摘要
在真实水下环境中,检测和识别水下目标一致是研究的重点。介绍了一种基于最优阈值分割算法的水下目标自动实时识别系统。首先运用去噪、图像均衡等方法对实时摄取的水下图像进行预处理,接着运用基于遗传算法优化的Otsu(即大津方法)最优阈值分割算法对所得图像进行区域分割,提取图像的特征向量,最后采用BP神经网络对提取的特征向量进行自动分类从而最终确定了水下目标的类型。水槽仿真试验表明系统能够在恶劣的环境下自动地检测水下目标,而且该方法具有较强的抗光线干扰能力和较高的准确度。
It is an important task to detect and recognize the underwater targets . An underwater objects real - time recognition system is presented, which is based on the expanded Otsu optimal threshold segmentation algorithm. In preprocessing, image enhancement techniques are employed to improve the image quality. After preprocessing, images are segmented with a region segmentation algorithm: Ostu optimal threshold algorithm based on GA algorithm, After the feature vector has been extracted from the segmented image, BP neural network is used for classifying the underwater objects ( mines, torpedoes etc). This system can automatically detect and recognize underwater objects. The effectiveness and robustness are tested by the experiments in flume.
出处
《计算机仿真》
CSCD
2005年第8期101-105,共5页
Computer Simulation
关键词
图像识别
目标识别
遗传算法
同态滤波
Image recognition
Target recognition
GA
Homomorphic filtering