The invention relates to the technical field of intelligent traffic, and particularly discloses a traffic speed data filling method based on space-time regularization tensor decomposition. The method monitors the road traffic condition in real time through an intelligent traffic system, and comprises the following steps: establishing a three-dimensional tensor based on road section detection points, the number of days and the time period of each day; detecting missing traffic speed data of pre-filling, and initializing a missing data set; carrying out truncated high-order singular value decomposition on the initialized tensor to respectively obtain truncated left singular value vectors of the data set in three dimensions, and calculating a core tensor; and constructing a missing data filling model, adding space-time regularization guidance, and continuously and circularly optimizing the model, thereby realizing filling of missing traffic speed data. According to the method, through a high-precision initialization strategy, the space-time characteristics of the traffic speed data are fully considered, and under the condition that the constructed real world is missing, the missing value is more accurately filled.
本发明涉及智能交通技术领域,具体公开了一种基于时空正则化张量分解的交通速度数据填补方法。该方法通过智能交通系统对道路交通情况进行实时监测,包括如下步骤:基于路段检测点、天数和每天的时间段建立三维张量;检测到预填充缺失的交通速度数据,对缺失的数据集进行初始化;对初始化张量采用截断高阶奇异值分解,分别得到数据集在三个维度上的截断左奇异值向量,计算核心张量;构建缺失数据填补模型,并加入时空正则化指导,不断循环优化模型,从而实现对缺失的交通速度数据的填补。本发明通过高精度的初始化策略,充分考虑交通速度数据的时空特性,在构造的类似真实世界的缺失情况下,更加精确的填补缺失值。
Traffic speed data filling method based on space-time regularization tensor decomposition
基于时空正则化张量分解的交通速度数据填补方法
2024-02-02
Patent
Electronic Resource
Chinese
Traffic flow data completion method and system based on tensor chain decomposition
European Patent Office | 2023
|Traffic data matrix filling method based on Hessian regular space-time low-rank constraint
European Patent Office | 2022
|