Research On Key Technologies And Applications Of High Speed Vision System For Dynamic Measurement

With the advent of the multimedia age, high pixel camera appeared. Image can record the location information of structure completely, so dynamic characteristics reap-pearance is available by effective digital image processing technology. Vision mea-surement technology has been widely draw many researchers’attention because of its advantages, such as high efficiency, high precision, non-contact and distance measure-ment. Nowadays, with the improvement of computer performance and development of related technologies, Vision measurement technology can supplement the deficiency of the conventional measurement methods so that it can obtained global application quickly.High speed vision measurement system represents faster sampling rate and higher computing speed. It can not only complete the measurement of geometric size and shape of the common attitude, but also realize the fast tracking of the moving objects using the technology of computer vision to analyze and process the video image. In this paper, the software and hardware design of the system is deeply studied. Especially for the software of the video image displacement extraction algorithms which solved the three key technologies of speed, accuracy and robustness in high speed vision measurement.First, this paper gave an analysis of the hardware selection. In order to increase the adaptability of the system, the whole system is composed of different combina-tions to complete the video capture of high speed moving objects in different situations. The pendulum experiment demonstrates the motion measurement process of whole high speed vision measurement system. Including the composition of the measurement sys-tem, the calibration of the system and the process of extracting the motion displacement using the related digital image processing technology.In order to develop the high performance of the image matching algorithms, this paper studied the image matching technology including matching template, correlation functions, displacement extraction algorithm and sub pixel accuracy localization. Com-pared with the performance and connection of the commonly correlation function, a combination algorithm to improve the measurement accuracy based on block matching and sub pixel location was proposed. The deficiency and application range of Lucas-Kanade based algorithm using in vision measurement were also pointed out.According to the research results of image correlation method and the character-istics of high sampling frequency of the high speed vision measurement, two template matching algorithms were proposed to estimate the target motion, which is not sensitive to illumination and noise, and has high efficiency and high location accuracy. One is the Local Search algorithm (LS), the other is the optimized particle swarm optimization algorithm (PSO). Local search template matching method is suitable for small displace-ment of the target between adjacent frames, which can achieve high computation effi-ciency under certain circumstances. The optimized PSO can extraction the displacement between adjacent frames by simulating the prey of birds, which is not affected by the scope of the search and can avoid local optima. Both template matching algorithms are based on NCC similarity criterion which can obtain subpixel accuracy without adding extra computation by combining surface fitting method. These two algorithms can be used as the common algorithm in high speed vision measurement system.High speed vision measurement system can not only achieve the conventional structure dynamic response measurement, but also can dig those motion signals which are not perceived, invisible and neglected by human vision. Those weak signals can be extracted and amplified to the order that human eye can perceive by Euler Video Magnification. Since the amplified video is somewhat different for the real movement, this paper introduces the SVD based displacement extraction algorithm which is simple, fast and reliable by reducing the information dimension of the image and revealing the hidden message behind the image. SVD based algorithm can extract the motion princi-pal components from the video transformed by EVM and can also avoid the redundant information introduced by EVM.Based on the efficient displacement extraction algorithm, this paper developed vi-sual measurement software on the computer platform. Using the software to complete the engineering measurement and solve practical engineering problem, such as vibration measurement and structure optimization of Truck steering wheel, in-situ measurement and model identification of wind-induced pulse response of sound barrier, response measurement and tightness judgment of bolted construction, vibration measurement of soft material excited by the sound and recovery of the sound from video. Results showed the actual measurement effect of high speed vision based measurement system and proved that the high-speed vision based measurement is an effective and reliable measurement method. It has important application value.