二值化特征在快速舰船目标预选中的应用

Improved binarized normed gradients for fast ship target pre-detection

为了快速准确地自动提取和识别海面舰船疑似目标,为舰船目标精检测提供可信的数据基础,采用了二值化特征进行舰船目标粗检测,并根据舰船窄而长的几何特征提出了改进算法,按照舰船目标不同的方向分别进行模板训练。实验表明,二值化特征可以有效地提取疑似舰船目标,并且改进算法可以在提取窗口数相同时,提高查全率,更利于进一步的精检测。

For accurate and efficient detection and recognition of suspected ship target,reliable data base is provided for fine detection.The binarized feature for the coarse detection of ship target is adopt.An improved algorithm is proposed based on the narrow and long geometric feature of ship.Different templates are trained on the basis of the direction of ships.Experiments demonstrate that the binarized feature performs well in the extraction of candidate ship windows,and the improved algorithm can obtain a higher recall rate than original binarized feature when extracting the same amount of windows,which is helpful for the more efficient and accurate fine detection.