Research On Spatio-temporal Data Visualization Based On Spatial Features Mining

Spatiotemporal data visualization is an important branch of big data visualization The technology would show complex high-dimensional time-varying data in a visual form through analyzing and extracting spatiotemporal data,helping users to mine the multi-variable data features.Spatio-temporal data involves human activities or changes in the natural world.Therefore,it has always been a research hotspot in data visualization and the cross-field of other domains.It is an important task to accurately restore the data features and enable users to accu-rately perceive the space,time and dynamic in the spatiotemporal data.However,due to the high-dimensional and discrete features of Spatio-temporal data,it is difficult to balance the data features visualization and the system complexity,which poses a great challenge to the Spatio-temporal data visualization design methods.In response to the above problems,this article divides the visualization of spatiotem-poral data into four levels according to the attributes of spatiotemporal data:intuitive spatial distribution;time-series relationship characteristics distribution;the dynamic in-formation in discrete data;the spatial-temporal tracking distribution of data.This article focuses on the above four topics,propose researches are as follows:(1)We proposed an algorithm for the spatial distribution visualization of spatiotempo-ral data in three-dimensional space.We proposed a method to replace the image data distribution to data space,and transformd the absolute spatial calculation prob-lem into another spatial distance problem of relative relationship.It can solve the difficulty of accurately classifying in the case of unmarked data.The visual form of this method balances the simplification of data and the feature details visualization.(2)We propose a method to visualize the temporal relationship mapping in spatiotem-poral information.The method converts the events relationship into a graphic dis-tribution mode,and shows the events sequence relationship through the image dis-tribution change.We combined the uncertain visualization and visual conding of factors,convert factors relationship into a causal sequence and designed many ele-ments and interactive methods.(3)We propose a method for dynamic information extraction and visualization of spa-tiotemporal data based on a generative model.This method uses density map as a medium,combined with generated model,which have the powerful generalization ability and generation effect.The model can obtains continuous data dynamic infor-mation through hidden space interpolation.This method can infer the most likely change process between the input data,combined with a variety of optimization algorithms in image domain,to achieve the discrete spatio-temporal data dynamic process reconstruction and visualization.(4)We propose a spatiotemporal data tracking framework based on optical flow calcu-lation and multi-target detection.This method uses density map-based visualization as a medium,and multiple objects detection as a means of extracting targets.After improving a variety of tracking algorithms,the model get the details of objects by using optical flow features.This method finally obtains the movement information of multiple objects in the area,and provides the tracking trajectories and boundary changes.The technologies above have solved the key problems in spatiotemporal data visu-alization,such as spatial visualization,timing visualization,dynamic visualization,and multiple objects tracking visualization.In this paper,these algorithms are integrated into applicable systems or frameworks to get a good result.