In order to obtain a reasonable and efficient sprayer UAV’s path in geometrically complex farmland, a new path planning approach was proposed based on the improved ant colony optimization. The 3D spray paths were first built up by discretizing the parallel scan lines inside the boundary and interpolating on the 3D terrain surface. A spray path generation algorithm was then designed to find a set of shortest and least spray paths with the optimized heading. Secondly, the ant colony optimization was improved to sort the 3D spray with the objective of getting the shortest transfer paths after adding the new function of recording starting point of each spray path. Furthermore, the prospered path planning approach was tested with the example. Different headings were calculated in the same farmland depending on whether the 3D terrain was considered, which illustrated the necessity of the 3D terrain in path planning problems of sprayer UAV. The improved ant colony optimization was compared with the greedy algorithm. Aiming at the shortest transfer paths, the two algorithms were used to sort the same set of spray paths. The total lengths of the transfer paths calculated by the improved ant colony optimization were 3% to 28% shorter than those results calculated by the greedy algorithm under the conditions of the same selected starting point. The minimum values obtained by the improved ant colony optimization and the greedy algorithm were 1661m and 1763m, respectively. The actual application was basically consistent with the trends and situations shown by the example, which showed the better optimization performance of the improved ant colony optimization. In addition, the boundaries and terrain of the two farmlands in the example and practical application were complex enough to indicate that the proposed path planning approach had practicality.