Computer Engineering and Applications ›› 2024, Vol. 60 ›› Issue (3): 119-128.DOI: 10.3778/j.issn.1002-8331.2304-0270

• Special Issue on Object Detection • Previous Articles     Next Articles

Remote Sensing Image Target Detection Algorithm Based on Improved YOLOv6

XU Degang, WANG Zaiqing, XING Kuijie, GUO Yixin   

  1. 1.School of Information Science and Engineering, Henan University of Technology, Zhengzhou 450001, China
    2.Key Laboratory of Grain Information Processing and Control, Ministry of Education, Henan University of Technology, Zhengzhou 450001, China
    3. School of Computer and Communication, Hunan Institute of Engineering, Xiangtan, Hunan 411104 , China
  • Online:2024-02-01 Published:2024-02-01



  1. 1.河南工业大学 信息科学与工程学院,郑州 450001
    2.河南工业大学 粮食信息处理与控制教育部重点实验室,郑州 450001
    3.湖南工程学院 计算机与通信学院,湖南 湘潭 411104

Abstract: Aiming at the low target detection accuracy caused by complex background of remote sensing images, generally small targets and multi-scale distribution of targets, a remote sensing image target detection algorithm based on improved YOLOv6 is proposed. Firstly, a coordinate attention module is introduced into the backbone network to improve the feature extraction ability and target location ability of the model under complex background. Secondly, a context enhancement module is proposed to enable the model to obtain rich context information, so as to improve the ability of model to extract multi-scale target details. Finally, in order to realize the adaptive fusion of different scale features, an adaptive spatial feature fusion is introduced into the neck network to improve the detection accuracy of multi-scale targets, especially small targets. The proposed algorithm is trained and tested on DOTA-v1.0, an open data set of remote sensing images. The experimental results show that the convergence speed and convergence accuracy of the improved algorithm are better than that of the original algorithm, and the AP value reaches 54.6%, which is 1.4 percentage points higher than that of the original algorithm. Meantime, compared with some other advanced target detection algorithms, accuracy and speed is improved which demonstrates the effectiveness of the improved algorithm.

Key words: remote sensing image, target detection, attention mechanism, multi-scale target, YOLOv6

摘要: 针对遥感图像背景复杂、目标普遍比较小且呈多尺度分布所导致的目标检测精度较低的问题,提出了一种改进YOLOv6的遥感图像目标检测算法。在骨干网络引入一种坐标注意力模块,以提高复杂背景下模型的特征提取能力和目标定位能力。提出一种上下文增强模块,使模型获取丰富的上下文信息,从而提升模型提取多尺度目标细节信息的能力。为了实现不同尺度特征的自适应融合,通过在颈网络引入一种自适应空间特征融合,提升了多尺度目标尤其是小目标的检测精度。将所提改进算法在遥感图像公开数据集DOTA-v1.0上进行训练并测试,实验结果表明,改进算法的收敛速度与收敛精度均优于原算法,其中AP值达到了54.6%,相比原算法提高了1.4个百分点,同时相比于一些其他目前先进的目标检测算法在精度和速度上均有提升,证明了改进算法的有效性。

关键词: 遥感图像, 目标检测, 注意力机制, 多尺度目标, YOLOv6