Study On The Key Technologies Of Automatic Identification For Cooperative Target On Spacecraft

With the development of human space technology, the space activities are more frequent. The measurement of relative position and pose between the tracking spacecraft and the target spacecraft is a key essential step for smooth completion of space tasks. In order to meet the demand of measurement position and pose,pre-designed cooperative target is usually installed to the target spacecraft, and the tracking spacecraft identifies the target to get the position and pose parameters.Therefore automatic identification of the spacecraft cooperative target is an important premise of stable system. The paper expends the research on the key technologies of cooperative target automatic identification during the space manipulator performing tasks, and includes the following major aspects.1. A three-dimensional target is designed combined with the characteristics of points, segment lines and circle based on a thorough analysis of target characteristics in application in space engineering at home and abroad. The position relationship of feature points meets the condition of pose measurement by P3 P algorithm, and is good to measure pose. The mutual restriction relation of circle and segment lines improves identification accuracy and the capability of anti-interference and antiocclusion in the complex scenes.2. In aerospace applications, the high-grade components usually have poorprocessing ability, little memory and slow speed. Therefore, the complex algorithms,such as Hough transform, pattern matching and so on, can not meet the application requirements. In this paper, a fast identification algorithm based on geometric features is proposed according to the shape characteristics of the target. The algorithm detects circles by the geometrical property that two different arcs on the same circle have the same center and radius and identifies the target by the position relation of circle and segment lines. The experimental results show that the algorithm is robust and can identify the target in many complex scenes, such as noise pollution,bright light, dim light, backlight, rotation, occlusion and so on. The algorithm is simple and fast, and the processing time is less than 125 ms which meets the requirements of aerospace real-time pose measuring at 8 frames.3. During the working process of space manipulator, the relative movement between hand-eye camera results in the motion blurred image. In order to recognize the target without the influence of motion blurring, the research focuses on the regularization blind deconvolution algorithms. A regularized model based 0-norm is set up to estimate blur kernel, and the piecewise function is used to instead of0-norm solution, avoiding solving the NP-hard problem of 0-norm. It simplifies the algorithm. After getting the accurate kernel, the restored image is obtained by Hyper-Laplacian prior. The experimental results show that the algorithm can get good restored image with less time, and can improve the ability of identification of the target with the influence of motion blurring.4. Hand-eye camera on the space manipulator uses a fixed focus system, so in the work range, hand-eye camera takes defocus blurred images with different degrees blur at different locations. The poor clarity of the circle and segment lines on the target leads to failed identification. In this paper, it finds ROIs(Regions of Interest) according to the characters of defocus target image, then restores and identifies in ROIs. In restoring algorithm, a FRFT(Fractional Fourier Transform)model is set up according to the features which FRFT corresponding to the Fresnel diffraction. The restoration is completed by continual FRFT and inverse FRFT.During the restoration, a fast algorithm is used to resolve the discrete computation.The experimental results show that the algorithm can overcome the affluence of defocus and improve the accuracy of the target identification under the direction of defocus.