Safe navigation and detailed perception in unfamiliar environments are a challenging activity for the blind people. This paper proposes a cloud and vision-based navigation system for the blind. The goal of the system is not only to provide navigation, but also to make the blind people perceive the world in as much detail as possible and live like a normal person. The proposed system includes a helmet molded with stereo cameras in the front, android-based smartphone, web application and cloud computing platform. The cloud computing platform is the core of the system, integrates object detection and recognition, OCR (Optical Character Recognition), speech processing, vision-based SLAM (Simultaneous Localization and Mapping) and path planning, which are all based on deep learning algorithm. The blind people interact with the system in voice. The cloud platform communicates with the smartphone through Wi-Fi or 4G mobile communication technology. For testing the system performance, two groups of tests have been conducted. One is perception and the other is navigation. Test results show that the proposed system can provide more abundant surrounding information and more accurate navigation, and verify the practicability of the newly proposed system.