Abstract: In order to solve the problem that surveillance videos were difficult to be unified in the image space and the crowd status was difficult to be perceived macroscopically and quantitatively, a new method was proposed based on GIS, which can be used to display, analyze and early warn crowd status through 2D map. Firstly, the surveillance video was accessed through the API of camera manufacturers or the standard video stream interface. Secondly, the video images were adjusted to a smaller size and the optical flow field in image space was calculated through some optical calculation algorithm. Finally, the optical flow field of each surveillance video was mapped to the corresponding geographic space. Then through GIS, the movement status of the crowd can be observed in the map. Two analysis methods, including scatter point interpolation and contour map, were put forward. The crowd alarm can be realized through relevant thresholds setting. Based on MATLAB, ArcGIS Engine and Microsoft Visual Studio C#, a prototype system was developed to validate the algorithm using several crowd videos. The results showed that the new method had the advantages of geographic location, measurement, macroscopic observation and early warning, etc. when compared with the previous methods.