Localization of Autonomous Vehicles: Proof of Concept for A Computer Vision Approach

TL;DR

This paper presents a visual-based localization system for autonomous vehicles using a single camera, combining image retrieval with Kalman filtering to achieve approximately 2-meter accuracy on the KITTI dataset.

Contribution

It introduces a novel two-step localization method that relies solely on visual data and a database of geotagged images, eliminating the need for complex hardware.

Findings

Achieved an average localization accuracy of 2 meters.

Successfully implemented the system using Python libraries.

Validated on the KITTI dataset with promising results.

Abstract

This paper introduces a visual-based localization method for autonomous vehicles (AVs) that operate in the absence of any complicated hardware system but a single camera. Visual localization refers to techniques that aim to find the location of an object based on visual information of its surrounding area. The problem of localization has been of interest for many years. However, visual localization is a relatively new subject in the literature of transportation. Moreover, the inevitable application of this type of localization in the context of autonomous vehicles demands special attention from the transportation community to this problem. This study proposes a two-step localization method that requires a database of geotagged images and a camera mounted on a vehicle that can take pictures while the car is moving. The first step which is image retrieval uses SIFT local feature descriptor to find an initial location for the vehicle using image matching. The next step is to utilize the Kalman filter to estimate a more accurate location for the vehicle as it is moving. All stages of the introduced method are implemented as a complete system using different Python libraries. The proposed system is tested on the KITTI dataset and has shown an average accuracy of 2 meters in finding the final location of the vehicle.