Trajectory representation method and system based on deep learning

Abstract

The invention relates to a trajectory representation method and system based on deep learning. The method comprises the following steps: coding sampling points in a trajectory by using a neural network; using convolution operation to extract spatial features of the trajectory; and using an LSTM model to extract time sequence features of the trajectory. Through combination of the neural network and the LSTM, the spatial-temporal features of the trajectory are coded into dense semantic vectors, and calculation of the semantic vectors is equivalent to calculation of the original trajectory, so that the problem that a traditional trajectory calculation method excessively depends on the quality of sampling points is solved, the method can be used as the basis of a machine learning model in high-precision map making, and the model performance can be improved.

Description

technical field [0001] The invention belongs to the technical field of high-precision map production, and in particular relates to a trajectory representation method and system based on deep learning. Background technique [0002] Guidance line generation is a pain point in the production of high-precision maps, serving intelligent driving. The traditional kinematics-based method generates guidelines through mathematical models, but does not consider the real scene information of the road in actual use. The actual road scene is complex and changeable, and the guidelines generated by mathematical models are difficult to apply. In crowdsourced high-precision maps, trajectory data-driven guidance line generation is a promising solution to the above problems, and one of the most critical technologies is trajectory representation. It requires the extracted trajectory representation vector to preserve the spatio-temporal features in the original trajectory. Contents of the inve...

AI Technical Summary

Problems solved by technology

[0003] In order to solve the problem that the traditional guideline generation process cannot adapt to the actual road scene and the spatio-temporal attributes of the guideline are not preserved, in the first aspect of the present invention, a trajectory representation method based on deep learning is provided, including: acquiring trajectory sampling point data , use the neural network model to encode the sampling points in the trajectory into vectors of fixed dimensions; for the encoded trajectory sampling points, use the convolution and pooling operations of the neural network model to extract the spatial feature vector; the spatial feature vector Input into the LSTM model, the LSTM model obtains the trajectory semantic vector with spatiotemporal features by extracting the temporal features in the spatial feature vector

Images