Fig 1.
MTDP-FD framework.
Fig 2.
Transform trajectory data into trajectory images.
The input is a GPS trajectory sequence Tr, and the output is a two-dimensional trajectory image Is.
Fig 3.
Frequency domain processing of trajectory image.
Fig 4.
Process of extracting deep features from trajectory image.
Fig 5.
(a) and (b) are two processes of convolution.
Fig 6.
Max pooling.
Fig 7.
The processes of RNN prediction.
Fig 8.
The prediction accuracy of different D0.
Fig 9.
The prediction accuracy of different D1.
Fig 10.
The prediction accuracy of different D2 and D3.
Fig 11.
The prediction accuracy of different input.
Fig 12.
The prediction accuracy of different combinations.
Table 1.
Comparison results.
Fig 13.
The spatial domain image and its deep features.
Fig 14.
The low-frequency representation of image and its deep features.
Fig 15.
The high-frequency representation of image and its deep features.
Fig 16.
The intermediate-frequency representation of image and its deep features.