计算机科学与探索 ›› 2020, Vol. 14 ›› Issue (6): 1017-1027.DOI: 10.3778/j.issn.1673-9418.1906038

• 人工智能 • 上一篇    下一篇

不安全越界行为的个性化实时检测

林强,张淋均,谢艾伶,王维兰   

  1. 1. 西北民族大学 数学与计算机科学学院,兰州 730124
    2. 西北民族大学 中国民族语言文字信息技术教育部重点实验室,兰州 730030
  • 出版日期:2020-06-01 发布日期:2020-06-04

Personalized Real-Time Detection of Unsafe Boundary Transgression

LIN Qiang, ZHANG Linjun, XIE Ailing, WANG Weilan   

  1. 1. School of Mathematics and Computer Science, Northwest Minzu University, Lanzhou 730124, China
    2. Key Laboratory of China's Ethnic Languages and Information Technology of Ministry of Education, Northwest Minzu University, Lanzhou 730030, China
  • Online:2020-06-01 Published:2020-06-04

摘要:

户外迷路甚至走失事件在老年人群体中多发频发,成为危及他们独立生活安全的突出问题之一。为防止老年人走离日常生活所在的安全区域,进而避免走失事件的发生,研究并提出基于个人出行轨迹的个性化安全地理围栏构建方法及面向越界行为发现的异常轨迹实时检测算法。首先,建模每个人的户外安全地理围栏为不规则多边形,其中顶点代表经常到访的物理位置,边代表连接物理位置之间的道路;其次,使用GPS轨迹实例化构建的安全地理围栏模型,包括相关区域的划分和轨迹的映射处理;再次,通过在传统点在多边形内部判定算法中融入异常轨迹跨越度的量化评价指标,提出不安全越界行为的个性化实时检测算法;最后,使用一组来自个人的GPS轨迹数据进行了实验验证。实验结果表明提出的方法在老年人边界越界行为的识别中是可行的,在所有数据集上获得的AUC值均高于0.995,该原型系统具有良好的响应时间和检测性能。

关键词: 地理围栏, GPS轨迹挖掘, 智能辅助技术, 实时检测算法

Abstract:

Older adults are more prone to experience disorientation or even get lost, which has a negative impact on their independent living. To prevent an older adult from walking away from the safety area where she or he lives and thus from getting lost, this paper proposes an approach that is able to construct personalized safety geofencing for each individual based on her or his outdoor traces. And a real-time detection algorithm is developed to recognize the outlying trajectories caused by boundary transgression behavior of the elderly. First, this paper models each individual's safety geofencing as an irregular polygon, in which every vertex denotes the physical location where she or he frequently visits and every edge denotes the path between any two physical locations. Second, this paper instantiates the constructed safety geofencing using each individual??s historical GPS trajectories via partitioning her or his region of interest into cells and mapping trajectories into the partitioned region of interest. Third, this paper proposes the real-time detection algorithm for differentiating boundary transgression behavior from the normal ones by introducing the deviation degree of outlying trajectories into the traditional point in polygon detection algorithm. At last, experimental results conducted on a group of individuals?? GPS trajectories demonstrate that the proposed method is feasible to recognize boundary transgression behavior of older adults, obtaining an AUC value of more than 0.995 for all datasets. The prototype system also has good response time and detection performance.

Key words: geofencing, GPS trajectory mining, intelligent assistive technology, real-time detection algorithm