Research On Micro-binocular Vision System Based On Non-cooperative Target

Nowadays, the technology of on-orbit servicing for non-cooperative spacecrafthas been wildly applicated in the field of the maintenance of spacecrafts, thecleaning up of debris, the space attack and defense, etc. The method based on stereovision is one of the key technology when the tracking spacecraft is approaching thetarget spacecraft. Taking the characters of small size and low power consumptioninto consideration, this paper designed a micro-binocular vision hardware systemand studied a relative algorithm of image processing.Firstly, this paper established a hardware system with two FPGAs and a DSP.The VmodCAM board from Digilent company in the USA was selected as the imageacquisition unit. One FPGA controled one channel respectively. An improved Cannyoperator was used in the FPGA to extract the edge pixels, and the Hough transformwas used in the DSP to extract the straight lines.Secondly, to meet the perfermance indicators of miniaturization and low powerconsumption, this paper selected the key components of the system reasonably, anddesigned the whole circuits. The A3PE1500-FG676was selected because of itsstructure based on Flash、low power consumption and high speed. The high speedTMS320C6416T was selected. Besides, this paper studied the current value ofSRAM and SDRAM when they worked in different mode, and reduced the powerconsumption by optimizing in software. Moreover, this paper selected switchingpower supply module because it has high transfer efficiency and takes up small area.This paper also designed the minimum system circuit of the FPGA and the DSP, anddesigned the interface circuit of the data and the program memory, etc.Thirdly, in the FPGA, some modules were generated reasonably based on theflow of the image data, such as the module of image-capture, the module of edgepixel extraction based on the improved Canny operator, the module of storage of theimage data and the coordinates of the edge pixels, etc. In the DSP, the fringe straightlines were extracted based on Hough transform. Besides, two improvements,including dividing the region reasonably and feedback from the former frame, wereapplied to decrease the data size to increase the data updating rate.At last, this paper verified the result of edge pixel extraction in the FPGAs byprocessing the images of a box of complex edges, and verified the result ofextraction and identification of the fringe straight lines based on Hough transfer byprocessing the images of the non-cooperative target spacecraft.Some experimentsare conducted to simulate the approaching process, and the performance of thesyetem is verified. The average data updating rate is6Hz at least. The power consumption when the system is working continuously is4W at most. Moreover,The size of the whole system is106mm×106mm×26mm only, and the weight is260g only. So the micro-binocular vision hardware system in this paper is ofengineering application value since its high performance of small size、low powerconsumption and high data updating rate.