Research On Non-Contact Measurement Method And Application Based On Machine Vision

Structural displacement measurement and experimental modal analysis are widely used in mechanical engineering,material engineering,aerospace and other fields.As the most direct reaction of structure under load,displacement has always been an important measurement content in engineering projects.At present,displacement sensor,acceleration sensor and other contact measurement technology is the most common technical means to measure the structural displacement,but in practical engineering practice,these methods are often limited by environmental conditions,and the additional mass brought by them will have a certain impact on the lightweight structure.The emergence of laser displacement sensor solves the problem of additional mass,but the measurement range is only limited to single point measurement,which cannot meet the requirements of the whole structure measurement.The non-contact measurement method based on machine vision has the advantages of full-field measurement,non-contact,high precision and low requirements for measurement environment,and has great application advantages in displacement measurement.In practical engineering applications,camera calibration error,camera pose and large amount of modal test data affect the measurement accuracy and calculation efficiency of this method.To solve the above problems,combined with the application research of structural static displacement measurement and modal test,the non-contact measurement method and application research based on machine vision is carried out in this paper.The specific work content includes the following parts:1.Aiming at the problem that traditional camera calibration optimization methods tend to fall into local optimal solution and affect the calibration accuracy,a camera internal parameter optimization method based on improved particle swarm optimization algorithm is proposed.The method can control the reprojection error within 0.05 pixel and reduce the deformation calculation error caused by the camera calibration error.At the same time,aiming at the problem that the optical axis of the camera is not perpendicular to the measured surface,the source of the error is analyzed from the imaging principle,and the coordinate transformation method based on the calibration plate is used to correct the position and pose of the camera coordinate system,and the effectiveness of the method for improving the measurement accuracy is verified by the pan-tilt translation experiment.2.Aiming at the problem of adding additional mass into the speed sensor in the contact modal test measurement,the non-contact modal test research based on the connected domain detection method and the digital image correlation method was carried out respectively.The connected domain detection method pastes the optical marker on the surface of the structure,and obtains the vibration displacement of the measured point by tracking the gray center of gravity of the marker.The test process is relatively simple and convenient.The digital image correlation method obtains the displacement data by matching the speckle sprayed on the structure surface.The method takes the pixel as the calculation point and can complete the measurement of the whole field by subdividing the calculation point grid.The two methods are used to carry out modal tests on the cantilever structure,and the accuracy of the method is verified by comparing the results with the contact modal measurement method based on acceleration sensor.3.Aiming at the problems of complicated calculation process and low efficiency of digital image correlation method in dynamic measurement,optimization is carried out from two aspects.On the one hand,an automatic speckle region extraction method based on image entropy is proposed to solve the problem of complex shape structure requiring manual division of calculation area.This method can extract speckle region efficiently and quickly,further realize the division of the whole field of calculation points,and help improve the automation performance of digital image correlation method.On the other hand,the calculation path and matching method of the calculation points are improved,and the frequency domain correlation matching method is combined with the reliable guidance path method as the pixel search method,which effectively improves the calculation accuracy and efficiency.The optimized method is used to carry out full-field modal test of the panel structure,and the modal frequency error is controlled within 4%.The vibration mode of the test model is highly matched with the simulation results,which has practical engineering significance.