This research introduces a trajectory smoothing technique based on convolution parameterization to enhance the performance of the widely used Dubins path in UAV trajectory planning. While traditional Dubins paths are favored for their simplicity and optimality in terms of the shortest path, they suffer from issues such as abrupt changes of angular velocity and discontinuities in radial acceleration. Our method employs a convolution-based transformation process on the original reference trajectory, smoothing high-order derivatives while maintaining the inherent advantages of the original trajectory. To ensure the effective application of the convolution operation, we selected the kernel width through analysis of three distinct window-sizes and preprocessed the path data, thereby mitigating acceleration discontinuities that may occur at the trajectory endpoints. The results show that our method effectively eliminates high-order derivative discontinuities while preserving the essential details of the original Dubins path, achieving smooth motion with continuous curvature.