In order to standardize parking, city managers mark parking areas with rectangular solid white lines on the pavement to remind riders to park their bicycles within the fence. A new method uses the K-means clustering principle for fence of shared bicycles is proposed. First, the location data of Gauss distribution are simulated. Second, to get multiple clusters, the data of each location are clustered, and then by means of continuous iteration, the final center of each cluster is determined. Finally, the data of the largest cluster of central points for entry fence detection are analyzed. The accurate rate of entry fence detection is verified by the actual measurement of location data of shared bicycles. The results show that the algorithm can accurately detect whether the shared bicycles enter the fence under the prescribed conditions, accuracy is as high as 80%-100%, and has high practical value.