This solution could stitch four images from wide-angle cameras of different locations, and achieve MOD, LDWS, BSD, ODA, and Stop and Go functions to improve driving safety and reduce possibility of accidents. According to lightweight and merging deep learning architectures, computer vision, computer graphics, heterogeneous multi-core scheduling, and memory access optimization technologies to solve challenges in image quality, variable driving environments, and hardware resource requirements. This solution could speed up time to marking, reduce R&D costs, and enhance product differentiation and competitiveness.