无线电工程

针对物流无人机尚未建立根据应用场景差异开展特性研究的问题，提出一种应用于西南口岸城市末端，即丘陵、山地地形下城郊与乡村地区物流无人机系统。基于物流配送需求，提出一种“路径规划+图像配准投递”的双目标模型。利用A^*算法与地面危险区域分类，采用一种保证飞行安全的“跳跃式”路径规划策略；利用“模板+SURF”算法开展主动判断，快速定位卸货点并完成卸货。基于实际场景的验证性试验，建立的路径规划模型在避让山火隐患区域的前提下，较人工配送效率提高50%以上；所建立图像配准模型在耗费14%机载电源的条件下可实现88%的配准准确率，相比传统SIFT算法耗电量增加7%、准确率提高57%;相比Desnet算法耗电量降低18%、准确率降低7%。所设计系统在丘陵与山区区域，以及房屋相似度高的区域可大大提高物流无人机的配送与投递准确度。

As the lack of research on scenario-specific characteristics of logistics drones, a last-mile drone model is proposed for application in southwest region, particularly in suburban and rural areas with hilly and mountainous terrains. For the needs of delivery process, the proposed dual-objective model introduces a “path planning + image registration delivery” approach. To ensure flight safety, the model uses the A^* algorithm and ground hazard zone classification for “jumping” path planning strategy. The “template + SURF” algorithm is employed for quick localization and efficient unloading. The proposed path planning model has been validated through tests in real scenarios while avoiding the fire hazard area. It demonstrates a more than 50% improvement in efficiency compared to manual delivery. The proposed model can achieve 88% alignment accuracy with 14% of power consumption, which is 7% higher than the traditional SIFT algorithm in terms of power consumption and 57% higher in terms of accuracy. Compared with the Desnet algorithm, it is 18% lower in terms of power consumption and 7% lower in terms of accuracy. The designed system can greatly improve the distribution and delivery accuracy of logistics UAVs in hilly and mountainous areas, as well as in areas with a high degree of housing similarity.