Intelligent sensors serve as crucial elements in the realm of smart car mobility solutions and urban sensing technology. This paper presents a novel automotive environment perception system that utilizes a binocular vision sensor. The binocular camera is employed to capture images and obtain cloud points for obstacle perception and environment positioning. The proposed system is built on a low-power embedded platform but maintains a high perception performance. It can accurately identify and locate obstacles, such as vehicles and pedestrians. The complete system is comprehensively described, encompassing the hardware structure, software architecture, and algorithm program. Furthermore, the process of the obstacle detection algorithm, which relies on disparity space and deep learning, is thoroughly presented. The feasibility of the fast stereo-matching algorithm is demonstrated theoretically and validated through experimental verification. Extensive experimental results indicate that the system is capable of delivering reliable and precise real-time environmental perception for intelligent vehicles. Consequently, the system can be readily implemented in commercial real-time intelligent driving applications. As a pertinent research in urban sensing applications, this system holds promise as a viable solution for enhancing smart mobility.