CN-122023395-A - Curtain wall connecting piece detection early warning method based on image analysis

CN122023395ACN 122023395 ACN122023395 ACN 122023395ACN-122023395-A

Abstract

S01, deploying high-definition camera equipment at a preset detection point position to acquire N connection piece setting attention areas detected in a single high-definition camera equipment acquisition picture so as to obtain a first data set in a complete two-dimensional coordinate; S02, extracting real-time images of each frame under the same time stamp based on the obtained first data set, searching and matching in the neighborhood of each concerned region by adopting a multi-feature fusion tracking algorithm, obtaining pixel coordinates of each connecting piece in the current image, calculating pixel displacement vectors and comprehensive displacement scalar of the connecting pieces relative to the reference coordinates, and S03, executing image global registration correction on the displacement data, wherein the image global registration correction comprises feature points. The invention realizes accurate monitoring and active early warning of the displacement of the curtain wall connecting piece.

Inventors

SHEN ZIZHEN
LU WENHAO

Assignees

浙江建设职业技术学院
浙江大合检测有限公司

Dates

Publication Date: 20260512
Application Date: 20260410

Claims (9)

1. The curtain wall connecting piece detection early warning method based on image analysis is characterized by comprising the following steps of: S01, deploying high-definition camera equipment at a preset detection point position to acquire N connection pieces detected in an acquisition picture of the high-definition camera equipment to set a concerned region so as to obtain a first data set in a complete two-dimensional coordinate; S02, extracting real-time images of each frame under the same time stamp based on the acquired first data set, searching and matching in the neighborhood of each concerned region by adopting a multi-feature fusion tracking algorithm, acquiring pixel coordinates of each connecting piece in the current image, and calculating a pixel displacement vector and a comprehensive displacement scalar of each connecting piece relative to a reference coordinate; s03, performing cleaning including image global registration correction based on feature points, outlier filtering based on statistics and data validity verification based on tracking confidence on the displacement data to obtain a cleaned second data set; S04, recording continuous points of the second data under continuous time, carrying out trend analysis and correlation analysis with environmental parameters to obtain displacement time sequence data, and finally calculating a global risk index Rg and a regional risk index Rz, wherein Rg is the proportion of the number M of connecting pieces with the displacement exceeding a dynamic safety threshold to the total monitoring points N; s05, matching the obtained Rg with a preset value, and preparing early warning information based on a matching result, wherein the early warning information comprises risk area positioning, overrun connector detail list and visual risk thermodynamic diagram.
2. The method for detecting and early warning the curtain wall connecting piece based on the image analysis according to claim 1, wherein in the step S02, the search matching is performed in the neighborhood of each region of interest by adopting a multi-feature fusion tracking algorithm, and the method comprises the following steps: S21, quick response is carried out in a prediction search area by adopting a discriminant correlation filter based on the directional gradient histogram characteristics, and an initial displacement vector delta Pi and a response peak confidence Cd thereof are obtained; s22, in the vicinity of the initial position, BRISK characteristic points with rotation invariance and partial scale invariance are extracted and matched with BRISK characteristic descriptors stored in a reference template, and PROSAC algorithm is utilized to robustly estimate affine transformation parameters of sub-pixel level, so that an accurate displacement vector delta Pi' and a matching interior point proportion confidence coefficient Cf are obtained; S23, calculating a dense light flow field in the attention area of the connecting piece, if the included angle between the average direction of the light flow vector and the direction of delta Pi' is smaller than 15 degrees and the consistency coefficient of the light flow direction is larger than 0.8, checking the light flow, otherwise, regarding the light flow as potential interference; S24, determining final displacement through a weighted fusion formula ΔPit=ω1.ΔPi+ω2.ΔPi', wherein ω1 and ω2 are weights and are dynamically distributed according to Cd, cf and an optical flow verification result, and ω1+ω2=1, and when the optical flow verification is not passed, ω1 is forcedly set to zero to depend on a feature matching result.
3. The curtain wall connecting piece detection and early warning method based on image analysis according to claim 1, wherein, In the step S04, the recording of the continuous point location of the second data under continuous time is performed, and the trend analysis and the correlation analysis with the environmental parameters are performed to obtain displacement time series data, which includes the following steps: S41, acquiring displacement data samples under the working condition of no wind or breeze and stable temperature, and calculating the integral standard deviation sigma g of the displacement amplitudes of all the connecting pieces; S42, converting the physical safety displacement limit value allowed by engineering specifications into a pixel reference value Db according to calibration parameters of an imaging system, and calculating a dynamic threshold value according to a formula Dd (T) =alpha.Db+beta.sigma g+gamma.delta T (T), wherein alpha and beta are weight coefficients set according to structural safety levels, delta T (T) is a difference value of the current temperature relative to the temperature at the time of reference image acquisition, gamma is a temperature compensation coefficient, and the dynamic threshold value is determined through linear regression analysis of displacement and temperature in historical data.
4. The method for detecting and early warning curtain wall connectors based on image analysis according to claim 1, wherein the step S04 further comprises performing spatial clustering analysis based on density on the connectors with the identified overrun, and the method comprises the following steps: S43, clustering all the connecting pieces with overrun displacement on the projection coordinates of the two-dimensional plane of the curtain wall by adopting a DBSCAN algorithm, wherein the neighborhood radius epsilon in the algorithm is adaptively set according to the physical size of the curtain wall plate, and the minimum point number MinPts is set to be 3; s44, each cluster identified is regarded as a local risk area.
5. The method for detecting and pre-warning the curtain wall connecting piece based on the image analysis according to claim 4, wherein the step S05 of matching the obtained Rg with a preset value comprises the following steps: S51, triggering II-level early warning when any local risk area is identified or Rz of any predefined logic area continuously exceeds 20%, and highlighting the risk area in a report for positioning; And S51, triggering a III-level alarm when two continuous monitoring periods of the global risk index Rg exceed 30 percent or two or more local risk areas which are not adjacent in space are identified, wherein the early warning information comprises an overrun connector detail list and a visual risk thermodynamic diagram.
6. The method for detecting and early warning curtain wall connectors based on image analysis according to claim 1, wherein the image global registration correction based on the feature points in the step S03 comprises the steps of extracting SURF feature points from a current frame and a reference frame, performing preliminary matching, then performing iterative screening by using a RANSAC algorithm to solve an optimal homography transformation matrix H, and performing inverse transformation on the calculated current coordinates of all connectors by using the H to correct global pixel offset caused by micro displacement or jitter of a camera.
7. The method for detecting and early warning curtain wall connectors based on image analysis according to claim 1, wherein the step S04 further comprises modeling a displacement time sequence of each connector after cleaning by applying a seasonal autoregressive integral moving average model to predict a displacement trend in a specific future period; if the predicted displacement value exceeds the dynamic safety threshold value Dd, a trend early warning notice is sent out in advance to prompt attention to the potential development risk of the connecting piece even if the current actually measured displacement is not exceeded.
8. A non-transitory computer readable storage medium having at least one instruction or at least one program stored therein, wherein the at least one instruction or the at least one program is loaded and executed by a processor to implement the method of curtain wall connector detection pre-warning method based on image analysis of any one of claims 1-7.
9. An electronic device comprising a processor and a memory, wherein the memory stores at least one instruction or at least one program, the at least one instruction or the at least one program loaded and executed by the processor to implement the method for image analysis based curtain wall connector detection pre-warning method of any one of claims 1-7.

Description

Curtain wall connecting piece detection early warning method based on image analysis Technical Field The invention relates to the technical field of computer processing, in particular to a curtain wall connecting piece detection and early warning method based on image analysis. Background The building curtain wall is used as a main stream peripheral protection structure of a modern high-rise building, and the safety of the building curtain wall is important. The curtain wall panels are connected to the main structure by a plurality of metal connectors (e.g., bolts, backbolts, hangers, etc.). Under the coupling effect of environmental factors such as long-term wind load, temperature stress, material aging, the connecting piece is liable to occur accumulative loosening and slipping, so that the panel is at risk of falling off, and public safety is seriously threatened. At present, the state monitoring of curtain wall connecting pieces mainly depends on manual regular inspection, and the method has low efficiency, high cost, strong subjectivity and incapability of realizing real-time early warning, and is difficult to meet the daily safety management requirement of a large-area curtain wall. With the development of computer vision technology, some image-based non-contact monitoring methods have emerged. For example, digital Image Correlation (DIC) is used to monitor global deformation, or target detection algorithms are used to identify panel cracks. However, these methods have the following limitations: Most concern about overall deformation or panel damage, and lack of special and refined monitoring of the key force transmission node of the connecting piece; the existing vision algorithm has insufficient stability under the influence of complex illumination change and weather, and is easy to generate false alarm; the method comprises the steps of calculating displacement at a plurality of points, and lacking statistical analysis on the behavior rule of a connector group and a systematic evaluation model from pixel change to structural risk; The single-point abnormality is judged by adopting a fixed threshold value, spatial distribution and time evolution information cannot be combined, and the early warning accuracy and engineering guiding value are limited. Disclosure of Invention Aiming at the technical problems, the technical scheme adopted by the invention is a curtain wall connecting piece detection early warning method based on image analysis, and the method comprises the following steps: S01, deploying high-definition camera equipment at a preset detection point position to acquire N connection pieces detected in an acquisition picture of the high-definition camera equipment to set a concerned region so as to obtain a first data set in a complete two-dimensional coordinate; S02, extracting real-time images of each frame under the same time stamp based on the acquired first data set, searching and matching in the neighborhood of each concerned region by adopting a multi-feature fusion tracking algorithm, acquiring pixel coordinates of each connecting piece in the current image, and calculating a pixel displacement vector and a comprehensive displacement scalar of each connecting piece relative to a reference coordinate; s03, performing cleaning including image global registration correction based on feature points, outlier filtering based on statistics and data validity verification based on tracking confidence on the displacement data to obtain a cleaned second data set; S04, recording continuous points of the second data under continuous time, carrying out trend analysis and correlation analysis with environmental parameters to obtain displacement time sequence data, and finally calculating a global risk index Rg and a regional risk index Rz, wherein Rg is the proportion of the number M of connecting pieces with the displacement exceeding a dynamic safety threshold to the total monitoring points N; s05, matching the obtained Rg with a preset value, and preparing early warning information based on a matching result, wherein the early warning information comprises risk area positioning, overrun connector detail list and visual risk thermodynamic diagram. Preferably, in the step S02, the search matching is performed in the neighborhood of each region of interest by using a multi-feature fusion tracking algorithm, including the following steps: S21, quick response is carried out in a prediction search area by adopting a discriminant correlation filter based on the directional gradient histogram characteristics, and an initial displacement vector delta Pi and a response peak confidence Cd thereof are obtained; s22, in the vicinity of the initial position, BRISK characteristic points with rotation invariance and partial scale invariance are extracted and matched with BRISK characteristic descriptors stored in a reference template, and PROSAC algorithm is utilized to robustly estimate affine transformation