As the existing algorithm has such problems as poor detection effects,low precision,and missed detection of small air targets a single target detection algorithm based on the bilinear interpolation is proposed.First,the backbone network is designed,and feature fusion is carried out with different meth-ods according to the characteristics of the feature map.Then,the bilinear interpolation algorithm is used for up-sampling during feature fusion.Finally,the priori box of the algorithm is designed to make the size of the priori box of the algorithm fit better with the size of the dataset target frame.The experimental re-sults show that the average detection accuracy of the proposed algorithm is improved by 10.29％compared with that of SSD.The computational complexity of this algorithm is reduced by 82.18％compared with that of Faster-RCNN.
关键词
空中小目标检测/双线性插值/特征融合/先验框
Key words
air small target detection/bilinear interpolation/feature fusion/priori box
维普
万方数据