Aerial image small object detection network based on multilevel feature aggregation center

Small object detection in aerial images is a critical component of object detection tasks and plays a vital role in the field of computer vision. Unlike natural scene images, objects in aerial images occupy extremely small pixel areas, making them susceptible to background noise interference. Additionally, the loss and degradation of information during feature extraction and fusion processes are particularly detrimental to small object detection. To address these challenges, this paper proposes a small object detection algorithm based on a Multi-level Feature Aggregation Center. The model innovatively designs a Multi-level Feature Aggregation Center network (MFAC), which establishes a cross-layer feature direct connection mechanism to efficiently acquire and fuse multi-scale contextual information, effectively mitigating the loss and degradation of small object features during transmission. Furthermore, the model introduces a Cross-level Semantic Enhancement Module (CSEM) to enrich the semantic information of small objects, making their feature information more prominent. This module synchronously optimizes classification confidence and bounding box regression accuracy, achieving simultaneous enhancement of classification and localization information in small object detection tasks. It improves the regression accuracy of bounding boxes while ensuring object recognition precision. This paper conducts experiments on the VisDrone2019 datasets to validate the performance and structural superiority of the proposed algorithm. The experimental results demonstrate that the proposed method can effectively improve the accuracy of small object detection and exhibits competitive performance compared to mainstream algorithms.