The aesthetic design of urban landscape decorations involves multiple complex design parameters and objectives, and it is difficult for traditional optimisation methods to effectively deal with highly nonlinear and multi-objective design problems. Therefore, the study combines genetic algorithm and particle swarm optimization algorithm to construct an aesthetic design system based on multilevel optimization algorithm for urban landscape decorations, which performs multidimensional synergistic optimization of morphology design, colour matching and structural stability. The experimental results show that GA-PSO is outstanding in terms of aesthetic design effect, convergence speed and computational efficiency. The aesthetic score after system optimisation is 92.5, which is significantly better than GA (89.0) and PSO (85.3), and the convergence time is 45.2 s, which is more computationally efficient than other algorithms. Among the multi-dimensional scores, the average scores of design aesthetics, functional rationality and environmental integration are 4.58, 4.40 and 4.45 respectively, showing good cross-disciplinary applicability. In particular, image processing experts rated the design aesthetics as high as 4.8, reflecting the advantages of the system in visual optimisation. The design system based on the multi-level optimisation algorithm can generate high-quality urban landscape design solutions quickly and stably while ensuring the global optimum, and can provide an effective solution for the aesthetic optimisation of urban landscape decorations.