In the Electro-Optical acquisition-tracking-pointing system (ATP), the optical components will be damaged with the several influencing factors. In this situation, the rate will increase sharply when the arrival of damage to some extent. As the complex processing techniques and long processing cycle of optical components, the damage will cause the great increase of the system development cost and cycle. Therefore, it is significant to detect the laser-induced damage in the ATP system. At present, the major research on the on-line damage detection technology of optical components is for the large optical system in the international. The relevant detection systems have complicated structures and many of components, and require enough installation space reserved, which do not apply for ATP system. To solve the problem mentioned before, This paper use a method based on machine vision to detect the damage on-line for the present ATP system. To start with, CCD and PC are used for image acquisition. Secondly, smoothing filters are used to restrain false damage points produced by noise. Then, with the shape feature included in the damage image, the OTSU Method which can define the best segmentation threshold automatically is used to achieve the goal to locate the damage regions. At last, we can supply some opinions for the lifetime of the optical components by analyzing the damage data, such as damage area, damage position. The method has the characteristics of few-detectors and simple-structures which can be installed without any changes of the original light path. With the method, experimental results show that it is stable and effective to achieve the goal of detecting the damage of optical components on-line in the ATP system.