With the increasing popularity of computer vision, in the context of the development of Computer Vision(CV) and robot technology today, computer vision has more applications in many fields, such as in space recognition technology, accurate space recognition is important for robot navigation、positioning、route planning. However, traditional computer vision-based spatial recognition relies heavily on multiple cameras or depth sensors, which is costly and complex. Therefore, this paper proposes a visual landmarks map-based spatial recognition using monocular camera which extracts the spatial information of objects in space through video, and perform operations such as noise reduction and de-duplication filtering on the data to improve the matching rate of landmarks and spaces and the accuracy of space recognition. In order to verify the performance of the proposed algorithm, a series of tests are carried out. The results show that the map landmarks can achieve a high degree of similarity to the real situation in real space, and compared with previous algorithms, the method improves the success accuracy to 87.5%.