An indoor structure point cloud contour line optimization method based on gestalt theory

Abstract

The application discloses an indoor structure point cloud contour line optimization method based on Gestalt theory, which comprises the following steps: (1) obtaining indoor point cloud data, rotating the point cloud to align with the coordinate system, and reducing the point cloud data amount through voxel sampling; (2) segmenting indoor structure plane point cloud, and then extracting indoor structure contour line based on the plane; (3) removing outlier line segments caused by occlusion in each plane based on the simple principle of Gestalt complete form rule; (4) fusing repeated extracted line segments based on the similarity principle and the adjacent principle, and then regularizing the contour line; (5) correcting topological error line segments (discontinuity and non-closure) in the contour line based on the continuity principle and the closure principle; (6) repairing and completing the floor contour and the wall contour in turn based on the symmetry principle; and (7) finally repeating steps (4) and (5) to obtain the final result. The application greatly improves the integrity, accuracy and readability of the contour line.

Description

Technical Field

[0001] This invention relates to the field of vector reconstruction technology for indoor scenes, and in particular to a method for optimizing the outline of indoor structural point clouds based on Gestalt theory. Background Technology

[0002] Line segments, as the most common basic element, play a crucial role in the vectorization reconstruction of indoor scenes. Currently, point cloud contour optimization methods include deep learning-based methods and traditional optimization methods. Deep learning-based methods optimize the local features of wireframe models by designing neural networks, but so far they can only achieve contour optimization consistent with labeled samples. However, real-world scenes are very complex, and limited artificially labeled samples are insufficient to optimize diverse contours. Wang et al. (2018) constructed a conditional generative adversarial network (cGAN) deep learning model to optimize wireframes, including structural completion, extrusion removal, and regularization, but this method does not fully consider the problems existing in line segment extraction and cannot cope with complex indoor scenes. (Reference 1: Cheng Wang, Shiwei Hou, Chenglu Wen, Zheng Gong, Qing Li, Xiaotian Sun, Jonathan Li, Semantic line framework-based indoor building modeling using backpacked laser scanning point cloud, ISPRS Journal of Photogrammetry and Remote Sensing, 2018, 143, 150.). Traditional optimization methods mainly optimize contours by using geometric and topological information in the contours. Lu et al. (2019) fused repeatedly extracted line segments from the contour and removed outliers, but did not consider missing structural line segments or topological errors such as unclosed and discontinuous sections in the contour. (Reference 3: Lu, X.; Liu, Y.; Li, K. Fast 3D line segment detection from unorganized point cloud. arXiv2019, arXiv:1901.02532.). Tian et al. (2022) achieved contour fusion, but lacked solutions for handling outliers, topological errors, and missing structural line segments, resulting in low accuracy of line segment extraction results. (Reference 4: Tian, ​​P.; Hua, X.; Tao, W.; Zhang, M. Robust Extraction of 3D Line Segment Features from Unorganized Building Point Clouds. Remote Sens. 2022, 14, 3279.).

[0003] In summary, current traditional contour optimization methods cannot effectively address issues such as missing line segments, numerous messy line segments, discontinuous contours, and unclosed contours, resulting in low accuracy and insufficient completeness of the extracted wireframes. Therefore, how to cleverly utilize the structural information contained in point cloud contours to compensate for information loss and other anomalies during the acquisition process remains a challenging research problem. Summary of the Invention

[0004] Due to factors such as target occlusion, viewing angle, and complex object surfaces, extracted structural line segments exhibit anomalies such as missing segments, topological errors, duplicate extractions, and incomplete or discontinuous segments. Existing contour optimization methods cannot effectively handle these situations, significantly reducing the accuracy and completeness of the extraction results. To address these issues, this invention combines Gestalt rules to optimize the contour lines of indoor architectural structures from the perspective of human visual cognition. It proposes a Gestalt-based method for optimizing indoor structural point cloud contour lines, aiming to optimize different types of abnormal line segments and obtain a complete and accurate 3D wireframe model.

[0005] To achieve the above objectives, this invention applies Gestalt rules to structural line segment optimization. Different Gestalt rules are used to design objective functions, and the geometric information of the interior structural outline is combined to complete and repair missing lines caused by occlusion and a large number of abnormal lines. Experiments show that the optimization results not only better conform to human visual perception but also greatly improve the integrity, accuracy, and readability of the outline. This method has higher precision and recall compared to existing methods. The main steps of the proposed Gestalt theory-based optimization method for interior structural point cloud outlines are as follows:

[0006] An optimization method for indoor structural point cloud contours based on Gestalt theory includes:

[0007] Step 1: Acquire indoor point cloud data and rotate the indoor point cloud data to align with the coordinate system;

[0008] Step 2: Segment the indoor structural planar point cloud data, and then extract the contour lines based on the planar data;

[0009] Step 3: Based on the simplicity principle in Gestalt gestalt rules, remove outlier segments caused by occlusion in each plane;

[0010] Step 4: Based on the similarity and proximity principles in Gestalt gestalt rules, repeatedly extracted line segments are merged, and then the contour lines are regularized.

[0011] Step 5: Correct topologically incorrect line segments in the contour line based on the continuity and closure principles in Gestalt gestalt rules;

[0012] Step 6: Based on the symmetry principle in Gestalt gestalt rules, repair and complete the partially and completely missing line segments in the outline. Repeat steps 4-5 to obtain the final indoor structure point cloud outline.

[0013] Further, step 3 includes:

[0014] The set of contour lines L of the plane 1 all The longest line segment is considered the main direction line segment, and L is... 1 all Line segments whose angle with the main direction line segment is less than the angle threshold ε1 are considered structural line segments. These structural line segments are then obtained. The remaining longest line segment is used again as the main direction line segment to find structural line segments. Line segments that are not in the main direction but are longer than the average length are also considered structural line segments. The remaining segments are considered outliers. After removing outliers, a new line segment set L is obtained. 2 all .

[0015] Further, step 4 includes:

[0016] From L 2 all Starting with the longest line segment, and combining the angles, distances, and spatial relationships between line segments, in set L... 2 all The search term is L, which represents the line segment that can be merged with the given line segment. 3 all The conditions for line segment fusion include: the angle between line segments is less than ε1, and the distance between line segments is less than the adaptive distance threshold. The projection of at least one endpoint of a line segment lies within another line segment.

[0017] Furthermore, the distance between the line segments is calculated according to the following formula:

[0018] d f =min{d s ,d m ,d e}

[0019] Where d s ,d m ,d e These are the distances from the start point, midpoint, and end point of the shorter line segment to the longer line segment, respectively;

[0020] The distance d from a point to a line is calculated using the following formula:

[0021]

[0022]

[0023] in It is the distance from the point to the starting point of the line segment. It is the distance from the point to the end of the line segment. It is the length of the line segment.

[0024] Furthermore, the adaptive distance threshold Calculate according to the following formula:

[0025]

[0026] Among them l a l1 is the average length of all line segments, and l2 is the length of the two line segments being processed.

[0027] Furthermore, the method of merging the line segments includes:

[0028] The contour lines of the planar edges are processed as follows: First, find the fusion reference point (x) of L1 and L2. b ,y b ,z b Divide L1 and L2 into two groups: the two closest endpoints are on different line segments, and the remaining two endpoints form another group. Find the endpoints in each group that are closest to x. b y b and z b The x, y, and z values ​​with the largest differences are used as the starting or ending points of the merged line segment; the merge reference point (x b ,y b ,z b The calculation formula is as follows:

[0029]

[0030] Where (x) min ,y min ,z min ) and (x max ,y max ,z max ) are L 2 all Minimum and maximum coordinates of the endpoints of the line segment, (x A ,y A ,z A ) and (x B ,y B ,z B () represents the coordinates of the two points that are furthest apart from the four endpoints of the line segment being processed;

[0031] To handle cases where an in-plane contour line is extracted into two line segments due to occlusion, the following method is used: From the four endpoints of the two line segments and the two perpendicular projection points of the shorter line segment onto the longer line segment, select the two points that are furthest apart as the start and end points of the new line segment after fusion.

[0032] Furthermore, in step 4, the regularization process for the contour lines includes:

[0033] Determine distance L 3 all The longest line segment l 3 1st The line segment closest to the two endpoints, and according to l 3 1st The angle between the nearest line segment and l 3 1st Perform parallel or orthogonal adjustments, and then adjust other line segments sequentially using the adjusted line segment as a reference until L is reached. 3 all All line segments in the middle were processed, and the adjusted line segments were placed into L. 4 all .

[0034] Furthermore, in step 5, for discontinuous line segments, first from L... 4 all Search for the shortest line segment l 4 1st The line segment closest to the endpoint is used as the baseline, and then l is calculated. 4 1st The perpendicular projection point of the endpoint and the adjacent line segment, the adjusted line segment l n Follow the formula below:

[0035]

[0036] Where ang(l,l) n () is used to calculate the angle of the line segment before and after adjustment, l nl For the adjusted line segment length, l l This is the original line segment length;

[0037] For unclosed line segments, the processing order is determined by the length of the unclosed line segment. First, find L. 4 all The shortest unclosed line segment l 4 tem An unclosed endpoint P 4 s Then find the distance P 4 s The line segment containing the three nearest endpoints and the distance l 4 temThe line segment containing the three nearest endpoints to the other endpoint is used to calculate the perpendicularity coefficient V. er If V er If it is greater than 0.05, then P will be... 4 s Change it to its nearest endpoint; the modified line segment is now l. 4 tem2 After all unclosed endpoints have been processed, the endpoints of the distance segment are iteratively removed. The unclosed line segment within is obtained as L. 5 all The vertical coefficient V is calculated according to the following formula. er :

[0038]

[0039] Where i is the distance P 4 s The three nearest endpoints of the line segment are related to l 4 tem The number of approximately perpendicular ones, l i θ represents the approximately perpendicular line segment mentioned above. ir It is l 4 tem With l i The angle between them, θ ic It is l 4 tem2 With l i The angle between them, where n represents the distance l 4 tem The other endpoint is one of the three nearest endpoints to l 4 tem The number of approximately perpendicular line segments, l n Representing the above line segment, θ nr It is l 4 tem With l n The angle between them, θ nc It is l 4 tem2 With l n The angle between them.

[0040] Further, determine according to the following formula

[0041]

[0042] in is the distance threshold, and l is the length of the line segment being processed.

[0043] Further, step 6 includes:

[0044] First, repair the missing lines in the floor using ceiling projection lines:

[0045] Find the ceiling wireframe top Center and floor frame l bot The median segment is a line segment l that lies in the same plane and is symmetrical in direction. sy , will l sy The z-value of the endpoint is modified to the average z-value of the endpoints of the floor segment, if l bot The endpoints of the line segment in l sy Adjacent or in l sy Above, then l bot Replace the line segment in the text with l sy Then, use the remaining ceiling segment lr to repair the entire missing segment in the floor, modifying the z-value of lr to the average z-value of the floor wireframe endpoints. bot If there exists an endpoint adjacent to the endpoint of lr, and lr is a boundary segment in the floor, then lr is considered the entire missing segment of the floor, and lr is added to l. bot middle;

[0046] Then, using the repaired floor outline, the missing lines on each wall surface are filled in:

[0047] If there is only one unclosed point p1, then p1 is adjusted to be the perpendicular projection point of the nearest line segment ln, and then the endpoint of ln is modified to be the projection point of the line segment containing p1. If there are two or more unclosed points, then the two closest points are considered as a group. If the line segments containing the two points are parallel and the plane is perpendicular, a line segment is constructed by projecting the point of the longer line segment onto the point of the shorter line segment, and then the point of the shorter line segment is modified to be the perpendicular projection point of the newly constructed line segment. If the line segments containing the two points are perpendicular, then the endpoint of the line segment containing one point in the group is modified to be the projection point of the line segment containing the other point, without destroying the original topology. Finally, the line segments with unclosed endpoints are removed to obtain the final result.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] This invention optimizes the extracted interior structural outlines, utilizing Gestalt rules and the implicit architectural structural information within the outlines to repair topological errors, local missing values ​​caused by occlusion, outliers due to surface holes, and missing structural segments. Experiments demonstrate that this invention significantly improves the completeness of the results, achieving both precision and recall rates of 92%, providing stronger support for subsequent vectorized reconstruction. Attached Figure Description

[0050] Figure 1 This is a flowchart illustrating an indoor structural point cloud contour optimization method based on Gestalt theory, according to an embodiment of the present invention.

[0051] Figure 2 The results of point cloud contour extraction and optimization for example scenes;

[0052] Figure 3 This is a schematic diagram of abnormal line segments in the outline;

[0053] Figure 4 This is a schematic diagram of the contour line to be optimized in the projection plane. Detailed Implementation

[0054] The present invention will be further explained below with reference to the accompanying drawings and specific embodiments:

[0055] like Figure 1 As shown, a method for optimizing the outline of indoor structural point clouds based on Gestalt theory includes:

[0056] Step S101: Download the S3DIS dataset. The S3DIS dataset contains six regions and 271 rooms. The following steps take the selected Area_3_lounge_1 indoor point cloud as an example. Figure 2 As shown in (a).

[0057] Step S102: Point cloud preprocessing. To maximize the use of the geometric and topological information of the point cloud, the original point cloud is rotated to make the ceiling and floor parallel to the XOY plane, while trying to make other planes parallel to the XOZ and YOZ planes.

[0058] Step S103: Use RANSAC to segment the interior structural plane (ceiling, floor, walls, beams, etc.).

[0059] Step S104: Project the point cloud of the segmented structure onto its best-fit plane to obtain two-dimensional projection points.

[0060] Step S105: Extract contour lines from the 2D projection. The α-shape algorithm is used to extract contour points from the 2D projection, and line fitting is performed using the RANSAC algorithm. The extracted contour lines are stored in L... 1 all In, such as Figure 2 As shown in (b).

[0061] Step S106: Optimize the anomalies in the contour lines based on Gestalt gestalt rules. This invention mainly addresses anomalies such as messy line segments, repeated extraction, non-regularization (non-parallel and non-perpendicular), topological errors (unclosed and discontinuous), and missing line segments. Figure 3 (As shown) Optimization processing will be performed. The processing solutions for different situations are as follows:

[0062] (1) Remove outlier line segments caused by occlusion in each plane based on the principle of simplicity. Interior structural outlines typically display structural information such as parallelism and orthogonality, but also contain line segments that do not conform to the Manhattan World Assumption (MWA). Therefore, remove cluttered line segments based on the plane. While non-MWA line segments in the ceiling can be filtered out, repeatedly extracted line segments in the vertical plane are retained. The set of outlines for the plane L 1 all The longest line segment is considered the principal direction line segment. Line segments with an angle less than ε1 between their principal directions in the plane are considered structural line segments; in this invention, ε1 is set to 10° (approximately parallel). After finding these line segments, the remaining longest line segment is used again as the principal direction line segment to find structural line segments. Furthermore, longer line segments are more likely to be correct. Therefore, non-principal directions but longer than the average length are also considered structural line segments, and other random line segments are considered outliers. After removing outliers, a new set of line segments L is obtained. 2 all The line segments in the set are sorted in descending order of length. The formula for calculating the angle θ1 between the line segments is as follows:

[0063]

[0064] Where (a1,b1,c1) and (a2,b2,c2) are the normalized direction vectors obtained from the endpoints of the two line segments, respectively.

[0065] (2) Merging of repeatedly extracted line segments based on the principles of similarity and proximity. From L 2 all The longest line segment l 2 1st Initially, considering the angles, distances, and spatial relationships between line segments, in set L... 2 all Search for line segments that can be merged with it. The result of merging all line segments is denoted as L. 3 all The merging conditions for line segments L1 and L2 are as follows:

[0066] 1) The two line segments need to be approximately parallel, that is, the angle between L1 and L2 is less than ε1.

[0067] 2) The distance d between L1 and L2 f Less than the adaptive distance threshold

[0068] 3) At least one projection point of the endpoint of L1 is located within L2.

[0069] This involves the distance d between two line segments. f The calculation formula is as follows:

[0070]

[0071] Where d s ,d m ,d e They are l 2 nst The starting point, midpoint, and ending point of (the shorter line segment) to line segment l 2 1st The distance from a point to a line is d. In this invention, the formula for calculating the distance d from a point to a line is as follows:

[0072]

[0073] in

[0074] (4)

[0075] in, It is the distance from the point to the starting point of the line segment. It is the distance from the point to the end of the line segment. It is the length of the line segment.

[0076] The adaptive distance threshold is obtained from the line segment length and the line segment difference, and its calculation formula is as follows:

[0077]

[0078] Among them l a L1 is the average length of all line segments, and L2 are the lengths of the two line segments being processed. There are two fusion methods for line segments L1 and L2 that meet the fusion criteria. The first method primarily processes the contour lines of planar edges (the length of L1 is less than twice the length of L2), aiming to restore the instance contour lines to the greatest extent possible, ensuring the accuracy of edge features during the fusion process. Specifically, this invention first finds the fusion reference point (x) of L1 and L2. b ,y b ,z b Divide L1 and L2 into two groups: the two closest endpoints (located on different line segments) and the remaining two endpoints into another group. Find the endpoint in each group that is closest to x. b y b and z b The x, y, and z values ​​with the largest differences are used as the starting or ending points of the merged line segment. Merging reference point (x... b ,y b ,z b The calculation formula is as follows:

[0079]

[0080] Where (x) min ,y min ,z min ) and (xmax ,y max ,z max ) are L 2 all Minimum and maximum coordinates of the endpoints of the line segment, (x A ,y A ,z A ) and (x B ,y B ,z B The coordinates of the two furthest points among the four endpoints of the line segment being processed are denoted as x. The second fusion method primarily addresses the issue of in-plane contour lines being extracted as two line segments due to occlusion. To preserve the original line segment information, this invention selects the two furthest points from six points—the four endpoints of the two line segments and the two perpendicular projection points of the shorter line segment onto the longer line segment—as the start and end points of the new line segment. The perpendicular projection points (x) during the fusion process... p ,y p ,z p The calculation formula is as follows:

[0081]

[0082] Where (x, y, z) are the points to be projected, and (a, b, c) are the normalized vectors of the direction vectors of the projection line segments. l ,y l ,z l () represents the starting coordinates of the projected line segment. The line segment fusion algorithm is shown in Table 1.

[0083] Table 1 Segment Merging

[0084]

[0085] (3) Regularization processing: Regularization processing is performed on approximately parallel / perpendicular contour lines. The distance L is determined according to formula (3). 3 all The longest line segment l 3 1st The line segment closest to the two endpoints, and the angle between them is used to determine l. 3 1st Perform parallel or orthogonal adjustments. Then, using the adjusted line segment as a reference, adjust the other line segments sequentially until L... 3 all All line segments in the middle were processed, and the adjusted line segments were placed into L. 4 all When adjusting in parallel, the formula for calculating the endpoints is as follows:

[0086]

[0087] Where (x) e ,ye ,z e (x) is the adjusted endpoint. s ,y s ,z s (a,b,c) is the starting point of the adjustment line segment, l is the length of the adjustment line segment, and (a,b,c) is the normalized vector of the direction vector of the baseline line segment. Vertical adjustment is similar to parallel processing, except that (a,b,c) in formula (8) is calculated by formula (9).

[0088]

[0089] Where (a) l ,b l ,c l (a) is the normalized vector after adjusting the direction vector of the line segment. b ,b b ,c b ) is the normalized vector of the baseline segment, d p It is the dot product of two vectors.

[0090] (4) Correct topological errors (discontinuities and unclosed segments) in the contour line based on continuity and closure. For discontinuous segments, first start from L... 4 all Search for the shortest line segment l 4 1st The line segment closest to the endpoint is used as the baseline, and then l is calculated using formula (7). 4 1st The perpendicular projection points of the endpoints and adjacent segments. To maintain the original structure, this invention proposes the standard of "maintaining topological integrity." Adjusted segment l n The following formula should be followed:

[0091]

[0092] Where ang(l,l) n () is used to calculate the angle of the line segment before and after adjustment, l nl For the adjusted line segment length, l l This is the original line segment length.

[0093] If topological integrity is not maintained, then the next nearest line segment should be processed in the manner described above.

[0094] For unclosed line segments, the processing order is determined by the length of the unclosed line segments, starting with the shortest unclosed line segment, using l. 4 tem L represents 4 all The shortest unclosed line segment in the equation is represented by P. 4 s Indicate l4 tem Given an unclosed endpoint, find the distance P. 4 s The line segment containing the three nearest endpoints and the distance l 4 tem The line segment containing the three nearest endpoints to the other endpoint (which may be a closed endpoint or an open endpoint) is then used to calculate the perpendicularity coefficient V. er If V er If it is greater than 0.05, then P will be... 4 s Modify it to (i.e., P) 4 s Let the nearest endpoint be the modified line segment l. 4 tem2 Vertical coefficient V er The calculation formula is as follows:

[0095]

[0096] Where i is the distance P 4 s The three nearest endpoints of the line segment are related to l 4 tem The number of approximately perpendicular numbers, denoted by l i To represent these approximately perpendicular line segments, θ ir It is l 4 tem With l i The angle between them, θ ic It is l 4 tem2 With l i The angle between them. Furthermore, n is similar to i, it represents l. 4 tem The other endpoint is one of the three nearest endpoints to l 4 tem The number of approximately perpendicular line segments, denoted by l. n To represent these line segments, θ nr It is l 4 tem With l n The angle between them, θ nc It is l 4 tem2 With l n The angle between them. If V er If it is greater than 0.05, then l 4 tem Change to l 4 tem2 After all unclosed endpoints have been processed, the segment endpoints near the iterations are removed. Line segments without any endpoints (i.e., unclosed line segments) are considered until all line segments are closed, resulting in L. 5 all . The calculation formula is as follows:

[0097]

[0098] Where l is the length of the line segment being processed.

[0099] (5) Repairing and completing missing line segments based on the principle of symmetry. This invention first repairs missing line segments in the floor using ceiling projection line segments, and then completes the missing line segments on each wall surface using the repaired floor outline. This maximizes the closure and continuity of line segments in each plane. The ceiling lineframe in the line segment... top and floor frame l bot It is divided using the best-fit plane of the ceiling and floor. This is based on the fact that most ceiling and floor segments lie on the same plane and are symmetrical. As shown in Table 2, the present invention finds l top Middle and l bot The line segment l whose median lies in the same plane and is symmetrical in direction (i.e., approximately parallel in the z-axis direction). sy l sy The z-values ​​of the endpoints are modified to the average z-values ​​of the endpoints of the floor segment. If l bot The endpoints of the line segment in l sy Adjacent or in l sy Above, then l bot Replace the line segment in the text with l sy Then, the remaining ceiling segment `lr` is used to repair the entire missing segment in the floor. The z-value of `lr` is also modified to the average z-value of the floor wireframe endpoints. If `l` bot If there exists an endpoint adjacent to the endpoint of lr, and lr is a boundary segment in the floor, then lr is considered the entire missing segment of the floor. It will be added to l. bot The formula for determining the floor boundary segment is as follows:

[0100] max{d pro1 ,d pro2}>d 12 (13)

[0101] Where d pro1 and d pro2 d is the distance from one end of the projected line segment to any two parallel lines. 12 It is the distance between these two line segments.

[0102] Table 2 Floor Repair

[0103]

[0104]

[0105] For missing line segments in the wall (such as...) Figure 4 As shown in the figure, this invention optimizes the contour lines on each plane based on the principle of symmetry, ensuring that the contour lines of each plane are closed and continuous to the greatest extent. First, the line segments are divided according to the best-fit plane. The set of unclosed points on the plane is determined by calculating the distance from the endpoint to the line segment. The repair algorithm is shown in Table 3. If there is only one unclosed point p1, then p1 is adjusted to the perpendicular projection point of the nearest line segment ln. Then, the endpoint of ln is modified to the projection point of the line segment where p1 is located. If there are two or more unclosed points, then the two closest points are considered as a group. If the line segments where the two points are located are parallel and the plane is a perpendicular plane (specifically, a plane perpendicular to the XOY plane), then a line segment that conforms to MWA is constructed, that is, the point of the longer line segment is projected onto the point of the shorter line segment to construct a line segment, and then the point of the shorter line segment is modified to the perpendicular projection point of the newly constructed line segment. If the line segments where the two points are located are perpendicular, then the endpoint is modified to the projection point of the line segment where the other point is located, without destroying the original topology. Finally, we remove the line segments with unclosed endpoints to obtain the final result.

[0106] Table 3 Repair of Missing Line Segments in Plane

[0107]

[0108]

[0109] Table 4 Constructs Manhattan Line Segments

[0110]

[0111] Step S107: Place L 5 all Repeat step S106 as input to obtain L. 6 all , will L 6 all The final result is obtained by removing unclosed line segments, such as... Figure 2 As shown in (c).

[0112] In summary, this invention optimizes the extracted interior structural outlines, utilizing Gestalt rules and the implicit architectural structural information within the outlines to repair topological errors, local missing values ​​caused by occlusion, outliers due to surface holes, and missing structural segments. Experiments demonstrate that this invention significantly improves the completeness of the results, achieving both precision and recall rates of 92%, providing stronger support for subsequent vectorized reconstruction.

[0113] The above description is only a preferred embodiment of the present invention. It should be noted that those skilled in the art can make several improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for optimizing the outline of indoor structural point clouds based on Gestalt theory, characterized in that, include: Step 1: Acquire indoor point cloud data and rotate the indoor point cloud data to align with the coordinate system; Step 2: Segment the indoor structural planar point cloud data, and then extract the contour lines based on the planar data; Step 3: Based on the simplicity principle in Gestalt gestalt rules, remove outlier segments caused by occlusion in each plane; Step 4: Based on the similarity and proximity principles in Gestalt gestalt rules, repeatedly extracted line segments are merged, and then the contour lines are regularized. Step 5: Correct topologically incorrect line segments in the contour line based on the continuity and closure principles in Gestalt gestalt rules; Step 6: Based on the symmetry principle in Gestalt gestalt rules, repair and complete the partially and completely missing line segments in the outline. Repeat steps 4-5 to obtain the final indoor structure point cloud outline. Step 3 includes: The set of contour lines L of the plane 1 all The longest line segment is considered the main direction line segment, and L is... 1 all Line segments whose angle with the main direction line segment is less than the angle threshold ε1 are considered structural line segments. These structural line segments are then obtained. The remaining longest line segment is used again as the main direction line segment to find structural line segments. Line segments that are not in the main direction but are longer than the average length are also considered structural line segments. The remaining segments are considered outliers. After removing outliers, a new line segment set L is obtained. 2 all ; Step 4 includes: From L 2 all Starting with the longest line segment, and combining the angles, distances, and spatial relationships between line segments, in set L... 2 all The search term is L, which represents the line segment that can be merged with the given line segment. 3 all The conditions for line segment fusion include: the angle between line segments is less than ε1, and the distance between line segments is less than the adaptive distance threshold. The projection of at least one endpoint of a line segment lies within another line segment; Step 6 includes: First, repair the missing lines in the floor using ceiling projection lines: Find the ceiling wireframe top Center and floor frame l bot The median segment is a line segment l that lies in the same plane and is symmetrical in direction. sy , will l sy The z-values ​​of the endpoints are modified to the average z-values ​​of the endpoints of the floor segment, if l bot The endpoints of the line segment in l sy Adjacent or in l sy Above, then l bot Replace the line segment in the text with l sy Then, use the remaining ceiling segment lr to repair the entire missing segment in the floor, modifying the z-value of lr to the average z-value of the floor wireframe endpoints. bot If there exists an endpoint adjacent to the endpoint of lr, and lr is a boundary segment in the floor, then lr is considered the entire missing segment of the floor, and lr is added to l. bot middle; Then, using the repaired floor outline, the missing lines on each wall surface are filled in: If there is only one unclosed point p1, then p1 is adjusted to be the perpendicular projection point of the nearest line segment ln, and then the endpoint of ln is modified to be the projection point of the line segment containing p1. If there are two or more unclosed points, then the two closest points are considered as a group. If the line segments containing the two points are parallel and the plane is perpendicular, a line segment is constructed by projecting the point of the longer line segment onto the point of the shorter line segment, and then the point of the shorter line segment is modified to be the perpendicular projection point of the newly constructed line segment. If the line segments containing the two points are perpendicular, then the endpoint of the line segment containing one point in the group is modified to be the projection point of the line segment containing the other point, without destroying the original topology. Finally, the line segments with unclosed endpoints are removed to obtain the final result.

2. The method for optimizing the outline of indoor structural point clouds based on Gestalt theory according to claim 1, characterized in that, The distance between the line segments is calculated according to the following formula: Where d s ,d m ,d e These are the distances from the start point, midpoint, and end point of the shorter line segment to the longer line segment, respectively; The distance d from a point to a line is calculated using the following formula: in It is the distance from the point to the starting point of the line segment. It is the distance from the point to the end of the line segment. It is the length of the line segment.

3. The method for optimizing the outline of indoor structural point clouds based on Gestalt theory according to claim 1, characterized in that, The adaptive distance threshold Calculate according to the following formula: Among them l a l1 is the average length of all line segments, and l2 is the length of the two line segments being processed.

4. The method for optimizing the outline of indoor structural point clouds based on Gestalt theory according to claim 1, characterized in that, The methods for merging line segments include: The contour lines of the planar edges are processed as follows: First, find the fusion reference point (x) of L1 and L2. b , y b , z b Divide L1 and L2 into two groups: the two closest endpoints are on different line segments, and the remaining two endpoints form another group. Find the endpoints in each group that are closest to x. b y b and z b The x, y, and z values ​​with the largest differences are used as the starting or ending points of the merged line segment; the merge reference point (x b , y b , z b The calculation formula is as follows: Where (x) min , y min , z min ) and (x max ,y max ,z max ) are L 2 all Minimum and maximum coordinates of the endpoints of the line segment, (x A ,y A ,z A ) and (x B ,y B ,z B () represents the coordinates of the two points that are furthest apart from the four endpoints of the line segment being processed; To handle cases where an in-plane contour line is extracted into two line segments due to occlusion, the following method is used: From the four endpoints of the two line segments and the two perpendicular projection points of the shorter line segment onto the longer line segment, select the two points that are furthest apart as the start and end points of the new line segment after fusion.

5. The method for optimizing the outline of indoor structural point clouds based on Gestalt theory according to claim 1, characterized in that, Step 4, the regularization process for the contour lines includes: Determine distance L 3 all The longest line segment l 3 1st The line segment closest to the two endpoints, and according to l 3 1st The angle between the nearest line segment and l 3 1st Perform parallel or orthogonal adjustments, and then adjust other line segments sequentially using the adjusted line segment as a reference until L is reached. 3 all All line segments in the middle were processed, and the adjusted line segments were placed into L. 4 all .

6. The method for optimizing the outline of indoor structural point clouds based on Gestalt theory according to claim 5, characterized in that, In step 5, for discontinuous line segments, first start from L 4 all Search for the shortest line segment l 4 1st The line segment closest to the endpoint is used as the baseline, and then l is calculated. 4 1st The perpendicular projection point of the endpoint and the adjacent line segment, the adjusted line segment l n Follow the formula below: Where ang(l, l) n () is used to calculate the angle of the line segment before and after adjustment, l nl For the adjusted line segment length, l l The length of the original line segment; For unclosed line segments, the processing order is determined by the length of the unclosed line segment. First, find L. 4 all The shortest unclosed line segment l 4 tem An unclosed endpoint P 4 s Then find the distance P 4 s The line segment containing the three nearest endpoints and the distance l 4 tem The line segment containing the three nearest endpoints to the other endpoint is used to calculate the perpendicularity coefficient V. er If V er If it is greater than 0.05, then P will be... 4 s Change it to its nearest endpoint; the modified line segment is now l. 4 tem2 After all unclosed endpoints have been processed, the endpoints of the distance segment are iteratively removed. The unclosed line segment within is obtained as L. 5 all The vertical coefficient V is calculated according to the following formula. er : Where i is the distance P 4 s The three nearest endpoints of the line segment are related to l 4 tem The number of approximately perpendicular ones, l i θ represents the approximately perpendicular line segment mentioned above. ir It is l 4 tem With l i The angle between them, θ ic It is l 4 tem2 With l i The angle between them, where n represents the distance l 4 tem The other endpoint is one of the three nearest endpoints to l 4 tem The number of approximately perpendicular line segments, l n Representing the above line segment, θ nr It is l 4 tem With l n The angle between them, θ nc It is l 4 tem2 With l n The angle between them.

7. The method for optimizing the outline of indoor structural point clouds based on Gestalt theory according to claim 6, characterized in that, Determined according to the following formula : in is the distance threshold, and l is the length of the line segment being processed.

Images