Feature Extraction Technology Of Spatial Non-cooperative Target Based On ZYNQ

With the development of human space exploration activities,more and more spacecraft are launched into the outer sky.Space on orbit services such as replenishment of early manned spacecraft and maintenance of failed spacecraft are usually completed by astronauts.With the development and progress of space technology,the realization of space on orbit services by fully autonomous system has become a research hotspot.Accurate recognition of spacecraft motion attitude is an important technical guarantee for space on orbit services.In the post Moore's law era,heterogeneous computing is an important means to improve the computing power of hardware platform.Therefore,this paper will study the feature extraction technology of spatial non-cooperative target based on ZYNQ multi-core heterogeneous So C.In this paper,firstly,a typical non-cooperative target ground simulation device is built,and the shape characteristics of the typical non-cooperative target are analyzed,and the binocular stereo vision scheme is used to identify the four vertices of the solar array.Then the distortion model of camera,the imaging model of binocular camera and the theoretical basis of image processing algorithm are analyzed and studied.The image recognition scheme based on clustering and polygon fitting is used to extract the features of spatial non-cooperative objects,and the recognition effect is analyzed.Then,using the binocular camera imaging model,the motion trajectories of four vertices of the uncooperative target in the spatial coordinate system are obtained.Secondly,image processing technology based on video stream is adopted,and image processing algorithm based on FPGA is developed by using C++ language with Vivado HLS high level synthesis tool.At the same time,the algorithm is parallelized with PIPELINE,DATAFLOW and UNROOL,and the algorithm is verified from three aspects: algorithm,RTL and physical test.In addition,the algorithm is optimized from the perspective of resource consumption,delay and throughput.Finally,it is encapsulated as HLS image processing IP core.Then,ZYNQ7 Processing System soft core is called to complete the configuration of processing system's resources,and IP cores such as camera data receiving,HLS image processing,video display are called to complete the construction of image acquisition and preprocessing circuit and image display circuit by using the combination of AXI4 bus and VDMA.Finally,after the ZYNQ hardware system is synthesized and implemented,the Vivado SDK development tool is invoked to create the software engineering and board support packet.At the same time,the initialization configuration of PL port's IP core,system related peripherals and image cache is completed,and the processing system image processing algorithm is developed.Experimental results show that image preprocessing with FPGA can effectively improve the real-time performance of image processing system,and ZYNQ So C has unique advantages in image processing architecture.