A cascade profile design method was proposed using the aerodynamic load and blade thickness distribution as the design constraints, which were correspondent to the demands from the aerodynamic characteristics and the ...A cascade profile design method was proposed using the aerodynamic load and blade thickness distribution as the design constraints, which were correspondent to the demands from the aerodynamic characteristics and the blade strength. These constraints,together with all the other boundary conditions , were involved in the stationary conditions ofa variational principle , in which the angle-function was employed as the unknown function.The angle-function ( i. e. , the circumferential angular coordinate) was defined in the image plane composed of the stream function coordinate ( circumferential direction ) and streamline coordinate. The solution domain, i.e., the blade-to-blade passage, was transformed into a square in the image plane, while the blade contour was projected to a straight line ; thus, the difficulty of the unknown blade geometry was avoided. The finite element method was employed to establish the calculation code. Applications show that this method can satisfy the design requests on the blade profile from both aerodynamic and strength respects. In addition, quite different from the most inverse-problem approaches that often encounter difficulties in the convergence of iteration, the present method shows a stable and fast convergence tendency. This will be significant for engineering applications.展开更多
We propose a cascade system of filters for realizing a non-uniform waveband separation for optical networks. The use of such separation is required at the DEMUX stage in a optical OXC switching wavebands. The design o...We propose a cascade system of filters for realizing a non-uniform waveband separation for optical networks. The use of such separation is required at the DEMUX stage in a optical OXC switching wavebands. The design of the system is based on optimized balanced tree, which minimizes the overall optical loss.展开更多
Recent learning-based approaches show promising performance improvement for the scene text removal task but usually leave several remnants of text and provide visually unpleasant results.In this work,a novel end-to-en...Recent learning-based approaches show promising performance improvement for the scene text removal task but usually leave several remnants of text and provide visually unpleasant results.In this work,a novel end-to-end framework is proposed based on accurate text stroke detection.Specifically,the text removal problem is decoupled into text stroke detection and stroke removal;we design separate networks to solve these two subproblems,the latter being a generative network.These two networks are combined as a processing unit,which is cascaded to obtain our final model for text removal.Experimental results demonstrate that the proposed method substantially outperforms the state-of-the-art for locating and erasing scene text.A new large-scale real-world dataset with 12,120 images has been constructed and is being made available to facilitate research,as current publicly available datasets are mainly synthetic so cannot properly measure the performance of different methods.展开更多
基金Foundation items: the National Natural Science Foundation of China (50136030) the Shanghai Major Subject Item: Power Source Island System Research
文摘A cascade profile design method was proposed using the aerodynamic load and blade thickness distribution as the design constraints, which were correspondent to the demands from the aerodynamic characteristics and the blade strength. These constraints,together with all the other boundary conditions , were involved in the stationary conditions ofa variational principle , in which the angle-function was employed as the unknown function.The angle-function ( i. e. , the circumferential angular coordinate) was defined in the image plane composed of the stream function coordinate ( circumferential direction ) and streamline coordinate. The solution domain, i.e., the blade-to-blade passage, was transformed into a square in the image plane, while the blade contour was projected to a straight line ; thus, the difficulty of the unknown blade geometry was avoided. The finite element method was employed to establish the calculation code. Applications show that this method can satisfy the design requests on the blade profile from both aerodynamic and strength respects. In addition, quite different from the most inverse-problem approaches that often encounter difficulties in the convergence of iteration, the present method shows a stable and fast convergence tendency. This will be significant for engineering applications.
文摘We propose a cascade system of filters for realizing a non-uniform waveband separation for optical networks. The use of such separation is required at the DEMUX stage in a optical OXC switching wavebands. The design of the system is based on optimized balanced tree, which minimizes the overall optical loss.
基金supported by the National Natural Science Foundation of China(62102418 and 62172415)the National Key R&D Program of China(2019YFB2204104)the Open Research Fund Program of State key Laboratory of Hydroscience and Engineering,Tsinghua University(sklhse-2020-D-07).
文摘Recent learning-based approaches show promising performance improvement for the scene text removal task but usually leave several remnants of text and provide visually unpleasant results.In this work,a novel end-to-end framework is proposed based on accurate text stroke detection.Specifically,the text removal problem is decoupled into text stroke detection and stroke removal;we design separate networks to solve these two subproblems,the latter being a generative network.These two networks are combined as a processing unit,which is cascaded to obtain our final model for text removal.Experimental results demonstrate that the proposed method substantially outperforms the state-of-the-art for locating and erasing scene text.A new large-scale real-world dataset with 12,120 images has been constructed and is being made available to facilitate research,as current publicly available datasets are mainly synthetic so cannot properly measure the performance of different methods.
