基于数学形态学的月海圆形撞击坑自动识别方法

Automatic identification of circular mare craters based on mathematical morphology

撞击坑是月球表面最为常见的地质单元,是研究月球地质演化历史的重要对象,也是月球地质定年的基本依据,因此撞击坑识别具有重要意义.本文根据嫦娥一号采集的月球CCD图像,基于数学形态学方法对撞击坑进行自动识别提取研究.在CCD图像中,撞击坑边缘的灰度变化明显,梯度较大,由此可以计算获取撞击坑的边缘形态;一般情况下,依据图像灰度梯度突变,通过边缘检测得到的撞击坑边缘比较粗糙、不连续,而且有断口和小洞.根据数学形态学的基本思想——用具有一定形态的结构元素去量度和提取图像中的对应形状,对识别出来的边缘作进一步处理,可得到较光滑、连续的撞击坑边缘弧,从而能方便地拟合出撞击坑边缘,并获得撞击坑的直径与位置.用数学形态学进行撞击坑识别与提取的主要步骤是:首先对CCD影像计算灰度的梯度,得到梯度图像,然后进行二值化,再使用数学形态学分离出边缘,最后用圆形对撞击坑进行拟合并计算出撞击坑的位置和直径.本文分别对月海和月陆地区进行撞击坑识别实验,结果表明,我们设计的算法能够识别的最小撞击坑直径为10个像素.其中月海区域撞击坑识别准确可靠;而月陆区域岩性差异大、地形起伏,造成CCD图像背景变化较大,其识别效果相对差一些,有待进一步改善.

Identification of lunar crater is very significant, because the crater is not only the most common geological unit on the surface of the moon, but also an important object to study lunar geological evolution history and the fundamental basis of lunar geological dating. This paper conducts to recognize and extract the craters automatically based on mathematical morphology, from the lunar CCD image acquired by the Chang＇e I satellite. The grayscale of the crater rims vary obviously in the CCD image, and their gradient are large, which can be calculated to obtain the shape of crater rim. In general, results of crater rims extracted by edge detection based on sharp grayscale gradient of image are discontinuous and rough, and also have gaps and holes. Some processings are further done to take the smooth and continuous arcs of crater rims, by measuring and extracting the corresponding shape from the image with a certain form of structural elements according to the basic idea of mathematical morphology. And then processed arcs may be used to fit the rims of craters conveniently and to obtain the diameters and locations of craters. The algorithm based on mathematical morphology for crater identification is as follows： firstly, we calculate the gradient of grayscale of the CCD image to form a gradient image, and binarize it; then separate the rims by mathematical morphology; finally extract the sizes and locations of craters by fitting rims with circles. In this paper, experiments on identification of craters in lunar mare and terra area show that the algorithm designed by authors may identify the crater with smallest diameter of 10 pixels. The identification of mare craters is accurate and reliable, while of terra craters is a little worse, due to big grayscale variety in CCD image background caused by big lithological differences and fluctuant topography, so we need a futher study on improvement of algorithm.