We are developing an SoC FPGA-based unmanned mobile vehicle for the FPGA design competition. For the vehicle to follow roads successfully, it must be able to detect not only straight lines but also curved lines accurately. Therefore, we implemented a lane detection algorithm that is robust not only against straight lines but also against curves to improve driving performance. We implemented an autonomous driving system employing this algorithm on Digilent Zybo Z7-20. We evaluated the lane detection algorithm based on simulations and showed that this algorithm can reduce false detection of lane features compared to the classical Canny filter.