ABSTRACT
The daily toll on society is high due to car accidents. A combination of careless driving and a lag in reaction time from emergency services is the most common cause. Reliable early warning and information exchange systems are life-saving for individuals injured in traffic accidents. An urgent system that notifies nearby emergency services of an accident's exact location must be put into place without delay. Multiple authors in scholarly journals have proposed different automatic accident detection algorithms. The four steps that make up the proposed method are data preparation, model training, feature extraction, and segmentation. Data cleansing and encoding technology make up preprocessing. Because of its strong association with accident frequency, the AADT is a crucial variable in road segmentation. Among the current features, Histogram of Oriented Gradient (HOG) offers the best performance. The usage of DML resulted in a more accurate model training process. A 95.40% accuracy rate was achieved using the suggested method.
