Insurance Fraud Identification using Computer Vision and IoT: A Study of Field Fires

Abstract

Abstract Insurance fraud detection has always been manual labor relegated to claim agents, who examine the facts and reach on an intuition- based conclusion. The following article proposes an automated solution to regulate the process of fraud detection of field fire-based insurance claims in the agricultural sector. The proposed work is an amalgam of computer vision, deep learning, and Internet of Things and aims to inculcate the positives of each of these technologies. To the best of our knowledge, a combination of the said technologies has never been used for insurance fraud analyses in the field of agriculture, making this a novel approach. The proposed model actively reads the input from the IR and temperature sensors of the IoT device, which further collects images of the field once the sensor values cross their respective thresholds. The collected images are then fed into a fire detection model trained using a variety of classifiers for performance comparisons. The results display that the proposed solution has an accuracy of 97%, which can be further increased with a refined dataset dedicated solely to fraud detection.