Guide Tracking Method Based On Particle Filter Fusion

Traditional localization techniques cannot meet the requirements of indoor localization accuracy. Although PDR can obtain high localization accuracy in a relatively short period of time, its localization error will gradually accumulate as the user's walking distance increases. Therefore, this paper proposes a particle filter fusion-based guided trajectory tracking method, which combines pedestrian heading estimation and convolutional neural network-based landmark detection method to achieve real-time tracking of position and trajectory. The article uses the PDR method for estimation, including the number of steps and step lengths, to achieve the calculation of the position, collects the data of pre-set landmark points in the guide path by inertial sensors, and realizes the landmark point recognition based on CNN. The article designs two sets of experiments to analyze the fusion localization results, and compared with the traditional PDR method. The actual experimental results show that the localization error of this method is less than 1m, which effectively reduces the cumulative error of PDR.