基于目标检测模型的无人机影像识别技术

UAV images recognition technology based on target detection model

YOLOv5具有较高的目标检测速度和检测精度,但在无人机影像小目标检测方面效果不太好。为解决在自然环境情况下小目标检测精度低及鲁棒性差等问题,本文以自然环境情况下无人机影像为研究对象,提出了一种改进的YOLOv5小目标检测模型。通过对特征图增加上采样处理,使特征图继续扩大,从而降低采样率和缩小感受野,提高模型对小目标的检测能力。改进的模型在天大无人机影像VisDrone数据集上进行了训练和测试。实验结果表明,改进YOLOv5的算法平均精度值为46.4%,与原YOLOv5模型相比,平均精度值提升了14.9%,改进YOLOv5在一定程度上改善了YOLOv5无人机影像识别率。

YOLOv5 has high target detection speed and detection accuracy, but it is not very effective in detecting small targets in UAV images. In order to solve the problems of low detection accuracy and poor robustness of small targets in natural environment, this paper takes UAV images in natural environment as the research object, and proposes an improved YOLOv5 small target detection model. By adding up-sampling processing to the feature map, the feature map continues to expand, thereby reducing the sampling rate and the receptive field, and improving the model’s ability to detect small targets. The improved model is trained and tested on the VisDrone dataset of UAV images. The experimental results show that the average accuracy of the improved YOLOv5 algorithm is 46.4%. Compared with the original YOLOv5 model, the average accuracy is increased by 14.9%. The improved YOLOv5 can improve the YOLOv5 UAV images recognition rate to a certain extent.