A building plan wall line recognition method based on faint line recall

Abstract

This invention provides a method for identifying wall lines in architectural floor plans based on faint line recall. Taking the basic wall line identification results as input, it performs supplementary extraction and incorporation correction on locally missed wall lines to obtain a more complete final wall line result. The invention includes the following steps: S1, acquiring input data; S2, forming and standardizing basic wall line results; S3, based on the wall line representation, determining whether there is a significant risk of underrecall in the current basic wall line results and locating suspected gaps; S4, generating local candidate regions; S5, performing image enhancement and faint line candidate mask extraction within the local candidate regions; S6, generating faint line candidate segments through the extracted faint line candidate masks and performing initial screening; S7, performing constraint determination on the faint line candidate segments; S8, performing sorting, elimination, and incorporation processing on valid faint line candidate segments; S9, updating the wall line results and determining whether there are still gaps requiring recall; S10, outputting the final wall line result.

Description

Technical Field

[0001] This invention relates to a method for identifying wall lines in architectural floor plans based on faint line recall. Background Technology

[0002] Regarding wall / wall line recognition in architectural floor plans, currently published patents mainly focus on four approaches: one is to directly identify the wall outline or wall thickness based on structured line segment data such as CAD / DXF; another is to perform threshold segmentation, morphological processing, edge detection, and line detection on full-page floor plans, floor plans, or PDF drawings to directly identify walls from the original image; another is to introduce deep learning models for direct wall detection, or to cross-validate model results with rule results; and the third is to gradually infer walls based on user interaction or specific component relationships.

[0003] In design, construction, and existing drawing management scenarios, problems often arise such as large CAD drawing file sizes, cluttered layers, loss of original files, and the difficulty in directly utilizing vector information from historical scans, screenshots, or exported bitmaps. In raster images of drawings, wall lines typically appear as pixel bands or grayscale bands with a certain length, thickness, and directional continuity. Existing pixel-based wall line recognition methods usually begin by converting the original image to grayscale, extracting directional structures, analyzing connected components, and geometrically aggregating the data to extract basic wall lines from the pixel-level structure. However, during the wall line recognition process in architectural floor plans, the basic wall line results often miss some true wall lines due to low contrast, light gray color, local breaks, scan compression, export distortion, or conservative selection in the preceding recognition stages. This leads to problems such as an insufficient number of wall lines, broken long walls, incomplete corner closures, and insufficient local connectivity.

[0004] Chinese patent CN103971098A discloses a method for recognizing the planar contour of a wall, comprising: reading all line segment data in a given building plan and storing these line segments in a line segment array V1; traversing all line segments in V1, marking two distinct line segments that meet the conditions for candidate contour edges, setting them as candidate contour edges, and storing these two line segments as a candidate contour in a candidate contour array Vcc; calculating the contour centerline of each candidate contour in Vcc, and storing these centerlines in an array Vml; traversing Vml, optimizing any two distinct centerlines mli and mlj (i≠j); and reconstructing the planar contour of the wall based on the optimized centerlines and their corresponding widths. Although this patent solution processes raster image input, it belongs to the category of whole... Figure 1 One-time wall recognition process.

[0005] Chinese patent CN109753918A discloses a method and system for identifying walls in floor plans. The components include load-bearing walls, non-load-bearing walls, and bay windows. The method first binarizes the input image, then designs rules to detect the positions of specific components within the image. While this patented solution features a two-stage process of "identifying one part first, then completing the other," its completion target is the non-load-bearing wall area in the binary image. Summary of the Invention

[0006] The purpose of this invention is to provide a method for identifying wall lines in architectural floor plans based on faint line recall. The method takes the basic wall line identification results as input and performs supplementary extraction and incorporation correction on the locally missed wall lines to obtain a more complete final wall line result.

[0007] To achieve the above objectives, the present invention provides a method for identifying wall lines in architectural floor plans based on faint line recall, comprising the following steps: Step S1: Obtain input data, which includes the original raster image of the building floor plan and the results of the foundation wall lines; Step S2: Form and standardize the basic wall line results, unifying the different forms of basic wall line results into a wall line representation; Step S3: Based on the wall line representation, determine whether there is a significant risk of underrecall in the current basic wall line results and locate the suspected gap; if it is determined that there is no significant risk of underrecall, proceed to step S10; if it is determined that there is a significant risk of underrecall, proceed to step S4. Step S4: Generate local candidate regions; Step S5: Perform image enhancement and light line candidate mask extraction within the local candidate region; Step S6: Generate candidate line segments for light lines using the extracted light line candidate mask and perform initial screening; Step S7: Perform constraint judgment on the candidate light line segment. If the judgment condition is met, the candidate light line segment is accepted and the process proceeds to step S8. If the judgment condition is not met, the candidate light line segment that does not meet the constraint is deleted and not included in the final wall line result. Step S8: Perform sorting, elimination, and merging processes on the valid light line candidate segments; Step S9: Based on the valid candidate light line segments processed in step S8, update the wall line results and determine whether there are still any missing segments that need to be recalled; if there are still any missing segments that need to be recalled, return to step S4; if there are no missing segments that need to be recalled, end the iteration and proceed to step S10. Step S10: Output the final wall line result.

[0008] Step S2 of the present invention specifically includes the following steps: The results of different forms of basic wall lines are uniformly converted into a computable wall line representation, and the main direction, endpoint position, length, thickness estimate, center line or boundary position, adjacency relationship and region of each wall line are unified; for results given in the form of masks or wall blocks, their central axis or boundary is extracted first, and then a unified line segment expression is formed.

[0009] Step S3 of the present invention specifically includes the following steps: Based on the unified wall line representation formed in step S2, the endpoints, continuity, and adjacency relationships of the basic wall lines are checked to determine whether there is an obvious risk of underrecall in the current basic wall line results. Suspected gaps are located based on wall line interruptions, exposed endpoints, unclosed endpoints, or abnormal blanks between wall lines in the same direction. When it is determined that there is no risk of underrecall or the risk of underrecall is low, local faint line recall is no longer performed, and the current wall line result is output in step S10. When it is determined that there is a risk of underrecall, the process proceeds to step S4 to generate local candidate regions.

[0010] Step S4 of the present invention specifically includes the following steps: For each suspected gap candidate obtained in step S3, instead of searching again on the entire drawing, a local candidate area is generated around the exposed endpoint, fracture boundary, the extension direction of adjacent wall segments, or the blank zone between adjacent wall lines in the same direction.

[0011] Step S5 of the present invention specifically includes the following steps: Step S5-1: Within the local candidate region obtained in step S4, perform grayscale conversion, local contrast enhancement, and weak noise suppression on the original image; Step S5-2: Form a candidate mask using relatively relaxed extraction conditions for candidates with weak grayscale, while retaining a weaker grayscale structure; Step S5-3: Output local light line candidate masks for use in subsequent step S6.

[0012] Step S6 of the present invention specifically includes the following steps: Step S6-1: Perform directional connectivity and connected component extraction on the local faint line candidate mask obtained in step S5, and retain only the linear structure that is consistent with the direction of the suspected gap or the direction of the adjacent foundation wall line. Step S6-2: Fit line segments to the retained linear structure and merge multiple candidate segments that can be continuously spliced ​​in the same direction into candidate line segments; Step S6-3: Perform preliminary screening based on the minimum length, thickness range, and length-to-thickness ratio of candidate line segments, and remove candidate line segments that are too short, have abnormal thickness, or do not have obvious linear characteristics. Step S6-4: Output the candidate light line segments that have passed the initial screening for constraint determination in step S7.

[0013] Step S7 of the present invention specifically includes the following steps: Step S7-1: Perform core constraint checks on the candidate line segments of the light line; core constraints include direction consistency checks, endpoint anchoring checks, and exclusion zone conflict checks.

[0014] Step S7-2: While performing step S7-1, perform supplementary constraint checks on the candidate line segments of the light line; the supplementary constraints include thickness compatibility checks, crossing distance checks, and connectivity or closure improvement checks.

[0015] Step S7-3: Determine whether the candidate line segment of the light line satisfies the core constraints and at least one of the supplementary constraints.

[0016] Step S7-4: If the judgment condition of step S7-3 is met, the candidate line segment of the light line is accepted, and the process proceeds to step S8 for sorting and merging. Step S7-5: If the judgment condition of step S7-3 is not met, delete the candidate line segments of the light line that do not meet the constraints and do not merge them into the final wall line result.

[0017] Step S8 of the present invention specifically includes the following steps: Step S8-1: Perform deduplication and overlap suppression on the valid light line candidate segments, and remove candidates that cross doorways or fall into text, size, or furniture exclusion areas; Step S8-2: Determine whether the light-colored candidate line segments processed in step S8-1 are still valid candidates; Step S8-3: If the light line candidate segment is no longer a valid candidate, then classify it as an incorrect candidate; Step S8-4: If the light line candidate segments are still valid candidates, sort them according to anchoring strength, direction matching, thickness compatibility and grayscale support. Step S8-5: According to the sorting results of step S8-4, merge the candidate light line segments into the nearest wall group and perform gap filling, extension or bridging. Step S8-6: Determine whether merging the candidate line segment of the light line causes abnormal backtracking, incorrect closure, or cross-regional misconnection; Step S8-7: If merging a candidate line segment causes an abnormal return, incorrect closure, or incorrect connection across regions, then classify it as an incorrect candidate and cancel the merging; if no such abnormality is caused, proceed to step S9.

[0018] Step S9 of the present invention specifically includes the following steps: Step S9-1: Based on the candidate merging results retained in step S8, update the endpoints, lengths, thicknesses, connectivity relationships, and source markers of the corresponding wall lines or wall groups; Step S9-2: Based on the updated wall line results, re-determine whether a new shortage has been formed and whether the shortage meets the conditions for local recall again; if there is still a shortage that meets the local recall conditions, return to step S4, regenerate the local candidate region around the shortage and perform the next round of light line recall; if there is no shortage that meets the local recall conditions, end the iteration and proceed to step S10.

[0019] Step S10 of the present invention specifically includes the following steps: outputting the final set of wall segments or wall network results after light line recall correction, and retaining candidate source markers to distinguish the basic identification results from the subsequent light line replenishment results.

[0020] Compared with the prior art, the advantages and positive effects of the present invention are as follows: (1) This invention mainly targets rasterized and pixelated inputs such as PNG, JPG, JPEG, PDF bitmap pages exported from CAD and scanned drawings to carry out wall line recognition and partial omission recognition and repair, without relying on native CAD vector line segment data.

[0021] (2) After the present invention is integrated into the existing wall line recognition results as an independent post-processing enhancement link, it has stronger adaptability and is more conducive to combined deployment with the existing upstream recognition system.

[0022] (3) The present invention only triggers faint line recall in the suspected missing area of ​​the basic wall line, rather than blindly performing weak line detection on the entire page of drawings. Therefore, it is easier to suppress the interference of non-wall fine lines such as drawing frame, title block, dimension line, furniture line and instruction line.

[0023] (4) The present invention uses a combination of constraints such as direction consistency, thickness compatibility, endpoint anchoring, spanning distance, exclusion zone conflict and connectivity topology improvement to conduct controlled screening of light line candidates, so that "recalling more candidates" and "keeping the results credible" can be taken into account at the same time. Attached Figure Description

[0024] Figure 1 This is a general flowchart of an embodiment of the present invention.

[0025] Figure 2 This is a flowchart of the light line candidate extraction process according to an embodiment of the present invention.

[0026] Figure 3 This is a flowchart illustrating the candidate constraint determination process for light lines according to an embodiment of the present invention.

[0027] Figure 4 This is a flowchart illustrating the candidate elimination and merging process in an embodiment of the present invention.

[0028] Figure 5This is a schematic diagram comparing the partial view of the basic wall line, the partial candidate area, and the light line restoration result of a drawing according to an embodiment of the present invention; in the figure, (a) is a partial view of the basic wall line, (b) is a partial candidate area view, and (c) is a light line restoration result view.

[0029] Figure 6 This is a schematic diagram comparing the result of the basic wall line in a drawing and the result after the light line is filled in, according to an embodiment of the present invention; in the figure, (a) is the result of the basic wall line and (b) is the result after the light line is filled in.

[0030] Figure 7 This is a schematic diagram comparing the result of the basic wall line and the result after the light line is filled in, which is another drawing of the embodiment of the present invention; in the figure, (a) is the result of the basic wall line and (b) is the result after the light line is filled in. Detailed Implementation

[0031] The technical solution described in this invention will be explained below with reference to the accompanying drawings.

[0032] A method for identifying wall lines in architectural floor plans based on faint line recall includes the following steps: Step S1: Obtain input data, which includes the original raster image of the building floor plan and the results of the foundation wall lines.

[0033] The original raster image can be a PNG, JPG, JPEG, PDF bitmap page or scanned image exported from CAD; the basic wall line result is preferably a set of wall line segments, which can be converted from any upstream wall line recognition module, manual correction module or existing recognition result; when the input result is a wall group, wall block, center line, parallel line pair or raster wall mask, it can be converted into a wall line segment representation before being input into this method.

[0034] It should be noted that the basic wall line results can be generated by any architectural plan wall line recognition method. This method does not limit the specific generation algorithm, model type, rule process, or data source. Regardless of whether the basic wall line results come from whole-map wall line detection, local wall extraction, mask segmentation, line segment aggregation, manual correction results, or a combination thereof, they can all be used as the pre-input for this method to perform local light line completion. Optionally, auxiliary constraint information such as the main plane range, manually confirmed range, drawing frame area, text area, dimension annotation area, table area, or furniture area, provided by the upstream module or preprocessed by this method, can also be obtained. When the above auxiliary information is missing, subsequent processes can still be executed solely based on the wall line geometric relationship and grayscale features.

[0035] Step S2: Generate and standardize the basic wall line results, converting different forms of basic wall line results into a unified, calculable wall line representation. This specifically includes the following steps: Different forms of basic wall line results are uniformly converted into a computable wall line representation, standardizing the main direction, endpoint positions, length, thickness estimation, centerline or boundary position, adjacency relationship, and region to which each wall line belongs. For results given in the form of masks or wall blocks, their central axis or boundary is extracted first, and then a unified line segment representation is formed.

[0036] This step enables subsequent gap identification, candidate extraction, constraint determination, and merging updates to be performed based on a unified and standardized structure.

[0037] Step S3: Based on the unified wall line representation formed in Step S2, determine whether there is a significant risk of underrecall in the current basic wall line results and locate potential gaps. This specifically includes the following steps: Based on the unified wall line representation formed in step S2, the endpoints, continuity, and adjacency relationships of the basic wall lines are checked to determine whether there is an obvious risk of underrecall in the current basic wall line results. Suspected gaps are located based on wall line interruptions, exposed endpoints, unclosed endpoints, or abnormal blanks between wall lines in the same direction. When it is determined that there is no risk of underrecall or the risk of underrecall is low, local faint line recall is no longer performed, and the current wall line result is output in step S10. When it is determined that there is a risk of underrecall, the process proceeds to step S4 to generate local candidate regions.

[0038] Step S4: Generate local candidate regions. This specifically includes the following steps: For each potential gap candidate obtained in step S3, instead of searching the entire drawing again, a local candidate region is generated around the exposed endpoint, fracture boundary, the extension direction of adjacent wall segments, or the blank band between adjacent wall lines in the same direction. The candidate region is preferably a rectangular region, a strip-shaped region, or a search region formed by combining multiple local strip-shaped regions. The width of the candidate region can be set to 0.5 to 3 times the estimated thickness of the adjacent wall, preferably 1 to 2 times; the length of the candidate region along the extension direction of the gap can be set to 1 to 8 times the wall thickness, or the upper limit of pixels converted according to the drawing resolution. If there is auxiliary exclusion area information such as the drawing frame area, title block, text description, dimension annotation, dense furniture area, and table area, areas that overlap too much with them should be eliminated or downgraded; if there is no corresponding auxiliary information, this constraint is omitted.

[0039] Step S5: Perform image enhancement and light line candidate mask extraction within the local candidate region. This specifically includes the following steps: Step S5-1: Only within the local candidate region obtained in step S4, perform grayscale conversion, local contrast enhancement, and weak noise suppression on the original image to enhance the wall line response of light gray, low contrast, or scan distortion.

[0040] Step S5-2: Based on step S5-1, a candidate mask is formed using a more relaxed extraction condition for candidates with light lines, while retaining a weaker grayscale structure.

[0041] Step S5-3: Output local light line candidate masks for use in subsequent step S6.

[0042] The conditions for extracting faint line candidates can be set independently of the basic wall line result generation stage, and can be adjusted according to the image grayscale distribution, the relative brightness relationship between the wall line and the background, and the local contrast. Specifically, a candidate mask can be formed using a faint line extraction threshold T.

[0043] For an 8-bit grayscale image, in an embodiment where the wall lines appear relatively dark and the background relatively bright after uniform polarity processing, a light line threshold T can be used to extract light line candidates. The light line threshold T is preferably between 225 and 245, more preferably 235. Pixels with grayscale values ​​not greater than the light line threshold T are identified as light line candidate pixels, and a candidate mask is formed from these light line candidate pixels. If the grayscale polarity of the wall lines and the background is opposite, grayscale inversion processing can be performed first, or an equivalent threshold range adapted to the grayscale polarity can be used. The above threshold ranges and extraction methods are preferred embodiments and do not constitute a limitation on the scope of protection of this invention. The output of this step is a local light line candidate mask or a set of weakly linear structures.

[0044] Step S6: Generate candidate line segments and perform initial screening. This specifically includes the following steps: Step S6-1: Perform directional connectivity and connected domain extraction on the local faint line candidate mask obtained in step S5, and retain only the linear structure that is consistent with the direction of the suspected gap or the direction of the adjacent foundation wall line.

[0045] Step S6-2: Perform line segment fitting on the linear structure retained in step S6-1, and merge multiple candidate segments that can be continuously spliced ​​in the same direction into computable candidate line segments or candidate groups.

[0046] Step S6-3: Perform preliminary screening based on the minimum length, thickness range, and length-to-thickness ratio of the candidate structures, and eliminate candidate structures that are too short, have abnormal thickness, or do not have obvious linear characteristics.

[0047] Step S6-4: Output the candidate light line segments or candidate groups that have passed the initial screening for constraint determination in step S7.

[0048] Step S7: Perform constraint judgment on the candidate light-colored line segments. If the judgment condition is met, accept the candidate light-colored line segment and proceed to step S8. If the judgment condition is not met, delete the candidate light-colored line segments that do not meet the constraints and do not incorporate them into the final wall line result. Specifically, this includes the following steps: Step S7-1: Perform core constraint checks on the candidate line segments for the light-colored lines. Core constraints include direction consistency checks, endpoint anchorage checks, and exclusion zone conflict checks.

[0049] Step S7-2: While performing S7-1, perform supplementary constraint checks on the candidate light-colored line segments. Supplementary constraints include thickness compatibility checks, spanning distance checks, and connectivity or closure improvement checks.

[0050] Step S7-3: Determine whether the candidate line segment of the light line satisfies the core constraints and at least one of the supplementary constraints.

[0051] Step S7-4: If the judgment condition of step S7-3 is met, the candidate line segment of the light line is accepted, and the process proceeds to step S8 for sorting and merging.

[0052] Step S7-5: If the judgment condition of step S7-3 is not met, delete the candidate line segments of the light line that do not meet the constraints and do not merge them into the final wall line result.

[0053] Step S8: Sort, eliminate, and merge valid light line candidate segments. This specifically includes the following steps: Step S8-1: Perform deduplication and overlap suppression on the valid light line candidate segments, and remove candidates that cross doorways or fall into text, size, or furniture exclusion areas.

[0054] Step S8-2: Determine whether the light-colored candidate line segment processed in step S8-1 is still a valid candidate.

[0055] Step S8-3: If the candidate line segment is no longer a valid candidate, it will be classified as an incorrect candidate and will not be included in this round of merging.

[0056] Step S8-4: If the candidate line segments are still valid candidates, sort them according to anchoring strength, direction matching, thickness compatibility and grayscale support.

[0057] Step S8-5: According to the sorting results of step S8-4, merge the candidate light line segments into the nearest wall group and perform gap filling, extension or bridging.

[0058] Step S8-6: Determine whether merging the candidate line segment of the light line causes abnormal backtracking, incorrect closure, or cross-regional misconnection.

[0059] Step S8-7: If merging a candidate line segment causes an abnormal return, incorrect closure, or incorrect connection across regions, then classify it as an incorrect candidate and cancel the merging; if no such abnormality is caused, proceed to step S9.

[0060] Step S9: Update the wallline results and determine if there are still recallable shortages. This includes the following steps: Step S9-1: Based on the candidate merging results retained in step S8, update the endpoints, lengths, thicknesses, connectivity relationships, and source markers of the corresponding wall lines or wall groups.

[0061] Step S9-2: Based on the updated wall line results, re-determine whether a new shortage has been formed and whether the shortage meets the conditions for local recall again; if there is still a shortage that meets the local recall conditions, return to step S4, regenerate the local candidate region around the shortage and perform the next round of light line recall; if there is no shortage that meets the local recall conditions, end the iteration and proceed to step S10.

[0062] Step S10: Output the final wall line result. This includes the following steps: The system outputs the final set of wall segments or wall network results after light line recall correction, and can retain candidate source markers to distinguish the basic identification results from subsequent light line completion results. When downstream modules need to be adapted, the final wall line results can be further converted into wall group structures, wall boundaries, wall centerlines, or wall mask representations.

[0063] The above description is merely a preferred embodiment of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A method for identifying wall lines in architectural floor plans based on faint line recall, characterized in that: Includes the following steps: Step S1: Obtain input data, which includes the original raster image of the building floor plan and the results of the foundation wall lines; Step S2: Form and standardize the basic wall line results, unifying the different forms of basic wall line results into a wall line representation; Step S3: Based on the wall line representation, determine whether there is a significant risk of underrecall in the current basic wall line results and locate the suspected gap; if it is determined that there is no significant risk of underrecall, proceed to step S10; if it is determined that there is a significant risk of underrecall, proceed to step S4. Step S4: Generate local candidate regions; Step S5: Perform image enhancement and light line candidate mask extraction within the local candidate region; Step S6: Generate candidate line segments for light lines using the extracted light line candidate mask and perform initial screening; Step S7: Perform constraint judgment on the candidate light line segment. If the judgment condition is met, the candidate light line segment is accepted and the process proceeds to step S8. If the judgment condition is not met, the candidate light line segment that does not meet the constraint is deleted and not included in the final wall line result. Step S8: Perform sorting, elimination, and merging processes on the valid light line candidate segments; Step S9: Based on the valid candidate light line segments processed in step S8, update the wall line results and determine whether there are still any missing segments that need to be recalled; if there are still any missing segments that need to be recalled, return to step S4; if there are no missing segments that need to be recalled, end the iteration and proceed to step S10. Step S10: Output the final wall line result.

2. The method for identifying wall lines in architectural floor plans based on faint line recall according to claim 1, characterized in that: Step S2 specifically includes the following steps: The results of different forms of basic wall lines are uniformly converted into a computable wall line representation, and the main direction, endpoint position, length, thickness estimate, center line or boundary position, adjacency relationship and region of each wall line are unified; for results given in the form of masks or wall blocks, their central axis or boundary is extracted first, and then a unified line segment expression is formed.

3. The method for identifying wall lines in architectural floor plans based on faint line recall according to claim 2, characterized in that: Step S3 specifically includes the following steps: Based on the unified wall line representation formed in step S2, the endpoints, continuity, and adjacency relationships of the basic wall lines are checked to determine whether there is an obvious risk of underrecall in the current basic wall line results. Suspected gaps are located based on wall line interruptions, exposed endpoints, unclosed endpoints, or abnormal blanks between wall lines in the same direction. When it is determined that there is no risk of underrecall or the risk of underrecall is low, local faint line recall is no longer performed, and the current wall line result is output in step S10. When it is determined that there is a risk of underrecall, the process proceeds to step S4 to generate local candidate regions.

4. The method for identifying wall lines in architectural floor plans based on faint line recall according to claim 3, characterized in that: Step S4 specifically includes the following steps: For each suspected gap candidate obtained in step S3, instead of searching again on the entire drawing, a local candidate area is generated around the exposed endpoint, fracture boundary, the extension direction of adjacent wall segments, or the blank zone between adjacent wall lines in the same direction.

5. The method for identifying wall lines in architectural floor plans based on faint line recall according to claim 4, characterized in that: Step S5 specifically includes the following steps: Step S5-1: Within the local candidate region obtained in step S4, perform grayscale conversion, local contrast enhancement, and weak noise suppression on the original image; Step S5-2: Use the light line extraction threshold T to perform threshold extraction on the local grayscale image after step S5-1, and determine the pixels that meet the light line extraction threshold T as light line candidate pixels and form a candidate mask. Step S5-3: Output local light line candidate masks for use in subsequent step S6.

6. The method for identifying wall lines in architectural floor plans based on faint line recall according to claim 1 or 5, characterized in that: Step S6 specifically includes the following steps: Step S6-1: Perform directional connectivity and connected component extraction on the local faint line candidate mask obtained in step S5, and retain only the linear structure that is consistent with the direction of the suspected gap or the direction of the adjacent foundation wall line. Step S6-2: Fit line segments to the retained linear structure and merge multiple candidate segments that can be continuously spliced ​​in the same direction into candidate line segments; Step S6-3: Perform preliminary screening based on the minimum length, thickness range, and length-to-thickness ratio of candidate line segments, and remove candidate line segments that are too short, have abnormal thickness, or do not have obvious linear characteristics. Step S6-4: Output the candidate light line segments that have passed the initial screening for constraint determination in step S7.

7. The method for identifying wall lines in architectural floor plans based on faint line recall according to claim 6, characterized in that: Step S7 specifically includes the following steps: Step S7-1: Perform core constraint checks on the candidate line segments of the light line; the core constraints include direction consistency constraints, endpoint anchoring constraints, and exclusion zone conflict constraints. Step S7-2: While performing step S7-1, perform supplementary constraint checks on the candidate light-colored line segments; supplementary constraints include thickness compatibility checks, spanning distance checks, and connectivity or closure improvement checks; Step S7-3: Determine whether the candidate line segment of the light line satisfies the core constraints and at least one of the supplementary constraints; Step S7-4: If the judgment condition of step S7-3 is met, the candidate line segment of the light line is accepted, and the process proceeds to step S8 for sorting and merging. Step S7-5: If the judgment condition of step S7-3 is not met, delete the candidate line segments of the light line that do not meet the constraints and do not merge them into the final wall line result.

8. The method for identifying wall lines in architectural floor plans based on faint line recall according to claim 7, characterized in that: Step S8 specifically includes the following steps: Step S8-1: Perform deduplication and overlap suppression on the valid light line candidate segments, and remove candidates that cross doorways or fall into text, size, or furniture exclusion areas; Step S8-2: Determine whether the light-colored candidate line segments processed in step S8-1 are still valid candidates; Step S8-3: If the light line candidate segment is no longer a valid candidate, then classify it as an incorrect candidate; Step S8-4: If the light line candidate segments are still valid candidates, sort them according to anchoring strength, direction matching, thickness compatibility and grayscale support. Step S8-5: According to the sorting results of step S8-4, merge the candidate light line segments into the nearest wall group and perform gap filling, extension or bridging. Step S8-6: Determine whether merging the candidate line segment of the light line causes abnormal backtracking, incorrect closure, or cross-regional misconnection; Step S8-7: If merging a candidate line segment causes an abnormal return, incorrect closure, or incorrect connection across regions, then classify it as an incorrect candidate and cancel the merging; if no such abnormality is caused, proceed to step S9.

9. The method for identifying wall lines in architectural floor plans based on faint line recall according to claim 8, characterized in that: Step S9 specifically includes the following steps: Step S9-1: Based on the candidate merging results retained in step S8, update the endpoints, lengths, thicknesses, connectivity relationships, and source markers of the corresponding wall lines or wall groups; Step S9-2: Based on the updated wall line results, re-determine whether a new shortage has been formed and whether the shortage meets the conditions for local recall again; if there is still a shortage that meets the local recall conditions, return to step S4, regenerate the local candidate region around the shortage and perform the next round of light line recall; if there is no shortage that meets the local recall conditions, end the iteration and proceed to step S10.

10. The method for identifying wall lines in architectural floor plans based on faint line recall according to claim 9, characterized in that: Step S10 specifically includes the following steps: outputting the final set of wall segments or wall network results after light line recall correction, and retaining candidate source markers to distinguish the basic identification results from the subsequent light line replenishment results.

Images