Non-cooperative Target Based On Zynq Platform Posture Measurement Methods

As space becomes a strategic resource,using rendezvous and proximity operations(RPO)technology to achieve space dominance has become an important research field in the national defense strategy and security of each spacefaring nation under the new military revolution.Common RPO tasks include de-orbiting hostile satellites to abandoned orbits,conducting close reconnaissance on hostile satellites,performing close hard-kill attacks,on-orbit maintenance,assembly,and refueling operations for our malfunctioning spacecraft.Often,these tasks involve non-cooperative targets in space.The key to completing these tasks is to develop pose measurement techniques that can accurately determine the relative position and orientation of non-cooperative targets.Due to its high measurement accuracy and non-contact characteristics,stereo vision technology is the preferred technology for measuring the pose of non-cooperative targets in space.However,this technology is usually based on general-purpose computer platforms,which have problems with large size and high power consumption when directly applied to space scenarios.Considering that both space and power are extremely precious resources,this article proposes a non-cooperative target pose measurement method based on the Zynq platform,which fully integrates the low power consumption,high performance,and highprecision binocular stereo vision of the Zynq platform,and effectively overcomes the limitations of stereo vision non-cooperative target pose measurement technology in space application scenarios by acquiring images and solving poses based on the Zynq platform.The main work is as follows:(1)The principle of binocular vision stereo algorithm was studied,and a binocular vision measurement system twin body was built on a digital twin software platform.Through simulation experiments,the influence of binocular baseline,illumination intensity,and distance between camera and satellite on the pose measurement accuracy in the binocular vision pose measurement system was investigated.Simulation experiments show that with the increase of binocular baseline(300-500mm),illumination intensity(1-5W/m2),and distance between camera and satellite(0.5-1.5m),the pose measurement accuracy decreases.(2)A non-cooperative target pose measurement system based on Zynq platform was built.According to the processing capacity of field programmable gate array(FPGA)and ARM,the image acquisition task was completed by FPGA,and the pose calculation task was completed by ARM processor.High-speed data transmission was achieved through AXI-DMA.Test results show that the image acquisition and transmission of the built system are stable,and the final pose calculation speed is 8fps.(3)A method for distortion correction of zoom lenses based on deep learning and digital twin was proposed.The method is based on digital twin technology,which uses a high-fidelity virtual measurement system to produce training data sets under the structural parameter conditions of the actual system.Based on the U-Net network of deep learning,a distortion correction method for zoom and fixed-focus cameras was established.The image enhancement and distortion correction were achieved through the deep learning network,solving the problem of image distortion correction in space pose measurement.Finally,the SSIM-struct evaluation index was proposed to evaluate the proposed method from the perspective of structural information similarity.The results show that the SSIM-struct of this method reaches 99.98%.(4)Finally,the factors affecting the pose measurement accuracy were analyzed on the physical simulation experimental platform for space pose.These factors include binocular baseline,illumination intensity,and distance between camera and satellite.The results show that the conclusions of the physical experiments are consistent with those of the simulation experiments.Meanwhile,the proposed measurement method was applied practically by building an actual pose attitude measurement system based on the Huawei Hi3559-AV100 platform.The correctness of the method was proven,which can reduce power consumption,decrease volume,and provide an effective way for the development of embedded pose measurement systems.