model according to the oblique image data, and generating a low-poly 3D model according to the initial real-scene 3D model;

[0014] S5, performing texture mapping on the low-poly 3D model according to the mapping data of the initial real-scene 3D model, and performing texture repair on the mapped low-poly 3D model to obtain a second real-scene 3D model;

[0015] S6, performing model fusion on the first real-scene 3D model and the second real-scene 3D model to obtain a target real-scene 3D model;

[0016] The oblique image data includes aerial images taken at different angles in various areas and the posture information when the aerial images were taken.

[0017] Preferably, in S1, the contour pixels of each remote sensing image are screened out according to the texture factor to construct a contour map, including:

[0018] S11, calculating a first texture feature value according to the brightness value of the first layer neighborhood range of each pixel point in each remote sensing image;

[0019] S12, calculating a second texture feature value according to the brightness value of the second layer neighborhood range of each pixel point in each remote sensing image;

[0020] S13, calculating a texture factor of the pixel point according to the first texture feature value and the second texture feature value;

[0021] S14, taking pixel points whose texture factors are greater than a preset threshold as contour pixel points, removing non-contour pixel points in each remote sensing image, and obtaining a contour map of each remote sensing image.

[0022] Preferably, calculating the texture factor of the pixel point according to the first texture feature value and the second texture feature value in S13 includes:

[0023] The texture factor of a pixel is calculated using the following formula:

[0024]

[0025] Wherein, η is the texture factor, L1 and L2 are the first texture feature value and the second texture feature value of the pixel point respectively.

[0026] Preferably, detecting matching point pairs between remote sensing images in S2 includes:

[0027] S21, taking each pixel point in the contour map of each remote sensing image as the center, and setting a detection window for each center;

[0028] S22, extracting the brightness characteristic coefficient and the contour ratio coefficient in each detection window, and constructing a feature vector;

[0029] S23, sorting all remote sensing images according to shooting time, and determining matching point pairs between remote sensing images with adjacent shooting times by calculating the similarity of feature vectors.

[0030] Preferably, S3 optimizes the registered image, including:

[0031] S31, calculating the brightness value of the overlapping part in the registered image:

[0032]

[0033] Among them, L i ' is the brightness value at the i-th overlapping position in the registered image, L 1,i is the brightness value of one of the remote sensing images at the i-th overlapping position in the registered image, L 2,i is the brightness value of another remote sensing image at the i-th overlapping position in the registered image;

[0034] S32, adjusting the brightness of the overlapping parts in the registered images to obtain an optimized registered image.

[0035] Preferably, in S4, generating an initial real-scene 3D model according to the oblique image data, and generating a low-poly 3D model according to the initial real-scene 3D model includes:

[0036] S41, collecting oblique image data, and constructing an initial real-scene 3D model using free network adjustment and constrained network adjustment;

[0037] S42, optimizing the number of faces and adjusting the wiring of the initial real-scene 3D model according to the face information of the initial real-scene 3D model, to generate an initial low-poly 3D model;

[0038] S43, repairing defective parts in the initial low-poly three-dimensional model according to the aerial image of the blocked area, to obtain a low-poly three-dimensional model.

[0039] Preferably, in S42, according to the surface information of the initial real-scene 3D model, the face number optimization and wiring adjustment are performed on the initial real-scene 3D model to generate an initial low-poly 3D model, including:

[0040] S421, drawing a strip composed of polygons on a regular surface of the initial real-scene three-dimensional model according to the surface information of the initial real-scene three-dimensional model;

[0041] S422, performing fixed-point production on the complex structural parts of the initial real-scene 3D model according to the surface information of the initial real-scene 3D model, and determining the contour of the complex structural parts;

[0042] S423: According to the surface information of the initial real-scene 3D model, a new structure is extended at the open edge of the initial real-scene 3D model, and wiring is performed to generate an initial low-poly 3D model.

[0043] Preferably, in S43, the defective parts in the initial low-poly three-dimensional model are repaired according to the aerial image of the occluded area to obtain the low-poly three-dimensional model, including:

[0044] S431, generating an occlusion area map of the occlusion area according to the aerial image of the occlusion area;

[0045] S432, repairing each occlusion area map to the defective part of the initial low-poly three-dimensional model according to the identifier of the occlusion area map, to obtain a low-poly three-dimensional model.

[0046] Preferably, in S5, texture mapping is performed on the low-poly 3D model according to the mapping data of the initial real-scene 3D model, including:

[0047] S51, according to the mapping coordinate information and mapping angle information of the initial real-scene 3D model, matching the corresponding mapping coordinate information and mapping angle information in the low-poly 3D model to obtain a target mapping in the low-poly 3D model;

[0048] S52, mapping the texture information of the initial real-scene 3D model into the corresponding target map to obtain a mapped low-poly 3D model.

[0049] Preferably, in S5, texture repairing is performed on the mapped low-poly 3D model to obtain a second real-scene 3D model, including:

[0050] The texture to be repaired of the mapped low-poly 3D model is repaired using software tools to obtain a second real-scene 3D model.

[0051] (III) Beneficial effects

[0052] Compared with the prior art, the real scene 3D modeling method provided by the present invention has the following beneficial effects:

[0053] 1) remote sensing images with different shooting angles are obtained, contour pixel points of each remote sensing image are screened out according to the texture factor, a contour map is constructed, matching point pairs between the remote sensing images are detected, and the remote sensing images are registered according to the matching points to obtain a registered image, the registered image is optimized, and a three-dimensional curved surface is reconstructed for the optimized registered image to obtain a first real scene three-dimensional model, so that accurate and efficient three-dimensional modeling of the real scene can be achieved according to the remote sensing image;

[0054] 2) generating an initial real-scene 3D model according to the oblique image data, and generating a low-poly 3D model according to the initial real-scene 3D model, performing texture mapping on the low-poly 3D model according to the mapping data of the initial real-scene 3D model, and performing texture repair on the mapped low-poly 3D model to obtain a second real-scene 3D model, thereby enabling accurate and efficient 3D modeling of the real scene to be achieved according to the oblique image data;

[0055] 3) The first real-scene 3D model and the second real-scene 3D model are fused to obtain a target real-scene 3D model. By fusing the first real-scene 3D model and the second real-scene 3D model, accurate and efficient 3D modeling of complex real scenes can be achieved, effectively ensuring the accuracy and integrity of the modeling, and more accurately displaying the 3D characteristics and spatial relationships of the land, bringing convenience to planning and decision-making work. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0057] Figure 1 It is a schematic diagram of the process of the present invention;

[0058] Figure 2 A schematic diagram of a process of constructing a first real-scene three-dimensional model according to remote sensing images in the present invention;

[0059] Figure 3 It is a schematic diagram of the process of constructing a second real-scene three-dimensional model according to the oblique image data in the present invention. DETAILED DESCRIPTION

[0060] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0061] A real scene 3D modeling method, such as Figure 1 and Figure 2 As shown, S1, remote sensing images with different shooting angles are obtained, contour pixels of each remote sensing image are screened out according to the texture factor, and a contour map is constructed, which specifically includes:

[0062] S11, calculating a first texture feature value according to the brightness value of the first layer neighborhood range of each pixel point in each remote sensing image;

[0063] S12, calculating a second texture feature value according to the brightness value of the second layer neighborhood range of each pixel point in each remote sensing image;

[0064] S13, calculating a texture factor of the pixel point according to the first texture feature value and the second texture feature value;

[0065] S14, taking pixel points whose texture factors are greater than a preset threshold as contour pixel points, removing non-contour pixel points in each remote sensing image, and obtaining a contour map of each remote sensing image.

[0066] Specifically, in S13, the texture factor of the pixel is calculated according to the first texture feature value and the second texture feature value, including:

[0067] The texture factor of a pixel is calculated using the following formula:

[0068]

[0069] Wherein, η is the texture factor, L1 and L2 are the first texture feature value and the second texture feature value of the pixel point respectively.

[0070] S2. Detect matching point pairs between remote sensing images, and register the remote sensing images according to the matching points to obtain registered images.

[0071] Specifically, detecting matching point pairs between remote sensing images includes:

[0072] S21, taking each pixel point in the contour map of each remote sensing image as the center, and setting a detection window for each center;

[0073] S22, extracting the brightness characteristic coefficient and the contour ratio coefficient in each detection window, and constructing a feature vector;

[0074] S23, sorting all remote sensing images according to shooting time, and determining matching point pairs between remote sensing images with adjacent shooting times by calculating the similarity of feature vectors.

[0075] S3, optimizing the registered image, and reconstructing the three-dimensional surface of the optimized registered image to obtain a first real-scene three-dimensional model.

[0076] Specifically, the registered images are optimized, including:

[0077] S31, calculating the brightness value of the overlapping part in the registered image:

[0078]

[0079] Among them, L i ' is the brightness value at the i-th overlapping position in the registered image, L 1,i is the brightness value of one of the remote sensing images at the i-th overlapping position in the registered image, L 2,i is the brightness value of another remote sensing image at the i-th overlapping position in the registered image;

[0080] S32, adjusting the brightness of the overlapping parts in the registered images to obtain an optimized registered image.

[0081] The above technical scheme obtains remote sensing images with different shooting angles, selects the contour pixel points of each remote sensing image according to the texture factor, constructs a contour map, detects the matching point pairs between the remote sensing images, registers the remote sensing images according to the matching points to obtain the registered images, optimizes the registered images, and reconstructs the three-dimensional surface of the optimized registered images to obtain the first real-scene three-dimensional model, thereby realizing accurate and efficient three-dimensional modeling of the real scene based on the remote sensing images.

[0082] like Figure 1 and Figure 3 As shown, S4, generating an initial real scene 3D model according to the oblique image data (including aerial images taken at different angles in each area and the posture information when the aerial images are taken), and generating a low-poly 3D model according to the initial real scene 3D model, specifically including:

[0083] S41, collecting oblique image data, and constructing an initial real-scene 3D model using free network adjustment and constrained network adjustment;

[0084] S42, optimizing the number of faces and adjusting the wiring of the initial real-scene 3D model according to the face information of the initial real-scene 3D model, to generate an initial low-poly 3D model;

[0085] S43, repairing defective parts in the initial low-poly three-dimensional model according to the aerial image of the blocked area, to obtain a low-poly three-dimensional model.

[0086] 1) In S42, according to the surface information of the initial real-scene 3D model, the number of faces of the initial real-scene 3D model is optimized and the wiring is adjusted to generate an initial low-poly 3D model, including:

[0087] S421, drawing a strip composed of polygons on a regular surface of the initial real-scene three-dimensional model according to the surface information of the initial real-scene three-dimensional model;

[0088] S422, performing fixed-point production on the complex structural parts of the initial real-scene 3D model according to the surface information of the initial real-scene 3D model, and determining the contour of the complex structural parts;

[0089] S423: According to the surface information of the initial real-scene 3D model, a new structure is extended at the open edge of the initial real-scene 3D model, and wiring is performed to generate an initial low-poly 3D model.

[0090] 2) In S43, the defective parts in the initial low-poly 3D model are repaired according to the aerial image of the occluded area to obtain a low-poly 3D model, including:

[0091] S431, generating an occlusion area map of the occlusion area according to the aerial image of the occlusion area;

[0092] S432, repairing each occlusion area map to the defective part of the initial low-poly three-dimensional model according to the identifier of the occlusion area map, to obtain a low-poly three-dimensional model.

[0093] S5. Perform texture mapping on the low-poly 3D model according to the mapping data of the initial real-scene 3D model, and perform texture repair on the mapped low-poly 3D model to obtain a second real-scene 3D model.

[0094] 1) In S5, texture mapping is performed on the low-poly 3D model according to the texture data of the initial real-scene 3D model, including:

[0095] S51, according to the mapping coordinate information and mapping angle information of the initial real-scene 3D model, matching the corresponding mapping coordinate information and mapping angle information in the low-poly 3D model to obtain a target mapping in the low-poly 3D model;

[0096] S52, mapping the texture information of the initial real-scene 3D model into the corresponding target map to obtain a mapped low-poly 3D model.

[0097] 2) In S5, texture repair is performed on the mapped low-poly 3D model to obtain a second real-scene 3D model, including:

[0098] The texture to be repaired of the mapped low-poly 3D model is repaired using software tools to obtain a second real-scene 3D model.

[0099] The above technical scheme generates an initial real-scene three-dimensional model based on oblique image data, and generates a low-poly three-dimensional model based on the initial real-scene three-dimensional model, texture mapping is performed on the low-poly three-dimensional model according to the mapping data of the initial real-scene three-dimensional model, and texture repair is performed on the mapped low-poly three-dimensional model to obtain a second real-scene three-dimensional model, thereby being able to achieve accurate and efficient three-dimensional modeling of the real scene based on oblique image data.

[0100] like Figure 1 As shown, S6, the first real-scene 3D model and the second real-scene 3D model are fused to obtain a target real-scene 3D model.

[0101] The above technical scheme fuses the first real-scene three-dimensional model and the second real-scene three-dimensional model to obtain a target real-scene three-dimensional model. By fusing the first real-scene three-dimensional model and the second real-scene three-dimensional model, accurate and efficient three-dimensional modeling of complex real scenes can be achieved, effectively ensuring the accuracy and integrity of the modeling, more accurately displaying the three-dimensional characteristics and spatial relationships of the land, and bringing convenience to planning and decision-making work.

[0102] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A real-scene three-dimensional modeling method, characterized in that: The following steps are involved: S1, obtaining remote sensing images with different shooting angles, screening out contour pixels of each remote sensing image according to texture factors, and constructing a contour map; S2, detecting the matching point pairs between the remote sensing images, and registering the remote sensing images according to the matching points to obtain the registered images; S3, optimizing the registered image, and reconstructing the three-dimensional curved surface of the optimized registered image to obtain a first real-scene three-dimensional model; S4, generating an initial real-scene 3D model according to the oblique image data, and generating a low-poly 3D model according to the initial real-scene 3D model; S5, performing texture mapping on the low-poly 3D model according to the mapping data of the initial real-scene 3D model, and performing texture repair on the mapped low-poly 3D model to obtain a second real-scene 3D model; S6, performing model fusion on the first real-scene 3D model and the second real-scene 3D model to obtain a target real-scene 3D model; The oblique image data includes aerial images taken at different angles in various areas and the posture information when the aerial images were taken.

2. The real scene 3D modeling method according to claim 1, characterized in that: In S1, the contour pixels of each remote sensing image are screened out according to the texture factor to construct a contour map, including: S11, calculating a first texture feature value according to the brightness value of the first layer neighborhood range of each pixel point in each remote sensing image; S12, calculating a second texture feature value according to the brightness value of the second layer neighborhood range of each pixel point in each remote sensing image; S13, calculating a texture factor of the pixel point according to the first texture feature value and the second texture feature value; S14, taking pixel points whose texture factors are greater than a preset threshold as contour pixel points, removing non-contour pixel points in each remote sensing image, and obtaining a contour map of each remote sensing image.

3. The real scene 3D modeling method according to claim 2, characterized in that: In S13, the texture factor of the pixel is calculated according to the first texture feature value and the second texture feature value, including: The texture factor of a pixel is calculated using the following formula: Wherein, η is the texture factor, L1 and L2 are the first texture feature value and the second texture feature value of the pixel point respectively.

4. The real scene 3D modeling method according to claim 2, characterized in that: In S2, matching point pairs between remote sensing images are detected, including: S21, taking each pixel point in the contour map of each remote sensing image as the center, and setting a detection window for each center; S22, extracting the brightness characteristic coefficient and the contour ratio coefficient in each detection window, and constructing a feature vector; S23, sorting all remote sensing images according to shooting time, and determining matching point pairs between remote sensing images with adjacent shooting times by calculating the similarity of feature vectors.

5. The real scene 3D modeling method according to claim 4, characterized in that: S3 optimizes the registered images, including: S31, calculating the brightness value of the overlapping part in the registered image: Among them, L i ' is the brightness value at the i-th overlapping position in the registered image, L 1,i is the brightness value of one of the remote sensing images at the i-th overlapping position in the registered image, L 2,i is the brightness value of another remote sensing image at the i-th overlapping position in the registered image; S32, adjusting the brightness of the overlapping parts in the registered images to obtain an optimized registered image.

6. The real scene 3D modeling method according to claim 1, characterized in that: In S4, an initial real-scene 3D model is generated according to the oblique image data, and a low-poly 3D model is generated according to the initial real-scene 3D model, including: S41, collecting oblique image data, and constructing an initial real-scene 3D model using free network adjustment and constrained network adjustment; S42, optimizing the number of faces and adjusting the wiring of the initial real-scene 3D model according to the face information of the initial real-scene 3D model, to generate an initial low-poly 3D model; S43, repairing defective parts in the initial low-poly three-dimensional model according to the aerial image of the blocked area, to obtain a low-poly three-dimensional model.

7. The real scene 3D modeling method according to claim 6, characterized in that: In S42, according to the surface information of the initial real-scene 3D model, the number of faces of the initial real-scene 3D model is optimized and the wiring is adjusted to generate an initial low-poly 3D model, including: S421, drawing a strip composed of polygons on a regular surface of the initial real-scene three-dimensional model according to the surface information of the initial real-scene three-dimensional model; S422, performing fixed-point production on the complex structural parts of the initial real-scene 3D model according to the surface information of the initial real-scene 3D model, and determining the contour of the complex structural parts; S423: According to the surface information of the initial real-scene 3D model, a new structure is extended at the open edge of the initial real-scene 3D model, and wiring is performed to generate an initial low-poly 3D model.

8. The real scene 3D modeling method according to claim 7, characterized in that: In S43, defective parts in the initial low-poly 3D model are repaired according to the aerial image of the blocked area to obtain a low-poly 3D model, including: S431, generating an occlusion area map of the occlusion area according to the aerial image of the occlusion area; S432, repairing each occlusion area map to the defective part of the initial low-poly three-dimensional model according to the identifier of the occlusion area map, to obtain a low-poly three-dimensional model.

9. The real scene 3D modeling method according to claim 6, characterized in that: In S5, texture mapping is performed on the low-poly 3D model according to the texture data of the initial real-scene 3D model, including: S51, according to the mapping coordinate information and mapping angle information of the initial real-scene 3D model, matching the corresponding mapping coordinate information and mapping angle information in the low-poly 3D model to obtain a target mapping in the low-poly 3D model; S52, mapping the texture information of the initial real-scene 3D model into the corresponding target map to obtain a mapped low-poly 3D model.

10. The real scene 3D modeling method according to claim 9, characterized in that: In S5, texture repair is performed on the mapped low-poly 3D model to obtain a second real-scene 3D model, including: The texture to be repaired of the mapped low-poly 3D model is repaired using software tools to obtain a second real-scene 3D model.

Images