基于改进YOLOv7的遥感光学图像飞机目标检测

打开文本图片集

中图分类号：TP391 文献标志码：A 文章编号：1673-5072（2025）04-0434-07

Aircraft Target Detection Based on Improved YOLOv7 on Remote Sensing Optical Images

ZHANG Yan-yue，DAI Xian-zhi

（SchoolofElectronicInformation Engineering，China WestNormal University，Nanchong Sichuan6379，China）

Abstract：Aircraft target detection isof great significance for aircraft control at airports.However，inremote sensing images，there are always some problems such as smal-sized or obscured aircrafts，which leads to missed detection due to the missng aircraft target features.In order to reduce the missed detectionrateof aircrafts inremotesensing images，animproved aircraft target detectionalgorithm forremote sensing images combined with theattention mechanism is proposed based on YOLOv7.The algorithm improves the Backbone structure to enhance the extraction ability of small targets’features.At the same time，the design of the detection head isadjusted to reduce the number of model parameters and increase the detection speed for the characteristics of smal targets in remote sensing images.In the public CORSADD dataset，ablation experiments are caried out according to diffrent improvement strategies.The experiments show that the refinement of Backbone and detection head atthe same time is best for the target detection.Compared with the improvement before，the average precision value reaches 85.12% ，which is an increase of 3.04% ，and the detection speed reaches 48.16 FPS，which is an increase of 14.3% . In conclusion，the improved algorithm effctively increases the detection accuracy and speed of aircraft small targets in remote sensing images，which reduces the missed detection rate of aircraft targets.

Keywords：YOLOv7;attention mechanism;remote sensing image;aircraft detection;smalltarget; CORSADD dataset

遥感图像飞机目标检测在机场交通管控等领域具有重要意义，在民事领域可以对飞机起降、停放和运行状态进行实时监控，优化航班调度、提高机场运营效率等;在军事领域通过遥感图像对机场飞机目标进行准确检测，帮助军事部门进行战略规划、情报分析和敌情监测等。（剩余9065字）