CN-121482041-B - Bridge structure vibration long-term monitoring algorithm and system based on computer vision

CN121482041BCN 121482041 BCN121482041 BCN 121482041BCN-121482041-B

Abstract

The invention provides a bridge structure vibration long-term monitoring algorithm and system based on computer vision, and relates to the technical field of beam monitoring. Aiming at the lack of a vision-based bridge monitoring algorithm with high precision, strong robustness and shielding self-recovery capability in the prior art. The method comprises the steps of calculating confidence of the matching points, removing the matching points with low confidence score, calculating displacement of the bridge at a plurality of characteristic points to generate vibration time sequence data, extracting natural frequency and corresponding mode shape, comparing the natural frequency with a reference frequency range and the reference mode shape to obtain a monitoring result, and conducting shielding self-recovery aiming at environmental interference influence, so that accuracy and reliability of bridge monitoring are improved, and robustness of bridge monitoring activities in complex scenes is improved.

Inventors

ZHU LI
LI JIAHUAN
Cui Zhiruo
ZHU YAOYU

Assignees

北京交通大学

Dates

Publication Date: 20260508
Application Date: 20260108

Claims (8)

1. A bridge structure vibration long-term monitoring algorithm based on computer vision, comprising: s1, responding to a trigger instruction generated by a timer of a singlechip, and controlling a camera to shoot a time sequence image of a target bridge, wherein the time sequence image comprises a reference image and a plurality of target images; s2, extracting a plurality of characteristic points in the reference image, and matching each characteristic point in the reference image with each target image to obtain a preparation matching point pair; s3, generating a confidence coefficient score of each matching point pair, and removing the matching point pairs with the confidence coefficient score lower than a first threshold value to obtain a target matching point pair set; s4, calculating the displacement of the target bridge at a plurality of characteristic points based on the target matching point pair set so as to generate vibration time sequence data of the target bridge; s5, extracting at least first-order natural frequency and corresponding mode shape of the target bridge based on the vibration time sequence data, wherein the mode shape is used for representing the spatial deformation form of the target bridge when vibrating under the corresponding natural frequency; S6, comparing the natural frequency and the mode shape with a preset reference frequency range and a preset reference mode shape to obtain a monitoring result; comparing the natural frequency and the mode shape with a preset reference frequency range and a preset reference mode shape to obtain a monitoring result, wherein the monitoring result comprises the following steps: calculating a mode guarantee criterion value of a current mode shape and the reference mode shape, and generating a mode shape abnormal sign when the mode guarantee criterion value is lower than a preset mode correlation threshold value; When the frequency abnormality mark and the vibration mode abnormality mark exist simultaneously, calculating a vibration mode curvature difference between the current mode vibration mode and the reference mode vibration mode at each characteristic point position, identifying a continuous area where the vibration mode curvature difference exceeds a preset curvature threshold value, determining the position corresponding to the maximum value of the vibration mode curvature difference in the continuous area as a core damage area, calculating relative deviation values of a plurality of continuous inherent frequencies, distributing weight coefficients according to the sensitivity differences of the inherent frequencies to damage, calculating to obtain comprehensive scores based on the relative deviation values of the inherent frequencies and the corresponding weight coefficients, determining damage severity level based on the correspondence between the comprehensive scores and the damage severity level, wherein the weight distributed by a low-order frequency is larger than the weight distributed by a high-order frequency, determining a structural risk level according to the position of the core damage area in a bridge structure, generating a maintenance priority based on the damage severity level and the structural risk level through a predefined decision matrix, and generating a monitoring result representing damage confirmation of the structure based on the position information of the core damage area, the damage severity level and the maintenance priority; when only the frequency abnormality flag is present: acquiring current environmental temperature data, and comparing and verifying the current natural frequency with an environmental temperature-frequency correlation model established based on historical monitoring data; When the frequency change accords with the temperature change rule and the vibration mode characteristic is kept stable, the camera is controlled to execute a preset swinging action to weaken the influence of environmental interference, and a monitoring result representing the influence of the environmental interference is generated.
2. The computer vision based bridge structure vibration long-term monitoring algorithm of claim 1, wherein the calculating the displacement of the target bridge at a plurality of feature points based on the set of target matching point pairs to generate vibration timing data of the target bridge comprises: s41, calculating pixel coordinate displacement vectors between the feature points in the reference image and the corresponding matching points in the target image based on the target matching point pair set; S42, performing sub-pixel precision optimization on the pixel coordinate displacement vector based on gray gradient change of the image sequence in the space-time dimension and/or through gray distribution of a neighborhood of the surface fitting feature points, and obtaining an image plane displacement of sub-pixel level precision; S43, converting the image plane displacement into physical displacement of the target bridge in a three-dimensional space through a coordinate conversion relation based on camera parameters calibrated in advance, wherein the camera parameters comprise camera internal parameters and camera external parameters relative to the target bridge; s44, combining the physical displacement into a displacement-time sequence according to the corresponding relation between the time stamp corresponding to the trigger instruction of the singlechip and each characteristic point; s45, aggregating the displacement-time sequences of all the characteristic points to obtain vibration time sequence data used for representing vibration responses of the target bridge at a plurality of discrete positions.
3. The computer vision based bridge construction vibration long-term monitoring algorithm of claim 1, wherein the controlling the camera to perform a preset swing action to attenuate environmental disturbance effects comprises: Determining the swing amplitude of the preset swing action based on the image quality of the time sequence image; determining the swing frequency of the preset swing action based on the accumulation speed of dust in the environment; determining the swing time of the preset swing action based on the target detection task requirement; And generating a control instruction of the preset swinging motion based on the swinging amplitude, the swinging frequency and the swinging time of the preset swinging motion.
4. The computer vision based bridge construction vibration long-term monitoring algorithm of claim 3, wherein the determining the swing frequency of the preset swing action based on the accumulation speed of dust in the environment comprises: acquiring the concentration of dust in the environment and the environment humidity and the characteristic parameters of a lens of the camera; determining a dust viscosity coefficient based on the ambient humidity and the concentration of dust in the environment; Determining the adhesion type of dust based on the dust viscosity coefficient and characteristic parameters of a lens of the camera; determining an accumulation rate of dust in the environment based on the adhesion type of the dust; and determining the swing frequency of the preset swing action based on the accumulation speed of dust in the environment.
5. The computer vision based bridge construction vibration long-term monitoring algorithm of claim 4, wherein the characteristic parameters of the camera lens include hydrophobic properties, antistatic properties, and surface roughness.
6. The long-term monitoring algorithm for bridge construction vibration based on computer vision according to claim 1, wherein when no frequency abnormality flag and no vibration mode abnormality flag are found within a preset number of detection times: confirming that the target bridge is in a health state, and generating a monitoring result for representing the health confirmation of the target bridge; Updating the reference frequency range and optimizing the reference mode shape based on the monitoring data in the preset detection times; And generating a trigger frequency adjustment instruction of a timer of the singlechip based on the duration of the target bridge in the healthy state.
7. The long-term monitoring algorithm for bridge structure vibration based on computer vision according to claim 1, wherein the matching each feature point in the reference map with each target image to obtain a preliminary matching point pair comprises: S21, extracting image features of the reference image and the target image, wherein the image features comprise feature point positions and corresponding high-dimensional feature descriptors; S22, based on the high-dimensional feature descriptors of the feature points of the reference image and the target image, forward matching is carried out from the reference image to the target image by calculating a similarity matrix between the feature descriptors, and candidate matching points are searched in the target image for each feature point in the reference image; S23, consistency test is carried out on results of forward matching and reverse matching, matching point pairs corresponding to each other in the forward matching and the reverse matching are reserved, and a preparation matching point pair is obtained.
8. A computer vision-based bridge construction vibration long-term monitoring system for implementing the computer vision-based bridge construction vibration long-term monitoring algorithm of any one of claims 1 to 7, comprising: The image generation unit is used for responding to a trigger instruction generated by a timer of the singlechip and controlling the camera to shoot a time sequence image of the target bridge, wherein the time sequence image comprises a reference image and a plurality of target images; The feature matching unit is connected with the image generating unit and is used for extracting a plurality of feature points in the reference image, and matching each feature point in the reference image with each target image to obtain a preparation matching point pair; The matching filtering unit is connected with the characteristic matching unit and is used for generating a confidence coefficient score of each matching point pair, and removing the matching point pair with the confidence coefficient score lower than a first threshold value to obtain a target matching point pair set; The displacement field reconstruction unit is connected with the matching filtering unit and is used for calculating the displacement of the target bridge at a plurality of characteristic points based on the target matching point pair set so as to generate vibration time sequence data of the target bridge; The modal parameter identification unit is connected with the displacement field reconstruction unit and is used for extracting at least first-order natural frequency and corresponding modal shape of the target bridge based on the vibration time sequence data, and the modal shape is used for representing the spatial deformation form of the target bridge when vibrating under the corresponding natural frequency; and the structural health diagnosis unit is connected with the modal parameter identification unit and is used for comparing the natural frequency and the modal shape with a preset reference frequency range and a preset reference modal shape to obtain a monitoring result.

Description

Bridge structure vibration long-term monitoring algorithm and system based on computer vision Technical Field The invention relates to the technical field of bridge monitoring, in particular to a bridge structure vibration long-term monitoring algorithm and system based on computer vision. Background With the rapid development of computer vision and image processing technology, the vision-based structural health monitoring method has great application potential and development prospect in the field of bridge engineering monitoring due to the unique advantages of non-contact, full-field measurement, flexible deployment and low cost. Compared with the traditional sensor, the vision method can reconstruct the full-field displacement response of the structure under the action of load by collecting the structure image sequence through the camera, so that the damage of the structure caused by installing the sensor is avoided, the space intensive data of far-ultra-point measurement can be obtained, and a rich information basis is provided for comprehensively evaluating the integral performance and damage identification of the structure. The technical route is hopeful to become a powerful supplement and even an important component part of the existing monitoring system, and is a key technical direction for realizing the intellectualization of the infrastructure and long-life fortune maintenance. However, existing bridge monitoring methods based on computer vision still have significant drawbacks, especially when facing long-term, automated monitoring tasks. Firstly, under a complex and changeable natural environment, the robustness of the existing method is insufficient, so that the stability and reliability of the image characteristic matching in a core link are obviously reduced, mismatching or matching failure is easy to generate, rough differences or interruption in subsequent displacement data are further caused, and the final precision and reliability of the whole monitoring system are reduced. Secondly, the prior art has insufficient capability of distinguishing environmental interference from real structural damage, and has insufficient shielding self-recovery capability for the real problems of image quality degradation caused by dust accumulation of a camera lens in long-term operation, thereby further reducing the final precision and reliability of the whole monitoring system. Therefore, the bridge structure vibration long-term monitoring algorithm and system based on computer vision, which are high in precision, strong in robustness and capable of shielding self-recovery, are developed, and have important significance for bridge structure vibration long-term monitoring. Disclosure of Invention Therefore, the invention provides a bridge structure vibration long-term monitoring algorithm and system based on computer vision, which are used for solving the problems of perception failure caused by poor feature matching robustness and diagnosis erroneous judgment caused by weak environment interference discrimination capability in the prior art. In order to achieve the above object, the present invention provides a bridge structure vibration long-term monitoring algorithm based on computer vision, comprising: s1, responding to a trigger instruction generated by a timer of a singlechip, and controlling a camera to shoot a time sequence image of a target bridge, wherein the time sequence image comprises a reference image and a plurality of target images; S2, extracting a plurality of characteristic points in the reference image, and matching each characteristic point in the reference image with each target image to obtain a preliminary matching point pair; s3, generating a confidence coefficient score of each matching point pair, and removing the matching point pairs with the confidence coefficient score lower than a first threshold value to obtain a target matching point pair set; S4, calculating the displacement of the target bridge at a plurality of characteristic points based on the target matching point pair set so as to generate vibration time sequence data of the target bridge; S5, extracting at least first-order natural frequency and corresponding mode shape of the target bridge based on vibration time sequence data, wherein the mode shape is used for representing the spatial deformation form of the target bridge when vibrating at the corresponding natural frequency; S6, comparing the natural frequency and the mode shape with a preset reference frequency range and a preset reference mode shape to obtain a monitoring result. Further, calculating displacements of the target bridge at the plurality of feature points based on the set of target matching point pairs to generate vibration timing data of the target bridge, comprising: S41, calculating pixel coordinate displacement vectors between the feature points in the reference image and the corresponding matching points in the target image based on the tar