In the structured operation environment, the application of warehouse automation equipment can greatly improve the operation efficiency. But in general, there is a lot of uncertainty in the warehouse. For example, the position and pose of the pallets in the warehouse have large uncertainty. Applying the RGB-D sensor and lidar sensor, the pallet pose and position can be estimated. After pallet recognition, forklift utilizing path planning technology can achieve picking up the pallet independently in dried fruit workspace, which improves the flexibility of the forklift. Taking the forklift with nonintegrity constraint characteristics as the research object, a pallet picking path planning method was proposed based on uniform cubic Bspline curve. Considering the multiple constraints in pallet picking process, such as minimum turning radius of the forklift, startpoint and endpoint constraints and the curvature continuity constraint, the objective function of path curvature minimization was established. The related curve parameters needed to be optimized were solved by the optimization toolbox in Matlab. The simulation results showed that for different pallet picking scenarios, the proposed method can obtain a feasible path with continuous curvature. The steering wheel angle of the planning path did not exceed the maximum of the forklift. In the warehouse, the path planning and tracking test were carried out. In the scenario where the driving distance was 6500mm, the lateral offset distance was 1500mm, and the angle offset was 15°, the results showed that at the end point of path, the transverse offset error of the path tracking was 4.71cm and the heading angle error was 9.6×10-3rad. The feasibility of the algorithm was verified, and it can be applied in similar large warehouse intelligent vehicles. The research can provide reference for the path planning of forklift in automated warehousing operation.