Remote sensing technology has made significant advancements, leading to the widespread adoption of unmanned aerial vehicles (UAVs) in critical tasks such as smart cities, traffic surveillance, and disaster assistance. However, the identification of objects from drone-view images poses substantial challenges: 1) small object; 2) scale variance; 3) non-uniform distribution; 4) occlusion. To address these challenges, this paper introduces a novel strategy for remote sensing data collection via multiple UAVs combination, which can obtain drone-view images from various scales, angles and multi-regions in real time. Then we perform small object detection on drone-view images via a novel network structure, called Query-YOLOX, which combines the Cascade Sparse Query mechanism and the YOLOX anchor-free detector. Our novel Query-YOLOX model demonstrates remarkable advancements in small object detection for drone-view images by substantially reducing computation costs while preserving high detection accuracy. The experimental results conducted on the VisDrone2019 dataset validate the superiority of our approach, which now stands as the state-of-the-art solution for efficient and precise small object detection in drone-view imagery.