Abstract: The method of projection is used to feature extraction and recogniton of the underwater acoustic images obtained by sonar imaging. After discussing the characteristics of Radon transform and fan-beam projection respectively, a new method using Radon transform for underwater acoustic image’s feature extraction and classification is proposed,which is non-sensitive to the ambiguous noise.After getting the close edge by morphological edge extraction and thinning operators ,we construct the edge moment invariants based on Radon projection and use them in classification of three kinds of underwater objects. Simulation results show the proposed method is robust and its effectiveness are better than Hu’s moment invariant and object’s plane moment invariance in Radon projective space according to the computing speed.

Key words: feature extraction, recognition, Radon transform, sonar image, fan-beam projection

摘要： 用投影变换的方法进行了成像声纳采集的水声图像特征提取与识别。阐述了Radon与扇束投影变换的特点与关系，并提出了一种基于图像边缘Radon变换的水声图像矩特征提取和识别方法。首先使用一种形态学边缘提取算子和细化算法提取二维图像中目标的轮廓，构造目标轮廓在 Radon变换空间的平移、比例和旋转矩不变量，并应用于三类水下物体的分类识别中，实验仿真结果表明该方法在运算速度上优于Hu’s不变矩和图像目标面Radon投影空间不变矩，具有很好的性能和较高的实用价值。

关键词: 特征提取, 识别, Radon变换, 声纳图像, 扇束投影