基于改进YOLOv8的无人机航拍检测算法

Drone aerial photography detection algorithm based on improved YOLOv8

摘要:

针对当前的目标检测算法对于无人机影像存在背景复杂、目标较小而造成的漏检误检等问题，提出了一种基于YOLOv8的无人机航拍检测改进算法YOLOv8-A；为了提高算法对于小目标的检测能力，使用全局通道-空间注意力模块来提高输入特征图的表达能力；为了进一步提高算法对于捕获输入数据的上下文信息和内在特征的能力，在颈部结构更换了RT-DETR算法的AIFI模块，有助于提取上下文的信息；为了提高算法对于困难样本的定位能力，采用了WIoU函数来提高算法对于困难样本的定位稳定性；在visdrone2019数据集上的实验表明，所改进的算法与原有算法相比，mAP提升了2.1%，相比其他主流的算法，该算法有更好的检测效果。

Abstract:

Aiming at the problems such as missed detection and false detection caused by complex background and small target in the current object detection algorithms for unmanned aerial vehicle (UAV) images, an improved UAV aerial photography detection algorithm based on YOLOv8, YOLOv8-A, is proposed; In order to improve the detection ability of the algorithm for small targets, a global channel-spatial attention module was used to enhance the expressive ability of the input feature map. In order to further improve the algorithm's ability to capture the context information and intrinsic features of the input data, the AIFI module of the RT-DETR algorithm was replaced in the neck structure, which is helpful for extracting the context information. In order to enhance the algorithm's positioning ability for difficult samples, the WIoU function is adopted to improve the algorithm's positioning stability for difficult samples. Experiments on the visdrone2019 dataset show that the improved algorithm has a 2.1% increase in mAP compared to the original algorithm. Compared with other mainstream algorithms, this algorithm has a better detection effect.