In order to improve the accuracy and robustness of the navigation system of forest orchard mobile robot, a real-time navigation method was proposed based on high and low frequency dual source information fusion between orchard rows based on laser radar three-dimensional point cloud. Firstly, the spray robot carried a 3D laser radar to collect the point cloud information of the fruit trees on both sides, and the original point cloud data were preprocessed such as through filtering, down sampling, and statistical filtering, so as to retain the point cloud of the fruit tree canopy in the region of interest. Then the inter row navigation lines fitted by the Newton interpolation algorithm based on high-frequency update and the non-linear support vector machine (NSVM) algorithm based on low-frequency update were complementary fused. Finally, when the navigation line was switched, the stability of the merged navigation line was optimized, and a cubic B-spline algorithm was used to smooth the navigation line. The experimental results showed that the maximum curvature of the fusion optimized navigation line was 0.048m-1, and the average curvature was 0.018m-1. The fusion optimized navigation line was tracked at the speed of 0.5m/s and 1.0m/s, respectively, and the maximum absolute lateral deviation was 0.104m and 0.130m, and the average value was 0.053m and 0.049m, respectively, which showed that this navigation method can meet the requirements of autonomous navigation of operating equipment between rows in the orchard, which provided a technical reference for autonomous navigation of spray robots in the orchard environment.