Displacement Measurement Method For Bridge Deformation Based On Image Sequence

As an important indicator of bridge safety assessment,bridge deformation monitoring has been paid more and more attention in the field of structural engineering.With the rapid development of artificial intelligence,the non-contact,high-precision and low-cost measurement method of image measurement is increasingly adopted by the engineering community.The traditional contact measurement method has certain limitations,and the image method based on photogrammetry is easy to operate,has a wide range of use and strong adaptability.This paper aims to study the method of bridge deformation monitoring based on image sequence.The method mainly captures the image sequence by shooting video,determines the processing scope by extracting the bridge ROI of the image sequence,uses the AKAZE algorithm and the ORB algorithm to detect and describe the feature points,performs feature matching through the two-way FLANN algorithm,and eliminates false matches by using the PROSAC algorithm.The pixel displacement of the best matching and deformation point is obtained by the consistency method of pixel displacement that based on neighborhood feature points proposed in this paper,and then the two-dimensional displacement value of the bridge is obtained in combination with the conversion coefficient.This research has important theoretical significance and engineering practical value.The main research contents of this paper are as follows:1.Analyze the characteristics of the image sequence of the bridge,and propose an image processing method based on the region of interest(ROI),which provides a starting point for real-time processing of video images.Based on the basic theory of scale space,analyzed AKAZE algorithm and the advantages and disadvantages of the ORB algorithm in feature detection and description: AKAZE algorithm performance is excellent,but the computational complexity,need fast hardware environment support,suitable for demanding applications in the running effect;The ORB algorithm is simple in principle,fast in computation and able to run in low speed hardware environment.It is suitable for applications requiring fast and real-time processing.Combining with the requirement of fast and stable processing of image sequence,the method of combining AKAZE feature detection with ORB descriptor is presented.Through experimental verification,the AKAZE test proposed in this paper combines the method of ORB description,with fast processing speed and strong robustness.2.The FLANN algorithm has the advantages of being able to automatically,quickly and stably match,and the matching accuracy is improved at the expense of a small amount of time by the bidirectional FLANN matching method.By using the PROSAC algorithm to eliminate the mismatches,the time spent in the bidirectional FLANN search matching method was made up,and the matching method was further optimized.In this paper,a method based on the consistency of pixel displacements of neighboring feature points is proposed to eliminate false matches completely,thus obtaining the best matching result.Through experimental verification,the processing speed and reliability of this method can meet the real-time processing requirements of bridge video images.3.Expand the principle of parallax method,and use the bridge's own structure to obtain the coefficient of conversion from drawing distance to actual distance.The checkerboard experiment was performed on a non-measurement camera,and the calibration parameters of the camera were obtained.Based on the previous research on feature extraction and feature matching of image sequences,a method for extracting bridge deformation displacement based on image sequence is proposed.The experimental results show that this method has high precision,strong reliability and fast processing speed,and can deal with video image in real time.Compared with the traditional bridge deformation monitoring methods,this method has the advantages of simple operation,low measurement cost,high efficiency,and real-time tracking and monitoring of monitoring points.In addition,this method can achieve multi-point synchronous measurement,depict the deflection curve of the entire bridge,provide more integrated and detailed information for bridge deformation monitoring,good for the bridge disaster detection and early warning analysis.4.According to the image processing algorithm in this paper,A bridge deformation displacement measurement system based on image sequence is designed and written.The system can automatically realize high precision and fast image feature extraction and matching,obtain the displacement of the bridge in static and dynamic,and track the monitoring points.It can also realize multi-point synchronous measurement and describe the overall deflection curve of the bridge.