Research On 3D Reconstruction Of Vibrating Objects By Machine Vision And The Modal Identification

With the continuous progress and innovation of modern science and technology, a lot of mechanical products, equipments and dynamic systems are gradually developing towards the direction of precision, automation and intelligence. With such a high standard and fierce competition, the dynamic characteristics of the structure of systems have to be optimized in order to achieve the purpose of controlling vibration and noise in aerospace, automotive, marine industry, architecture and other fields. Thus, vibration testing has generally become an important issue that should be seriously addressed and studied in the field of engineering technology. Vibration testing methods can be divided into two types, one is the contact measurement method, and the other is non-contact measurement method. The contact measurement method mainly depends on sensors and strain gauges, which must be set on the surface of the structure, to collect the vibration data. Sometimes the sensors have to be set on the all surface of the structure for the completeness of data, which consumes a lot of manpower. For flexible and lightweight structures, the data transmission lines and the mass of loads introduced into structures affect greatly the measurement results. The non-contact measurement method is mainly based on the technology of optoelectronics and electromagnetism and the measurement system does not need to contact with the measured object. Typical non-contact measurement methods include laser triangulation method, ultrasonic measurement method, eddy current method and machine vision method, but most of them are difficult to obtain the three-dimensional vibration data of structure at the same time. In the era of high speed industrial camera and digital image processing technology, machine vision technology has been combining with vibration testing. The high speed cameras based on the binocular system are used to simulate the real-time human vision system to capture the vibration of structure and the three-dimensional vibration data of systems can be reconstructed according to the theory of binocular stereo vision. By combining these vibration data with the corresponding modal theory, the system mode can be analyzed. So it is very meaningful to study the combination of modal theory and the vibration data of system obtained by using machine vision technology for modal analysis.In this paper, the machine vision technology based on binocular stereo vision theory is used in vibration testing. The binocular vision system is used to collect the time domain data of vibration system and the mode of system is analyzed combining with modal theory. In order to achieve the purpose of the study, firstly, the binocular high speed cameras should be calibrated by applying checkerboard plane calibration method proposed by Zhang Zhengyou. Secondly, the Harris corner detection algorithm combining with cross marks should be used to track the real-time vibration structure and the detection accuracy should be improved to sub-pixel level. The three-dimensional vibration time domain data is able to be reconstructed. Finally, the modal analysis of system is completed according to the vibration condition. In order to verify the feasibility of the research idea, the appropriate high speed cameras, light source, image acquisition card, signal generator, computer and vibration testing system should be prepared for vibration testing system based on binocular stereo vision. Under unknown excitation conditions, the cantilever beam and the pipeline of axial flow fan are chosen as vibration experiment objects. The ITD method and the fast Fourier transform based on Bayesian theory are used for modal identification under free vibration and random vibration conditions. For the clamped beam, the camera can’t collect all its vibration data once and for all. After the clamped beam is divided into five regions, its mode shapes are assembled by least squares. The experimental results show that this method is feasible, which lays a solid foundation for the further development of the vibration testing technology with machine vision.