An industrial vision-based building foundation detection system for elevator retrofitting

Abstract

The application discloses an elevator installation building foundation detection system based on industrial vision, and relates to the technical fields of building structure detection and industrial image processing.The system comprises a surface feature acquisition module, a visual anomaly identification module, a thermal response analysis module, a sound wave imaging detection module, a structural integrity evaluation module, and an installation path optimization and risk avoidance module.The surface feature acquisition module collects building foundation surface images and spatial data, uses an industrial camera and a structured light scanning device to perform high-resolution image acquisition and spatial contour scanning, and constructs a surface feature reference atlas.The application realizes a closed-loop detection process of anomaly identification and structural health evaluation by fusing surface modeling, thermal analysis and sound wave imaging, improves detection accuracy and stability, optimizes the installation path based on the health index, avoids hidden danger areas, significantly enhances the safety and intelligent level of elevator installation engineering, and has good popularization and application prospects.

Description

TECHNICAL FIELD

[0001] The application relates to the field of building structure detection and industrial image processing technology, and particularly relates to an elevator installation building foundation detection system based on industrial vision. BACKGROUND

[0002] The elevator installation building foundation detection system based on industrial vision refers to a detection device integrating building three-dimensional visual recognition and intelligent structure analysis, which is applied to the foundation structure evaluation scene before elevator installation of existing old residential buildings. The system performs image acquisition and point cloud modeling on the building foundation area by deploying high-precision industrial cameras and laser scanning devices, constructs a digital three-dimensional model of the building structure in combination with a BIM model, identifies the foundation size, morphological features and possible structural defects, simultaneously links a distributed fiber sensor array to dynamically detect stress distribution, settlement trend and humidity state, and assists in evaluating the foundation bearing capacity and stability. The detection results output by the system provide input basis for modular foundation size optimization, adaptive adjustment strategy formulation and robot installation path planning, realize the full-process data closed loop from building status recognition to structure response prediction, and provide high-precision and low-error foundation adaptation guarantee for old residential elevator installation projects.

[0003] The prior art has the following disadvantages:

[0004] In the building foundation detection process, if there are shallow cavities, peeling and erosion layers or early carbonization and other internal deterioration phenomena of concrete below the foundation surface, since such defects usually do not form obvious geometric distortion, color difference features or texture abnormalities on the surface, their image performance is highly similar to that of normal structure regions, and they are difficult to be accurately distinguished by conventional recognition algorithms relying on surface visual texture analysis or three-dimensional profile reconstruction, and are easily misjudged as structure intact regions. Most of the existing detection methods based on industrial vision focus on processing the image features or laser point cloud data of the building foundation surface, and lack deep perception ability of the structure state below the surface, especially in the case that the defect distribution depth is shallow, the boundary transition is fuzzy or the carbonization has not caused obvious optical changes, the recognition accuracy is greatly reduced, and there is a serious risk of missed detection. If the module foundation installation or elevator structure installation operation is directly performed in such un-identified hidden danger regions, local bearing imbalance, stress concentration or continuous settlement are easily caused, and then serious structure safety problems such as foundation cracking, shaft deviation and wall detachment are caused, which has high engineering accident risk and use hidden danger.

[0005] The above information disclosed in the background section is only used to enhance the understanding of the background of the present disclosure, and therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0006] The application aims to provide an industrial vision-based building foundation detection system for elevator installation to solve the problems in the background art.

[0007] To achieve the above-mentioned purpose, the application provides the following technical solution: an industrial vision-based building foundation detection system for elevator installation, comprising a surface feature acquisition module, a visual anomaly recognition module, a thermal response analysis module, a sound wave imaging detection module, a structural integrity evaluation module, and an installation path optimization and risk avoidance module:

[0008] The surface feature acquisition module collects building foundation surface images and spatial data, uses an industrial camera and a structured light scanning device to perform high-resolution image acquisition and spatial contour scanning, and constructs a surface feature reference atlas.

[0009] The visual anomaly recognition module performs abnormal area screening processing on the reference atlas, calculates the brightness gradient change rate, texture boundary continuity, and geometric curvature mutation value, and outputs a spatial annotation layer.

[0010] The thermal response analysis module performs thermal response inversion processing based on the spatial annotation layer, applies low-frequency infrared excitation, records the thermal diffusion path and temperature rise amplitude change, and constructs a thermal inertia indication atlas.

[0011] The sound wave imaging detection module performs sound wave time difference imaging processing based on the thermal inertia indication atlas, applies directional low-frequency sound wave pulses, collects reflected echo delay data, and constructs a sound-heat cooperative anomaly positioning map.

[0012] The structural integrity evaluation module performs integrity discrimination based on the sound-heat cooperative anomaly positioning map, counts the number of overlapping areas and spatial distribution of sound-heat abnormal areas, outputs a structural health index, and realizes quantitative grading.

[0013] The installation path optimization and risk avoidance module optimizes the installation scheme according to the structural health index result, adjusts the foundation layout position and support column landing point, avoids high-risk areas, and forms a closed-loop process of identification, early warning, and avoidance.

[0014] Preferably, the surface feature acquisition step is as follows:

[0015] By setting a multi-angle high-resolution industrial camera and a structured light projection device, high-precision images and structured light stripe reflection images of the building foundation surface are collected.

[0016] Start the synchronous working mechanism of the industrial camera and the structured light device, perform multi-angle multi-frame image acquisition, and obtain texture images and spatial coding patterns of the target area.

[0017] Perform three-dimensional reconstruction processing based on image and depth map data, construct a spatial reconstruction point cloud model and a high-fidelity texture layer.

[0018] A unified coordinate system is established to form a surface feature reference atlas by integrating the gray scale, texture direction, spatial depth and curvature information of each pixel.

[0019] Preferably, the visual anomaly recognition step is as follows:

[0020] The brightness value of each pixel in the building foundation surface feature reference atlas is extracted, the brightness gradient change rate in the horizontal and vertical directions is calculated, a brightness change rate atlas is generated, and a brightness anomaly candidate region is marked;

[0021] The gray level co-occurrence matrix and the direction gradient histogram are calculated in the candidate region, the texture direction continuity is analyzed, and the texture boundary discontinuous region is marked;

[0022] The principal curvature and Gaussian curvature of the spatial contour point cloud data are calculated to identify the curvature mutation region as a potential structural anomaly point;

[0023] The brightness, texture and curvature anomaly regions are fused to construct an anomaly confidence atlas and generate a spatial annotation layer.

[0024] Preferably, the thermal response analysis step is as follows:

[0025] The position of the anomaly region is calibrated according to the spatial annotation layer, the excitation parameters are set, and low-frequency infrared thermal excitation is applied, and the temperature rise image sequence in the thermal diffusion process of the anomaly region is collected;

[0026] The pixel temperature response curve is extracted based on the temperature rise image sequence, the initial response time, the maximum temperature rise value and the thermal equilibrium time consumption are calculated, and a thermal diffusion time characteristic parameter set is generated;

[0027] The temperature response sequence is subjected to Fourier transform, the frequency domain phase and amplitude characteristics of each pixel are extracted, and a thermal conduction lag index map is constructed;

[0028] The phase lag map and the temperature rise distribution map are fused to generate a thermal inertia indication atlas and map it into the building foundation unified coordinate system.

[0029] Preferably, the acoustic imaging detection step is as follows:

[0030] The position of the thermal conduction lag region is calibrated according to the thermal inertia indication atlas, the acoustic excitation path is set, and low-frequency acoustic pulse excitation is applied;

[0031] The acoustic reflection echo delay data is collected, a reflection echo delay distribution map is generated, and the acoustic propagation path and time domain characteristics are recorded;

[0032] Based on the acoustic propagation model, the reflection data is subjected to inversion analysis to deduce the position and boundary form of the sound speed mutation point;

[0033] The acoustic velocity sudden change point is aligned and fused with the thermal conduction lag region in coordinates, and an acoustic-thermal cooperative anomaly positioning map is constructed.

[0034] Preferably, the structural integrity evaluation step is as follows:

[0035] The area covered by the acoustic-thermal cooperative anomaly positioning map is divided into detection units with spatial boundaries and unique identifiers;

[0036] The acoustic wave reflection delay and thermal diffusion lag response data in each detection unit are extracted, and the overlap degree and spatial continuity of the abnormal area are counted;

[0037] According to the intensity, coincidence rate and distribution characteristics of the abnormal response, the structural health index is calculated and the health grade is given;

[0038] The structural health grade is mapped to the building foundation three-dimensional model to form a structural health grade map and color identification expression.

[0039] Preferably, the installation path optimization and risk avoidance step is as follows:

[0040] Based on the structural health index map, the building foundation area is spatially partitioned and marked as an optimal area, a warning area and an avoidance area;

[0041] According to the structural layout requirements, set the candidate layout scheme, eliminate the support point positions in the avoidance area and optimize the landing point combination;

[0042] The optimized support point layout scheme is subjected to structural mechanics simulation analysis to verify that the stress distribution and deformation response meet the bearing requirements;

[0043] The final layout scheme and the health response map layer are integrated to generate a construction layout reference map for construction landing point verification and risk avoidance guidance.

[0044] In the above technical solution, the technical effects and advantages provided by the present application are:

[0045] The present application constructs a full-process diagnosis link from anomaly area identification, physical response verification to structural health quantitative evaluation through the linkage processing mechanism of surface feature modeling, thermal response analysis and acoustic imaging detection, significantly improves the detection accuracy and recognition robustness. At the same time, combined with the quantitative output results of the structural health index, the system can actively adjust the foundation component layout position and installation path to avoid high-risk hidden danger area, realize the closed-loop control from "defect identification" to "risk avoidance". The system not only improves the intelligent level of building foundation detection and installation safety, but also provides a scientific, reliable and engineering-based decision-making basis for old residential elevator installation engineering, has significant technical popularization value and engineering application prospect. BRIEF DESCRIPTION OF DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed in the embodiments. Obviously, the drawings described below only show some embodiments of the present application, and other drawings can also be obtained by those skilled in the art based on these drawings.

[0047] Figure 1 A module schematic diagram of a building foundation detection system based on industrial vision for elevator installation. DETAILED DESCRIPTION

[0048] Example implementations will now be described more fully with reference to the accompanying drawings. Example implementations may, however, be implemented in many different forms and should not be construed as limited to the examples set forth herein; rather, these example implementations are provided so that this disclosure will be thorough and complete, and will fully convey the inventive concept to those skilled in the art.

[0049] The present application provides a building foundation detection system based on industrial vision for elevator installation as shown in Figure 1 The building foundation detection system based on industrial vision for elevator installation comprises a surface feature acquisition module, a visual anomaly recognition module, a thermal response analysis module, a sound wave imaging detection module, a risk assessment module and an installation strategy optimization module.

[0050] The surface feature acquisition module collects building foundation surface images and spatial data, uses an industrial camera and a structured light scanning device to collect high-resolution images and scan the spatial profile of the building foundation area, and constructs a building foundation surface feature reference atlas.

[0051] The starting stage of the building foundation detection work is image collection and spatial geometric data acquisition of the building foundation area, to realize high-precision visualization and digital expression of the building foundation surface structure state. In this stage, by fusing high-resolution visual imaging equipment and structured light ranging equipment, a joint representation model of foundation surface texture and geometric features is constructed, and a feature reference atlas for subsequent structure defect recognition tasks is generated. The following steps are included:

[0052] Before the building foundation detection, the detection area where the target building foundation is located is selected by presetting the scanning area boundary of the construction site. A multi-angle high-resolution industrial camera is erected around the selected area, which has an imaging accuracy better than 0.2mm / pixel and a low-distortion correction optical lens, and can truly restore the micro-texture details of the concrete surface layer. At the same time, a structured light projection device is arranged above the imaging area, which projects a coded grating pattern to form a projection pattern with spatial coding characteristics for subsequent depth calculation. During the collection process, a constant ambient illumination control device is set to avoid the influence of natural light interference on image quality, and to ensure the stability and consistency of image texture features.

[0053] Industrial cameras that can be used for high-precision image acquisition of building foundation surfaces include the Basler ace series (such as acA5472-17um), the FLIR Blackfly S series (such as BFS-U3-200S6C-C), the Allied Vision Alvium1800 series, the Teledyne DALSA Genie Nano series, and the IDS UI series, etc. These cameras all have a resolution of 5 million pixels to over 20 million pixels, are equipped with low-noise high-dynamic CMOS sensors, support USB3.0 or GigE interfaces, can achieve sub-millimeter level image acquisition accuracy, and are suitable for capturing the fine texture and crack features of the concrete foundation surface. When used in combination with low-distortion industrial lenses, they can ensure that high-contrast, low-distortion image data is obtained at medium and close distances, meeting the high-precision requirements of surface reference atlas construction and subsequent anomaly detection.

[0054] The industrial camera and the structured light projection device are started to work synchronously, multi-angle and multi-frame image acquisition tasks are performed according to the set time interval and exposure parameters, and multi-dimensional image data covering the upper surface of the building foundation, the edge connecting surface, and the contact area with the ground are obtained. During the acquisition process, the structured light projection device projects a specific spatial coding grating pattern, and the industrial camera synchronously captures the grating deformation image reflected by the surface. By means of the geometric distortion information of the reflected fringe, the spatial profile change of the target surface is inversely calculated in real time, and a dense depth map data set is formed. The depth map can be used to describe the small concave-convex features, peeling marks, and joint changes of the concrete surface layer, and assist in subsequent topographic anomaly recognition.

[0055] The obtained image data and depth map data are fused by a multi-view stereo reconstruction algorithm, and a global bundle adjustment method is used to realize the three-dimensional registration and geometric correction of the images of the building foundation surface from different angles. Through this process, a spatial reconstruction point cloud model and a high-fidelity texture layer of the foundation area are constructed, realizing the integration and coupling of surface texture information and spatial geometric information. In order to improve the clarity of the texture expression, multi-scale edge enhancement and brightness gradient equalization processing are performed on the reconstruction results, so that in the subsequent image anomaly area detection process, the micro-features such as boundary fracture, texture discontinuity, and topographic mutation can be accurately extracted.

[0056] Based on the completed image and spatial geometry fusion results, a surface feature reference atlas in a unified coordinate system is established. This atlas covers the gray value, texture direction, spatial depth, and curvature information of each pixel point, and is aligned with the reference boundary in the original building structure design model to obtain a complete, accurate, and reference-standard building foundation surface digital expression template. This reference atlas serves as the input reference data for subsequent anomaly area positioning, thermal response analysis, and acoustic imaging, and provides a unified and standard feature expression basis for the entire building foundation detection process.

[0057] The main role of this step is to provide a high-precision, standardized, and full-scene coverage initial digital expression benchmark for the subsequent identification and analysis of the building foundation structure state. Specifically, the high-resolution images obtained by the industrial camera can record the visual details of the building foundation surface, such as gray scale, texture, and edge morphology. The structured light scanning device further provides the three-dimensional profile and micro-topography information of the building foundation surface, including flatness variation, local relief, and crack depression. The building foundation surface feature benchmark atlas constructed by combining the two can not only have accurate spatial registration capability but also express the geometric features and texture features of the structure simultaneously, thereby forming a "reference sample" for subsequent abnormal area identification. By comparing the real-time collected data with this atlas, potential abnormal areas such as texture discontinuity and morphology mutation can be effectively identified, providing a high-precision spatial positioning basis and feature calibration reference for subsequent thermal response analysis, acoustic imaging identification, and structure integrity evaluation. Therefore, this step plays a dual foundational role of "digital modeling + abnormal judgment benchmark" in the overall detection process and is a key pre-step for realizing high-precision building foundation detection closed loop.

[0058] The visual anomaly recognition module performs visual anomaly area screening processing on the surface feature benchmark atlas, identifies potential abnormal areas by calculating the brightness gradient change rate, texture boundary continuity, and geometric curvature mutation value, and outputs a spatial annotation layer.

[0059] The visual anomaly area screening processing is performed on the building foundation surface feature benchmark atlas to achieve early identification and spatial positioning of potential structural abnormal areas. Since some shallow structural degradation such as carbonization, erosion, or hollowing has not yet formed significant geometric distortion or color abnormalities on the surface, conventional visual algorithms cannot identify them. Therefore, a composite index calculation is proposed that combines the brightness gradient change rate, texture boundary continuity, and geometric curvature mutation value to significantly improve the sensitivity and accuracy of abnormal screening. The specific steps include:

[0060] After the building foundation surface feature benchmark atlas is constructed, the brightness value sequence corresponding to each pixel point is extracted, and the first and second order brightness gradients in the horizontal and vertical directions are calculated to form a brightness change rate atlas. The local maximum value and change rate gradient of the brightness change are calculated by local window convolution, and the mutation degree of the brightness change of adjacent areas is compared. If the brightness gradient change rate in a local area is significantly higher than the average value of the surrounding areas and forms a continuous edge feature, it is marked as a brightness abnormal candidate area.

[0061] Further texture boundary continuity is calculated for the candidate region, and a gray level co-occurrence matrix, a histogram of oriented gradient (HOG), and other texture description operators of each pixel point are extracted to analyze the texture direction continuity of the image at a specific scale. If the texture direction of a certain region changes abnormally, or the gray level correlation of the boundary region decreases significantly, it indicates that the region may have non-natural texture transition, i.e., potential peeling, erosion or material debonding signs. Combined with the texture direction offset vector of the spatial adjacent pixels, the discontinuity of the boundary can be further quantified, and distinguished from the normal region.

[0062] The texture description operator is a mathematical tool used in image processing and computer vision to quantify the texture pattern, structure direction, roughness or repetitive distribution characteristics in the image region. Its role is to convert the "texture information" perceptible by the human eye into calculable and comparable numerical features to realize texture discrimination in image classification, target recognition, defect detection and other tasks. In building foundation surface detection, the texture description operator can be used to identify areas with peeling, erosion, roughness and unevenness different from the normal concrete surface texture. Commonly used texture description operators include gray level co-occurrence matrix (GLCM), local binary pattern (LBP), histogram of oriented gradient (HOG), Gabor filter, etc. The extraction method usually includes: first, converting the image to a grayscale image, then in the local area (such as window or block) of the image, the spatial relationship or direction change rule between pixel values is counted to generate a description matrix or feature vector. For example, the gray level co-occurrence matrix counts the co-occurrence frequency of gray level pairs to reflect the direction and contrast information, while the LBP encodes the relative brightness of the local neighborhood to form a binary pattern, thereby capturing texture details. These texture features can be used as a key basis for subsequent image segmentation and abnormal region identification.

[0063] Further geometric curvature mutation analysis is performed on the marked abnormal region. Using the spatial profile point cloud data obtained by the structured light scanning, the continuous surface model of the building foundation surface is reconstructed, and the high-order curvature information of the surface is calculated, including the principal curvature, average curvature and Gaussian curvature distribution. By comparing the curvature smoothness and continuity features of the normal region surface, if the curvature value of a certain region changes sharply within a small scale range (such as local concave, convex or edge tearing), the region is identified as a high-risk point with potential structural deformation or material delamination. The geometric curvature mutation can be spatially aligned with the aforementioned brightness and texture abnormal information to realize consistent positioning of multi-source features.

[0064] The three characteristic indexes are superimposed to construct a comprehensive abnormal confidence atlas, and a spatial annotation layer is generated based on threshold segmentation and region growing algorithm. The layer is based on the building foundation coordinate system, and each image segment identified as abnormal is spatially coded and classified and annotated, and the output layer contains the location index of the abnormal area, the corresponding abnormal type reference value and the morphological feature abstract. This layer will serve as the positioning input for subsequent thermal response inversion and acoustic travel time imaging, effectively improving the targeting and detection efficiency of thermal excitation and acoustic scanning.

[0065] The purpose of this step is to identify and spatially locate possible structural abnormal areas through in-depth analysis of the building foundation surface feature reference atlas, providing high-confidence detection targets for subsequent thermal inversion, acoustic imaging and integrity evaluation. Specifically, because the deterioration of the building foundation surface (such as shallow hollowing, erosion, carbonization) often lacks significant geometric distortion or color difference changes in the early stages of imaging, conventional image recognition methods are difficult to accurately detect. Therefore, this step introduces three types of visual analysis indicators: first, the brightness gradient change rate is used to identify local brightness change abnormalities and capture subtle gray transition changes caused by material density or reflectivity differences; second, the texture boundary continuity is quantified by a texture description operator to identify possible locations of peeling, interface delamination or roughness abnormalities; third, the geometric curvature mutation value is obtained by local surface fitting and curvature analysis of three-dimensional data obtained by structure light scanning to find non-continuous topographic features such as surface depressions, protrusions and crack edges. The spatial anomaly index system formed by the three can focus and screen potential structural hazards at the visual level and generate a spatial annotation layer that digitally annotates the location, shape and confidence of each high-risk area, providing input guidance for the multi-modal detection system. This step effectively realizes the transition from "full-image recognition" to "local key detection", improving the efficiency and accuracy of subsequent detection, and is a key pre-determination step in the entire structural health assessment process.

[0066] The thermal response analysis module performs thermal response inversion processing based on the spatial annotation layer, applies low-frequency infrared excitation to the abnormal area, records the thermal diffusion path and surface temperature rise amplitude change, identifies the thermal conduction lag area, and constructs a thermal inertia indicator atlas;

[0067] The purpose of performing thermal response inversion processing based on the spatial annotation layer is to identify changes in thermal conductivity caused by a decrease in density, an increase in porosity, or interface debonding in the material by actively exciting the abnormal area of the building foundation surface and recording the thermal diffusion process, thereby realizing non-contact quantitative analysis of shallow structural defects. By combining infrared thermal imaging and low-frequency periodic excitation, a reproducible, high-resolution thermal diffusion behavior measurement process is constructed, and a thermal inertia indicator map with deep response capability is finally output. Specifically, the following steps are included:

[0068] According to the spatial annotation layer generated in the previous stage, the specific location and range of the abnormal area to which the thermal excitation needs to be applied are determined. Through the two-dimensional coordinate index provided in the layer and the three-dimensional registration model of the building foundation surface, the space is back projected to physically calibrate the detection area on the real surface. The thermal excitation scope is set at the center point of each abnormal area, and the heating time, frequency, and power parameters are set in combination with the area size to ensure that the excitation range completely covers the target to be detected, while not causing thermal interference to the surrounding structure. In the selection of the infrared excitation source, a resistive film heater with a surface array heating characteristic and adjustable response time is used to ensure that the excitation frequency is controlled in the low-frequency range of 0.05-0.5 Hz, which is beneficial to the detection of the dynamic behavior of thermal diffusion.

[0069] Start the thermal excitation device and apply thermal input to the selected area within the set excitation period, while real-time acquisition of temperature distribution image sequences is performed by a high-sensitivity infrared imaging device at a rate of 10 frames / second or higher. During the acquisition process, the entire process from the heating start time to the natural diffusion equilibrium of heat is recorded, covering multiple stages such as thermal front advance, heat distribution homogenization, and cooling decay. On the data processing side, the temperature rise curve is extracted to calculate the initial response time, maximum temperature rise value, and heat balance time of each pixel point, thereby constructing a set of time-domain characteristic parameters reflecting the thermal conductivity of the material in the region.

[0070] Based on the collected time-temperature data sequence, the Fourier transform infrared thermal imaging analysis method is used to deconstruct the temperature rise response in the frequency domain to identify the response phase and amplitude characteristics of each pixel point to the excitation frequency. Through phase delay calculation, the thermal conduction inertia index of the region can be obtained, where the more delayed the phase is, the weaker the thermal diffusion capability of the position is, which is usually related to structural voids, pores, carbonization, or erosion layers. In order to eliminate the interference of non-structural factors such as surface texture and reflectivity, the results are normalized with a reference map to improve the structural correlation and discrimination ability of the thermal conduction parameters.

[0071] Fourier Transform Infrared Thermography (FTIR or FTT) is a non-destructive testing method that combines infrared thermography with periodic thermal excitation to reveal the thermal response characteristics of materials through frequency domain analysis. The basic principle is to continuously record the infrared image sequence of the target area during the process of applying low-frequency periodic thermal excitation (such as sinusoidal heating), and then perform Fourier transform on the temperature waveform of each pixel point in the time sequence to convert the thermal response from the time domain to the frequency domain, and extract the amplitude and phase information of each frequency component. Among them, the phase angle is not sensitive to the reflectivity, texture and other interference of the material surface, and can more stably reflect the thermal diffusion characteristics of the material internal; while the amplitude component can represent the heat transfer rate and absorption intensity. This method can be used to identify the thermal inertia changes caused by voids, debonding, delamination, carbonization, etc., and is especially suitable for identifying the deteriorated areas that are difficult to observe directly from the surface in the shallow structure of building foundation. Through frequency domain feature analysis, Fourier transform infrared thermography can effectively improve the signal-to-noise ratio of defect identification and reduce the influence of surface texture interference, with high sensitivity and robustness, and is suitable for deep thermal response analysis and precise positioning of potential abnormal areas in building foundation detection.

[0072] The phase lag map obtained by analysis and the maximum temperature rise distribution map are weighted and fused to construct a complete thermal inertia indicator atlas, and mapped back to the unified coordinate system of the building foundation. The atlas expresses the thermal diffusion lag level of each pixel point in a color-coded manner, clearly identifies the spatial distribution and boundary profile of the thermal response abnormal area, and has the functions of visualization, quantification and spatial alignment. The thermal inertia indicator atlas will be used as the positioning input for the next stage of acoustic time difference imaging processing, significantly improving the effective focusing rate of acoustic excitation and the efficiency of structural abnormal linkage identification.

[0073] The purpose of this step is to further reveal the differences in thermal conductivity of the shallow structure below the building foundation surface by implementing directional thermal excitation on the identified abnormal area and observing its thermal diffusion behavior, thereby identifying possible voids, delamination, carbonization or other material degradation areas. Since many shallow structure defects do not exhibit obvious abnormalities in visual images and three-dimensional profiles, changes in their physical properties, especially a decrease in thermal diffusion capacity, become an important basis for judging structural integrity. Therefore, this step applies low-frequency periodic infrared excitation to the high-suspected area located in the spatial annotation layer, and simultaneously uses a high-sensitivity infrared thermal imaging device to record the surface temperature rise over time, obtaining multi-dimensional information such as thermal diffusion path, temperature rise rate, and heat decay curve. Combined with the Fourier transform infrared thermal imaging analysis method, the thermal response data is converted from the time domain to the frequency domain, further extracting the response amplitude and phase characteristics of each pixel. Compared to traditional thermal imaging, which only focuses on surface temperature distribution, this method focuses more on the dynamic process of temperature change and deep propagation characteristics, especially the thermal inertia of thermal diffusion, which is particularly sensitive to issues such as reduced density, delamination of adhesive layers, or increased porosity in structures. The final thermal inertia indicator map visually marks each pixel's thermal response capacity in the image with color coding, providing a clear spatial risk distribution map that provides a depth of structural reference in the thermal dimension for subsequent acoustic detection and foundation reinforcement strategies. This step realizes the cross-modal recognition upgrade from surface visual judgment to thermal physical property analysis, and is a key intermediate link for non-contact detection of shallow structure.

[0074] The acoustic imaging detection module performs acoustic time difference imaging processing based on the thermal inertia indicator map, applies directional low-frequency acoustic wave pulses to the thermal conduction lag area, collects acoustic reflection echo delay distribution data, identifies acoustic velocity mutation points, and constructs an acoustic-thermal cooperative abnormal positioning map.

[0075] The purpose of performing acoustic time difference imaging processing based on the thermal inertia indicator map is to use the thermal conduction lag areas identified in the thermal inversion stage as the focus of acoustic detection, by applying directional low-frequency acoustic excitation to these areas and collecting the time delay characteristics of their reflection echoes, to further identify physical abnormalities such as delamination, cavities, debonding, and low-density areas in the shallow structure of the concrete foundation, achieving acoustic dimension verification and reinforcement of the internal integrity of the structure. This method is based on multi-physical field fusion and couples thermal information with acoustic propagation characteristics for analysis, enabling three-dimensional imaging of the structure below the surface without destruction. Specifically, the following steps are included:

[0076] According to the thermal inertia indication atlas marked in the thermal conduction lag area, the action range of the sound wave excitation is determined. Through the pixel space coordinate conversion in the atlas to the actual detection area of the building foundation, and the establishment of the sound wave excitation point array consistent with the three-dimensional coordinate system. In each thermal lag area, the sound wave transmitter and receiver pair are arranged to form a directional excitation path for the area. In order to ensure the effective propagation of sound waves in concrete structures, the sound wave excitation source selects low frequency (such as 10-50 kHz) pulse waveform, which is lower than the conventional ultrasonic detection frequency, so as to improve the penetration depth and echo signal strength in the coarse aggregate concrete medium.

[0077] In the sound wave excitation stage, the control signal source applies pulse excitation strictly defined by time window to the target area, and the synchronous receiving device records the reflected wave, scattered wave and multiple reflection signals in real time at high sampling rate. In the concrete medium, different density or material boundary will cause the change of sound velocity and the change of sound wave reflection path, which shows the difference of echo delay time. The difference between the arrival time of the first reflection wave and the excitation time is analyzed in the collected data, and the reflection echo delay distribution diagram is generated, which records the sound wave propagation path and its time domain characteristics of each detection point.

[0078] For the collected reflection delay data, the sound wave propagation model is used for inversion analysis to calculate the spatial distribution of sound velocity change in the target area. By comparing the sound wave transmission time under ideal state with the actual measured time, the position and boundary form of the sound velocity mutation point are derived. The sound velocity mutation is common in interface layering, hollow or material heterogeneity area, and the sound wave propagation speed in these areas is obviously lower than that in healthy concrete, and the reflection energy is stronger. In order to improve the identification accuracy, statistical filtering and multi-path signal reconstruction algorithm are introduced to eliminate the interference of pseudo reflection signals caused by surface roughness, boundary diffraction and other factors.

[0079] The sound wave propagation model is a physical and mathematical model that describes the behavior of sound waves in a specific medium. It is used to characterize the propagation speed, path, attenuation characteristics, and reflection and refraction behavior of sound waves, which change with space, material properties, and structural boundaries. In building foundation testing, the sound wave propagation model establishes the propagation relationship between the sound wave excitation source and the receiving point. By analyzing the propagation time delay and waveform changes of sound waves in different materials such as concrete, pores, and layers, the physical state and spatial distribution characteristics of the structure inside can be inferred. Specifically, the inversion analysis process includes: first, based on the known sound velocity parameters of the material, the propagation path and time distribution under ideal conditions are constructed; then, the actual recorded sound wave reflection time is compared with the model prediction value, and the delay deviation is analyzed; finally, through inversion algorithms such as time difference tomography, finite element simulation, and wave velocity imaging inversion, the sound velocity distribution map that makes the prediction result close to the measured data is solved, so as to identify the location and range of abnormal structures. This method can non-destructively detect the sound velocity mutation caused by the decrease of density, the existence of voids or the interface debonding in building foundation, and is an important theoretical support means for quantitative analysis and positioning diagnosis of shallow structure state.

[0080] The spatial position of the sound velocity mutation point is aligned and superimposed with the thermal inertia indication map to construct a sound-heat cooperative abnormal positioning map. This map integrates the thermal diffusion lag area and the sound velocity abnormal area, forming a unified coordinate system for abnormal area visualization expression layer, with color coding, boundary encapsulation, and confidence level identification functions. The sound-heat cooperative abnormal positioning map not only improves the spatial accuracy of structure abnormal identification, but also provides clear avoidance basis for subsequent structure integrity evaluation and foundation installation path planning, ensuring the accuracy and safety of foundation installation operation.

[0081] The purpose of this step is to further identify the location and physical characteristics of shallow structural defects below the building foundation surface by implementing directional acoustic excitation in the thermal inertia indication map marked thermal conduction lag area, and to achieve acoustic dimension verification and spatial fine positioning of potential cavities, delamination, debonding and other deterioration areas. Thermal inertia map can only reveal the abnormal behavior of regional heat diffusion, but cannot accurately determine the depth, boundary profile and physical nature of the defect, so it is necessary to introduce acoustic travel time imaging technology to capture the sound speed mutation point in the structure by using the speed difference and reflection characteristics of sound waves in different structural media, and to identify material interfaces or integrity abnormal areas. In this step, the acoustic excitation source acts on the selected area in the form of low-frequency pulse to ensure strong penetration and echo response capability, while recording the delay distribution data of the acoustic reflection echo simultaneously. By comparing the acoustic propagation time and path difference with the theoretical propagation model, the sound speed variation information is extracted, and the local sound speed field distribution map is constructed by combining the inversion algorithm to identify internal density mutations, voids or crack channels and other structural abnormalities. Finally, the obtained sound speed anomaly distribution and the original thermal inertia map are aligned and fused in coordinates to generate a sound-heat collaborative anomaly positioning map, realizing multi-dimensional mapping and reinforcement of thermal perception and acoustic inversion in the same spatial framework. This collaborative map not only provides high-confidence, multi-physical-source-supported positioning results for internal defects, but also provides a quantifiable, visual and traceable data basis for subsequent structural integrity assessment and installation point avoidance strategy formulation, which is an important intermediate link for accurate identification of shallow hidden dangers and construction safety assurance.

[0082] The structural integrity assessment module activates the regional integrity discrimination process based on the sound-heat collaborative anomaly positioning map, counts the number and spatial distribution of overlapping areas in acoustic and thermal responses of abnormal segments, and outputs the structural health index of each detection unit to realize quantitative classification of potential hazard areas.

[0083] The purpose of performing the regional integrity discrimination process based on the sound-heat collaborative anomaly positioning map is to perform multi-dimensional analysis and quantitative evaluation of the shallow structure state below the building foundation surface based on the fusion of thermal and acoustic identification results, so as to realize the classification and risk grading of potential hazard areas and provide visual and decision-making basis for subsequent structural installation layout strategy. This step analyzes the spatial overlap of acoustic reflection delay information and thermal diffusion lag area, calculates the structural health index of each detection area based on the density and continuity of abnormal responses, and uses it for integrity grading and visual expression. This process includes the following steps:

[0084] According to the acoustic-thermal synergistic anomaly positioning map, the building foundation to be detected area is regularly divided, and a unified spatial detection grid is established. Each detection unit corresponds to a fixed area on the building foundation, has a clear spatial boundary and a unique identifier, and can accurately correspond to the thermal and acoustic detection results. In the division process, considering the scale characteristics of the changes in the details of the building foundation surface, it is recommended to control the scale of the detection unit within tens of millimeters to accurately capture small structural anomalies.

[0085] Extract the image data related to acoustic and thermal anomalies in each detection unit and count the internal anomaly feature performance. In the thermal dimension, identify areas with temperature rise delay, slow heat diffusion, or phase response anomalies, and in the acoustic dimension, identify areas with abnormal sound wave reflection time, energy echo enhancement, or significant propagation speed anomalies. Then, count the degree of overlap of the areas with anomalies in the two dimensions in space, including the number of abnormal pixels, spatial distribution continuity, and the proportion of abnormal areas. If a detection unit shows high-intensity anomalies in both thermal and acoustic responses and overlaps, it is judged to be a region with low structural integrity.

[0086] Based on the statistical data of multiple anomalies, the structural health status of each detection unit is comprehensively analyzed. Units with high health usually show low acoustic-thermal response consistency, small abnormal coverage area, and clear boundaries; while low health regions often show high acoustic-thermal anomaly overlap, high response amplitude, blurred boundaries, and strong spatial continuity. By comparing the response performance of all detection units in the thermal and acoustic dimensions, each unit is assigned a structural health level and classified into high integrity, medium integrity, or low integrity.

[0087] Map the structural health level to the three-dimensional model of the building foundation to form a structural health level map and visualize it through color zoning. The health level map not only shows the spatial position and structural state of each detection unit, but also can be connected with the subsequent construction process, providing accurate avoidance basis for elevator foundation installation path planning, support point arrangement optimization, etc. High integrity areas can be used as priority points, medium integrity areas may need to be reinforced, and low integrity areas should be completely avoided to ensure the structural safety and long-term operation stability of the elevator after installation.

[0088] The purpose of this step is to perform multi-source data fusion discrimination on the identified potential abnormal area, quantify the area structure state from two dimensions of acoustics and thermodynamics, output the structure health index, and thus realize the risk level classification of the shallow structure of the building foundation. The thermal response reflects the hysteresis of the material to thermal diffusion, which can reveal the decline in heat conduction ability caused by hollowing, debonding, and increased porosity; the acoustic response reflects the change in material density and interface state, revealing delamination, cracks, or density mutation phenomena through sound wave reflection delay or energy change. However, a single physical channel is often limited by material properties, environmental interference, or response sensitivity, and may have misjudgment or missed judgment risks. Therefore, this step analyzes the overlap degree, distribution density, and spatial continuity of acoustic and thermal abnormal areas in the same spatial unit to comprehensively judge the structural integrity of the area. In specific operations, the detection area is divided into equal-scale detection units, the thermal response hysteresis area and the acoustic velocity mutation area in each unit are statistically compared, the overlapping area that simultaneously exhibits abnormal characteristics is identified, the confidence of the structural abnormality is judged, and the structure health index is assigned accordingly. The health index, as a continuous index, quantitatively reflects the integrity level of each detection unit, and the lower the value, the higher the potential risk of the structure. The final output of the structure health index map can be used to mark different levels of hidden danger areas and as key input data for subsequent foundation layout, support point positioning and installation avoidance strategies to ensure accurate, safe and reliable construction. Therefore, this step not only improves the accuracy and stability of abnormal identification, but also provides a quantitative structure safety evaluation basis for engineering implementation, which is a key link to realize structure state closed-loop perception and intelligent decision-making.

[0089] The installation path optimization and risk avoidance module relies on the structure health index results to optimize the foundation installation scheme, dynamically adjusts the foundation module layout position and support column landing point, and makes the support system actively avoid high-risk hidden danger areas, thereby ensuring the overall installation safety and long-term operation stability, forming a complete closed-loop process of identification, early warning, and avoidance.

[0090] The structure health index results are used to optimize the building foundation installation scheme, which aims to directly convert the shallow structure risk identification results into decision-making basis for construction layout strategy, actively avoid high-risk areas, and thus improve the safety, durability, and long-term stability of the foundation structure during elevator installation. This step is based on the structure health index map obtained by multi-source data fusion in the early stage, dynamically adjusts the spatial layout, landing position, and connection method of the support component, and constructs a closed-loop control process composed of "identification, early warning, and avoidance". The process includes the following steps:

[0091] According to the generated structural health index map, the building foundation area is spatially divided, and the entire area within the construction range is marked according to the health level of the detection unit. The lower the health index, the higher the probability of structural abnormalities in the detection unit, usually manifested as areas with severe acoustic reflection delay, significant thermal diffusion lag, or high coincidence of the two. Through the visual expression of the health level, the foundation construction area is divided into three spatial tags: preferred area, warning area, and avoidance area, and a construction safety map with spatial hierarchy is constructed, providing a clear decision-making reference for subsequent landing point selection.

[0092] In the foundation layout planning stage, according to the requirements of the elevator installation structure for the number, spacing, and mechanical transmission path of the bearing points, a set of feasible foundation layout scheme candidates is set. In each candidate scheme, the spatial relationship between the proposed support points and the identified high-risk areas is analyzed, and the landing point combination in the area with higher health level is preferentially selected, and the layout points with structural deterioration risk are excluded. If necessary, geometric adjustment is made to the overall layout scheme, and the avoidance ability of high-risk areas is improved by increasing or decreasing the number of support points, optimizing the layout spacing, or changing the direction of connecting members, etc., to ensure that each support point has good structural integrity.

[0093] After the layout optimization is completed, the support point landing simulation and mechanical response simulation process is started. Combined with the three-dimensional structure model of the building foundation, the structural mechanics analysis of the adjusted support point scheme is carried out to verify its stress distribution, deformation response, and load redundancy under static and dynamic loads. For areas with medium health level, the structural response changes under different reinforcement strategies (such as expanded foundation, vibration isolation pad, etc.) can be simulated to evaluate whether they can be included in the construction range. Through simulation feedback, the landing point scheme is further corrected to improve the stability and construction feasibility of the overall installation structure.

[0094] The optimized landing point layout is integrated with the construction drawings to output the final construction layout reference map and structural health response map. During construction, construction personnel can verify the layout position on site according to this layer information to ensure that the actual landing point is consistent with the planned area, and the warning area should start manual review or further detection measures. This process not only ensures construction accuracy and structural safety, but also provides a traceability basis for post-construction monitoring and maintenance, realizes the direct driving of detection results on construction behavior, and constructs a closed-loop structural optimization process from identification, analysis, evaluation to avoidance control.

[0095] The role of this step is to directly convert the structural health index results obtained based on multi-physical sensing means (such as thermal response analysis and acoustic time difference imaging) in the early stage into the core basis for support member layout and landing point decision-making during building foundation installation, thereby realizing active avoidance of potential risk areas of shallow structures and optimization of construction paths. Traditional building foundation installation often relies on the experience of construction personnel or static structural drawings for layout, making it difficult to consider the inhomogeneity and local deterioration of the foundation in real time, which can easily lead to the landing of support points in areas with hidden problems such as cavities, stratification, and debonding, thereby causing uneven bearing, settlement misplacement, and even foundation cracking and other safety risks. This step analyzes the spatial information of the structural health index map, divides the foundation construction area into three regions of high integrity, medium integrity, and low integrity, and dynamically adjusts the layout position and landing path of the support member accordingly. On this basis, each candidate support point is evaluated for integrity matching, with high-health-index areas being preferred as bearing nodes, while low-health-index areas are avoided or reinforced, ensuring that the force transmission path is within a stable and structurally safe area. In addition, this step can be linked with subsequent structural simulation analysis to verify the stress and evaluate the response of the optimized layout scheme, improving the reliability and engineering feasibility of the decision-making. Ultimately, with this structure-state-driven layout adjustment mechanism, a logical closed loop is achieved from "identifying potential problems" to "dynamic construction decision-making", which not only reduces the probability of construction safety accidents caused by misjudgment of hidden problems, but also improves the operational stability and service life of the elevator foundation system, with significant engineering practical value and intelligent construction guiding significance.

[0096] Through the above-mentioned industrial vision-based building foundation detection system for elevator installation, multi-dimensional perception and depth analysis of the building foundation surface and its shallow structure state can be achieved, breaking through the limitations of traditional methods that rely only on image features or point cloud contours for judgment, effectively improving the identification ability of early carbonization, shallow cavities, and erosion layers. The system constructs a full-process diagnosis link from abnormal area identification, physical response verification to structural health quantitative evaluation through the linkage processing mechanism of surface feature modeling, thermal response analysis, and acoustic imaging detection, significantly improving the detection accuracy and identification robustness. At the same time, combined with the quantitative output results of the structural health index, the system can actively adjust the layout position and installation path of the foundation member to avoid high-risk hidden trouble areas, realizing closed-loop control from "defect identification" to "risk avoidance". This system not only improves the intelligent level of building foundation detection and installation safety, but also provides a scientific, reliable, and engineering-based decision-making basis for elevator installation in old residential buildings, with significant technical promotion value and engineering application prospects.

[0097] The above formulas are all de-dimensioned to calculate the numerical values, the formulas are obtained by collecting a large amount of data to simulate a formula of the most recent real situation, and preset parameters in the formulas are set by a person skilled in the art according to actual conditions.

[0098] The above merely describes certain exemplary embodiments of the present application by way of illustration, and it is needless to say that those skilled in the art can modify the described embodiments in various manners without departing from the spirit and scope of the present application. Therefore, the above drawings and descriptions are merely illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present application.

[0099] It should be noted that, in this document, relational terms such as first and second and the like can be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises a... " does not, without more constraints, exclude the existence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0100] It should be understood that, in various embodiments of the present application, the size of the sequence number of each process described above does not mean the order of execution, and the execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0101] Those skilled in the art can appreciate that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0102] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system, device and unit described above can refer to the corresponding process in the foregoing method embodiments, which will not be described here.

[0103] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Part or all of the units can be selected to achieve the purpose of the embodiment scheme according to actual needs.

[0104] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit.

[0105] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited to this. Any skilled person in the art can easily think of changes or replacements within the technical range disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0106] The above only describes some exemplary embodiments of the present application by way of illustration, and it is needless to say that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present application. Therefore, the above figures and descriptions are illustrative in nature and should not be understood as limiting the scope of protection of the claims of the present application.

Claims

1. An industrial vision-based building foundation detection system for elevator retrofits, characterized by The surface feature acquisition module, the visual anomaly recognition module, the thermal response analysis module, the acoustic imaging detection module, the structural integrity evaluation module, and the installation path optimization and risk avoidance module are included. The surface feature acquisition module collects building foundation surface images and spatial data, uses industrial cameras and structured light scanning devices for high-resolution image acquisition and spatial contour scanning, and constructs a surface feature reference atlas. The visual anomaly recognition module performs abnormal area screening on the reference atlas, calculates the brightness gradient change rate, texture boundary continuity, and geometric curvature mutation value, and outputs a spatial annotation layer. The thermal response analysis module performs thermal response inversion processing based on the spatial annotation layer, applies low-frequency infrared excitation, records the thermal diffusion path and temperature rise amplitude change, and constructs a thermal inertia indication atlas. The acoustic imaging detection module performs acoustic time difference imaging processing based on the thermal inertia indication atlas, applies directional low-frequency acoustic pulses, collects reflected echo delay data, and constructs an acoustic-thermal cooperative anomaly positioning map. The structural integrity evaluation module performs integrity discrimination based on the acoustic-thermal cooperative anomaly positioning map, counts the number of overlapping areas and spatial distribution of acoustic-thermal anomalies, outputs a structural health index, and realizes quantitative classification. The installation path optimization and risk avoidance module optimizes the installation scheme based on the structural health index results, adjusts the foundation layout position and support column landing point, avoids high-risk areas, and forms a closed-loop process of identification, early warning, and avoidance. The thermal response analysis steps are as follows: According to the spatial annotation layer, the abnormal area position is calibrated, the excitation parameters are set, and low-frequency infrared thermal excitation is applied. The temperature rise image sequence during the thermal diffusion process of the abnormal area is collected. Based on the temperature rise image sequence, the pixel temperature response curve is extracted, the initial response time, maximum temperature rise value, and thermal equilibrium time are calculated, and a set of thermal diffusion time characteristic parameters is generated. Perform Fourier transform on the temperature response sequence, extract the frequency domain phase and amplitude characteristics of each pixel, and construct a thermal conduction lag index map. Fuse the phase lag map and temperature rise distribution map to generate a thermal inertia indication atlas and map it to the building foundation unified coordinate system. The acoustic imaging detection steps are as follows: According to the thermal inertia indication atlas, the thermal conduction lag area position is calibrated, the acoustic excitation path is set, and low-frequency acoustic pulse excitation is applied. Collect acoustic reflection echo delay data, generate a reflection echo delay distribution map, record the acoustic propagation path and time domain characteristics. Based on the acoustic propagation model, the reflection data is analyzed by inversion, and the position and boundary form of the acoustic velocity mutation point are derived. Align the coordinates of the acoustic velocity mutation point and the thermal conduction lag area and fuse the annotations to construct an acoustic-thermal cooperative anomaly positioning map.

2. The building foundation detection system for elevator installation based on industrial vision according to claim 1, characterized in that, The surface feature acquisition steps are as follows: By setting a multi-angle high-resolution industrial camera and a structured light projection device, high-precision images and structured light stripe reflection images of the building foundation surface are collected. Start the industrial camera and structured light device synchronous working mechanism, perform multi-angle multi-frame image acquisition, and obtain the texture image and spatial coding pattern of the target area. Based on the image and depth map data, perform three-dimensional reconstruction processing to construct a spatial reconstruction point cloud model and a high-fidelity texture layer. A unified coordinate system is established to build a model of the atlas, which integrates the gray scale, texture direction, spatial depth and curvature information of each pixel to form a surface feature reference atlas.

3. The building foundation detection system for elevator installation based on industrial vision according to claim 1, characterized in that, The visual anomaly recognition procedure is as follows: The brightness value of each pixel in the building foundation surface feature reference atlas is extracted, and the brightness gradient change rate in the horizontal and vertical directions is calculated to generate a brightness change rate atlas and mark the brightness abnormal candidate region; The gray level co-occurrence matrix and direction gradient histogram are calculated in the candidate region to analyze the texture direction continuity and mark the texture boundary discontinuity region; The surface principal curvature and Gaussian curvature are calculated based on the spatial contour point cloud data to identify the curvature mutation region as a potential structural anomaly point; The brightness, texture and curvature abnormal regions are fused to construct an abnormal confidence atlas and generate a spatial annotation layer.

4. The building foundation detection system for elevator installation based on industrial vision according to claim 1, characterized in that, The structural integrity evaluation procedure is as follows: The area covered by the acoustic-thermal cooperative anomaly positioning map is divided into detection units with spatial boundaries and unique identifiers; The acoustic wave reflection delay and thermal diffusion lag response data in each detection unit are extracted, and the overlap degree and spatial continuity of the abnormal region are calculated; According to the intensity, coincidence rate and distribution characteristics of the abnormal response, the structure health index is calculated and the health level is assigned; The structure health level is mapped to the building foundation three-dimensional model to form a structure health level map and perform color identification expression.

5. The building foundation detection system for elevator installation based on industrial vision according to claim 1, characterized in that, The installation path optimization and risk avoidance procedure is as follows: Based on the structure health index map, the building foundation area is spatially partitioned and marked as an optimal zone, a warning zone and an avoidance zone; According to the structure layout requirements, candidate layout schemes are set up, the support point positions in the avoidance zone are eliminated, and the landing point combination is optimized; The optimized support point layout scheme is subjected to structural mechanics simulation analysis to verify that the stress distribution and deformation response meet the bearing requirements; The final layout scheme and health response layer are integrated to generate a construction layout reference map, which is used for construction landing point verification and risk avoidance guidance.

Images