GPGPU Based ADAS application Framework on Smartphone

초록

This paper proposes a novel framework to utilize both CPU and GPU resource to facilitate the execution of ADAS (Advanced Driver Assistance System) application on a mobile platform. In this framework, the OpenGL ES 2.0 based shading language is used for video processing, where each frame of an input video is loaded as texture and the result of execution is displayed from a frame buffer, so that the image data-transferring is simplified. The framework also includes the usage of various information from different sensors such as GPS, camera, accelerometer, gyroscope, etc. The proposed framework is developed for Android smart phones and demonstrates acceptable processing speed for a personalized ADAS system. The developed framework is expected to be further adopted in real-time detection of moving vehicles and pedestrians for intelligent vehicles using mobile platforms.