Develop Of Concrete Vibration Monitoring System Based On Stereo Vision

With the rapid increase in the amount of infrastructure and urban construction,the construction progress is getting faster and faster,and the requirements for the quality of concrete pouring have become increasingly stringent.Vibration is the key process in the concrete pouring process.The monitoring and control of vibration quality is an important part in the construction quality control.Therefore,it is of great significance to the intelligent monitoring and evaluation of vibration quality.At present,constructors mainly control vibration quality according to their experience.It is difficult to quantitatively evaluate the effect of vibration.Inadequate vibration,excessive vibration and vibration leak will cause quality defects and cannot be known in time,leaving significant safety hazards.The existing vibrating quality monitoring systems often require significant improvement of the vibrators,and GPS base stations need to be established on the construction site.The procedures are cumbersome,the operation is complicated,and the cost is hi gh,which cannot be universally applied.According to the construction scene and vibration process requirements,the hardware parameters of the vibrating quality monitoring system are calculated,selected and constructed.Zhang Zhengyou calibration method and three-dimensional calibration algorithm are used to calibrate the binocular camera,and the conversion relationship between the image coordinates and the three-dimensional physical coordinates is obtained;Based on the comparison of the Intersection-over-Union(IOU)and the center coordinate deviation,four kinds of motion tracking algorithms,Adaptive Scaled Mean Shift(ASMS),Kernerlized Correlation Filter(kcf),Efficient Convolution Operators(ECO),and Siamese Fully convolutional(Siam FC),are analyzed.The ECO algorithm with the best robustness,highest precision,and highest efficiency is selected to perform vibrating stick motion tracking.A new tracking confidence measure is designed in combination with the maximum response value and the average peak-to correlation energy(APEC).Study the deep learning target detection algorithm YOLO v2 algorithm for the detection of the vibrator rod position in the initial frame and the first frame after dropping.Based on parallax errors and algorithmic efficiency,three matching algorithms,namely feature point matching,block matching,and template matching,are compared.The optimal template matching algorithm is used to perform the parallax calculation.The triangulation principle is used to measure the position of the vibrator and the depth of vibration.Analyze the vibrating rod's motion state and working state to design vibrating time measurement algorithm,and construct interactive interface based on Qt.Based on machine vision and artificial intelligence,the vibrating rod tracking algorithm is constructed.The IOU of the vibrator rod tracking algorithm is maintained at about 0.9,the pixel distance of the center coordinate deviation is less than 0.02 mm,and the three-dimensional physical position error of the vibrating point is less than 2cm in the x and z directions;the vibrating depth error is less than 1cm;the vibrating time error is less than 1 second.