Deep Learning-Based Road Accident Severity Detection and Recommendation System Using Image Analysis

Abstract

Road accidents are one of the leading causes of fatalities and injuries worldwide, making timely detection and severity assessment crucial for saving lives. This project proposes a deep learning-based system that automatically detects road accident severity from images and provides appropriate recommendations. The system integrates computer vision techniques with machine learning and deep learning models to classify accident severity into different categories and assist in decision-making. The proposed system utilizes a Convolutional Neural Network (CNN) model trained on a dataset of accident images to perform classification. The dataset is preprocessed through resizing, normalization, and labeling to improve model accuracy. In addition to CNN, traditional machine learning algorithms such as Support Vector Machine (SVM), Decision Tree, and Random Forest are also implemented to compare performance metrics like accuracy, precision, recall, and F1-score. The application is developed using the Flask web framework, enabling users to upload accident images through a user-friendly interface. Once an image is uploaded, it is processed and passed through the trained CNN model to predict the type of accident. The system further analyzes the severity of injuries by detecting red regions in the image using HSV color space, which often corresponds to critical damage or injuries. Based on this analysis, the system categorizes the severity into “Minor” or “Major.” An additional feature of the system is the recommendation module, which provides safety measures and emergency responses based on the predicted accident class. This enhances the system’s practical usability by assisting first responders and authorities in taking appropriate action. The experimental results demonstrate that the CNN model outperforms traditional machine learning models in terms of accuracy and reliability. Visualization techniques such as confusion matrices and performance graphs are used to evaluate and compare model effectiveness. Overall, the proposed system provides an efficient and automated approach to accident severity detection, reducing response time and potentially saving lives. It can be extended further by integrating real-time video analysis and IoT-based alert systems for smarter transportation and emergency response solutions.