Vision-Based Parking Assist System with Bird- Eye Surround Vision for Reverse Bay Parking Maneuver Recommendation

摘要

In recent years, the automotive industry has seen rapid growth in autonomous vehicle and advanced driver- assistance system (ADAS) technologies. One of the key aspects of this technology is the parking assist system. The core function of a parking assist system is to help drivers in parking their vehicles safely and efficiently. This paper proposes a method for low-cost real-time semi-autonomous vehicle parking in which a Vision-Based Parking Assist System (VPAS) gives the driver maneuver recommendations for reverse bay parking. The proposed method integrates wide-angle lens correction, bird-eye surround view, and user-guided vision-based parking line detection as ways to calculate recommendations for the driver. VPAS aims for affordability and generality, thus it emphasizes streamlining most of the calculations to compensate for realtime needs while maintaining reliability in a relatively low-cost system that can be run on an entry-level traditional hatchback car. To verify the performance of the proposed methods, experiments on the integrated system are conducted in a controlled environment. From experimental results, we demonstrate the performance of VPAS in terms of real-time and accuracy aspects while running on a 4-cores 3.5 GHz processor powered by the car battery.