基于改进谱聚类和交叉标注的多人员指纹定位方法
Method of multi-personnel fingerprint positioning in mine tunnel based on Optimized Spectral Clustering and cross-tagging
-
摘要: 在目前矿井集中式人员定位方法精度与运行效率较低的背景下, 文中分析了分布式矿井定位系统模型构建的必要性与算法要求, 将现今较为先进的机器学习谱聚类算法使用图论的方法进行了算法改进, 分析了运用切比雪夫多项式估计方法完成图谱滤波器与特征值的运算简化过程, 构建了一套与矿井定位要求相匹配的改进谱聚类算法。结合矿井具体定位应用过程, 分析了传统定位指纹数据收集方式的局限性, 提出多人员定位指纹数据交叉标注技术, 在数据采集与算法运行两个层面提高了矿井人员定位系统的整体精度, 从根本上改进了定位时效性。Abstract: Centralized positioning system has been widely used in mining localization, but it shows a low accuracy and efficiency compared to distributed method.Because of the influential position in mining positioning, the authors analyze the necessity and algorithm requirements of the distributed mine positioning system modeling.This model is based on a determined algorithm in machine learning called Optimized Spectral Clustering (OSC) which is developed from Spectral Clustering (SC) introduced in the paper.The OSC comes from traditional SC, but has been improved using Chebyshev estimation to simplify the calculation process in SC, and brings a high efficiency that traditional SC cannot reach.In the period of location fingerprint gathering, the authors put forward a new method called cross-tagging method for multi-entities tracking.Combined with OSC, a whole localization system is set up in the study and these two techniques bring a higher efficiency and accuracy in mining localization.