The widespread deployment of wireless devices in indoor environments has made location-based services a hot research topic. Indoor positioning based on received signal strength is the key to providing accurate location services among them. But it needs to build a fingerprint database for the sensing area. Therefore, whether it is to establish or update the fingerprint map, a large number of RSS values of the reference points need to be sampled. This process is time-consuming and laborintensive, with a huge amount of work. To solve this problem, researchers collect RSSs of some reference points and use them to interpolate other points to form a map of the entire area. This method can effectively reduce the time to create and update the map, but it also reduces the positioning accuracy. According to the propagation characteristics of wireless signals, the signal of the insertion points are formed by the superposition of multiple directional signals. Therefore, the correlation of neighboring points should not be considered only, but should be expanded in different directions. According to the propagation characteristics of actual signals, this paper designs a method based on multi-chain interpolation, which combines the influence of different propagation links on the insertion point to evaluate the signal strength. The basic idea of this method is to perform interpolation calculation in different directions under the given sampling rules. Then the predicted values of insertion points are obtained by using inverse distance weighting. Next, the corresponding signal attenuation Models are obtained by fitting in each direction and the errors are calculated as the direction weights. Finally, the estimation values of the insertion points are obtained. Through repeated iteration, the fingerprint database composed of real points and virtual points is finally formed. Two sampling models are used in this paper. And the sampling rates are 25% and 50% of that of full sampling, which means that the workload of map construction is reduced by 75% and 50% respectively. According to large-scale experiments, the positioning accuracy of the two MCI sampling methods is 13.58% and 4.74% higher than that of the full sampling method respectively. Compared with the classical interpolation method, the MCI method has better stability. Especially when the sampling amount is small, the advantage is more obvious. When the sampling amount is only 25%, the average accuracy is 18.50% higher than that of the full sampling method.