Road Accident Detection using SVM and Learning: A Comparative Study

Everyday, a great deal of children and young adults (aged five to 29) lives are lost in road accidents. The most frequent causes are a driver’s behavior, the streets infrastructure is of lower quality and the delayed response of emergency services especially in rural areas. There is a need for automatics road accident systems detection that can assist in recognizing road accidents and determining their positions. This work reviews existing machine learning approaches for road accidents detection. We propose three distinct classifiers: Convolutional Neural Network CNN, Recurrent Convolution Neural Network R-CNN and Support Vector Machine SVM, using a CCTV footage dataset. These models are evaluated based on ROC curve, F1 measure, precision, accuracy and recall, and the achieved accuracies were 92%, 82%, and 93%, respectively. In addition, we suggest using an ensemble learning strategy to maximize the strengths of individual classifiers, raising detection accuracy to 94%.

@article{Qanouni2024,
  title     = {Road Accident Detection using SVM and Learning: A Comparative Study},
  journal   = {International Journal of Advanced Computer Science and Applications},
  volume    = {15},
  number    = {5},
  year      = {2024},
  publisher = {The Science and Information Organization},
  author    = {Fatima Qanouni and Hakim El Massari and Noreddine Gherabi and Maria El Badaoui},
  doi       = {10.14569/IJACSA.2024.0150565},
  url       = {https://doi.org/10.14569/IJACSA.2024.0150565}
}