基于链码的合成孔径声纳图像目标识别算法研究*

为了进一步提高识别速度、增大识别效率，基于图像边缘的链码表示，将微积分中连续曲线曲率的定义推广到离散域，提出了链码离散曲率算法。通过利用链码计算图像边缘的离散曲率，结合特定的函数进行图像匹配，实现了以合成孔径声纳为代表的一类高分辨率、低信噪比的水声遥感图像的目标识别。实验结果表明，该算法计算复杂度较低，较之传统的基于特征提取的目标识别算法具有更高的识别效率。

Algorithm study of object identification of SAS based on chain-code

In order to improve both the speed and efficiency of the identification, this paper put forward the chain-code disperse curvature algorithm with the generalization of the definition of curvature in calculus into the disperse domain based on the chain-code representation of the edge of images. Then obtained the disperse curvature of the contour of images through chain-code, which could be used along with certain functions used for image matching, to identify objects in some fields that featured underwater remote-sensing images with high resolution and low SNR such as SAS. The experimental result shows that this algorithm has better identification efficiency with relatively low complexity than traditional ones that are based on feature extraction.