Automatic recognition of infrared warship target is an important content for weapon's track and guidance under the sea-sky background. For the infrared images of low Signal-to-Noise Ratio and low contrast, mean smoothing filter and exponential enhancement are selected to preprocess the infrared images, and then least squares fit out the sky-sea line iteratively on the basis of detected iterative using Sobel operator. Image is segmented with the method of region growing and combination to get the typical shape and grey features representing the warship target, and automatic recognition of potential targets is carried out by fusing weighted features. Experimental results show that this algorithm can recognize the warship target in real time with high accuracy.%海天背景下红外舰船目标的自动识别是武器跟踪导引的重要内容.针对低信噪比和低对比度的红外图像,首先采取均值平滑滤波和指数增强对红外图像进行预处理,然后在Sobel算子检测梯度的基础上,采用最小二乘迭代拟合出海天线.通过区域生长与合并分割图像,提取出表征舰船目标的典型形状与灰度特征,并进行加权融合以实现对潜在目标的自动识别.实验结果表明该算法能以较高准确率实时识别出舰船目标.
展开▼
