To satisfy the real-time precise positioning requirement of single-system single-baseline with dual-frequency data, a novel approach based on dual-frequency uncombined data is proposed. The single-system single-baseline and dual-frequency uncombined Real Time Kinematic (RTK) is realized based on inter-station-inter-satellite carrier phase and pseudorange measurement model. Based on the analysis of the residual errors of the double difference model, the state vector is established, and then the state prediction equation and the measurement equation are derived. The real-time transformation matrix of the state vector is established at each epoch, and the weights of observations including pseudorange and carrier phase are adjusted by a random model. Then extended Kalman filter is used to estimate the position of the user receiver. The approach is tested with several groups of medium-range and long-range single-baseline data which are the actual collection of satellites in the experiment. The effectiveness of the approach is verified by examining the three-dimensional errors of the positioning result as well as the success rate of the ambiguities. It is shown that when the baseline is 135.6 km, the positioning accuracy of the proposed approach can reach the centimeter level with the 97.3% success rate of the ambiguities, and the CEP95 positioning errors of east and north and sky directions are 1.35 cm, 1.84 cm, 7.08 cm respectively. This approach makes full use of the advantages of differential technology to eliminate the relative errors independent of distance and effectively avoid the noise caused by the linear combination of the multiple frequency measurements and can satisfy real-time precise positioning for medium and long single-baseline.