Mileage Extraction From Odometer Pictures for Automating Auto Insurance Processes

Abstract

For an insurance company is a priority to supply customers with an easy and streamlined way to provide all the information needed when reporting a claim or asking for a quote. A simple and efficient process to do so improves customer experience, reduces human error and accelerates the information collection process. An accurate mileage reading is a key piece of information that is relevant for auto insurance quotes and claims processing. The vehicle mileage can be combined with the License Plate number and the Vehicle Identification Number (VIN) to get a complete overview of the information needed for many insurance processes and workflows. In this paper, we describe a novel solution for extracting vehicle mileage from odometer images taken by mobile devices. There are many available low-cost commercial solutions for both License plate recognition and VIN recognition from images, however, to the best of our knowledge, this is not existing commercial solutions for odometer mileage extraction from images. Our proposed system mainly consists of two parts: (a) identifying the odometer display and, (b) extracting characters inside the display. We leverage existing state-of-the-art object detection deep learning architectures to solve each part and design a post-processing algorithm to identify mileage from the extracted characters. We tested empirically our proposed system in unseen odometer images taken in the wild. We achieve satisfactory performance that meets the requirements needed for real-life applications in the insurance industry.