GNSS has important applications in land vehicle positioning services. However, in special harsh scenarios such as urban canyon with strong occlusion, the stability of the GNSS signal will be significantly reduced, and the positioning result usually has a lower update rate, such as 1HZ. Therefore, GNSS can be combined with the inertial navigation system (INS) to achieve higher update rate and more stable positioning. GNSS and INS can complement each other in signal frequency and positioning accuracy, so well-positioned results can be obtained even with less expensive devices. An integration positioning algorithm that can suppress the accumulation of positioning errors in the case of GNSS outage is proposed in this paper. The algorithm is based on a technique called Partial-ZUPT and a sliding window polynomial predictor. Traditional loosely coupled integration is used in the algorithm. The experimental results show that the quadratic positional error growth can be suppressed to the first-order growth by applying the algorithm when the GNSS positioning information is invalid. In terms of positioning accuracy, the integration algorithm only shows a cumulative position error of 0.5m after about 5 seconds of the GNSS outage, which is a significant improvement compared with the RMS error of about 1.5m in the traditional loosely coupled integration algorithm.