Design And Implementation Of Data Visualization For Experimental Data Cloud Platform

In the information era, the emergence of cloud storage makes the volume of data expand quickly. Obtaining useful information from the mass of data is an urgent need. Data visualization techniques, meanwhile, as a means of displaying data intuitively, has become one of the research hotspots in big data era. In recent years, the research on data visualization technology has accumulated a lot of experience, and there were many effective visualization tools developed especially in medical filed. But most of studies in the past, both offline tool and web-based visualization techniques, have processed data from database or a single file, which has much difference in quantity and form with processing data stored in cloud platform.With the continuous development of information and intelligence in aerospace fields, time sequence data generated in experiments has the characteristics of variety in kinds, surge in quantity and different size in file. In response to this situation, scientific methods should be applied to manage these data efficiently and uniformly. Experimental data cloud platform is a cloud-based data management platform designed and implemented under this background. In addition to the quick storage ability, a reasonable solution for a variety of data visualization is also needed.One of the focuses of this paper is to overcome the difficulty in drawing curve with large amounts of data in cloud platform, which requires users to be able to use the client to observe data in views with different precision, and the display of the overall view should basically keep the original trend. In addition, the ability of showing data orderly stored in the platform is important, as well as quickly locating and viewing desired content.Firstly, several most commonly used algorithms about piecewise linear representation of time series has been discussed. We analyzed the difference among limiting the number of segments algorithm, limiting segment error algorithm and feature points based algorithm. Then, according to the characteristics of experimental data cloud platform, we selected the piecewise aggregate approximation algorithm and trend analysis algorithm for performing several sets of experiment to determine the reasonable parameters. While these two algorithms still existed drawbacks applied to the platform under the optimal parameters, we proposed an effective solution for sampling with fixed step and Minmax value. Compared with the first two algorithms, the method we proposed was more suitable for the platform considering the fitting error, compression rate, time and the capability of maintaining the trend of curve.We applied the sample method we proposed in experimental data cloud platform, which ensured the reasonable density of extracting data within the effective time during data plotting. Finally, the visual part of GUI client was well designed and implemented: the navigation tree could display the directories and files stored in the platform orderly; users could obtain tabular data, images and video easily through the client; graphics-related functions were also well implemented. Finally, after the functional and performance testing, the GUI client met all visual needs and had a good user experience.