为进一步提高动态目标室内可见光定位追踪系统性能, 提出了一种基于随机森林(RF)算法的室内可见光指纹定位方法。利用发光二极管(LED)的光强信号作为特征构建指纹数据库, 应用指纹库中的数据训练决策树, 引入RF算法进行初始定位, 再通过卡尔曼滤波对初始位置估计进行优化, 从而获得更准确的定位轨迹。仿真结果表明: 在5 m×5 m×3 m的室内场景下, 通过所提定位方法能获得大部分采样点误差分布在4 cm之内的定位效果; 此外, 通过与不同室内可见光定位算法的性能进行对比, 验证了所提算法的技术优势。

In order to further improve the performance of dynamic target indoor visible light location and tracking system, a indoor visible light fingerprint location method based on random forest(RF) algorithm was proposed. The light intensity signal of light-emitting diode(LED) was used as the feature to build a fingerprint database, and the data in the fingerprint database was used to train the decision tree. RF algorithm was introduced for initial positioning, and then Kalman filter was used to optimize the initial position estimation, so as to obtain a more accurate positioning trajectory. The simulation results show that in the indoor scene of 5 m×5 m×3 m, the proposed positioning method can obtain the positioning effect that most sampling points error distribution is within 4 cm. In addition, this paper verifies the technical advantages of the proposed algorithm by comparing the performance of different indoor visible light positioning algorithms.