Abstract: It has become an important means to execute the mission of intelligence and reconnaissance by carrier unmanned aerial vehicle.Image segmentation is critical to feature extraction，recognition and tracking of UAV infrared detection ship target.Because of low contrast，blurry edges and low SNR of infrared image，this paper proposes an improved segmentation method of local entropy based transition region extraction.Experimental results show that compared with the one-dimensional and two-dimensional OTSU algorithm，the proposed method can obtain good segmentation results under meeting real-time requirement.

Key words: Unmanned Aerial Vehicle（UAV）, local entropy, transition region, ship target, infrared image segmentation

摘要： 利用舰载无人机执行情报、侦察任务已经成为海上信息化作战中的重要手段。红外图像分割是无人机探测红外舰船目标特征提取、识别与跟踪的基础。针对红外图像普遍存在目标与背景对比度较低、目标边缘模糊和信噪比低等特点，提出了改进的基于局部熵过渡区提取的分割方法。实验结果表明，与一维和二维最大类间方差法的分割性能相比，该方法在满足实时性要求下可以获得良好的分割结果。

关键词: 无人机, 局部熵, 过渡区, 舰船目标, 红外图像分割