基于深度学习的复杂背景雷达图像多目标检测

Multi-objective detection of complex background radar image based on deep learning

针对传统雷达图像目标检测方法在海杂波及多种干扰物组成的复杂背景下目标分类识别率低、虚警率高的问题,提出将当前热点研究的深度学习方法引入到雷达图像目标检测。首先分析了目前先进的YOLOv3检测算法优点及应用到雷达图像领域的局限,并构建了海杂波环境下有干扰物的舰船目标检测数据集,数据集包含了不同背景、分辨率、目标物位置关系等条件,能够较完备地满足实际任务需要。针对该数据集包含目标稀疏、目标尺寸小的特点,首先利用K-means算法计算适合该数据集的锚点坐标;其次在YOLOv3的基础上提出改进多尺度特征融合预测算法,融合了多层特征信息并加入空间金字塔池化。通过大量对比实验,在该数据集上,所提方法相比原YOLOv3检测精度提高了6.07%。

In the complex background of sea clutter and various interfering objects, the traditional radar image object detection method has a low detection precision and high false alarm rates. The proposed deep learning method of the current hotspot research, is introduced into radar image target detection. Firstly, the advantages of the current advanced YOLOv3 detection algorithm and the limitations applied to the radar image field are analyzed , and the ship target detection data set with interference objects in the sea clutter environment is constructed . The data set contains various backgrounds, resolution, target position relations and other conditions, which can meet the actual task needs more completely. In allusion to the target sparseness and small target size, the K-means algorithm is used to calculate the anchor coordinates suitable for the dataset. Secondly, based on YOLOv3, an improved multi-scale feature fusion prediction algorithm is proposed, which fuses the multi-layer features information and the spatial pyramid pooling. Through a large number of comparative experiments, the mAP of the proposed method is improved by 6.07% compared with the original YOLOv3.