Comprehensive Anomaly Detection in Ghana: Addressing Social, Economic, and Security Challenges Through Transfer Learning in Traffic Surveillance
Isaac Opei, Isaac Azure, Anas Musah · SCIREA Journal of Traffic and Transportation · 2019
The abstract presents a comprehensive overview of the social and economic challenges faced by Ghanaians, including high inflation, currency depreciation, and a surge in violent crimes such as armed robberies and murders. It highlights the alarming increase in road accidents, emphasizing their significant impact on mortality rates and the country's GDP. The document expresses concern about the poor response time of security agencies to these anomalies, leading to reputational consequences and a potential deterrent for foreign investors. The Inspector General of Police aims to reduce response time, but existing solutions focus on specific types of anomalies. The abstract proposes using transfer learning, specifically the DenseNet121 method, to detect various anomalies in traffic surveillance videos. The study details the training process, dataset augmentation, and model evaluation. While the proposed model shows positive results in identifying anomalies in certain conditions, the abstract suggests the need for improved models to handle long-distance, poor coverage, and hazy environment scenarios in future research.