Traffic monitoring system with emergency support using SOM

• Main Author: Tan, Hoai Thang
• Format: Final Year Project / Dissertation / Thesis
• Published: 2020
• Subjects: Q Science (General)
• Online Access: http://eprints.utar.edu.my/3917/1/16ACB02229_FYP.pdf; http://eprints.utar.edu.my/3917/
• Institution: Universiti Tunku Abdul Rahman

With the advance of science and technology, everyone will buy a car for convenience. With more cars on the road, emergency vehicles such as ambulances are having a hard time bypassing busy lane. Therefore, with the help of the vehicle emergency alarm system, the driver can stay alert and move sideways, enabling the emergency vehicle to reach the destination as soon as possible. This paper proposed a system of emergency notification to alert the driver. By using SelfOrganizing Map technique, the ambulance siren can be localized based on the detected ambulance siren. Also, Support vector machines were used to classify the presence of ambulance siren and further support self-organizing maps. Next, A mobile app has been developed and installed on a smartphone to forward the results to drivers on the other side of the road. Several processes will implement such pre-processing, as well as feature extraction, for better classification and localization process. In the classification section, all the processed parameters are input into the classification algorithm to classify the groups to which the input parameters belong. Lastly, the classification results are divided into two groups: whether there is an ambulance siren or not. If there is an ambulance, siren localization will be carried out and output the distance of the ambulance siren. In addition, the results are uploaded to an online database and notified to drivers on the road. To examine the reliability of the system, the system was tested on the St. John ambulance siren dataset, which is the real-world ambulance siren collected at outdoors. Based on the test results, the application was able to perform localization on the St. John ambulance dataset with an average accuracy of 98.0%. Lastly, to test the performance of the system to detect the presence of ambulance sirens, the Kaggle online dataset was used for testing and with an average accuracy of 96%.