中文摘要：

非合作目标不提供已知的目标标识器和敏感器件，无疑增加了检测识别的难度。CCD相机能够提供大量包含有非合作目标的尺寸、形状、相对位置和相对姿态等信息的高分辨率图像，用常规方法进行非合作目标图像分割时，由于空间环境的影响，使得目标的轮廓线非常零碎，常掩埋在杂乱的背景分割线中。为了克服这种问题，提出了一种保持边界的整体变分方法和数学形态学相结合的方法。先用整体变分方法对图像进行平滑的同时可以最大限度地保留图像中的轮廓、边缘等特征信息以增强目标的边缘，然后采用数学形态学的方法对其进行膨胀、腐蚀以滤除众多的背景噪声。经大量的实验验证，这种方法可以有效地提取出非合作目标的边缘信息，并可滤除大量噪声，为进一步开展非合作目标识别工作奠定基础。

英文摘要：

Unlike the cooperative targets, non-cooperative targets do not provide known optical feature points or sensors. Detection and recognition of non-cooperative targets are rarely difficult. Images from the CCD camera can provide abundant information about non-cooperative targets such as size, shape, relative position and attitude. However, the contour line detected by traditional method become bits and pieces owing to complex environment of outer space. In order to overcome the shortcomings of traditional method in edge detection, Total Variation Method was adopted to remove noise effectively and preserve some key details such as edges and contours, and then, some methods of Mathematical Morphology were used to dilate and erode the image which was processed by means of Total Variation Method. The results show that the algorithm detects image edges successfully,simultaneously filters a large amount of noises, and it also can be the basis of further works for identification of non-cooperative targets.