基于SSD的实时轻量级无人机检测算法

针对当前无人机检测算法普遍不能做到快速准确检测的问题,提出了一种基于SSD的改进实时轻量级无人机检测算法--TSSD。首先,针对SSD算法的骨干网络权重参数量大的问题,改进得到一种轻量级的骨干网络。其次,针对SSD只利用多层特征图进行多尺度预测,而特征之间的联系没有被很好地融合利用,加入了一种特征增强模块来提高检测能力。在自建无人机数据集中进行的实验结果表明,提出的算法检测速度达到125f/s,远高于原始SSD的检测速度,且准确率比原始SSD也有所提升。

Real-time Lightweight UAV Detection Method Based on SSD Algorithm

In order to solve the problem that detection algorithm of unmanned aerial vehicle (UAV) can not detect UAV quickly and accurately, an improved real-time lightweight UAV detection algorithm based on SSD is proposed, which is called TSSD. Firstly, a lightweight backbone network is improved to solve the problem of large amount of weights parameters in the backbone network of SSD algorithm. Secondly, SSD only uses multi-layer feature map for multi-scale prediction, but the relationship between features is not well utilized, and a feature enhancement module is added to improve the detection ability. The experimental results of the self-built UAV dataset show that the detection speed of the proposed algorithm is 125f/s, which is much higher than that of the original SSD, and the accuracy of the proposed algorithm is also higher than that of the latter.