A traclet is an image representation of a trajectory. This representation is indicative of the mobility patterns of the moving objects. TraClets need to efficiently visualize and capture two key features that characterize the trajectory patterns of moving objects: i) the shape of the trajectory which indicates the way the object moves in space, and ii) the speed that indicates how fast the object moves in space.
python traclet.py --d [dataset_path] --s [size of the resulting images]
python traclet.py --d dataset.csv --s 224
After the execution, a folder called traclets is created that contains one folder per label in the dataset. Each label folder contains the respective trajectory images.
- numpy
- pandas
- opencv-python
- bresenham
The repository also contains a file that uses the RandomForests classifier to perform k-fold cross-validation on the generated traclets. The features used for the classification are:
- Color histogram
- Hu invariant moments
python classifer.py --d [dataset_path] --f [number of folds]
python classifer.py --d traclets --f 5
- numpy
- opencv-python
- imutils
- scikit-learn
1. I. Kontopoulos, A. Makris, D. Zissis, K. Tserpes, "A computer vision approach for trajectory classification", 22nd IEEE International Conference on Mobile Data Management (MDM), 2021.
2. I. Kontopoulos, A. Makris, K. Tserpes, "A Deep Learning Streaming Methodology for Trajectory Classification", ISPRS International Journal of Geo-Information, Volume 10, Issue 4, 2021.