High-resolution space observation is of great importance for scientific and military use. To get higher resolution, a larger imaging aperture is highly required. For example, to obtain one-meter ground sampling distance (GSD) in visible band on geostationary orbit (GEO), the pupil diameter of space telescope is around 25 meters. Trying to fabricate and launch so large monolithic mirror will meet many unconquerable obstacles. A feasible scheme is using sparse aperture imaging technique based on small satellites formation. This paper is focused on a sparse aperture telescope consisting of small sub-telescopes to form a Fizeau imaging interferometer. Each sub-telecope is based on a small satellite. Imaging performance of an annular structure consisting of 25 sub-apertures is evaluated by simulation. The influence of phasing error (including piston and tip/tilt) of subapertures on image quality is evaluated. The co-phasing error budget of sparse aperture telescope dependent on field of view is also analyzed. A co-phasing error detection and correction method based on wavefront sensorless adaptive optics (WSLAO) is proposed at last.
It is indispensable to obtain more information such as the 3D structure of the space target by detecting and identifying the target, when complete the on-orbit servicing and on-orbit control tasks. Both lidar and binocular stereo vision can provide three dimensional information of the environment. But it is very sensitive to the illuminance of environment and difficult to image registration at weak texture region, when we are using the binocular stereo vision in space. And lidar also has some defects such as the lidar data is sparse and the scanning frequency is low. So lidar and binocular stereo vision should be used together. The data of the lidar and binocular stereo vision are fused to make up for each others flaws.
In this paper, uniform point drift registration method is used in the fusion of point cloud which is sampled by lidar and binocular stereo vision. In this method, the two groups of point cloud are considered as one which submit to mixed probability distribution and the other one which is sampled from the points submit to mixed probability distribution. The transformation estimation between the two groups of the point cloud is maximum likelihood estimation. The transformation is required to take overall smoothness. In other words, the point clouds should be uniformed. The uniform point drift method can solve the registration problem efficiently for 3D reconstruction. Usually the time can be compressed by 10%.
One of the important part of object target recognition is the feature extraction, which can be classified into feature extraction and automatic feature extraction. The traditional neural network is one of the automatic feature extraction methods, while it causes high possibility of over-fitting due to the global connection. The deep learning algorithm used in this paper is a hierarchical automatic feature extraction method, trained with the layer-by-layer convolutional neural network (CNN), which can extract the features from lower layers to higher layers. The features are more discriminative and it is beneficial to the object target recognition.
An automatic target detection method used in long term infrared (IR) image sequence from a moving platform is proposed. Firstly, based on POME(the principle of maximum entropy), target candidates are iteratively segmented. Then the real target is captured via two different selection approaches. At the beginning of image sequence, the genuine target with litter texture is discriminated from other candidates by using contrast-based confidence measure. On the other hand, when the target becomes larger, we apply online EM method to estimate and update the distributions of target's size and position based on the prior detection results, and then recognize the genuine one which satisfies both the constraints of size and position. Experimental results demonstrate that the presented method is accurate, robust and efficient.