基于特征的江面轮船识别算法

摘要

This paper proposed a feature method for ship recognition. First, preprocessed the image of the boats, and then put up two-dimensional wavelet transform on the image, abstracted the marginal profile and separated the target from the background. Combined with FE algorithm, got four features of the boats, according to the features, built a knowledge base. Then chose the correct knowledge and used set of generative rules to distinguish the objects, eliminated the interfering objects, identified the ship target. Finally, from the experiment carried on some extracted pictures, and compared with the former single feature mothed and AdaBoost method. It shows that this method can recognize ships more efficiently than single feature method and more effectively than AdaBoost method.%提出了一种江面轮船目标的特征识别方法.首先对江面轮船图像进行预处理,然后通过二维小波变换提取出边缘轮廓,将目标物体与背景分离开来.结合提出的FE(feature extract)算法提取出轮船图像的四个特征,根据这些特征建造一个知识库,通过选取适当的知识,采用产生式规则对目标物体进行判别,排除干扰目标,从而识别出轮船目标.最后从图片库中抽取几张图片进行实验,相比于之前的单特征方法和AdaBoost方法,该方法在识别率上要高于单特征方法,在识别速度上要快于AdaBoost方法.