选区相关跟踪技术研究

To compensate the image movement introduced by atmosphere disturbance, earth vibration or satellite attitude, the application of Correlation Tracker (CT) is necessary. For the ordinary method, the field of view (FOV) of the CT detector is very small compared with it of the main detector. Moreover, the FOV of CT is fixed, i.e., it can't change automatically. As we know, the precision of the displacement calculated by cross-correlation technique is variable with the contents of the image. The situation becomes worse when the content of the image evolve with time, or in other words, the tracking target have shorter lifetime. For example, the mean lifetime of granulation which is often used as the tracking target of CT is about 8 minutes. Using the ordinary method, the precision of the CT will change with time and contents of image, and may different between the two orthogonal directions of the movement. Therefore, the performance of the CT system becomes unstable and somewhat unpredictable. These characteristics of cross-correlation algorithm and solar target will degrade the image resolution, especially for some applications which need long time exposure and high space resolution. To solve the problem, we propose a modified method, named region-select correlation tracker. Firstly, the system grabs an image with full resolution, which has almost the same FOV as main FOV. Secondly, the system chooses a suitable window or several windows for tracking using some criterion based on the properties of the solar target. Then the system fixes the windows and works like the ordinary one. By adding the FOV search procedure, the system can change and choose the content of the tracking image by itself, and therefore expect more stable performance, especially for the target on the solar which have certain lifetime. The studies on region-select CT will benefit the Deep-space Solar Telescope and other solar telescope.

为克服大气扰动、地基振动或卫星姿态抖动对成像的影响，在地基或空基高分辨率太阳光学望远镜中，相关跟踪器的应用是必然的。目前使用的相关跟踪器探测视场相对成像视场小且固定，探测视场（图像）无法主动选择。随探测图像内容的不同，由探测图像互相关得到的光轴偏移量计算精度也会有变化，尤其是对于有一定寿命的太阳目标（如米粒组织等）变化会更大。由此导致不同视场、不同时刻的稳像精度不同，并且两个垂直方向上的稳像精度也时常各不相同，最终降低成像质量，对于需长时间积分的情况影响会更严重。为了解决这一问题，我们提出了选区相关跟踪技术：增加跟踪窗口搜索模式，即先在整个成像视场（图像）中搜索适合做跟踪的探测视场（图像），然后将探测窗口设定在该区域，再进入如常规相关跟踪一样的稳像跟踪模式。该方法可以有效减少由于图像不同而导致跟踪精度不稳定的情况，同时通过选择具有更长寿命的目标做为跟踪目标，可显著提高长时跟踪的精度。

为克服大气扰动、地基振动或卫星姿态抖动对成像的影响，在地基或空基高分辨率太阳光学望远镜中，相关跟踪器的应用是必然的。目前使用的相关跟踪器探测视场相对成像视场小且固定，探测视场（图像）无法主动选择。随探测图像内容的不同，由探测图像互相关得到的光轴偏移量计算精度也会有变化，尤其是对于有一定寿命的太阳目标（如米粒组织等）变化会更大。由此导致不同视场、不同时刻的稳像精度不同，并且两个垂直方向上的稳像精度也时常各不相同，最终降低成像质量，对于需长时间积分的情况影响会更严重。为了解决这一问题，本项目从目标适用性、目标选择算法、目标搜索策略、任意开窗探测单元、选区计算控制单元等方面进行选区相关跟踪技术的研究，建立选区相关跟踪技术验证实验系统，提供经实验验证的行之有效的选区相关跟踪技术方法。研究结果证明：选区相关跟踪技术通过在大视场内搜索并优选跟踪目标提高了跟踪精度的一致性，并将因跟踪目标的寿命问题导致的影响降至最低，有效的解决了常规相关跟踪技术中探测视场固定、探测图像无法选择带来的跟踪精度随图像内容改变而改变的问题，为地基、空基太阳高分辨率光学观测提供了更好的稳像技术方法。