A building hollowing automatic detection positioning method and system based on UWB positioning

Abstract

This invention discloses an automatic detection and location method and system for building hollow areas based on UWB positioning, comprising: deploying UWB base stations according to the building CAD plan or facade unfolding drawing, and obtaining drawing mapping parameters; acquiring spatial coordinates, vibration amplitude sequence, and audio spectrum to identify valid impact events; constructing a hollow defect level identification model to determine the hollow defect level and defect location results; converting the spatial coordinates of defect points to drawing coordinates, grouping defect points by DBSCAN density clustering to obtain a defect cluster set; calculating the hollow boundary mutation coefficient, extracting boundary control points and generating defect cluster edge contours to obtain the repair area boundary; establishing a structured defect database, and generating a repair location map and a digital inspection report of building hollow areas. The technical method provided by this invention can significantly improve the accuracy of continuous hollow area identification, the accuracy of repair boundary delineation, and the traceability of inspection results, thereby improving the efficiency of building hollow area detection and defect remediation.

Description

Technical Field

[0001] This invention relates to the field of building engineering testing technology, and in particular to an automatic detection and positioning method and system for building hollow areas based on UWB positioning. Background Technology

[0002] Hollow spots in building finishes are a common quality defect in construction projects, widely found in wall finishes, floor finishes, ceiling finishes, and exterior facade finishes. Hollow spots are typically caused by factors such as poor adhesion between the substrate and the finishing material, improper construction techniques, material shrinkage and deformation, or long-term environmental effects. When hollow spots develop to a certain extent, they can easily lead to cracking and peeling of the finish, affecting not only the building's quality and lifespan but also potentially posing safety hazards. Therefore, hollow spot detection is of great importance in the quality acceptance of construction projects, maintenance of existing buildings, and safety inspections. Currently, hollow spot detection mainly uses manual tapping. Inspectors tap the finish layer point by point with a hollow spot hammer or testing rod, judging the location and degree of hollow spots based on changes in the sound. While this method is simple to operate, it is highly dependent on the experience and subjective judgment of the inspectors, leading to discrepancies in results between different personnel. Furthermore, existing testing processes typically use chalk marking, photographic recording, or manual recording to preserve results, which suffers from low positioning accuracy, incomplete information, and difficulties in later retrieval. With the development of building information technology (BIM), some inspection systems have begun to record inspection results electronically. However, most existing solutions can only record the approximate location of defects, making it difficult to achieve a precise correspondence between the inspection location and architectural drawings. In complex scenarios such as building facades and high-rise walls, accurate location and reproduction of defects are even more difficult, hindering subsequent repair work. Furthermore, actual hollow defects usually exist in continuous areas, while existing technologies often present inspection results as discrete defect points. During continuous tapping, factors such as human tapping position deviation, positioning errors, and gradual changes in sound at the hollow boundary can cause the same hollow area to be recorded as multiple adjacent defect points, making it difficult to accurately determine the defect area boundary and leading to inaccurate delineation of the repair area. Simultaneously, existing inspection results lack a unified data organization and correlation management mechanism, making it difficult to achieve unified management of inspection data, defect levels, location information, audio recordings, and component information, which is detrimental to subsequent re-inspection, maintenance tracking, and operation and maintenance management. Therefore, how to achieve automatic detection, precise location, automatic identification of continuous defect areas, and digital management of inspection results for building hollow defects has become an urgent technical problem to be solved in the field of building quality inspection. Summary of the Invention

[0003] To address the problems existing in the prior art, this invention provides an automatic detection and positioning method and system for building hollow areas based on UWB positioning.

[0004] The first aspect of this invention provides an automatic detection and location method for building hollow areas based on UWB positioning, mainly comprising: Based on the architectural CAD floor plan or exterior facade unfolding, extract the boundary contour and spatial range data of the finishing area, deploy UWB base stations, establish a unified spatial coordinate system, and obtain the drawing mapping parameters; By periodically locating signals from the UWB tag on the handle of the hollow drum hammer, collecting data from the vibration sensor and microphone, spatial coordinates, vibration amplitude sequences and audio spectra are obtained, valid knocking events are identified, and knocking event data packets are acquired. Based on the tapping event data packet, audio segments are extracted and frequency processing and frame decomposition are performed to obtain multi-dimensional sound feature vectors. A hollow drum defect level recognition model is constructed to determine the hollow drum defect level and defect location result at the detection location. Based on the drawing mapping parameters and defect location results, the spatial coordinates of the defect points are converted into drawing coordinates and matched with component information. Defect points are grouped by DBSCAN density clustering to obtain a set of defect clusters. Based on the defect cluster set, information on defect points and adjacent points within the defect cluster is obtained, the abrupt change coefficient of the hollow boundary is calculated, boundary control points are extracted and the edge contour of the defect cluster is generated, and the boundary of the rework area is obtained. Based on the defect database file, a structured defect database is established, and a repair location map and a digital inspection report of building hollow areas are generated.

[0005] Furthermore, the step of extracting the boundary contour and spatial range data of the finishing area based on the architectural CAD plan or facade unfolding drawing, deploying UWB base stations, establishing a unified spatial coordinate system, and obtaining drawing mapping parameters includes: Based on the CAD plan or exterior unfolded drawing of the building to be inspected, the boundary contour information of the building's surface area is extracted, and the spatial range data of the surface area is obtained. The surface area includes wall surface layer, floor surface layer, ceiling surface layer, or building exterior surface layer. Four UWB base stations are installed in a rectangular or vertically aligned arrangement, and the three-dimensional installation coordinates of each base station are obtained. A unified time reference is obtained by sending periodic synchronization pulse signals through the base station main control module. The distance matrix between base stations is obtained based on the TOF ranging data between the base stations, and the base station coordinates are corrected by the least squares positioning algorithm to obtain a unified spatial coordinate system. The drawing coordinate data is obtained based on the imported CAD plan or exterior unfolded drawing, and the actual spatial coordinates of a preset number of reference points are obtained using on-site acquisition equipment. The correspondence between the drawing coordinates and the spatial coordinates is calculated through an affine transformation matrix to obtain the drawing mapping parameters. Based on the drawing mapping parameters, the spatial coordinates of the hollow hammer are converted into drawing coordinates.

[0006] Furthermore, the process involves acquiring spatial coordinates, vibration amplitude sequences, and audio spectra by periodically positioning signals from the UWB tag on the hollow drum hammer handle, collecting data from a vibration sensor and microphone, identifying valid striking events, and obtaining striking event data packets, including: A UWB tag installed on the handle of the hollow drum hammer periodically transmits positioning signals to obtain timestamp data received by each base station. The spatial coordinates of the hollow drum hammer are calculated using the TDOA positioning algorithm. Triaxial acceleration data is obtained from the internal vibration sensor of the hollow drum hammer, and peak detection is performed on the acceleration data using a sliding time window to obtain a vibration amplitude sequence. A sound energy sequence is obtained from the raw audio signal collected by the microphone, and the audio spectrum is calculated using short-time Fourier transform to obtain the frequency energy distribution result. Valid striking events are determined by synchronously comparing the vibration amplitude sequence and frequency energy distribution result with preset vibration thresholds and preset short-time energy thresholds. The corresponding timestamp is obtained for each valid striking event, and the current spatial coordinates and audio buffer segments are saved using a unified time reference to obtain a striking event data packet. Based on the striking event data packet, audio segments are extracted and subjected to frequency processing and frame decomposition to obtain multi-dimensional sound feature vectors. A hollow drum defect level identification model is constructed to determine the hollow drum defect level and defect location result at the detection location.

[0007] Further, the step of extracting audio segments based on the tapping event data packet and performing frequency processing and frame decomposition to obtain multi-dimensional sound feature vectors, constructing a hollow drum defect level recognition model, and determining the hollow drum defect level and defect location result at the detection location includes: The original audio segments are obtained from the tapping event data packets. A bandpass filter is used to filter the audio segments by frequency, obtaining valid sound data within a preset frequency range. Audio frame sequences are calculated through pre-emphasis processing and Hamming window framing to obtain standardized analysis data. Based on the standardized analysis data, short-time energy, zero-crossing rate, spectral centroid, spectral peak value, Mel frequency cepstral coefficients, and envelope decay time parameters are obtained. A multi-dimensional sound feature vector is constructed using feature splicing and stored in the tapping sound feature database. Historical data of the multi-dimensional sound feature vectors and corresponding hollow defect level labels are obtained from the tapping sound feature database. A random forest algorithm is used to train the model, constructing a hollow defect level recognition model. Hollow defect levels include no hollow, slight hollow, moderate hollow, and severe hollow. Based on the real-time acquired multi-dimensional sound feature vector data of the detection location, the hollow defect level recognition model is used to determine the hollow defect level and corresponding timestamp of the detection location. A timestamp matching method is used to associate spatial coordinates with the recognition results to obtain the defect location result.

[0008] Furthermore, based on the drawing mapping parameters and defect location results, the spatial coordinates of the defect points are converted into drawing coordinates and matched with component information. Defect points are then grouped using DBSCAN density clustering to obtain a defect cluster set, including: Based on the drawing mapping parameters and defect location results, the spatial coordinates of defect points in the defect location results are converted into drawing coordinates using an affine transformation matrix to obtain the drawing defect coordinates. Based on the drawing defect coordinates, the corresponding drawing area is matched using the CAD layer index, or the component to which the defect point belongs is determined by the component's geometric boundary in the BIM model, and the corresponding BIM component ID is associated to obtain the defect component information. Based on the defect component information and defect level data, a drawing annotation object is generated using color coding, and a defect mark is drawn at the corresponding coordinate position on the drawing. The associated audio path, inspector number, inspection time, defect point spatial coordinates, and drawing annotation object are encapsulated into a JSON format inspection record and stored in a defect database file. Based on the defect database file, the spatial coordinates and defect level data of the defect points are obtained, and the DBSCAN density clustering algorithm is used to cluster defect points with a spatial distance less than a preset clustering distance threshold to obtain a defect cluster set.

[0009] Further, based on the defect cluster set, information on defect points and adjacent points within the defect cluster is obtained, the abrupt change coefficient of the hollow boundary is calculated, boundary control points are extracted, and the edge contour of the defect cluster is generated to obtain the boundary of the rework area, including: Based on the defect cluster set, defect point data within each defect cluster is obtained, and a spatial neighborhood search method is used to determine the set of adjacent defect points for each defect point. The void level difference and spatial distance between adjacent defect points are then obtained based on the adjacent defect point sets, and a void boundary abrupt change evaluation formula is used. Calculate the abrupt change coefficient at the boundary of the hollow area. ,in, This indicates the hollowness level value corresponding to the current defect point. This indicates the hollowness level value corresponding to adjacent defect points. This represents the spatial distance between the current defect point and its adjacent defect points. Based on the abrupt change coefficient of the hollow boundary, defect points with a change coefficient greater than a preset boundary threshold are extracted as boundary control points. Based on the boundary control points, the spatial coordinates of the boundary control points are obtained, and the edge contours of each defect cluster are generated by sequentially connecting the boundary points. Based on the edge contours, the set of contour vertices is obtained, and the Douglas-Peucker polygon simplification algorithm is used to eliminate redundant boundary points, thereby obtaining the hollow repair boundary data, and marking the repair area boundary on the drawing.

[0010] Furthermore, the process of obtaining all inspection records through the defect database file, establishing a structured defect database, and generating a rework location map and a digital inspection report on building hollowness includes: By acquiring all inspection record information from the defect database file, a time series of the inspection process is established using a chronological reconstruction method to obtain the inspection history dataset. Based on the inspection history dataset, the corresponding audio file index, location coordinate data, hollow defect level, and BIM component ID are obtained, and a structured defect database is established using a unified primary key association method. Through the structured defect database, the rework area and defective component information are obtained, and a rework location map is generated using drawing rendering. The corresponding audio record, time record, and BIM component number are associated with the defect number to obtain rework guidance data. Based on the rework guidance data, the inspection task number and inspection personnel identity data are obtained, and a digital inspection report of building hollowness containing defect coordinates, defect level, inspection time, audio index, BIM component number, and rework area boundary is generated and written to the inspection archive database using a server archiving method.

[0011] A second aspect of the present invention provides an automatic detection and positioning system for building hollow areas based on UWB positioning, mainly comprising: The base station calibration module is used to extract the boundary contour and spatial range data of the finishing area based on the architectural CAD plan or facade unfolding drawing, deploy UWB base stations, establish a unified spatial coordinate system, and obtain drawing mapping parameters. The hollow hammer trigger module is used to acquire spatial coordinates, vibration amplitude sequence and audio spectrum by periodically positioning signal through UWB tag of hollow hammer handle, vibration sensor and microphone, identify valid knocking events and acquire knocking event data packets; The hollow drum level recognition module is used to extract audio segments based on the tapping event data packet, perform frequency processing and frame decomposition, obtain multi-dimensional sound feature vectors, construct a hollow drum defect level recognition model, and determine the hollow drum defect level and defect location result at the detection location. The defect clustering module is used to convert the spatial coordinates of defect points into drawing coordinates and match them with component information based on the drawing mapping parameters and defect location results. It then uses DBSCAN density clustering to group defect points and obtain a set of defect clusters. The rework boundary generation module is used to obtain information on defect points and adjacent points within defect clusters based on a defect cluster set, calculate the abrupt change coefficient of the hollow boundary, extract boundary control points, and generate the edge contour of the defect cluster to obtain the rework area boundary. The hollow area detection report generation module is used to establish a structured defect database based on the defect database file, and generate a repair location map and a digital detection report of building hollow areas.

[0012] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention provides an automatic detection and positioning method and system for building hollow areas based on UWB positioning. By mapping the spatial coordinates of the building's surface area to the coordinates of CAD or facade drawings, this invention achieves automatic correspondence between the detection location and the drawing location. Simultaneously, it integrates UWB positioning, vibration, and audio signal acquisition to intelligently identify valid tapping events and generate complete data packets, ensuring that operational errors and positioning jitter during continuous tapping do not affect the final result. This invention integrates tapping points that may have duplicate records or irregular boundaries into continuous defect clusters through defect point clustering analysis and hollow boundary mutation assessment. Combined with the hollow boundary mutation assessment mechanism, it analyzes the changes in hollow level and spatial distribution relationships between adjacent defect points, thereby solving the problem of inaccurate determination of rework areas in continuous hollow areas due to point-by-point tapping position deviations, positioning errors, and gradual changes in hollow boundaries. This achieves automatic generation and unified delineation of rework area boundaries. Furthermore, this invention establishes a structured defect database based on defect information, spatial coordinates, hollow level, audio recordings, and BIM component information, and generates digital inspection archives and rework positioning maps, enabling traceable management and visual presentation of inspection results. The automatic detection and location method for building hollow areas based on UWB positioning provided by this invention can realize automatic detection, accurate location, intelligent labeling and digital management of building hollow defects, significantly improve the accuracy of continuous hollow area identification, the accuracy of repair boundary delineation and the traceability of detection results, and improve the efficiency of building hollow detection and defect management. Attached Figure Description

[0013] Figure 1 This is a flowchart of an automatic detection and positioning method for building hollow areas based on UWB positioning according to the present invention; Figure 2 This is a schematic diagram of an automatic detection and positioning method for building hollow areas based on UWB positioning according to the present invention; Figure 3 This is a schematic diagram of an automatic detection and positioning system for building hollow areas based on UWB positioning according to the present invention. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0015] like Figure 1-2 This embodiment of an automatic detection and positioning method and system for building hollow areas based on UWB positioning may specifically include: Step S101: Based on the architectural CAD plan or facade unfolding drawing, extract the boundary contour and spatial range data of the finishing area, deploy UWB base stations, establish a unified spatial coordinate system, and obtain the drawing mapping parameters.

[0016] Based on the CAD floor plan or unfolded facade of the building to be inspected, the boundary contour information of the building's surface area is extracted, and the spatial range data of the surface area is obtained. The surface area includes wall surface layer, floor surface layer, ceiling surface layer, or building facade surface layer. Four UWB base stations are installed in a rectangular or vertically aligned arrangement, and the three-dimensional installation coordinates of each base station are obtained. A unified time reference is obtained by sending periodic synchronization pulse signals through the base station main control module. The distance matrix between base stations is obtained based on the TOF ranging data between the base stations, and the base station coordinates are corrected using a least squares positioning algorithm to obtain a unified spatial coordinate system. The drawing coordinate data is obtained based on the imported CAD floor plan or unfolded facade. The actual spatial coordinates of a preset number of reference points are obtained using on-site acquisition equipment. The correspondence between the drawing coordinates and the spatial coordinates is calculated using an affine transformation matrix to obtain the drawing mapping parameters. Based on the drawing mapping parameters, the spatial coordinates of the hollow hammer are converted into drawing coordinates.

[0017] For example, in an automatic detection scenario for hollow walls in an interior room, the object of detection is the wall finish on the north side of the living room of a residence. This wall is 6m long and 2.8m high. Before detection, the CAD floor plan of the room is imported, and four UWB base stations are placed at the four corners of the room. Base station 1 is installed at coordinates (0m, 0m, 1.5m), base station 2 at coordinates (6m, 0m, 1.5m), base station 3 at coordinates (6m, 5m, 1.5m), and base station 4 at coordinates (0m, 5m, 1.5m), thus establishing a two-dimensional positioning field covering the entire room. Each base station sends a synchronization pulse signal every 10ms through the main control module, and coordinate correction is performed using TOF ranging data to obtain a unified spatial coordinate system. Then, the interior CAD floor plan is imported, and corner A and corner B are selected as reference points. The actual spatial coordinates of corner A are (0m, 0m), corresponding to the coordinates (100mm, 100mm) on the drawing. The actual spatial coordinates of corner B are (6m, 0m), corresponding to the coordinates on the drawing (700mm, 100mm). An affine transformation matrix is ​​used to establish the mapping relationship between the drawing coordinates and the spatial coordinates, yielding the drawing mapping parameters. When the inspector uses a hollow-sounding hammer to strike the wall, the system calculates the real-time spatial coordinates of the hammer as (2.40m, 0.10m, 1.60m), where 2.40m represents the horizontal distance from the left edge of the wall, 0.10m represents the distance between the hammer and the wall, and 1.60m represents the height from the ground. After converting these spatial coordinates to drawing coordinates using the drawing mapping parameters, the corresponding drawing position is (340mm, 260mm). In a three-dimensional positioning and detection scenario for hollow areas on a building facade, when the object being detected is the stone cladding layer on the south facade of an office building, the facade is 30m wide and 60m high. Before testing, two UWB base stations were deployed on both sides of the building's rooftop. Base station 1 was installed on the left side of the rooftop at coordinates (0m, 0m, 60m), and base station 2 was installed on the right side at coordinates (30m, 0m, 60m). Simultaneously, two base stations were deployed at corresponding locations on the ground: base station 3 at coordinates (0m, 0m, 0m) and base station 4 at coordinates (30m, 0m, 0m), thus establishing a three-dimensional positioning field containing height information. After time synchronization and coordinate correction of each base station, the unfolded drawing of the building's facade was imported, and the lower left corner (A), lower right corner (B), and upper left corner (C) of the facade were selected as reference points. The actual spatial coordinates of point A are (0m, 0m, 0m), corresponding to (0mm, 0mm) on the drawing. The actual spatial coordinates of point B are (30m, 0m, 0m), corresponding to (3000mm, 0mm) on the drawing. The actual spatial coordinates of point C are (0m, 0m, 60m), which correspond to the coordinates on the drawing (0mm, 6000mm). The drawing mapping parameters are obtained by calculating the affine transformation matrix.When inspectors used a long-handled hollow-sounding hammer to inspect the stone cladding layer at a height of 35m above the ground, the UWB positioning system calculated the real-time spatial coordinates of the hammer as (12.50m, 0.30m, 35.20m). Here, 12.50m represents the horizontal distance from the left edge of the building along the facade, 0.30m represents the distance between the hammer and the facade, and 35.20m represents the actual height above the ground. After converting these spatial coordinates according to the drawing mapping parameters, the corresponding unfolded facade coordinates are obtained as (1250mm, 3520mm).

[0018] Step S102: By using the periodic positioning signal of the UWB tag on the hollow drum hammer handle, the vibration sensor and the microphone to collect data, spatial coordinates, vibration amplitude sequence and audio spectrum are obtained, valid knocking events are identified, and knocking event data packets are obtained.

[0019] A UWB tag installed on the handle of the hollow drum hammer periodically transmits positioning signals to obtain timestamp data received by each base station. The spatial coordinates of the hollow drum hammer are calculated using the TDOA positioning algorithm. Triaxial acceleration data is obtained from the internal vibration sensor of the hollow drum hammer, and peak detection is performed on the acceleration data using a sliding time window to obtain a vibration amplitude sequence. A sound energy sequence is obtained from the raw audio signal collected by the microphone, and the audio spectrum is calculated using short-time Fourier transform to obtain the frequency energy distribution result. Valid striking events are determined by synchronously comparing the vibration amplitude sequence and frequency energy distribution result with preset vibration thresholds and preset short-time energy thresholds. The corresponding timestamp is obtained for each valid striking event, and the current spatial coordinates and audio buffer segments are saved using a unified time reference to obtain a striking event data packet.

[0020] For example, when detecting hollow spots on the north wall of a residential living room, the inspector holds a hollow spot hammer. A UWB tag installed inside the hammer periodically sends positioning signals to four base stations at the four corners of the room every 20 milliseconds. The base stations record the signal reception timestamps: base station 1 receives the signal at 10.120s, base station 2 at 10.123s, base station 3 at 10.118s, and base station 4 at 10.121s. Using the TDOA positioning algorithm, the spatial coordinates of the hollow spot hammer are calculated as (X=2.35m, Y=0.05m, Z=1.60m), where X represents the horizontal distance from the left edge of the wall, Y represents the distance of the hollow spot hammer from the wall, and Z represents the height from the ground. When the inspector taps the wall at this location, the vibration sensor inside the hammer records peak triaxial accelerations of 3.2g (X-axis), 2.8g (Y-axis), and 0.5g (Z-axis). The peak sequence is calculated using a sliding time window. Simultaneously, the microphone collected 0.85 Pa² of sound energy. Short-time Fourier transform analysis revealed that the frequency energy distribution ranged from 500 Hz to 3 kHz, with a peak frequency of 1.2 kHz. The vibration amplitude sequence and audio energy sequence were synchronously compared with preset thresholds to determine that the strike was a valid event. The preset thresholds included a preset vibration threshold of 2.5 g and a preset short-time audio energy threshold of 0.6 Pa². Subsequently, the timestamp of the strike (10.350 s), the spatial coordinates of the hollow hammer (2.35 m, 0.05 m, 1.60 m), and the audio buffer segment were saved as a complete strike event data packet.

[0021] Step S103: Based on the tapping event data packet, extract audio segments and perform frequency processing and frame decomposition to obtain multi-dimensional sound feature vectors, construct a hollow drum defect level recognition model, and determine the hollow drum defect level and defect location result at the detection location.

[0022] The original audio segments are obtained from the tapping event data packets. A bandpass filter is used to filter the audio segments by frequency, obtaining valid sound data within a preset frequency range. Audio frame sequences are calculated through pre-emphasis processing and Hamming window framing to obtain standardized analysis data. Based on the standardized analysis data, short-time energy, zero-crossing rate, spectral centroid, spectral peak value, Mel frequency cepstral coefficients, and envelope decay time parameters are obtained. A multi-dimensional sound feature vector is constructed using feature concatenation and stored in the tapping sound feature database. Historical data of the multi-dimensional sound feature vectors and corresponding hollow defect level labels are obtained from the tapping sound feature database. A random forest algorithm is used to train the model, constructing a hollow defect level recognition model. Hollow defect levels include no hollow, slight hollow, moderate hollow, and severe hollow. Based on the real-time acquired multi-dimensional sound feature vector data of the detection location, the hollow defect level recognition model is used to determine the hollow defect level and corresponding timestamp of the detection location. A timestamp matching method is used to associate the spatial coordinates with the recognition results to obtain the defect location result.

[0023] For example, a raw audio segment with a duration of 0.8 seconds and a frequency range of 300Hz to 5kHz is obtained from the tapping event data packet. First, the audio segment is frequency-filtered using a bandpass filter, retaining valid sound data within the 300Hz to 5kHz range. Then, the filtered audio is pre-emphasized, and a Hamming window is used to calculate an audio frame sequence with a frame length of 25ms and a frame shift of 10ms, thus obtaining standardized analysis data. Based on the standardized analysis data, the short-time energy of each frame is extracted (maximum value 0.92 Pa²), zero-crossing rate (maximum value 220 times / second), spectral centroid (1.2kHz), spectral peak value (1.5kHz), Mel-frequency cepstral coefficients (average amplitude of the first 13 coefficients 0.36), and envelope decay time (time required to decay to baseline 0.12 seconds). These features are then concatenated to form a multi-dimensional sound feature vector. The vector was stored in a striking sound feature database and combined with historical multidimensional vector data and corresponding hollow defect level labels. The level was 1 for no hollow sound, 2 for slight hollow sound, 3 for moderate hollow sound, and 4 for severe hollow sound. A random forest algorithm was used to train the model, constructing a hollow defect level recognition model. The training model used 100 decision trees, with a maximum tree depth of 15 and a minimum number of splits of 5. The Gini coefficient was selected as the splitting criterion, and the model was trained using a historical dataset containing 2000 striking samples. Cross-validation yielded a model accuracy of 92%. In real-time detection, the multidimensional sound feature vector collected at the current position of the hollow hammer showed the following characteristics: short-time energy 0.87 Pa², zero-crossing rate 210 times / second, spectral centroid 1.18 kHz, spectral peak 1.45 kHz, average of the first 13 Mel coefficients 0.34, and envelope decay time 0.11 seconds. After calculation by the recognition model, the hollow sound level at this position was determined to be moderate, with a timestamp of 10.350 seconds. Subsequently, the hollowness level result was matched with the optimized spatial coordinates (X=2.35m, Y=0.05m, Z=1.60m) of the hollowness hammer based on the timestamp to obtain the final defect location result.

[0024] Step S104: Based on the drawing mapping parameters and defect location results, convert the spatial coordinates of the defect points into drawing coordinates and match them with component information. Group the defect points using DBSCAN density clustering to obtain a defect cluster set.

[0025] Based on the drawing mapping parameters and defect location results, the spatial coordinates of defect points in the defect location results are converted into drawing coordinates using an affine transformation matrix, thus obtaining the drawing defect coordinates. Based on the drawing defect coordinates, the corresponding drawing area is matched using the CAD layer index, or the component to which the defect point belongs is determined by the component's geometric boundary in the BIM model, and the corresponding BIM component ID is associated to obtain the defective component information. Based on the defective component information and defect level data, a drawing annotation object is generated using color coding, and a defect mark is drawn at the corresponding coordinate position on the drawing. The associated audio path, inspector number, inspection time, defect point spatial coordinates, and drawing annotation object are encapsulated into a JSON format inspection record and stored in a defect database file. Based on the defect database file, the spatial coordinates and defect level data of the defect points are obtained, and the DBSCAN density clustering algorithm is used to cluster defect points with a spatial distance less than a preset clustering distance threshold, resulting in a defect cluster set.

[0026] For example, based on the hollow defect location result obtained in the previous step and the drawing mapping parameters, the spatial coordinates (X=2.35m, Y=0.05m, Z=1.60m) of a defect point are converted into drawing coordinates (X... g =340mm, Y g =260mm). Based on the coordinates on the drawing, the CAD layer index is located to confirm that the point belongs to the finishing layer area of ​​the north wall of the living room. Alternatively, the component's geometric boundary in the BIM model can be used to determine that the defect point belongs to the wall panel component ID W-102, thus generating defect component information. Based on this defect component information and the detected medium hollowness level (level 3), a red marker is used to generate a drawing annotation object, and a defect marker is drawn at the corresponding coordinate position on the CAD drawing. Simultaneously, the associated audio path, inspector number, inspection time, defect point spatial coordinates, and drawing annotation object are encapsulated into a JSON format inspection record and stored in the defect database file. In the database, the spatial coordinates and level information of this point are read along with other continuous inspection points. The DBSCAN density clustering algorithm is used to cluster defect points with a spatial distance of less than 0.3m, ultimately forming a defect cluster set containing 12 adjacent defect points, forming the defect cluster for the hollow area in the middle section of the wall.

[0027] Step S105: Based on the defect cluster set, obtain the defect point and adjacent point information within the defect cluster, calculate the abrupt change coefficient of the hollow boundary, extract the boundary control points and generate the edge contour of the defect cluster to obtain the boundary of the rework area.

[0028] Based on defect cluster sets, defect point data within each defect cluster is obtained, and a spatial neighborhood search method is used to determine the set of neighboring defect points for each defect point. The void level difference and spatial distance between adjacent defect points are then obtained from the neighboring defect point sets, and a void boundary abrupt change evaluation formula is used. Calculate the abrupt change coefficient at the boundary of the hollow area. ,in, This indicates the hollowness level value corresponding to the current defect point. This indicates the hollowness level value corresponding to adjacent defect points. This represents the spatial distance between the current defect point and its adjacent defect points. Based on the abrupt change coefficient of the hollow boundary, defect points with a change coefficient greater than a preset boundary threshold are extracted as boundary control points. Based on these boundary control points, their spatial coordinates are obtained, and the edge contours of each defect cluster are generated by sequentially connecting the boundary points. A set of contour vertices is obtained from the edge contours, and redundant boundary points are eliminated using the Douglas-Peucker polygon simplification algorithm to obtain the hollow repair boundary data. The repair area boundaries are then marked on the drawings.

[0029] For example, the previous step uses the DBSCAN density clustering algorithm to form a defect cluster containing 12 defect points, where the hollowness level distribution of the defect points includes 2 severe hollowness points, 5 moderate hollowness points, and 5 slight hollowness points. A spatial neighborhood search method is used to determine the set of neighboring defect points for each defect point, and the hollowness boundary abrupt change evaluation formula is used. Calculate the abrupt change coefficient at the boundary of the hollow area. ,in, This indicates the hollowness level value corresponding to the current defect point. This indicates the hollowness level value corresponding to adjacent defect points. This represents the spatial distance between the current defect point and its adjacent defect points. When the preset boundary threshold is set to 5, a total of 6 boundary control points are selected, with spatial coordinates of (2.10m, 1.45m), (2.25m, 1.62m), (2.52m, 1.70m), (2.83m, 1.68m), (2.95m, 1.48m), and (2.76m, 1.32m). These boundary points are connected in spatial order to form the initial edge contour, resulting in a closed polygon with 6 vertices. The Douglas-Peucker polygon simplification algorithm is used to process the contour. When the simplification tolerance is set to 0.03m, the vertical distances from the second and fifth boundary control points to their adjacent edges are 0.012m and 0.015m, respectively, both less than the simplification tolerance. Therefore, they are judged as redundant boundary points and deleted. Finally, 4 key boundary vertices are retained, generating a hollow repair boundary area with an area of ​​0.62㎡. The repair area is then marked in the corresponding position on the CAD drawing.

[0030] Step S106: Obtain all inspection records through the defect database file, establish a structured defect database, and generate a repair location map and a digital inspection report on building hollowness.

[0031] By retrieving all inspection records from the defect database file, a time series of the inspection process is established using a chronological reconstruction method, resulting in an inspection history dataset. Based on this dataset, corresponding audio file indexes, location coordinate data, hollow defect levels, and BIM component IDs are obtained. A structured defect database is then established using a unified primary key association method. From this database, information on the repair area and defective components is obtained. A repair location map is generated using drawing rendering. Repair guidance data is obtained by associating the defect number with the corresponding audio record, time record, and BIM component number. Based on this guidance data, the inspection task number and inspector identification data are obtained. A digital inspection report on building hollowness, containing defect coordinates, defect level, inspection time, audio index, BIM component number, and repair area boundaries, is generated and written to the inspection archive database using a server archiving method.

[0032] For example, in the detection of hollow spots on the north wall of a residential living room, a total of 186 valid tapping events were collected, resulting in 42 defect records. First, all detection record information in the defect database file was read and reorganized according to the detection time (from 09:15:23 to 09:42:18 on May 20, 2023), forming 186 time-series records, thus obtaining a complete historical detection dataset. Based on the historical detection dataset, the audio file index, location coordinates, hollow spot defect level, and BIM component ID corresponding to each defect record were extracted. For example, defect record D-018 corresponds to the audio file AUDIO_018.wav, with spatial coordinates of (12.50m, 0.30m, 35.20m), a hollow spot level of severe hollow spot, and the corresponding BIM component ID F-305. The above information was stored using a unified primary key to establish a structured defect database. Furthermore, information on the repair area and defective components is retrieved from the structured defect database. For example, 12 defect points were detected within the F-305 area of ​​the facade component. After clustering and boundary calculation, a repair area of ​​1.36㎡ was formed. The coordinates of the unfolded diagram corresponding to the boundary vertices of the repair area are (1180mm, 3410mm), (1340mm, 3390mm), (1420mm, 3570mm), and (1260mm, 3690mm). This area is automatically rendered onto the facade unfolded diagram with a red boundary, and a repair location map is generated. At the same time, based on the defect number D-018, its corresponding audio recording AUDIO_018.wav, the detection time 09:26:42, and the BIM component number F-305 are automatically associated to form repair guidance data, enabling maintenance personnel to accurately locate the defect location and play back the original tapping sound for verification. Subsequently, based on the rework guidance data, the inspection task number TASK-0520-001 and the inspector number INS-008 were retrieved, and a digital inspection report on building hollowness was automatically generated. This report records the defect point's drawing coordinates (1250mm, 3520mm), spatial coordinates (12.50m, 0.30m, 35.20m), hollowness level as severe hollowness, inspection time as May 20, 2023, 09:26:42, associated audio file as AUDIO_018.wav, BIM component number as F-305, and the boundary information and location map of the rework area. Finally, the generated PDF inspection report, along with the corresponding audio file, drawing annotations, and structural defect data, was uploaded to the server for archiving and written into the inspection archive database, forming a building hollowness inspection archive.

[0033] like Figure 3 This embodiment describes an automatic detection and positioning system for building hollow areas based on UWB positioning, which may specifically include: The base station calibration module is used to extract the boundary contour and spatial range data of the finishing area based on the architectural CAD plan or facade unfolding drawing, deploy UWB base stations, establish a unified spatial coordinate system, and obtain drawing mapping parameters.

[0034] The hollow hammer trigger module is used to acquire spatial coordinates, vibration amplitude sequence and audio spectrum by periodically positioning signal from the UWB tag of the hollow hammer handle, collecting data from vibration sensor and microphone, identifying valid knocking events and acquiring knocking event data packets.

[0035] The hollow drum level recognition module is used to extract audio segments based on the tapping event data packet, perform frequency processing and frame decomposition, obtain multi-dimensional sound feature vectors, construct a hollow drum defect level recognition model, and determine the hollow drum defect level and defect location result at the detection location.

[0036] The defect clustering module is used to convert the spatial coordinates of defect points into drawing coordinates and match them with component information based on the drawing mapping parameters and defect location results. It then uses DBSCAN density clustering to group the defect points and obtain a set of defect clusters.

[0037] The rework boundary generation module is used to obtain information on defect points and adjacent points within a defect cluster based on a defect cluster set, calculate the abrupt change coefficient of the hollow boundary, extract boundary control points, and generate the edge contour of the defect cluster to obtain the rework area boundary.

[0038] The hollow area detection report generation module is used to establish a structured defect database based on the defect database file, and generate a repair location map and a digital detection report of building hollow areas.

[0039] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application 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 concept of this application. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A method for automatic detection and location of building hollow areas based on UWB positioning, characterized in that, The method includes: Based on the architectural CAD floor plan or exterior facade unfolding, extract the boundary contour and spatial range data of the finishing area, deploy UWB base stations, establish a unified spatial coordinate system, and obtain the drawing mapping parameters; By periodically locating signals from the UWB tag on the handle of the hollow drum hammer, collecting data from the vibration sensor and microphone, spatial coordinates, vibration amplitude sequences, and audio spectra are obtained, valid knocking events are identified, and knocking event data packets are acquired. Based on the tapping event data packet, audio segments are extracted and frequency processing and frame decomposition are performed to obtain multi-dimensional sound feature vectors. A hollow drum defect level recognition model is constructed to determine the hollow drum defect level and defect location result at the detection location. Based on the drawing mapping parameters and defect location results, the spatial coordinates of the defect points are converted into drawing coordinates and matched with component information. Defect points are grouped by DBSCAN density clustering to obtain a set of defect clusters. Based on the defect cluster set, information on defect points and adjacent points within the defect cluster is obtained, the abrupt change coefficient of the hollow boundary is calculated, boundary control points are extracted and the edge contour of the defect cluster is generated, and the boundary of the rework area is obtained. Based on the defect database file, a structured defect database is established, and a repair location map and a digital inspection report of building hollow areas are generated.

2. The method according to claim 1, wherein, The process involves extracting the boundary contour and spatial range data of the finishing area based on the architectural CAD plan or facade unfolding drawing, deploying UWB base stations, establishing a unified spatial coordinate system, and obtaining drawing mapping parameters, including: Based on the CAD plan or exterior unfolded drawing of the building to be inspected, the boundary contour information of the building's surface area is extracted, and the spatial range data of the surface area is obtained. The surface area includes wall surface layer, floor surface layer, ceiling surface layer, or building exterior surface layer. Four UWB base stations are installed in a rectangular or vertically corresponding arrangement, and the three-dimensional installation coordinates of each base station are obtained. A unified time reference is obtained by sending periodic synchronization pulse signals through the base station main control module. The distance matrix between base stations is obtained based on the TOF ranging data between the base stations, and the base station coordinates are corrected by the least squares positioning algorithm to obtain a unified spatial coordinate system. The drawing coordinate data is obtained based on the imported CAD plan or exterior unfolded drawing, and the actual spatial coordinates of a preset number of reference points are obtained using on-site acquisition equipment. The correspondence between the drawing coordinates and the spatial coordinates is calculated through an affine transformation matrix to obtain the drawing mapping parameters. Based on the drawing mapping parameters, the spatial coordinates of the hollow hammer are converted into drawing coordinates.

3. The method according to claim 1, wherein, The process involves periodically locating signals from the UWB tag on the hollow drum hammer handle, collecting data via a vibration sensor and microphone to obtain spatial coordinates, vibration amplitude sequences, and audio spectra, identifying valid striking events, and acquiring striking event data packets, including: A UWB tag installed on the handle of the hollow drum hammer periodically transmits positioning signals to obtain timestamp data received by each base station. The spatial coordinates of the hollow drum hammer are calculated using the TDOA positioning algorithm. Triaxial acceleration data is obtained from the internal vibration sensor of the hollow drum hammer, and peak detection is performed on the acceleration data using a sliding time window to obtain a vibration amplitude sequence. A sound energy sequence is obtained from the raw audio signal collected by the microphone, and the audio spectrum is calculated using short-time Fourier transform to obtain the frequency energy distribution result. Valid striking events are determined by synchronously comparing the vibration amplitude sequence and frequency energy distribution result with preset vibration thresholds and preset short-time energy thresholds. The corresponding timestamp is obtained for each valid striking event, and the current spatial coordinates and audio buffer segments are saved using a unified time reference to obtain a striking event data packet. Based on the striking event data packet, audio segments are extracted and subjected to frequency processing and frame decomposition to obtain multi-dimensional sound feature vectors. A hollow drum defect level identification model is constructed to determine the hollow drum defect level and defect location result at the detection location.

4. The method according to claim 1, wherein, The process involves extracting audio segments from the tapping event data packet, performing frequency processing and frame decomposition to obtain multi-dimensional sound feature vectors, constructing a hollow drum defect level recognition model, and determining the hollow drum defect level and defect location result at the detection location, including: The original audio segments are obtained from the tapping event data packets. A bandpass filter is used to filter the audio segments by frequency, obtaining valid sound data within a preset frequency range. Audio frame sequences are calculated through pre-emphasis processing and Hamming window framing to obtain standardized analysis data. Based on the standardized analysis data, short-time energy, zero-crossing rate, spectral centroid, spectral peak value, Mel frequency cepstral coefficients, and envelope decay time parameters are obtained. A multi-dimensional sound feature vector is constructed using feature splicing and stored in the tapping sound feature database. Historical data of the multi-dimensional sound feature vectors and corresponding hollow defect level labels are obtained from the tapping sound feature database. A random forest algorithm is used to train the model, constructing a hollow defect level recognition model. Hollow defect levels include no hollow, slight hollow, moderate hollow, and severe hollow. Based on the real-time acquired multi-dimensional sound feature vector data of the detection location, the hollow defect level recognition model is used to determine the hollow defect level and corresponding timestamp of the detection location. A timestamp matching method is used to associate spatial coordinates with the recognition results to obtain the defect location result.

5. The method according to claim 1, wherein, The process involves converting the spatial coordinates of defect points into drawing coordinates based on the drawing mapping parameters and defect location results, matching them with component information, and then grouping the defect points using DBSCAN density clustering to obtain a defect cluster set, including: Based on the drawing mapping parameters and defect location results, the spatial coordinates of defect points in the defect location results are converted into drawing coordinates using an affine transformation matrix to obtain the drawing defect coordinates. Based on the drawing defect coordinates, the corresponding drawing area is matched using the CAD layer index, or the component to which the defect point belongs is determined by the component's geometric boundary in the BIM model, and the corresponding BIM component ID is associated to obtain the defect component information. Based on the defect component information and defect level data, a drawing annotation object is generated using color coding, and a defect mark is drawn at the corresponding coordinate position on the drawing. The associated audio path, inspector number, inspection time, defect point spatial coordinates, and drawing annotation object are encapsulated into a JSON format inspection record and stored in a defect database file. Based on the defect database file, the spatial coordinates and defect level data of the defect points are obtained, and the DBSCAN density clustering algorithm is used to cluster defect points with a spatial distance less than a preset clustering distance threshold to obtain a defect cluster set.

6. The method according to claim 1, wherein, The process involves obtaining information on defect points and adjacent points within a defect cluster based on a defect cluster set, calculating the abrupt change coefficient of the hollow boundary, extracting boundary control points, and generating the edge contour of the defect cluster to obtain the boundary of the rework area, including: Based on defect cluster sets, defect point data within each defect cluster are obtained, and a spatial neighborhood search method is used to determine the set of adjacent defect points for each defect point. The void level difference and spatial distance between adjacent defect points are then obtained based on the adjacent defect point sets, and a void boundary abrupt change evaluation formula is used. Calculate the abrupt change coefficient at the boundary of the hollow area. ,in, This indicates the hollowness level value corresponding to the current defect point. This indicates the hollowness level value corresponding to adjacent defect points. This represents the spatial distance between the current defect point and its adjacent defect points. Based on the abrupt change coefficient of the hollow boundary, defect points with a change coefficient greater than a preset boundary threshold are extracted as boundary control points. Based on the boundary control points, the spatial coordinates of the boundary control points are obtained, and the edge contours of each defect cluster are generated by sequentially connecting the boundary points. Based on the edge contours, the set of contour vertices is obtained, and the Douglas-Peucker polygon simplification algorithm is used to eliminate redundant boundary points, thereby obtaining the hollow repair boundary data, and marking the repair area boundary on the drawing.

7. The method according to claim 1, wherein, The process of establishing a structured defect database based on the defect database file and generating a repair location map and a digital inspection report for building hollow areas includes: All inspection record information is obtained from the defect database file. The inspection process time series is established by reorganizing the data in chronological order to obtain the inspection history dataset. Based on the inspection history dataset, the corresponding audio file index, positioning coordinate data, hollow defect level and BIM component ID are obtained. A structured defect database is established by using a unified primary key association method. By using a structured defect database, information on repair areas and defective components is obtained. A repair location map is generated using drawing rendering. Repair guidance data is obtained by associating the defect number with the corresponding audio record, time record, and BIM component number. Based on the repair guidance data, the inspection task number and inspection personnel identity data are obtained. A digital inspection report on building hollowness, including defect coordinates, defect level, inspection time, audio index, BIM component number, and repair area boundary, is generated and written to the inspection archive database using server archiving.

8. An automatic detection and positioning system for building hollow areas based on UWB positioning, which is implemented based on the automatic detection and positioning method for building hollow areas based on UWB positioning as described in any one of claims 1-7, characterized in that, The system includes the following modules: The base station calibration module is used to extract the boundary contour and spatial range data of the finishing area based on the architectural CAD plan or facade unfolding drawing, deploy UWB base stations, establish a unified spatial coordinate system, and obtain drawing mapping parameters. The hollow hammer trigger module is used to acquire spatial coordinates, vibration amplitude sequence and audio spectrum by periodically positioning signal through UWB tag of hollow hammer handle, vibration sensor and microphone, identify valid knocking events and acquire knocking event data packets; The hollow drum level recognition module is used to extract audio segments based on the tapping event data packet, perform frequency processing and frame decomposition, obtain multi-dimensional sound feature vectors, construct a hollow drum defect level recognition model, and determine the hollow drum defect level and defect location result at the detection location. The defect clustering module is used to convert the spatial coordinates of defect points into drawing coordinates and match them with component information based on the drawing mapping parameters and defect location results. It then uses DBSCAN density clustering to group defect points and obtain a set of defect clusters. The rework boundary generation module is used to obtain information on defect points and adjacent points within defect clusters based on a defect cluster set, calculate the abrupt change coefficient of the hollow boundary, extract boundary control points, and generate the edge contour of the defect cluster to obtain the rework area boundary. The hollow area detection report generation module is used to establish a structured defect database based on the defect database file, and generate a repair location map and a digital detection report of building hollow areas.

Images