Subtle Motion Magnification And Its Application In Monitoring Of Hydropower Engineering

Subtle motion magnification technology can reveal the subtle changing in videos that are difficult or if not impossible to see with the naked eye, and it has potential applications in many fields. Eulerian video magnification(EVM) technology is an attractive technology that proposed in recent years for subtle motion amplification, and four key steps of it are spatial decomposition, temporal bandpass filtering, amplification and reconstruction. The existing Eulerian technology can be classified into the following two classes: linear Eulerian video magnification(linear EVM) and phase-based Eulerian video magnification(phase-based EVM).In this thesis, EVM is investigated in detail, especially the key technologies are analyzed. For the computation-intensive and time-consuming, a new EVM approach in the Y color channel is proposed to speed up the system. Taking existing EVM method for processing video sequences for each pixel of every frame for the magnification of motion targets into account, and it amplifies the changes in video will also amplify the noise, and it can’t be visually select region of interest in the video, another improvement method is proposed. Finally, the improvement method was applied to the monitoring of hydropower engineering. The specific research contents include the following aspects:Firstly, color space processing and multi-resolution decomposition were discussed in this paper. It proposed an improvement scheme based on Y color channel method for the motion magnification, and it reduces the complexity of calculation in order to make the speed of the algorithm increased. Experiments show that proposed method makes the video amplification processing speed significantly improved.Secondly, video noise will become larger with the magnification factor increases, and it cannot be visually select region of interest in the video. In view of the problem, the approach to retain out the region of motion information in video, by the foreground detection method, is proposed. Depend on the description and analysis of the motion, the motion region can be detected. Then processing it with morphological filter, and we can magnify the motion in video finally. For foreground constraint of the motion region, this paper compared several foreground detection algorithm, and the detection performance is relatively good, the ViBe algorithm, the same as the EVM based on the pixel, is selected as the method for foreground constraint. Experiments show that the value of the peak signal to noise ratio(PSNR) of the improved method for the original image is larger than EVM’s,so its fidelity is relatively good. And the improved method can be targeted to magnify the motion region, which not only makes video amplification processing speed improved, but also makes the monitored targets are more likely to highlight motion characteristics to see directly for the people.Finally, this paper proposes that the EVM technology can be used for the monitoring of hydropower engineering, and it provides a new research perspective. In the first place,the framework of the monitoring system is presented, and a brief analysis is given. In the next place, the difficulty and the importance of data collection are gave, and also talk about the extra matters of data collection. In the last place, we conducted testing and analysis.The results of ordinary bridge, large-scale bridge with the cable are affected by the horizontal wind or vehicle and the monitoring of dam’s construction and maintenance were discussed. In addition, experimental video of Ge He Yan dam was analyzed, and the test result was also discussed. Experiments show that, according to EVM method processing,the shaking of the bridge, the cable of the bridge, the crane and the dam can be clearly observed. And it is valuable to auxiliary determine the deformation characteristics of it and take further measures for early warning and risk handling. As well as it provides a new reference for the safety inspection of the target of monitoring.To sum up, the thesis researches EVM technology systematically. And the Y channel only based and the region of motion based foreground-constrained motion magnification methods are proposed separately. Finally, the EVM technology was used for the monitoring of hydropower engineering, and the corresponding experimental analysis and discussion ware carried out.