Satellites for observation missions, or imagery satellites, have increased drastically in number and performances since the beginning of the space age. Recent Earth observation satellites are now equipped with new instruments that allow image processing in real-time. Issues such as ground target tracking, moving or not, can now be addressed by controlling precisely the satellite attitude. The satellite "camera" can be used as an input sensor for real-time attitude control process. This can be addressed thanks to a closed loop control scheme that includes the image acquisition and image processing parts. Real-time attitude control using such sensors will then allow the tracking of static or moving ground targets. In this paper, we propose to consider this problem using a visual servoing (VS) approach. This work is thus focused on establishing a visual control law that allows to precisely control a low orbit Earth observation starer satellite attitude using images provided by its matrix sensor. The goal is to perform acquisition missions devoted to gaze on an object of interest that is visible in the image before the VS starts. The visual sensor is fixed to the satellite, and we have full control over its three rotational degrees of freedom subject to dynamic constraints, while the satellite is moving on an orbit that only influences its position (that is not controlled by our VS scheme). Compensating for the target motion in the image by explicitly embedding it in the control scheme becomes essential when it is significant. In our case, the satellite orbit is known, so we can determine accurately its translational motion, and compensate for it in the control law. When it comes to target motion, we propose to decompose it into a known displacement caused by Earth's dynamics and a residual motion due to its potential own motion. The contribution of this paper is a visual servoing scheme able to control the attitude of an agile Earth observation satellite for target tracking. Three visual features are selected for controlling the 3 attitude parameters, for achieving a centering task, and an orientation task. The control law allows for dealing with the satellite's high translational velocity induced by its orbit and other external motion including Earth's rotation and target own motion. A rate saturation algorithm is also proposed dealing with dynamic constraints. Simulations and experiments on an actual robot will be presented.