Texture and Color Descriptor Features-based Vacant Parking Space Detection using K-Nearest Neighbors

The importance of the detection of vacant parking spaces is increasing gradually. A system capable of detecting vacant parking spaces in real-time can play an important role in saving valuable time for motorists, decreasing traffic jams, and reducing air pollution. Vision-based parking space detection methods are advantageous in terms of installation and maintenance as existing security cameras in a parking area can be used without the requirement of additional hardware and the detection program can be run on a local or a remote server. One major problem of the vision-based detection method in this context is making the model generalized for detection in various weather conditions. This research proposes a hybrid method to detect vacant parking spaces that use texture and color descriptors. A weighted KNN is used for the classification of parking spaces. The proposed method experimented on PKLot, a large benchmarking dataset that contains images of three parking areas in three weather conditions. The proposed model achieves an accuracy of 99.47% on average while training with 10-fold cross-validation and achieves an accuracy of 99.41% accuracy on average while testing with unseen data. The model shows robustness and better performance in terms of accuracy and processing speed. Several comparisons are also done to show how well it performs with methods found in previous research.

@article{Saif2024,
  title     = {Texture and Color Descriptor Features-based Vacant Parking Space Detection using K-Nearest Neighbors},
  journal   = {International Journal of Advanced Computer Science and Applications},
  volume    = {15},
  number    = {2},
  year      = {2024},
  publisher = {The Science and Information Organization},
  author    = {A F M Saifuddin Saif and Zainal Rasyid Mahayuddin},
  doi       = {10.14569/IJACSA.2024.0150203},
  url       = {https://doi.org/10.14569/IJACSA.2024.0150203}
}