Peysakhovich Vsevolod, Hurter Christophe
ISAE-SUPAERO , France.
ENAC, ToulouseFrance.
J Eye Mov Res. 2018 Jan 8;10(5). doi: 10.16910/jemr.10.5.9.
We demonstrate the use of different visual aggregation techniques to obtain non-cluttered visual representations of scanpaths. First, fixation points are clustered using the mean-shift algorithm. Second, saccades are aggregated using the Attribute-Driven Edge Bundling (ADEB) algorithm that handles a saccades direction, onset timestamp, magnitude or their combination for the edge compatibility criterion. Flow direction maps, computed during bundling, can be visualized separately (vertical or horizontal components) or as a single image using the Oriented Line Integral Convolution (OLIC) algorithm. Furthermore, cosine similarity between two flow direction maps provides a similarity map to compare two scanpaths. Last, we provide examples of basic patterns, visual search task, and art perception. Used together, these techniques provide valuable insights about scanpath exploration and informative illustrations of the eye movement data.
我们展示了使用不同的视觉聚合技术来获得扫视路径的无杂乱视觉表示。首先，使用均值漂移算法对注视点进行聚类。其次，使用属性驱动的边捆绑（ADEB）算法对扫视进行聚合，该算法将扫视方向、起始时间戳、幅度或它们的组合用于边兼容性标准。在捆绑过程中计算的流方向图可以单独可视化（垂直或水平分量），也可以使用定向线积分卷积（OLIC）算法作为单个图像进行可视化。此外，两个流方向图之间的余弦相似度提供了一个相似度图来比较两条扫视路径。最后，我们提供了基本模式、视觉搜索任务和艺术感知的示例。综合使用这些技术，可以提供有关扫视路径探索的宝贵见解以及眼动数据的信息性插图。
