A trajectory recovery method based on deep learning and Kalman filter correction

Abstract

The invention discloses a trajectory recovery method based on deep learning and Kalman filter correction, and the specific steps include: S1 discretization of trajectory points; S2 cyclic neural network and trajectory modeling; S3 trajectory recovery; S4 utilizing spatiotemporal attention mechanism, Obtain an attention model based on the sequence-to-sequence model; S5 combines the Kalman filter and the cyclic neural network, introduces the Kalman filter to optimize the mean square error, and jointly trains the Kalman filter and the attention model to obtain the final model. The present invention provides a trajectory recovery method based on deep learning and Kalman filter correction, which uses a recurrent neural network to model the transfer law between trajectory points, uses the attention mechanism in deep learning to help track recovery, and finally introduces Kalman filtering is used to model the movement of objects in time and space, which reduces the uninterpretable and error of the deep learning model, has stronger interpretability, and reduces the error of trajectory recovery.

Description

technical field [0001] The present invention relates to the field of trajectory data mining, and more specifically relates to a trajectory recovery method based on deep learning and Kalman filter correction. Background technique [0002] The biggest difference between the current solution to the trajectory recovery problem and the previous trajectory recovery scheme is that the existing solutions are all data-driven. By mining and analyzing a large amount of historical data, it is possible to restore the trajectory from a low sampling rate to a high sampling rate very accurately. [0003] However, the existing schemes are still not enough for the utilization of data. Most of the solutions stop at simply counting the data, and then search for the optimal trajectory through complex heuristic search schemes. These schemes cannot capture to complex transfer laws between trajectory points. [0004] On the other hand, with the rise of deep learning, although the deep learning mo...

AI Technical Summary

Problems solved by technology

[0003] However, the existing schemes are still not enough for the utilization of data. Most of the solutions stop at simply counting the data, and then search for the optimal trajectory through complex heuristic search schemes. These schemes cannot capture to complex transfer laws between trajectory points

[0004] On the other hand, with the rise of deep learning, although the deep learning model can capture the complex laws of the data and can integrate multiple information to help, the behavior of the deep learning model is difficult to explain, and it cannot show the modeling Spatiotemporal information, which can also cause problems when modeling trajectories

Images