Compared with structured environments, unstructured environments are often more complex, and narrow passages, as a smaller subset of the complex environment, have weak connectivity. In this paper, a grid bridge narrow passage detection algorithm is proposed to solve the problems of collision detection, time consuming and low sampling efficiency of the current widely used bridge detection method, and on this basis, a free-space roadmap construction method and a path adjustment strategy are designed. In the experimental tests of two different scenarios, the path planning time is in milliseconds and 10 milliseconds respectively, demonstrating the real-time implementation value of the proposed method. Meanwhile, the algorithm proposed in this article also shows some superiority in path construction success rate and path optimization.