Three-dimensional point cloud texture mapping method for electric power facilities

Abstract

The invention provides a three-dimensional point cloud texture mapping method for electric power facilities. The method comprises the following steps: acquiring multi-view RGB-D sequence data of the electric power facilities to generate a three-dimensional point cloud model; extracting geometric features of the surface fragments, matching texture coordinates of the RGB image, and generating a U / V-direction mapping field; calculating and generating a texture distribution field and a fusion weight field based on the projection deviation; and obtaining an optimized mapping result through multi-scale fusion of adjacent fragment texture data. According to the method, projection deviations of adjacent fragments are processed through Gaussian weighted summation, fusion weights are set by utilizing curvature feature partitions, texture fusion is realized by combining bilateral filtering and normalization processing, and finally, current and adjacent fragment textures are subjected to weighted fusion to generate an optimization result. The texture mapping precision and the visual consistency of the complex curved surface area can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of three-dimensional reconstruction and computer vision, and more particularly, to a power facility three-dimensional point cloud texture mapping method. BACKGROUND

[0002] In the current operation and maintenance management of power facilities, the application of three-dimensional point cloud technology is increasingly widespread. By obtaining multi-view RGB-D sequence data of power facilities, an accurate three-dimensional point cloud model can be generated, providing important technical support for the visualization, monitoring and maintenance of power facilities. However, in the process of mapping the texture of an RGB image to a three-dimensional point cloud model, existing technologies face many challenges. Traditional texture mapping methods usually use a simple projection method to directly map the RGB image to the three-dimensional point cloud model. Although this method is relatively simple to implement, it often results in texture distortion, stretching and color discontinuity when dealing with complex geometric shapes of power facilities. Especially when there are high-curvature areas on the surface of a power facility, the traditional mapping method has difficulty accurately mapping the texture to these areas, resulting in poor visual effects of the mapping results and making it difficult to meet the requirements of fine management of power facilities.

[0003] In addition, existing technologies often ignore the texture fusion problem between different viewpoints when processing RGB image data from multiple viewpoints. When the texture data from multiple viewpoints is projected onto the same three-dimensional point cloud model, due to differences in viewing angles and lighting conditions, there may be significant deviations between the texture data from different viewpoints, resulting in color inconsistencies and blurred boundaries in the fused texture. These problems not only affect the visual effects of the three-dimensional point cloud model, but also may mislead subsequent analysis and application based on the three-dimensional point cloud model.

[0004] In the implementation of the embodiments of the present application, there are at least the following problems or defects in the prior art: the prior art cannot effectively handle the texture mapping problem of complex geometric shapes in the three-dimensional point cloud model of power facilities, resulting in texture distortion and color discontinuity; at the same time, in the multi-view texture fusion process, the deviation between the texture data from different viewpoints cannot be accurately handled, resulting in color inconsistencies and blurred boundaries in the fused results. SUMMARY

[0005] The present application provides a power facility three-dimensional point cloud texture mapping method, comprising:

[0006] obtaining multi-view RGB-D sequence data of a power facility to generate a three-dimensional point cloud model;

[0007] extracting the geometric features of each surface patch in the three-dimensional point cloud model and matching the texture coordinates of the corresponding RGB image to generate U-direction and V-direction mapping fields for each surface patch;

[0008] Calculating the texture projection deviation based on the U-direction mapping field and the V-direction mapping field to generate a texture distribution field for each surface fragment and mapping to generate a texture fusion weight field;

[0009] The texture fusion weight field of each surface slice is fused with the texture data of adjacent surface slices at multiple scales to generate an optimized texture mapping result.

[0010] Furthermore, the texture projection deviation calculation based on the U-direction mapping field and the V-direction mapping field includes:

[0011] Performing Gaussian weighted summation on the projection deviation value of each vertex in the U-direction mapping field of each surface slice and the U-direction projection deviation value of the vertex corresponding to the adjacent surface slice to obtain the U-direction projection deviation of each vertex in each surface slice;

[0012] The projection deviation value of each vertex in the V-direction mapping field of each surface slice is Gaussian-weighted summed with the V-direction projection deviation value of the vertex corresponding to the adjacent surface slice to obtain the V-direction projection deviation of each vertex in each surface slice.

[0013] Furthermore, generating the texture distribution field of each surface slice includes:

[0014] Based on the U-direction projection deviation and the V-direction projection deviation of each vertex in each surface fragment, each vertex is divided into a high curvature feature area and a plane area through a preset threshold and a U-direction fusion weight value and a V-direction fusion weight value are set;

[0015] Calculating the texture fusion weight value of each vertex based on the U-direction fusion weight value and the V-direction fusion weight value of each vertex;

[0016] The texture fusion weight value of each vertex is subjected to bilateral filtering to obtain the texture distribution value of each vertex of each surface fragment to form the texture distribution field of each surface fragment.

[0017] Furthermore, the step of dividing each vertex into a high curvature feature area and a plane area based on the U-direction projection deviation and the V-direction projection deviation of each vertex in each surface fragment by a preset threshold and setting the U-direction fusion weight value and the V-direction fusion weight value includes:

[0018] When the absolute value of the U-direction projection deviation of the vertex in each surface slice is greater than or equal to the U-direction feature threshold, the vertex is determined to be a U-direction high curvature feature point, and the U-direction fusion weight value of the vertex is set to the U-direction feature threshold;

[0019] When the absolute value of the U-direction projection deviation of the vertex in each surface slice is less than the U-direction feature threshold, the vertex is determined to be a U-direction plane point, and the U-direction fusion weight value of the vertex is set to the absolute value of the U-direction projection deviation of the vertex;

[0020] When the absolute value of the V-direction projection deviation of the vertex in each surface patch is greater than or equal to the V-direction feature threshold, the vertex is determined as a V-direction high-curvature feature point, and the V-direction fusion weight value of the vertex is set as the V-direction feature threshold;

[0021] When the absolute value of the V-direction projection deviation of the vertex in each surface patch is less than the V-direction feature threshold, the vertex is determined as a V-direction planar point, and the V-direction fusion weight value of the vertex is set as the absolute value of the V-direction projection deviation of the vertex.

[0022] Further, the texture fusion weight value of each vertex is calculated based on the U-direction fusion weight value and the V-direction fusion weight value of each vertex using the following formula:

[0023]

[0024] Wherein, T weights (u,v) is the texture fusion weight value of the vertex (u,v); W u (u,v) is the U-direction fusion weight value of the vertex (u,v); W v (u,v) is the V-direction fusion weight value of the vertex (u,v); ΔC is the color difference value between adjacent surface patches; σ is the color difference adjustment coefficient.

[0025] Further, the mapping generates a texture fusion weight field, including: the texture fusion weight value of each vertex of the texture distribution field is normalized using the following formula to obtain the texture fusion weight value of each vertex:

[0026] W final (u,v) = T smooth (u,v) / (k1*S max +k2*C max )

[0027] Wherein, W final (u,v) is the texture fusion weight value of the vertex (u,v); T smooth (u,v) is the texture distribution value of the vertex (u,v); S max is the maximum curvature value of the surface patch; C max is the maximum color difference value of the adjacent patch; k1, k2 are the balance coefficients of curvature and color difference.

[0028] Further, the projection deviation value of each vertex in the U-direction mapping field of each surface patch is Gauss-weighted and summed with the U-direction projection deviation value of the corresponding vertex of the adjacent surface patch to obtain the U-direction projection deviation of each vertex in the surface patch using the following formula:

[0029] D u (u,v) = α*Gσ(du (u,v)) + (1 - a) * D' u (u,v)

[0030] wherein, D u (u,v) is the U-projection deviation of vertex (u,v) in the current surface patch; d u (u,v) is the projection deviation value of vertex (u,v) in the current U-mapping field; D' u (u,v) is the U-projection deviation of vertex (u,v) in the adjacent surface patch; a is the spatial continuity weight coefficient; Gσis the Gaussian kernel function.

[0031] Further, the projection deviation value of each vertex in the V-mapping field of each surface patch is Gaussian-weighted summed with the corresponding vertex V-projection deviation value of the adjacent surface patch, to obtain the V-projection deviation of each vertex in the surface patch, using the following formula:

[0032] D v (u,v) = β * Gσ(d v (u,v)) + (1 - β) * D' v (u,v)

[0033] wherein, D v (u,v) is the V-projection deviation of vertex (u,v) in the current surface patch; d v (u,v) is the projection deviation value of vertex (u,v) in the current V-mapping field; D' v (u,v) is the V-projection deviation of vertex (u,v) in the adjacent surface patch; β is the geometric consistency weight coefficient.

[0034] Further, the texture fusion weight field of each surface patch is multi-scale fused with the texture data of the adjacent surface patch to generate an optimized texture mapping result, and the formula is as follows:

[0035] I fuse (u,v) = W final (u,v) * I current (u,v) + (1 - W final (u,v)) * I adjacent (u,v)

[0036] wherein, W final (u,v) is the texture fusion weight value of vertex (u,v); I current (u,v) is the texture color value of vertex (u,v) in the current surface patch; I adjacent (u,v) is the texture color value of vertex (u,v) in the adjacent surface patch.

[0037] Further, the generating of the U-mapping field and the V-mapping field of each surface patch comprises:

[0038] Based on the projection relationship between the surface patch of the three-dimensional point cloud model and the corresponding RGB image, the projection coordinate deviation of the vertex of each surface patch in the texture space is calculated using an improved ICP algorithm;

[0039] The projection coordinate deviation is decomposed into a U-deviation component and a V-deviation component, and the U-deviation component and the V-deviation component are respectively subjected to Laplace smoothing processing to obtain the U-mapping field and the V-mapping field of each surface patch.

[0040] The above embodiments of the present application have at least the following beneficial effects:

[0041] 1. The present application establishes a high-precision U / V-mapping field by multi-view RGB-D sequence data acquisition and three-dimensional point cloud modeling, combined with the accurate matching of surface patch geometric features and texture coordinates, effectively solves the texture misplacement problem caused by the complex geometric structure of power facilities, realizes the accurate alignment of texture and geometric model, and improves the geometric fidelity of texture mapping.

[0042] 2. The projection deviation calculation method based on Gaussian weighting comprehensively considers the geometric consistency of adjacent surface patches, divides the high-curvature region and the flat region through an adaptive feature threshold, and dynamically adjusts the fusion weight, thereby inhibiting the texture distortion and stretching phenomenon caused by the curvature mutation of the surface of the power facility, and ensuring the natural transition of the texture in the complex curved surface region.

[0043] 3. Through a multi-scale texture fusion mechanism, the texture distribution field is optimized in combination with bilateral filtering, and the influence of curvature and color difference is balanced by using normalization processing, realizing seamless fusion of adjacent patch textures, eliminating the texture discontinuity problem caused by the difference in viewing angle or changes in illumination in traditional methods, and finally generating a high-quality texture mapping result with consistent vision and rich details. BRIEF DESCRIPTION OF DRAWINGS

[0044] The above and other objects, features and advantages of the exemplary embodiments of the present application will be more apparent from the following detailed description taken in conjunction with the accompanying drawings, in which:

[0045] Figure 1 A flowchart of a power facility three-dimensional point cloud texture mapping method provided by an embodiment of the present application is shown. DETAILED DESCRIPTION

[0046] The principles and spirits of the present application will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are given only to enable those skilled in the art to better understand and implement the present application, and do not limit the scope of the present application in any way. On the contrary, these embodiments are provided to make the present application more thorough and complete, and to fully convey the scope of the present application to those skilled in the art.

[0047] Those skilled in the art know that the embodiments of the present application can be implemented as a system, device, apparatus, method or computer program product. Therefore, the present application can be embodied in the form of a complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0048] It should be noted that any number of elements in the drawings is used for illustration only and not limitation, and any naming is only for distinction and does not have any limiting meaning.

[0049] The principles and spirits of the present application will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are given only to enable those skilled in the art to better understand and implement the present application, and do not limit the scope of the present application in any way. On the contrary, these embodiments are provided to make the present application more thorough and complete, and to fully convey the scope of the present application to those skilled in the art. Figure 1 Figure 1 A flowchart of a power facility three-dimensional point cloud texture mapping method provided by an embodiment of the present application is shown in FIG. 1. As shown in FIG. 1, a power facility three-dimensional point cloud texture mapping method includes the following steps. Figure 1

[0050] S1, acquiring power facility multi-view RGB-D sequence data to generate a three-dimensional point cloud model;

[0051] S2, extracting the geometric features of each surface patch in the three-dimensional point cloud model and matching the texture coordinates of the corresponding RGB image to generate a U-direction mapping field and a V-direction mapping field of each surface patch;

[0052] S3, performing texture projection deviation calculation based on the U-direction mapping field and the V-direction mapping field to generate a texture distribution field of each surface patch and a texture fusion weight field;

[0053] S4, performing multi-scale fusion of the texture fusion weight field of each surface patch and the texture data of adjacent surface patches to generate an optimized texture mapping result.

[0054] ​​It should be noted that the multi-view RGB-D sequence data of power facilities refers to the data sequence containing color information RGB and depth information D obtained by shooting power facilities from different angles. The three-dimensional point cloud model is a model that represents the three-dimensional shape of an object through a collection of points, where each point contains spatial position information and color information. The process of generating a three-dimensional point cloud model is to fuse these multi-view RGB-D data to construct the three-dimensional structure of the power facility. Specifically, the depth information in the RGB-D data is used to determine the position of each pixel point in the three-dimensional space, while the color information is used to give the corresponding color attribute to the point cloud model, thereby generating a three-dimensional point cloud model with rich details. In actual operation, RGB-D data can be obtained through various sensors, such as using a depth camera or through structured light scanning and other methods. After preprocessing, such as removing noise, aligning, etc., the three-dimensional point cloud is constructed to ensure the accuracy and integrity of the model.

[0055] Specifically, the three-dimensional point cloud model is divided into multiple surface patches, each representing a local region on the model. The geometric features refer to the shape, curvature, and other attributes of the surface patches, which are used to determine how to correctly map the texture of the RGB image onto the surface patches. The texture coordinates refer to the coordinate positions in the RGB image corresponding to the points on the surface patches. The U-mapping field and the V-mapping field are the mapping relationships in two perpendicular directions, which define the distribution of texture coordinates on the surface patches. U and V usually correspond to the horizontal and vertical directions in the texture coordinate system, and through these two mapping fields, the texture of the RGB image can be accurately mapped onto each surface patch of the three-dimensional point cloud model. In actual operation, geometric features such as normal vectors and curvatures of the surface patches can be extracted by calculating, and then these features are matched with the RGB image to obtain the U-mapping field and the V-mapping field of each surface patch. For example, the improved ICP, Iterative Closest Point algorithm, can be used to calculate the projection coordinate deviation of the surface patch vertices in the texture space, and decompose it into U-direction deviation component and V-direction deviation component, and then generate the mapping field.

[0056] Preferably, the texture projection deviation calculation is done by comparing the texture coordinates of the current surface patch with the texture coordinates of the adjacent surface patch. The texture distribution field is calculated based on the texture projection deviation, which reflects the distribution of the texture on the surface patch. The texture fusion weight field is generated based on the texture distribution field, which is used to determine the weight of each surface patch when fusing the texture data of multiple viewpoints. When calculating the texture projection deviation, factors such as spatial continuity and geometric consistency can be considered, and the influence of these factors on the deviation calculation can be balanced by setting appropriate weight coefficients. The generation process of the texture fusion weight field can include normalizing the texture distribution field to ensure that the weight values are within a reasonable range, thereby providing accurate weight basis for subsequent texture fusion. For example, the U and V projection deviations of each vertex can be calculated using the method of Gaussian weighted summation, and then the fusion weights are set according to these deviation values. The specific calculation process can be: for each vertex in a surface patch, the projection deviation value in the U direction mapping field of the vertex is Gaussian weighted summed with the U direction projection deviation value of the corresponding vertex of the adjacent surface patch, to obtain the U direction projection deviation of the vertex; similarly, the V direction projection deviation is calculated. In this way, the texture fusion weight value of each vertex can be obtained, and the texture fusion weight field of the entire surface patch can be generated.

[0057] In some embodiments, the texture projection deviation calculation based on the U direction mapping field and the V direction mapping field comprises:

[0058] Gaussian weighted summing the projection deviation value of each vertex in the U direction mapping field of the surface patch with the U direction projection deviation value of the corresponding vertex of the adjacent surface patch to obtain the U direction projection deviation of each vertex in the surface patch;

[0059] Gaussian weighted summing the projection deviation value of each vertex in the V direction mapping field of the surface patch with the V direction projection deviation value of the corresponding vertex of the adjacent surface patch to obtain the V direction projection deviation of each vertex in the surface patch.

[0060] It should be noted that the texture projection deviation calculation based on the U direction mapping field and the V direction mapping field is to more accurately determine the mapping effect of the texture on the surface patch of the three-dimensional point cloud model. The U direction mapping field and the V direction mapping field are the mapping relationship in two perpendicular directions, which define the distribution of the texture coordinates on the surface patch. The texture projection deviation refers to the deviation of the texture coordinates in the mapping process due to factors such as geometric shape, viewing angle difference, etc. By calculating these deviations, the texture mapping can be better adjusted to reduce visual distortion and discontinuity, thereby improving the quality of texture mapping.

[0061] Specifically, the U and V directions in the U and V mapping fields refer to two orthogonal directions in the texture coordinate system, usually corresponding to the horizontal and vertical directions of the texture image. Each vertex in the mapping field has corresponding U and V coordinates, which determine how the pixels on the texture image are mapped to the surface patches of the three-dimensional point cloud model. The texture projection deviation calculation involves processing the U and V projection deviation values of each vertex, which reflect the errors caused by geometric shape and viewing angle differences during texture mapping. During the calculation, a Gaussian weighting parameter can be set to balance the deviation influence of the current vertex and adjacent vertices, ensuring the smoothness and consistency of texture mapping. For example, the Gaussian weighting parameter can be set according to the spatial distance between vertices, and the closer the vertices, the greater the influence on the current vertex.

[0062] Preferably, the texture projection deviation calculation can be refined by the following steps: first, for each vertex in the surface patch, extract its U projection deviation value in the U mapping field, and obtain the U projection deviation values of the corresponding vertices in the adjacent surface patches. Then, using the method of Gaussian weighted summation, the U projection deviation value of the current vertex is weighted and summed with the U projection deviation values of the adjacent vertices. The Gaussian weighting function can set the weight according to the spatial distance between vertices, ensuring that the influence of adjacent vertices on the current vertex decreases with the increase of distance. Similarly, the same processing is performed on the projection deviation values in the V mapping field. In this way, the U and V projection deviation values of each vertex can be obtained, which will be used for the generation of the texture distribution field and the texture fusion weight field. This Gaussian weighting method can effectively smooth the discontinuity in texture mapping, while maintaining the details and accuracy of the texture.

[0063] In some embodiments, the generating the texture distribution field of each surface patch comprises:

[0064] Based on the U and V projection deviation values of each vertex in the surface patch, the vertices are divided into high-curvature feature areas and flat areas by a preset threshold, and U and V fusion weight values are set;

[0065] Based on the U and V fusion weight values of each vertex, a texture fusion weight value of each vertex is calculated;

[0066] The texture fusion weight values of each vertex are bilateral filtered to obtain the texture distribution values of each vertex of the surface patch, which constitute the texture distribution field of the surface patch.

[0067] It should be noted that the texture distribution field of each surface patch is generated to distinguish high-curvature feature areas and flat areas according to the U-direction and V-direction projection deviations of the vertices, and to set appropriate blending weight values for each area. The texture distribution field reflects the distribution of the texture on the surface patch. By setting different blending weight values, the effect of texture mapping can be optimized, and the problems of texture distortion and color discontinuity can be reduced. The generation process of the texture distribution field includes calculating the U-direction and V-direction blending weight values of the vertices, and smoothing the texture distribution through bilateral filtering, so as to obtain more natural texture mapping results.

[0068] Specifically, the high-curvature feature area refers to the area with relatively large curvature in the surface patch, which usually corresponds to the complex geometric shape of the power facility, such as the edge, corner or curved part. The flat area refers to the relatively flat area in the surface patch. The U-direction blending weight value and the V-direction blending weight value are parameters set according to the U-direction and V-direction projection deviations of the vertices, which are used to control the weight of different areas in the texture blending process. The preset threshold is a parameter used to distinguish high-curvature feature areas and flat areas. When the absolute value of the projection deviation is greater than or equal to the threshold, it is determined to be a high-curvature feature point; otherwise, it is determined to be a flat point. The setting of these parameters can be adjusted according to the specific geometric shape of the power facility and the requirements of texture mapping to achieve the best texture mapping effect.

[0069] Preferably, the process of generating the texture distribution field can be further refined into the following steps: first, according to the preset U-direction and V-direction feature threshold, classify the vertices in each surface patch. For the vertices whose absolute value of U-direction projection deviation is greater than or equal to the U-direction feature threshold, they are determined to be U-direction high-curvature feature points, and their U-direction blending weight values are set to the U-direction feature threshold; for the vertices whose absolute value of U-direction projection deviation is less than the U-direction feature threshold, they are determined to be U-direction flat points, and their U-direction blending weight values are set to the absolute value of the U-direction projection deviation of the vertex. Similarly, the V-direction projection deviation is processed in the same way. Then, based on the U-direction and V-direction blending weight values of each vertex, the texture blending weight value of each vertex is calculated. Finally, the texture blending weight value of each vertex is processed by bilateral filtering to smooth the texture distribution and reduce the discontinuity in texture mapping. Bilateral filtering is an algorithm that can smooth the texture while preserving the edge details. By setting appropriate filtering parameters, the generation of the texture distribution field can be effectively optimized.

[0070] In some embodiments, the step of distinguishing the high-curvature feature areas and the flat areas and setting the U-direction blending weight value and the V-direction blending weight value based on the U-direction projection deviation and the V-direction projection deviation of each vertex in each surface patch comprises:

[0071] When the absolute value of the U-direction projection deviation of the vertex in each surface patch is greater than or equal to the U-direction feature threshold, the vertex is determined to be a U-direction high-curvature feature point, and the U-direction fusion weight value of the vertex is set to the U-direction feature threshold;

[0072] When the absolute value of the U-direction projection deviation of the vertex in each surface patch is less than the U-direction feature threshold, the vertex is determined to be a U-direction planar point, and the U-direction fusion weight value of the vertex is set to the absolute value of the U-direction projection deviation of the vertex.

[0073] When the absolute value of the V-direction projection deviation of the vertex in each surface patch is greater than or equal to the V-direction feature threshold, the vertex is determined to be a V-direction high-curvature feature point, and the V-direction fusion weight value of the vertex is set to the V-direction feature threshold.

[0074] When the absolute value of the V-direction projection deviation of the vertex in each surface patch is less than the V-direction feature threshold, the vertex is determined to be a V-direction planar point, and the V-direction fusion weight value of the vertex is set to the absolute value of the V-direction projection deviation of the vertex.

[0075] It should be noted that the high-curvature feature area and the planar area are distinguished according to the U-direction and V-direction projection deviations of the vertex, and the corresponding fusion weight values are set. This process is a key step in texture mapping, and the purpose is to better handle the texture mapping problem of different geometric feature areas in the texture fusion process. By setting different fusion weight values, the effect of texture mapping can be optimized, the problems of texture distortion and color discontinuity can be reduced, and the overall quality of texture mapping can be improved.

[0076] Specifically, the U-direction projection deviation and the V-direction projection deviation refer to the deviation between the texture coordinates of the vertex and the ideal mapping position in the U-direction and V-direction mapping field. These deviation values reflect the errors caused by geometric shape and viewing angle difference in the texture mapping process. The high-curvature feature area refers to the area with large curvature in the surface patch, which usually corresponds to the complex geometric shape of the power facility, such as the edge, corner or curved part. The planar area refers to the relatively flat area in the surface patch. The U-direction feature threshold and the V-direction feature threshold are preset parameters for distinguishing the high-curvature feature area and the planar area. When the absolute value of the U-direction or V-direction projection deviation of the vertex is greater than or equal to the corresponding feature threshold, the vertex is determined to be a high-curvature feature point, and the fusion weight value of the vertex is set to the feature threshold; otherwise, the vertex is determined to be a planar point, and the fusion weight value of the vertex is set to the absolute value of the projection deviation. The setting of these parameters can be adjusted according to the specific geometric shape of the power facility and the requirements of texture mapping to achieve the best texture mapping effect.

[0077] Preferably, the process of setting the U-direction and V-direction fusion weight values can be further refined into the following steps: first, for each vertex in each surface patch, calculate its U-direction and V-direction projection deviation absolute values. Then, according to the preset U-direction feature threshold and V-direction feature threshold, classify each vertex. If the U-direction projection deviation absolute value of the vertex is greater than or equal to the U-direction feature threshold, it is determined as a U-direction high-curvature feature point, and its U-direction fusion weight value is set as the U-direction feature threshold; otherwise, it is determined as a U-direction planar point, and its U-direction fusion weight value is set as the U-direction projection deviation absolute value of the vertex. Similarly, the same processing is performed on the V-direction projection deviation. In this way, appropriate fusion weight values can be assigned to each vertex, so that the texture mapping problem in the subsequent texture fusion process can be better handled in the high-curvature feature area and the planar area. This classification and weight setting method can effectively improve the accuracy and visual effect of texture mapping, especially when dealing with complex geometric shapes of power facilities.

[0078] In some embodiments, the texture fusion weight value of each vertex is calculated based on the U-direction fusion weight value and the V-direction fusion weight value of each vertex using the following formula:

[0079]

[0080] wherein T weights (u, v) is the texture fusion weight value of vertex (u, v); W u (u, v) is the U-direction fusion weight value of vertex (u, v); W v (u, v) is the V-direction fusion weight value of vertex (u, v); ΔC is the color difference value between adjacent surface patches; σ is the color difference adjustment coefficient.

[0081] It should be noted that the calculation of the texture fusion weight value of each vertex is completed by comprehensively considering the U-direction fusion weight value and the V-direction fusion weight value. This process also introduces the color difference value and the color difference adjustment coefficient to further optimize the effect of texture fusion. The calculation of the texture fusion weight value aims to balance the fusion weight contribution in different directions and consider the influence of color difference on texture fusion, so as to generate more natural and continuous texture mapping results.

[0082] Specifically, the U-directional fusion weight value and the V-directional fusion weight value are parameters set based on the U-directional and V-directional projection deviation of the vertex, used to control the weight contribution of different directions in the texture fusion process. The texture fusion weight value is obtained by calculating the square root of the square sum of the U-directional and V-directional fusion weight values, and multiplying by a decay factor related to the color difference. The color difference value is calculated by comparing the color difference between adjacent surface patches, and the color difference adjustment coefficient is used to control the degree of influence of the color difference on the texture fusion weight value. The setting of these parameters can be adjusted according to the specific texture mapping requirements to achieve the best texture fusion effect. For example, if the color difference between adjacent surface patches is large, the influence of this difference on texture fusion can be reduced by appropriately increasing the color difference adjustment coefficient.

[0083] Preferably, the process of calculating the texture fusion weight value can be further refined into the following steps: first, calculate the square root of the square sum of the U-directional fusion weight value and the V-directional fusion weight value of each vertex to obtain a basic weight value. Then, calculate the color difference value between adjacent surface patches, which can be done by comparing the color information of adjacent vertices. Next, calculate the decay factor according to the color difference value and the color difference adjustment coefficient, which is usually a value between 0 and 1, used to reduce the influence of color difference on texture fusion. Finally, multiply the basic weight value by the decay factor to obtain the final texture fusion weight value. In this way, it can be ensured that the texture fusion weight value not only considers the geometric features, but also considers the color information, so as to generate more natural and continuous texture mapping results.

[0084] In some embodiments, the mapping generates a texture fusion weight field, including: normalizing the texture distribution value of each vertex of the texture distribution field using the following formula to obtain the texture fusion weight value of each vertex:

[0085] W final (u,v) = T smooth (u,v) / (k1*S max +k2*C max )

[0086] where W final (u,v) is the texture fusion weight value of vertex (u,v); T smooth (u,v) is the texture distribution value of vertex (u,v); S max is the maximum curvature value of the surface patch; C max is the maximum color difference value of the adjacent patch; k1, k2 are the balance coefficients of curvature and color difference.

[0087] It should be noted that the normalization process is to adjust the texture distribution values of each vertex in the texture distribution field to a unified standard range, and the purpose is to obtain the texture fusion weight value of each vertex. This process considers the maximum curvature value of the surface patch and the maximum color difference value of the adjacent patch, which reflect the geometric complexity and color variation degree of the surface patch respectively. Through normalization processing, it can ensure that the texture fusion weight value is within a reasonable range, avoid the problem of texture mapping caused by too large or too small weight value, and generate a more stable and reliable texture fusion weight field, thereby optimizing the effect of texture mapping.

[0088] Specifically, the texture distribution value is calculated in the previous step, which reflects the distribution of texture on the surface patch. The texture fusion weight value is obtained by normalization processing, which aims to adjust the weight value to a reasonable range, usually between 0 and 1. The maximum curvature value of the surface patch refers to the maximum value of the curvature of all vertices in the surface patch, which reflects the geometric complexity of the surface patch. The maximum color difference value of the adjacent patch refers to the maximum value of the color difference between adjacent surface patches, which reflects the degree of color variation. The balance coefficients k1 and k2 are parameters used to adjust the influence degree of curvature and color difference on the texture fusion weight value, which can be set according to the specific texture mapping requirements. For example, if the geometric shape of the surface patch is relatively complex, the value of k1 can be appropriately increased; if the color difference is large, the value of k2 can be appropriately increased.

[0089] Preferably, the specific steps of normalization processing can be further refined as follows: first, calculate the texture distribution value of each vertex, which is the value in the texture distribution field obtained in the previous step. Then, determine the maximum curvature value of the surface patch and the maximum color difference value of the adjacent patch, which can be obtained by analyzing the geometric characteristics and color information of the surface patch. Next, set the balance coefficients k1 and k2, which can be adjusted according to the specific texture mapping requirements. Finally, use the normalization method to calculate the texture fusion weight value of each vertex, that is, divide the texture distribution value of each vertex by the normalization factor composed of the balance coefficients, the maximum curvature value of the surface patch and the maximum color difference value of the adjacent patch. In this way, it can ensure that the texture fusion weight value is within a reasonable range, thereby providing accurate weight basis for subsequent texture fusion. This normalization processing method can effectively balance the influence of geometric characteristics and color difference on texture fusion, and improve the quality and stability of texture mapping.

[0090] In some embodiments, the projection deviation value of each vertex in the U direction mapping field of each surface patch is Gaussian weighted and summed with the U direction projection deviation value of the corresponding vertex of the adjacent surface patch to obtain the U direction projection deviation of each vertex in the surface patch:

[0091] Du (u,v) = a * Gσ(d u (u,v)) + (1-a) * D' u (u,v)

[0092] where D u (u,v) is the U-projection deviation of vertex (u,v) in the current surface patch; d u (u,v) is the projection deviation value of vertex (u,v) in the current U-mapping field; D' u (u,v) is the U-projection deviation of vertex (u,v) in the adjacent surface patch; a is a spatial continuity weight coefficient; Gσ is a Gaussian kernel function.

[0093] It should be noted that the U-projection deviation of each vertex in each surface patch is calculated by Gaussian weighted summation. This method takes into account the projection deviation value of the current vertex and the projection deviation value of the corresponding vertex in the adjacent surface patch. Weighting with the Gaussian kernel function can assign weights according to the spatial distance between vertices, so that the closer the vertices, the greater the influence on the current vertex. The spatial continuity weight coefficient is used to balance the contribution of the projection deviation values of the current vertex and the adjacent vertex to the final result, so as to more accurately reflect the spatial continuity in texture mapping and reduce the problem of texture discontinuity caused by the difference in viewing angle.

[0094] Specifically, the U-projection deviation refers to the deviation between the texture coordinates of a vertex and the ideal mapping position in the U-mapping field. Gaussian weighting is a method of assigning weights according to the spatial distance between vertices, and the current vertex projection deviation value can be weighted by the Gaussian kernel function. The spatial continuity weight coefficient is a parameter between 0 and 1, which is used to adjust the relative importance between the projection deviation value of the current vertex and the projection deviation value of the adjacent vertex. For example, when the weight coefficient is large, the contribution of the projection deviation value of the current vertex to the final result is greater; when the weight coefficient is small, the contribution of the projection deviation value of the adjacent vertex to the final result is greater. By reasonably setting the weight coefficient, the spatial continuity of texture mapping can be optimized.

[0095] Preferably, the process of calculating the U-projection deviation can be refined as follows: First, determine the projection deviation value of the vertex in the U-mapping field of the current surface patch. Next, obtain the U-projection deviation value of the corresponding vertex in the adjacent surface patch. Then, select a suitable spatial continuity weight coefficient, which can be adjusted according to the specific requirements of texture mapping. Finally, use the Gaussian weighting method to weight the projection deviation value of the current vertex, and combine the adjusted projection deviation value of the adjacent vertex with the weight coefficient to obtain the final U-projection deviation. This method can effectively handle the spatial continuity problem in texture mapping and improve the quality and visual effect of texture mapping.

[0096] In some embodiments, the projection deviation value of each vertex in the V-direction mapping field of each surface patch is Gaussian-weighted summed with the V-direction projection deviation value of the corresponding vertex of the adjacent surface patch using the following formula to obtain the V-direction projection deviation of each vertex in the surface patch:

[0097] D v (u,v) = β * Gσ(d v (u,v)) + (1 - β) * D′ v (u,v)

[0098] where D v (u,v) is the V-direction projection deviation of vertex (u,v) in the current surface patch; d v (u,v) is the projection deviation value of vertex (u,v) in the current V-direction mapping field; D′ v (u,v) is the V-direction projection deviation of vertex (u,v) in the adjacent surface patch; and β is the geometric consistency weight coefficient.

[0099] It should be noted that the V-direction projection deviation of each vertex in each surface patch is calculated by Gaussian-weighted summation. This process is similar to the calculation of the U-direction projection deviation, but focuses on the deviation in the V-direction mapping field. The geometric consistency weight coefficient is used to balance the contribution of the V-direction projection deviation values of the current vertex and the adjacent vertex to the final result, ensuring the continuity of texture mapping on the geometric shape. In this way, the geometric consistency in texture mapping can be more accurately reflected, reducing the problem of texture discontinuity caused by the difference in viewing angle.

[0100] Specifically, the V-direction projection deviation refers to the deviation between the texture coordinates of a vertex and the ideal mapping position in the V-direction mapping field. The geometric consistency weight coefficient is a parameter between 0 and 1, used to adjust the relative importance between the V-direction projection deviation value of the current vertex and the V-direction projection deviation value of the adjacent vertex. The Gaussian kernel function is a commonly used weighting function that can assign weights according to the spatial distance between vertices, so that the closer the vertices, the greater the influence on the current vertex. For example, when the geometric consistency weight coefficient is large, the V-direction projection deviation value of the current vertex contributes more to the final result; when the coefficient is small, the V-direction projection deviation value of the adjacent vertex contributes more to the final result. By reasonably setting the geometric consistency weight coefficient, the geometric consistency of texture mapping can be optimized.

[0101] Preferably, the process of calculating the V-projection deviation can be refined into the following steps: first, determine the projection deviation value in the V-mapping field of the vertex in the current surface patch. Next, obtain the V-projection deviation value of the corresponding vertex in the adjacent surface patch. Then, select a suitable geometric consistency weight coefficient, which can be adjusted according to the specific needs of texture mapping. Finally, use the Gaussian weighting method to weight the V-projection deviation value of the current vertex, and combine the V-projection deviation value of the adjacent vertex adjusted by the geometric consistency weight coefficient, to obtain the final V-projection deviation. This method can effectively handle the geometric consistency problem in texture mapping, and improve the quality and visual effect of texture mapping.

[0102] In some embodiments, the multi-scale fusion of the texture fusion weight field of each surface patch with the texture data of adjacent surface patches generates an optimized texture mapping result, which is formulated as follows:

[0103] I fuse (u,v) = W final (u,v) * I current (u,v) + (1 - W final (u,v) ) * I adjacent (u,v)

[0104] where W final (u,v) is the texture fusion weight value of vertex (u,v); I current (u,v) is the texture color value of the current surface patch vertex (u,v); I adjacent (u,v) is the texture color value of the adjacent surface patch vertex (u,v).

[0105] It should be noted that the multi-scale fusion of the texture fusion weight field of each surface patch with the texture data of adjacent surface patches is to generate an optimized texture mapping result. This process balances the texture data of the current surface patch and the adjacent surface patch by considering the texture fusion weight value, thereby reducing texture discontinuity and color difference, and improving the overall quality of texture mapping. Multi-scale fusion means that texture data is fused at different resolutions or scales to ensure good texture mapping effect at different levels of detail.

[0106] Specifically, the texture fusion weight field is calculated based on the previous steps, which reflects the importance of each vertex in the texture fusion process. The texture color value of the current surface patch refers to the texture color of the current vertex on the current surface patch, and the texture color value of the adjacent surface patch refers to the texture color of the adjacent vertex on the adjacent surface patch. The texture fusion weight value is a parameter between 0 and 1, which is used to adjust the relative contribution of the texture data of the current surface patch and the adjacent surface patch in the fusion process. For example, if the texture fusion weight value of a vertex is close to 1, the texture data of the current surface patch will dominate in the fusion process; if the weight value is close to 0, the texture data of the adjacent surface patch will dominate. By reasonably setting the texture fusion weight value, smooth texture transition between different surface patches can be achieved.

[0107] Preferably, the specific steps of multi-scale fusion can be refined as follows: first, determine the texture fusion weight value of each vertex, which is calculated by the previous steps. Next, obtain the texture color value of each vertex in the current surface patch, and the texture color value of the corresponding vertex in the adjacent surface patch. Then, according to the texture fusion weight value, the texture data of the current surface patch and the adjacent surface patch is weighted and fused. This process can be carried out at multiple different scales, for example, first at a lower resolution to process larger texture features, and then at a higher resolution to process smaller texture details. Finally, the fusion results at different scales are combined to obtain the final optimized texture mapping result. In this way, smooth fusion of texture data at different levels of detail can be achieved, improving the overall quality and visual effect of texture mapping.

[0108] In some embodiments, the generating the U-mapping field and the V-mapping field of each surface patch comprises:

[0109] Based on the projection relationship between the surface patches of the three-dimensional point cloud model and the corresponding RGB image, the projection coordinate deviation of each surface patch vertex in the texture space is calculated using an improved ICP algorithm;

[0110] The projection coordinate deviation is decomposed into U-direction deviation component and V-direction deviation component, and the U-direction deviation component and the V-direction deviation component are respectively subjected to Laplacian smoothing processing to obtain the U-mapping field and the V-mapping field of each surface patch.

[0111] It is noted that the U-mapping field and the V-mapping field of each surface patch are generated by an improved ICP algorithm. ICP algorithm is a commonly used point cloud registration algorithm, which is used to calculate the best match between two point clouds. In this method, the improved ICP algorithm is used to calculate the projection relationship between the surface patches of the three-dimensional point cloud model and the corresponding RGB image, so as to obtain the projection coordinate deviation of each surface patch vertex in the texture space. These deviations are decomposed into U-direction deviation components and V-direction deviation components, and the final U-mapping field and V-mapping field are generated by Laplace smoothing processing. This method can effectively handle complex geometric shape and texture mapping problems, and improve the accuracy and stability of texture mapping.

[0112] Specifically, the improved ICP algorithm is a point cloud registration-based algorithm, which minimizes the distance between two point clouds through iteration to find the best matching relationship. In this method, the algorithm is used to calculate the projection relationship between the surface patches of the three-dimensional point cloud model and the corresponding RGB image. The projection relationship refers to how the points on the three-dimensional point cloud model are mapped to the two-dimensional RGB image. Through the improved ICP algorithm, the projection coordinate deviation of each surface patch vertex in the texture space can be obtained. These deviations are decomposed into U-direction deviation components and V-direction deviation components, wherein the U-direction and the V-direction correspond to the horizontal and vertical directions in the texture coordinate system respectively. Laplace smoothing processing is a commonly used smoothing algorithm, which is used to reduce noise and discontinuity in the mapping field, thereby generating smoother U-mapping field and V-mapping field.

[0113] Preferably, the process of generating U-mapping field and V-mapping field can be refined into the following steps: first, the projection relationship between the surface patches of the three-dimensional point cloud model and the corresponding RGB image is calculated using the improved ICP algorithm. This process involves preprocessing of point cloud data and RGB image data, such as noise removal, alignment, etc. Then, the projection coordinate deviation of each surface patch vertex in the texture space is calculated by the improved ICP algorithm. These deviations reflect the errors caused by geometric shape and viewing angle differences during texture mapping. Then, these deviations are decomposed into U-direction deviation components and V-direction deviation components. Finally, the U-direction and V-direction deviation components are respectively subjected to Laplace smoothing processing to generate the final U-mapping field and V-mapping field. In this way, noise and discontinuity in the mapping field can be effectively reduced, and the accuracy and visual effect of texture mapping can be improved.

[0114] The above-mentioned various embodiments of the present application have the following beneficial effects:

[0115] 1、The application establishes a high-precision U / V mapping field by multi-view RGB-D sequence data acquisition and three-dimensional point cloud modeling, combining the accurate matching of surface patch geometric features and texture coordinates, effectively solving the texture misplacement problem caused by the complex geometric structure of power facilities, realizing the accurate alignment of texture and geometric model, and improving the geometric fidelity of texture mapping.

[0116] 2、The projection deviation calculation method based on Gaussian weighting comprehensively considers the geometric consistency of adjacent surface patches, divides the high-curvature area and flat area through an adaptive feature threshold, and dynamically adjusts the fusion weight, thereby suppressing the texture distortion and stretching phenomenon caused by the curvature mutation of the surface of the power facility, and ensuring the natural transition of the texture in the complex curved surface area.

[0117] 3、Through a multi-scale texture fusion mechanism, the texture distribution field is optimized in combination with bilateral filtering, and the influence of curvature and color difference is balanced by using normalization processing, seamless fusion of adjacent patch textures is realized, the texture discontinuity problem caused by the difference in viewing angle or changes in illumination in the traditional method is eliminated, and finally a high-quality texture mapping result with visual consistency and rich details is generated.

[0118] Further, the storage medium of the embodiments of the application stores program instructions capable of implementing all the methods described above, wherein the program instructions can be stored in the storage medium in the form of a software product, including a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods described in various embodiments of the application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media capable of storing program codes, or a computer, a server, a mobile phone, a tablet, and other terminal devices.

[0119] The above description is only some of the preferred embodiments of the application and an explanation of the technical principles applied. Those skilled in the art should understand that the scope of the application involved in the embodiments of the application is not limited to the technical solutions formed by the specific combinations of the above technical features, and should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above inventive concept. For example, the above features and the technical features disclosed in the embodiments of the application (but not limited to) having similar functions are replaced with each other to form a technical solution.

Claims

1. A three-dimensional point cloud texture mapping method for power facilities, characterized in that: The steps include: Acquire multi-view RGB-D sequence data of power facilities and generate a 3D point cloud model; Extracting geometric features of each surface slice in the three-dimensional point cloud model and matching the texture coordinates of the corresponding RGB image to generate a U-direction mapping field and a V-direction mapping field of each surface slice; Calculating the texture projection deviation based on the U-direction mapping field and the V-direction mapping field to generate a texture distribution field for each surface fragment and mapping to generate a texture fusion weight field; The texture fusion weight field of each surface slice is fused with the texture data of adjacent surface slices at multiple scales to generate an optimized texture mapping result.

2. The method according to claim 1, characterized in that The calculating of the texture projection deviation based on the U-direction mapping field and the V-direction mapping field includes: Performing Gaussian weighted summation on the projection deviation value of each vertex in the U-direction mapping field of each surface slice and the U-direction projection deviation value of the vertex corresponding to the adjacent surface slice to obtain the U-direction projection deviation of each vertex in each surface slice; The projection deviation value of each vertex in the V-direction mapping field of each surface slice is Gaussian-weighted summed with the V-direction projection deviation value of the vertex corresponding to the adjacent surface slice to obtain the V-direction projection deviation of each vertex in each surface slice.

3. The method according to claim 2, characterized in that Generating the texture distribution field of each surface slice includes: Based on the U-direction projection deviation and the V-direction projection deviation of each vertex in each surface fragment, each vertex is divided into a high curvature feature area and a plane area through a preset threshold and a U-direction fusion weight value and a V-direction fusion weight value are set; Calculating the texture fusion weight value of each vertex based on the U-direction fusion weight value and the V-direction fusion weight value of each vertex; The texture fusion weight value of each vertex is subjected to bilateral filtering to obtain the texture distribution value of each vertex of each surface fragment to form the texture distribution field of each surface fragment.

4. The method according to claim 3, characterized in that The method of dividing each vertex into a high curvature feature area and a plane area based on the U-direction projection deviation and the V-direction projection deviation of each vertex in each surface fragment by a preset threshold and setting the U-direction fusion weight value and the V-direction fusion weight value includes: When the absolute value of the U-direction projection deviation of the vertex in each surface slice is greater than or equal to the U-direction feature threshold, the vertex is determined to be a U-direction high curvature feature point, and the U-direction fusion weight value of the vertex is set to the U-direction feature threshold; When the absolute value of the U-direction projection deviation of the vertex in each surface slice is less than the U-direction feature threshold, the vertex is determined to be a U-direction plane point, and the U-direction fusion weight value of the vertex is set to the absolute value of the U-direction projection deviation of the vertex; When the absolute value of the V-direction projection deviation of the vertex in each surface slice is greater than or equal to the V-direction feature threshold, the vertex is determined to be a V-direction high curvature feature point, and the V-direction fusion weight value of the vertex is set to the V-direction feature threshold; When the absolute value of the V-direction projection deviation of the vertex in each surface fragment is less than the V-direction feature threshold, the vertex is determined to be a V-direction plane point, and the V-direction fusion weight value of the vertex is set to the absolute value of the V-direction projection deviation of the vertex.

5. The method according to claim 3 or 4, characterized in that: The texture fusion weight value of each vertex is calculated based on the U-direction fusion weight value and the V-direction fusion weight value of each vertex using the following formula: Among them, T weights (u,v) is the texture fusion weight value of vertex (u,v); W u (u,v) is the U-direction fusion weight value of vertex (u,v); W v (u,v) is the V-direction fusion weight value of vertex (u,v); ΔC is the color difference value between adjacent surface fragments; σ is the color difference adjustment coefficient.

6. The method according to claim 5, characterized in that The mapping generates a texture fusion weight field, including: normalizing the texture distribution value of each vertex of the texture distribution field using the following formula to obtain a texture fusion weight value of each vertex: W final (u,v)=T smooth (u,v) / (k1*S max +k2*C max ) Among them, W final (u,v) is the texture fusion weight value of vertex (u,v); T smooth (u,v) is the texture distribution value of vertex (u,v); S max is the maximum curvature value of the surface patch; C max is the maximum color difference between adjacent slices; k1 and k2 are the balance coefficients between curvature and color difference.

7. The method according to claim 2, characterized in that: The following formula is used to perform Gaussian weighted summation of the projection deviation value of each vertex in the U-direction mapping field of each surface slice and the U-direction projection deviation value of the vertex corresponding to the adjacent surface slice to obtain the U-direction projection deviation of each vertex in each surface slice: D u (u,v)=α*Gσ(d u (u,v))+(1-α)*D′ u (u,v) Among them, D u (u,v) is the U-direction projection deviation of the vertex (u,v) in the current surface fragment; d u (u, v) is the projection deviation value of the vertex (u, v) in the current U-direction mapping field; D ′ u (u, v) is the U-direction projection deviation of the vertex (u, v) in the adjacent surface fragments; α is the spatial continuity weight coefficient; Gσ is the Gaussian kernel function.

8. The method according to claim 7, characterized in that: The following formula is used to perform Gaussian weighted summation of the projection deviation value of each vertex in the V-direction mapping field of each surface slice and the V-direction projection deviation value of the vertex corresponding to the adjacent surface slice to obtain the V-direction projection deviation of each vertex in each surface slice: D v (u,v)=β*Gσ(d v (u,v))+(1-β)*D′ v (u,v) Among them, D v (u,v) is the V-direction projection deviation of the vertex (u,v) in the current surface fragment; d v (u, v) is the projection deviation value of the vertex (u, v) in the current V-direction mapping field; D ′ v (u, v) is the V-direction projection deviation of the vertex (u, v) in adjacent surface fragments; β is the geometric consistency weight coefficient.

9. The method according to claim 1, characterized in that: The texture fusion weight field of each surface slice is multi-scale fused with the texture data of the adjacent surface slices to generate the optimized texture mapping result, and the formula is as follows: I fuse (u,v)=W final (u,v)*I current (u,v)+(1-W final (u,v))*I adjacent (u,v) Among them, W final (u,v) is the texture fusion weight value of vertex (u,v); I current (u,v) is the texture color value of the vertex (u,v) of the current surface fragment; I adjacent (u,v) is the texture color value of the adjacent surface patch vertex (u,v).

10. The method according to claim 1, characterized in that: Generating the U-direction mapping field and the V-direction mapping field of each surface slice includes: Based on the projection relationship between the surface fragments of the three-dimensional point cloud model and the corresponding RGB image, the improved ICP algorithm is used to calculate the projection coordinate deviation of each surface fragment vertex in the texture space; The projection coordinate deviation is decomposed into a U-direction deviation component and a V-direction deviation component, and Laplace smoothing is performed on the U-direction deviation component and the V-direction deviation component respectively to obtain a U-direction mapping field and a V-direction mapping field of each surface slice.

Images