Building engineering quality defect detection and evaluation method based on AI image recognition

Abstract

The application relates to the technical field of construction quality monitoring, and discloses a building engineering quality defect detection and evaluation method based on AI image recognition, which comprises a cross-domain dynamic acquisition correlation module, a multi-modal AI recognition and prediction module, a dynamic coefficient calculation and analysis module, a whole-cycle evaluation threshold module and a repair closed-loop twin module; the building outer facade and internal structure images are collected, noise is eliminated in real time and contrast is enhanced, and the data integrity and reliability are ensured; the image recognition technology is used to realize non-contact detection, the safety risks of traditional drilling or high-altitude operation are avoided, and the building structure is not damaged; the AI algorithm can identify defects and calculate defect coefficients, subjective errors are reduced, and the method is safer and more efficient; based on defect coefficient threshold comparison and weight distribution, a comprehensive emergency index is generated, repair priorities and time suggestions are automatically output, decision-making is changed from 'experience-driven' to 'data-driven', and the problem of risk response lag in the traditional method is solved.

Description

Technical Field

[0001] This invention relates to the field of construction quality monitoring technology, and more specifically to a method for detecting and evaluating quality defects in building engineering based on AI image recognition. Background Technology

[0002] With the acceleration of urbanization, the demand for building structural safety monitoring is growing exponentially.

[0003] Traditional methods for detecting quality defects in construction projects include visual inspection, measurement, and testing. Visual inspection relies on human senses to judge surface smoothness, cracks, color, and other external defects; it checks the adhesion of plaster layers, detects surface sanding, identifies hollow areas by sound when tapped, and uses mirrors or lights to inspect hidden areas. Measurement methods use tools for quantitative testing: a straightedge checks flatness, a plumb line and fillet to measure verticality, and measuring dimensional deviations, mortar fullness, and the squareness of corners using measuring tapes, grids, and feeler gauges. Testing methods utilize scientific instrumental analysis, such as rebound hammer testing of concrete strength, ultrasonic testing for internal defects, core drilling to verify concrete compressive strength, pull-out tests to check rebar anchorage, water tightness tests to check for leaks, and static load tests to verify foundation bearing capacity. The following defects are present:

[0004] Data collection is inefficient and unreliable: relying on traditional methods such as manual visual inspection and tape measure measurement, there are risks of data recording delays, high error rates and data loss. The coverage of the sample is insufficient, making it difficult to fully reflect the building quality. Hidden works are more likely to form blind spots in the inspection.

[0005] The detection methods are both destructive and high-risk: traditional methods often require drilling and sampling and scaffolding, which not only damages the integrity of the building structure, but also easily causes safety accidents such as falling from heights and falling objects. When inspecting old buildings, the risk of structural damage is further aggravated, and some areas cannot even be inspected due to safety restrictions.

[0006] Defect assessment is highly subjective and lacks precision: key indicators such as crack width and hollow area rely on the experience of the inspectors, and the results are easily affected by light and viewing angle. Different personnel have large differences in their assessments, and complex defects lack quantitative standards, making it difficult to accurately compare with the standard thresholds.

[0007] Risk decision-making is lagging and lacks dynamism: Traditional processes require manual data processing → meeting discussions → developing remediation plans, with response cycles lasting several weeks. This delays the handling of high-risk defects, and prioritization relies on qualitative experience, failing to quantify the cumulative effects and time decay factors of different defects.

[0008] Therefore, methods that enable efficient and reliable data acquisition, safe and non-destructive testing, accurate and comprehensive defect identification, and dynamic risk decision-making are needed to solve the above problems. Summary of the Invention

[0009] In order to overcome the above-mentioned deficiencies of the prior art, the present invention provides a method for detecting and evaluating the quality defects of building engineering based on AI image recognition, so as to solve the problems existing in the background art.

[0010] To achieve the above objectives, the present invention provides the following technical solution: a method for detecting and evaluating quality defects in building engineering based on AI image recognition, comprising:

[0011] S1. Deploy multi-source image acquisition equipment through cross-domain dynamic acquisition and association modules to acquire images of the building facade, interior, and components;

[0012] S2. The multimodal AI recognition and prediction module performs noise reduction and contrast enhancement processing on the acquired images and identifies the data contained in the images;

[0013] S3. The dynamic coefficient calculation and analysis module calculates the collected multi-module data to generate multiple defect coefficients;

[0014] S4. Set thresholds for each defect coefficient through the full-cycle evaluation threshold module and compare them. Combine the defect category weights to generate a comprehensive emergency index.

[0015] S5. By repairing the closed-loop twin module, the comprehensive emergency index is graded and a digital detection report including repair priority and repair time is output.

[0016] This invention acquires images of building facades, interiors, and components using multimodal data acquisition equipment. After processing by a multimodal AI recognition and prediction module, the image data is identified. A dynamic coefficient calculation and analysis module establishes a mathematical model to quantify defect characteristics and generates seven-dimensional coefficients for cracks, holes, exposed reinforcement, leakage, hollow areas, axis, and dimensions. An evaluation module calculates a comprehensive emergency index based on preset thresholds and weights. Finally, a repair closed-loop twin module outputs a digital report containing repair priorities and time recommendations, achieving fully automated detection from data acquisition to decision support.

[0017] The technical effects and advantages of this invention are as follows:

[0018] 1. This invention automatically collects images of building facades and internal structures through a cross-domain dynamic acquisition and association module, and a multi-modal AI recognition and prediction module eliminates noise and enhances contrast in real time, avoiding the problems of data lag, typos or loss in traditional manual recording, and ensuring data integrity and reliability.

[0019] 2. This invention utilizes image recognition technology to achieve non-contact detection, avoiding the safety risks of traditional drilling or high-altitude operations, while not damaging the building structure, making it suitable for old or complex projects;

[0020] 3. The AI ​​algorithm of this invention can identify seven types of defects, including cracks, holes, and exposed reinforcement, and quantify and calculate the defect coefficient, replacing the traditional fuzzy judgment that relies on experience, reducing subjective errors, and is safer and more efficient, especially in high-altitude or concealed areas.

[0021] 4. This invention generates a comprehensive urgency index based on defect coefficient threshold comparison and weight allocation, automatically outputting repair priority and time recommendations, shifting decision-making from "experience-driven" to "data-driven," and solving the problem of delayed risk response in traditional methods. Attached Figure Description

[0022] Figure 1 This is a structural block diagram of the present invention;

[0023] Figure 2 This is a flowchart of the method of the present invention;

[0024] Figure 3 This is a flowchart of the defect coefficient calculation process of the present invention. Detailed Implementation

[0025] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. In addition, the forms of the various structures described in the following embodiments are merely illustrative. The automatic unloading device for rotary kiln with self-cooling function involved in the present invention is not limited to the structures described in the following embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Reference Figure 1 This invention provides a method for detecting and evaluating quality defects in building engineering based on AI image recognition, including a cross-domain dynamic acquisition and association module, a multimodal AI recognition and prediction module, a dynamic coefficient calculation and analysis module, a full-cycle evaluation threshold module, and a repair closed-loop twin module.

[0027] Reference Figure 2 The specific implementation steps of the present invention include the following steps:

[0028] S1. Deploy multi-source image acquisition devices through cross-domain dynamic acquisition and association modules to acquire images of the building facade, interior, and components.

[0029] It should be specifically noted that the multimodal image acquisition equipment includes a wide-angle industrial camera, an infrared thermal imager, a linear scan camera, a 3D structured light scanner, a macro camera, and an ultrasonic flaw detector, as well as a dynamic environment acquisition unit, a structural stress acquisition unit, a concealed area acquisition unit, and a data association unit.

[0030] Wide-angle industrial cameras and infrared thermal imagers are used to collect images of the exterior facade of building projects, linear scan cameras and 3D structured light scanners are used to collect images of the internal structure of building projects, and macro cameras and ultrasonic flaw detectors are used to detect component images.

[0031] The dynamic environment acquisition unit is equipped with a temperature and humidity sensor with an accuracy of ±0.5℃ / ±3%RH and an ultraviolet intensity sensor with a measurement range of 0-1500μW / cm², which collects temperature, humidity and ultraviolet intensity data of the defect area in real time.

[0032] The structural stress acquisition unit consists of fiber optic stress sensors with an accuracy of ±1με attached to the surface of key components such as beams and columns to acquire real-time stress data of the components.

[0033] The concealed area acquisition unit uses a drone equipped with a 1080P resolution miniature endoscope (depth of field 5-50cm) and a laser positioning module to achieve "aerial positioning + close-range imaging" of concealed areas such as elevator shafts and pipe shafts.

[0034] The data association unit adds tags of "three-dimensional spatial coordinates (accuracy ±5mm) + timestamp (accuracy ±1s) + device ID" to all collected data, establishes a five-dimensional data association library of "image-environment-stress-location-time", and links it with the preset building defect history knowledge base, which includes data on similar building defects, full-cycle project data and disaster case data.

[0035] S2. The multimodal AI recognition and prediction module performs noise reduction and contrast enhancement processing on the acquired images and identifies the data contained in the images.

[0036] The multimodal AI recognition and prediction module includes: a multimodal fusion recognition submodule: employing an "image-environment-stress" multimodal fusion Transformer model, inputting preprocessed images, environmental time-series data, and stress curves, focusing on key defect areas through spatial attention and strengthening cross-domain data correlation features through channel attention, outputting defect type, size, and a "defect-environment-stress" correlation score (0-100 points); and a defect evolution prediction submodule: based on a time-series LSTM-Transformer hybrid model, inputting a three-month historical defect image sequence and environmental / stress data, outputting... The module displays the defect evolution trend over the next 1-6 months (including quantitative prediction values ​​and visualized trend curves), for example, "The current crack width is 0.2mm, and it will increase to 0.35mm in the next 3 months under the conditions of an average daily temperature of 35℃ and a stress of 150MPa"; the fine-grained classification submodule further subdivides the original 7 types of defects, such as cracks being classified into temperature cracks (irregular mesh texture), load cracks (longitudinal straight lines with stress concentration), and shrinkage cracks (short, thin, and dispersed); leakage being classified into pipe leakage (linear temperature difference zone) and wall penetration (sheet-like increase in humidity); and exposed rebar being classified into exposed rebar due to protective layer detachment, exposed rebar due to rebar corrosion, and exposed rebar due to construction omissions.

[0037] It should be explained in detail that, firstly, image filtering techniques, such as Gaussian filtering or median filtering, are used to eliminate noise interference in the acquired image and improve image clarity; secondly, contrast enhancement algorithms, such as histogram equalization or gamma correction, are applied to adjust the image brightness and color distribution to enhance defect feature recognition; then, AI pre-trained deep learning models, such as YOLO or convolutional neural networks, are used to perform defect recognition and feature extraction on the processed image and extract defect feature data.

[0038] It should be explained that the defect feature data specifically includes the perimeter, area and thickness of the component, the length, width and depth of the crack, the area and number of holes, the exposed area and degree of corrosion of the reinforcing bars, the area and temperature difference of the leakage area, the area and height of the hollow area, the position information of the axis and the size information of the component.

[0039] S3. The dynamic coefficient calculation and analysis module calculates the collected multi-module data to generate multiple defect coefficients.

[0040] Reference Figure 3 The defect coefficient is calculated as follows:

[0041] The specific crack defect coefficient is:

[0042] ;

[0043] Where F is the crack defect coefficient, which reflects the degree to which crack defects in concrete structures weaken the structural bearing capacity in building engineering.

[0044] It should be explained that L is the crack length, i.e. the geometric dimension of the crack extension. The longer the length, the wider the stress concentration area, and the higher the risk of overall structural failure. W is the maximum crack width, i.e. the geometric dimension of the crack opening at its widest point. k is W raised to the power of W, which is a weighted average of the maximum crack width, reflecting the accelerating effect of increased width on the risk of building damage. The value of k ranges from 1.2 to 1.5. S is the cross-sectional area of ​​the component where the crack is located, reflecting the buffering capacity of the component size. The larger the area, the smaller the relative harm of the crack.

[0045] D represents the depth ratio, which is the ratio of crack depth to component thickness. It reflects the extent to which the crack propagates into the component. The deeper the crack, the weaker the load-bearing capacity of the component. T represents the crack type correction coefficient. Temperature cracks are caused by uneven expansion and contraction of materials due to temperature changes. They have little impact on structural safety and are set to 1.0. Load cracks are caused by external loads that exceed the load-bearing capacity limit of the component. They reflect insufficient structural load-bearing capacity or stability defects and require an increased risk level. Therefore, the value range is between 1.5 and 2.0.

[0046] K env Values ​​are determined based on temperature and humidity: 1.2 for temperature > 35℃, 1.3 for humidity > 80%RH, and 1.1 for temperature < 0℃; K stress The value is determined based on the ratio of real-time stress to design stress: 1.5 when the ratio is >0.8, 1.2 when the ratio is 0.6-0.8, and 1.0 when the ratio is <0.6.

[0047] It should be specifically noted that the porosity defect coefficient is as follows:

[0048] ;

[0049] Where K is the porosity defect coefficient, which reflects the porosity of concrete in building construction.

[0050] It needs to be explained that S k S represents the total area of ​​voids, reflecting the degree of accumulation of void defects. A larger value indicates more severe voids inside the concrete. z It provides a baseline scale for the percentage of defects in the total area to be inspected, ensuring that the calculation results are independent of the size of the construction.

[0051] N represents the number of pores, indicating the dispersion of the pore defect distribution; a higher number indicates poorer material uniformity. x The limit for the number of holes is N / N x To reduce the number of pores to a reasonable range, avoid directly multiplying the numbers to prevent excessive amplification of the result. When the number of pores exceeds the limit, the uniformity of the concrete decreases significantly. x The value ranges from 10 to 20, depending on the size of the component.

[0052] It should be specifically noted that the exposed rebar defect coefficient is as follows:

[0053] ;

[0054] Where J is the exposed rebar defect coefficient, which reflects the severity of exposed and rusted rebar defects in building construction.

[0055] It needs to be explained that S l S represents the area of ​​exposed rebar, reflecting the extent of the defect; a larger value indicates more severe rebar exposure. g λ is the reference area of ​​the component, which is the product of the component's cross-sectional perimeter and the inspection length. Exposed reinforcement defects usually extend longitudinally along the component surface. Their severity is directly related to the coverage ratio of the exposed area on the component surface. The component reference area is directly taken as the standardized exposed ratio to eliminate the influence of size differences. λ is the exposed area coefficient, which weakens the influence of extremely large exposed areas and has a value range between 0.8 and 0.9.

[0056] γ is the corrosion grade coefficient, reflecting the degree to which the type of corrosion weakens the steel reinforcement section. 0 represents no corrosion, 0.3-0.4 represents surface yellow spots (floating rust), 0.6-0.7 represents reddish-brown raised areas (granular rust), and 1.0 represents fish-scale peeling (flaky rust). X represents the corrosion coverage rate, i.e., the proportion of corrosion diffusion. The exponential function is set such that when the corrosion diffusion range exceeds 50% of the construction surface, the risk increases exponentially, and corrosion spreads faster. When X≤0.5, the exponential term is assigned a value, and the amplification effect is gradual. When X>0.5, the exponential term is positive, and the risk increases. Dividing the denominator by 2 suppresses the growth rate of the exponential function to avoid excessively large values ​​that could cause distortion.

[0057] H represents the design protective layer thickness, which is the original rust prevention design requirement of the structure; △H represents the protective layer loss, which is the deviation between the design protective layer thickness and the current thickness, and is directly related to the degree of failure of the steel reinforcement's rust prevention capability.

[0058] It should be specifically noted that the leakage defect coefficient is as follows:

[0059] ;

[0060] Where S is the leakage defect coefficient, which reflects the severity of leakage defects in building engineering.

[0061] It needs to be explained that S s The area of ​​the seepage zone represents the extent of water diffusion and reflects the spatial scale of the leakage impact; the larger the area, the higher the risk. △T is the infrared temperature difference, which is the temperature difference between the seepage zone and the normal zone, reflecting the activity level of the leakage; the larger the temperature difference, the more serious the leakage. The product of the two reflects the degree of damage caused by the leakage.

[0062] R represents the remaining thickness of the waterproof layer, which is the effective thickness of the protective layer and reflects the structure's own protective capability. The smaller the thickness, the weaker the resistance to leakage. When the leakage area is large and the remaining thickness of the waterproof layer is small, the ratio increases, highlighting the combined effect of the expansion of the leakage range and the decline of the protective capability. If the thickness of the waterproof layer is small, even if the leakage area is small, it will lead to an increase in risk. When the remaining thickness of the waterproof layer is large, even if the leakage area is large, the ratio will not be large, indicating that the waterproof layer can still buffer some of the damage.

[0063] It should be specifically noted that the hollow defect coefficient is as follows:

[0064] ;

[0065] Where G is the hollow defect coefficient, which reflects the severity of hollow defects and assesses the potential risks to the safety and function of building projects.

[0066] It needs to be explained that S g The hollow area represents the actual area of ​​the hollow area within the inspection zone, that is, the area where the plaster layer or finishing material separates from the structural layer. It reflects the spatial extent of the hollow area. The larger the area, the more serious the adhesion failure, the higher the risk of stress concentration, leading to tile detachment or crack propagation; S q It indicates the overall area of ​​the wall surface being inspected, including hollow and intact areas, providing a benchmark for comparison, standardizing the impact of hollow areas, and the ratio reflects the proportion of hollow defects to the whole. The higher the proportion, the greater the hollow defect coefficient, indicating that the risk is more likely to spread to other areas; the square root calculation is used to minimize the impact of small-area hollows and highlight the severe impact of large-area hollows.

[0067] Hg represents the hollow height, indicating the vertical distance of the hollow area from the ground. It reflects the risk of detachment; the higher the height, the stronger the impact force during detachment, and the more serious the safety hazard. h is the height limit, which is related to Hg's vertical height. g After normalization, values ​​exceeding this range indicate high risk; the value of h is between 2.5 and 3.

[0068] It should be specifically noted that the axis deviation coefficient is as follows:

[0069] ;

[0070] Z is the axis deviation coefficient, which quantifies the degree of axis position deviation in building engineering. It reflects the ratio of the average deviation between the actual construction axis and the design axis to the allowable value in the specification, reflects the construction accuracy, and ensures the structural safety and stability.

[0071] It needs to be explained that △x i and △y iThe deviation values ​​for each monitoring point represent the horizontal and vertical deviation values ​​of the i-th measuring point, respectively. This is the algebraic difference between the measured position and the designed position. The absolute value of the difference, whether positive or negative, represents the offset of a single point, reflecting local construction errors. n represents the total number of monitoring points. By averaging the axial deviation of all points, the overall evaluation accuracy is avoided from being affected by individual outliers. The more monitoring points there are, the more representative the global deviation situation becomes. △L is the allowable deviation specified by the engineering standard.

[0072] It should be specifically noted that the dimensional deviation coefficient is as follows:

[0073] ;

[0074] Where C is the dimensional deviation coefficient, which reflects the accuracy defects in the dimensions of the component by the severity of the overall dimensional deviation and the dispersion of the component deviation distribution.

[0075] It should be explained that m represents the number of detection points, that is, the number of detection points evenly distributed on the surface of the component, to ensure comprehensive coverage of the component samples; △d j The measured deviation value is the difference between the actual size and the design size of the i-th detection point, reflecting the dimensional deviation at a local location; d s To standardize permissible deviations, the maximum acceptable error value specified by national standards is used as a benchmark threshold; the average deviation is calculated using the square root operation, and the larger the value, the larger the dimensional defect.

[0076] σ d σ represents the standard deviation of the deviation, which is the statistical value of the dispersion of the deviation values ​​across all detection points. It reflects the dispersion of dimensional fluctuations. d The large component indicates that it is twisted or uneven; The average of the absolute values ​​of the deviations, i.e., the average of the absolute values ​​of the deviations at all detection points, represents the overall offset level, and σ d The ratio is used to jointly correct for discrete effects and avoid interference from extreme values. The larger the ratio, the greater the local deviation, the more obvious the dimensional defects, and the greater the impact on the safety of the building project.

[0077] S4. The threshold for each defect coefficient is set and compared through the full-cycle evaluation threshold module, and a comprehensive emergency index is generated by combining the defect category weight.

[0078] It should be specifically noted that the comprehensive emergency index is as follows:

[0079] ;

[0080] A represents the comprehensive urgency index, which assesses the urgency of quality defect risks in building construction projects by integrating defect severity, category weights, and time effects.

[0081] Among them B q B is the measured coefficient of the q-th type of defect, reflecting the severity of a single defect; t This is the primary threshold for this type of defect, also known as the safety baseline. It is the critical value for determining whether intervention is necessary due to a defect, and is derived from national standards or engineering specifications; B max This is the secondary threshold for this type of defect, also known as the critical value of danger, which is the limit state that will cause structural failure and is calculated by structural safety. The measured defect coefficient is mapped to the 0-1 interval to indicate the degree of deviation from the threshold range.

[0082] w q The weighting coefficients for defect categories distinguish the importance of different defect types and are determined based on the importance of the safety components of a building project. Structural defects directly affect the load-bearing capacity and overall stability of a building, and are irreversible and high-risk, therefore they have the highest weight. These include cracks, holes, and exposed rebar, with values ​​ranging from 0.4 to 0.45. Functional defects affect the normal use of the building; although they do not directly endanger structural safety, they can cause interruptions in use and increased maintenance costs, so they have the next highest weight. These include leakage and hollow areas, with values ​​ranging from 0.3 to 0.35. Dimensional defects affect installation accuracy or trigger cascading problems, but can be adjusted locally, so they have a medium weight. These include axial deviations and component dimensional deviations, with values ​​ranging from 0.25 to 0.35.

[0083] t represents the number of weeks since the defect was discovered. The longer the defect has been present, the more urgent it becomes. The exponential function reflects the non-linear growth of defect risk over time. When t < 2, it is the initial stage of risk accumulation, with a small function value and a gradual increase in urgency. When t = 4, the risk begins to accelerate, the function value becomes large, and the urgency increases sharply to the critical value. When t > 6, it is a high-risk plateau period, with the function value approaching 1.

[0084] K phase Values: 0.8 for the construction phase (before delivery), 1.0 for the operation and maintenance phase (within 5 years after delivery), and 1.5 for the aging phase (>15 years after delivery); K scene Values: 1.6 for core areas (hospital operating rooms, load-bearing columns), 1.0 for general areas (residential living rooms), and 0.6 for non-load-bearing areas (partition walls);

[0085] To further verify the feasibility of the method of the present invention, the following mathematical modeling data for defect detection is provided in conjunction with a specific engineering scenario, and is presented in tables below:

[0086]

[0087] S5. By repairing the closed-loop twin module, the comprehensive emergency index is classified and a digital detection report containing repair priority and repair time is output.

[0088] The repair closed-loop twin module includes:

[0089] Repair scheme generation unit: Based on the defect type and scenario, the repair process is adapted, such as adapting "exposed reinforcement in the core area" to "rust removal + epoxy coating (thickness ≥ 0.2mm) + recast protective layer", and outputting cost-benefit analysis of multiple schemes (e.g., scheme A: cost 20,000 yuan / cycle 3 days, scheme B: cost 15,000 yuan / cycle 5 days).

[0090] Repair effect verification unit: Automatically generates retest plans (7 days and 30 days after repair), compares the changes in defect coefficients before and after repair, tracks recurrence for 1-2 years, and triggers scheme optimization when the recurrence rate is >5%;

[0091] Digital twin integration unit: Maps defect data (location, coefficient, priority) to the building digital twin model in real time, distinguishes priorities by color (red: first priority, yellow: second priority, blue: third priority, green: fourth priority), and supports multi-terminal (mobile APP, web) interactive viewing and repair simulation.

[0092] It should be specifically noted that the classification of the comprehensive emergency index is as follows:

[0093] When A≥a1, it has the highest priority.

[0094] When a2≤A<a1, it is the second priority;

[0095] When a3≤A<a2, it is the third priority;

[0096] When A < a3, it is the fourth priority;

[0097] The value of a1 is between 0.7 and 0.8; the value of a2 is between 0.5 and 0.6; and the value of a3 is between 0.2 and 0.3.

[0098] It should be noted that for the first priority defect category, it is recommended to complete the repair within 24 hours, and work should be stopped immediately and emergency reinforcement should be carried out.

[0099] For the second priority defect category, it is recommended to complete the repair within 3 days, which requires a special construction plan.

[0100] For the third priority defect category, it is recommended to complete the repair within 7 days and perform routine maintenance.

[0101] For the fourth priority defect category, it is recommended to complete the repair within 30 days and include it in the regular maintenance plan.

[0102] It should be noted that the report generation algorithm converts the graded results into a standardized digital format, which includes a defect location distribution map, defect type, repair priority, repair time recommendation and its detailed data table. The output is a PDF report, which can be exported to the engineering management system with one click, facilitating team collaboration and tracking, and ensuring defect response efficiency.

[0103] Through the above description of the embodiments, those skilled in the art can clearly understand that the various embodiments of this application can be implemented by means of software or software combined with necessary general-purpose hardware platforms, and of course, they can also be implemented by hardware functions; based on this understanding, the technical solution of this application, in essence or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions to cause a computer device, such as including but not limited to a personal computer, server, or network device, to execute all or part of the steps of the method described in any embodiment of this application.

[0104] The foregoing describes exemplary embodiments of this application. It should be understood that the above exemplary embodiments are not restrictive but illustrative, and the scope of protection of this application is not limited thereto. It should be understood that those skilled in the art can make modifications and variations to the embodiments of this application without departing from the spirit and scope of this application, and such modifications and variations should be within the scope of protection of this application.

Claims

1. A method for detecting and evaluating quality defects in building engineering based on AI image recognition, characterized in that, Specifically, it includes: S1. Deploy multi-source image acquisition equipment through cross-domain dynamic acquisition and association modules to acquire images of the building facade, interior, and components; S2. The multimodal AI recognition and prediction module performs noise reduction and contrast enhancement processing on the acquired images and identifies the data contained in the images; S3. The dynamic coefficient calculation and analysis module calculates the collected multi-module data to generate multiple defect coefficients; S4. Set thresholds for each defect coefficient through the full-cycle evaluation threshold module and compare them. Combine the defect category weights to generate a comprehensive emergency index. S5. By repairing the closed-loop twin module, the comprehensive emergency index is graded and a digital detection report including repair priority and repair time is output.

2. The method for detecting and evaluating quality defects in building engineering based on AI image recognition according to claim 1, characterized in that: The defect coefficients include crack defect coefficient, hole defect coefficient, exposed reinforcement defect coefficient, leakage defect coefficient, hollow defect coefficient, axis deviation coefficient, and dimensional deviation coefficient.

3. The method for detecting and evaluating quality defects in building engineering based on AI image recognition according to claim 2, characterized in that: The specific crack defect coefficient is: ； Where F is the crack defect coefficient, reflecting the degree to which cracks in concrete structures weaken the structural bearing capacity; L is the crack length; W is the maximum crack width; k is the power of W; S is the cross-sectional area of ​​the component where the crack is located; D is the depth percentage; and T is the crack type correction coefficient K. env Values ​​are determined based on temperature and humidity: 1.2 for temperature > 35℃, 1.3 for humidity > 80%RH, and 1.1 for temperature < 0℃; K stress The value is determined based on the ratio of real-time stress to design stress: 1.5 when the ratio is >0.8, 1.2 when the ratio is 0.6-0.8, and 1.0 when the ratio is <0.

6. The specific porosity defect coefficient is: ； Where K is the porosity defect coefficient, reflecting the degree of looseness of concrete in building construction; S k S represents the total area of ​​the hole. z Let N be the total area of ​​the detection area and N be the number of holes. x Limit on the number of holes; The exposed rebar defect coefficient is specifically: ； Where J is the exposed rebar defect coefficient, reflecting the severity of exposed and corroded rebar defects in building construction; S l S represents the area of ​​the exposed rebar. g λ is the reference area of ​​the component, γ is the product of the component's cross-sectional perimeter and the inspection length, λ is the exposed area coefficient, γ is the corrosion grade coefficient, X is the corrosion coverage, H is the design protective layer thickness, and ΔH is the protective layer loss. The leakage defect coefficient is specifically: ； Where S is the leakage defect coefficient, reflecting the severity of leakage defects in building engineering; S s Where ΔT is the area of ​​the seepage zone, ΔT is the infrared temperature difference, and R is the remaining thickness of the waterproof layer. The hollow defect coefficient is specifically: ； Where G is the hollow defect coefficient, reflecting the severity of hollow defects and assessing the potential risks to the safety and function of the building project; S g S represents the area of ​​the hollow area. q H represents the total area of ​​the wall surface being measured. g Where h is the hollow height, and h is the height limit; The axis deviation coefficient is specifically: ； Where Z is the axis deviation coefficient, which quantifies the degree of axis position offset in a building project; △x i and △y i Let be the deviation value of each detection point, representing the deviation value of the i-th detection point in the horizontal and vertical directions respectively, n representing the total number of detection points, and ΔL representing the allowable deviation according to the standard. The dimensional deviation coefficient is specifically: ； Where C is the dimensional deviation coefficient, reflecting the accuracy defects in the component's dimensions; m is the number of inspection points, and Δd j The measured deviation value is the difference between the actual size and the design size at the i-th inspection point, d. s To standardize the allowable deviation, σ d The standard deviation is the deviation. The mean of the absolute values ​​of the deviations.

4. The method for detecting and evaluating quality defects in building engineering based on AI image recognition according to claim 1, characterized in that: The comprehensive emergency index is specifically as follows: ； Where A is the comprehensive urgency index, which assesses the urgency of the risk of quality defects in building projects; B q B is the measured coefficient of the q-th type of defect, reflecting the severity of a single defect; t This is the primary threshold for this type of defect, a critical value for determining whether intervention is necessary; B max This is the secondary threshold for this type of defect, also known as the critical danger value, w q The defect category weighting coefficient is used to distinguish the importance of different defect types and is determined based on the importance of the safety components of a building project. t represents the number of weeks since the defect was discovered. The longer the defect has been present, the more urgent it becomes. K phase Values: 0.8 for the construction phase, 1.0 for the first 5 years after delivery, and 1.5 for more than 15 years after delivery; K scene Values: 1.6 for core areas including hospital operating rooms and load-bearing columns, 1.0 for general areas including residential living rooms, and 0.6 for non-load-bearing areas.

5. The method for detecting and evaluating quality defects in building engineering based on AI image recognition according to claim 1, characterized in that: The specific steps for classifying the comprehensive emergency index are as follows: When A≥a1, it has the highest priority. When a2≤A<a1, it is the second priority; When a3≤A<a2, it is the third priority; When A < a3, it is the fourth priority.

Images