An Online Learning System for Wireless Charging Alignment Using Surround-View Fisheye Cameras

Electric Vehicles are increasingly common, with inductive chargepads being considered a convenient and efficient means of charging electric vehicles. However, drivers typically poor at aligning the vehicle to necessary accuracy for charging, making automated alignment two plates desirable. In parallel electrification vehicular fleet, parking systems that make use surround-view camera becoming popular. this work, we propose system based on architecture detect, localize, automatically align chargepad. The visual design is not standardized necessarily known beforehand. Therefore, relies offline training will fail in some situations. Thus, self-supervised online learning method leverages driver’s actions when manually chargepad combine it weak supervision from semantic segmentation depth learn classifier auto-annotate video further training. way, faced previously unseen chargepad, driver needs only single time. As flat ground, easy detect distance. using Visual SLAM pipeline landmarks relative enable greater range. We demonstrate working an as illustrated https://youtu.be/_cLCmkW4UYo . To encourage research, share dataset used work (an initial version shared xmlns:xlink="http://www.w3.org/1999/xlink">https://drive.google.com/drive/folders/1KeLFIqOnhU2CGsD0vbiN9UqKmBSyHERd here).