Creating a Surround View Algorithm with Implementation on a Real Embedded ADAS Platform

Computer Vision (CV) has seen significant advances, and autonomous driving technologies have been transforming the automotive industry in recent years. Autonomous vehicles require robust CV algorithms, while Advanced Driver Assistance Systems (ADAS) help the driver when driving and increase traffic safety. In this paper, the algorithm for generating a Surround View (SV) of the vehicle is proposed and implemented on a real embedded ADAS platform to test its performance in real hardware scenarios. The algorithm implements geometric operations like image distortion correction, perspective transformation, and image stitching to create a comprehensive and connected surround view. The subjective image quality evaluation results showed that the proposed algorithm achieves a fairly high quality of the final top-view image. The experiments also show that its operating speed when running on a real ADAS platform is suitable for cases when the vehicle speed is not high, such as when parking.