Zhang, Y., Yin, Y., Zimermann, R., & Yu, Z. (2023). A Structure-Aware Method for Crowd-Sourced Sparse-to-Dense GPS Trajectory Image Generation. 2023 IEEE International Conference on Multimedia and Expo Workshops (ICMEW). https://doi.org/10.1109/icmew59549.2023.00092
Abstract:
Crowdsourced vehicle trajectories are valuable resources to interpret city road networks in a fresh and live
manner. Recently, trajectories have been popularly used together with other image data to facilitate various downstream location-based services, among which the very first step is to convert the trajectories data to an image layer. To achieve satisfying results, all these works require densely sampled GPS trajectories which is however sometimes unavailable. So this study explores the possibility to generate dense trajectory images from sparse trajectory data. In order to address the unique features of the trajectory images, we introduce a spatial structure-aware model with a novel triplet-based training strategy and a global-local loss. Experimental results on real-world trajectories show that by using just 1% of trajectory data we can restore most of the original information.
License type:
Publisher Copyright
Funding Info:
This research / project is supported by the Ministry of Education - Academic Research Fund Tier 1
Grant Reference no. : T1 251RES2029
This work was supported by the National Natural Science Foundation of
China (No. 61960206008, 62272390, 62025205, 62072375, 62032020)