论文部分内容阅读
Landmark recognition has been the study object of many institution and organizations all over the world for its comprehensive application in military and civilian.The technology of landmark recognition system is widely used in robot guiding travel field,vehicle guiding travel field and electronic travel aids for the blind field and so on. Character recognition for landmark system is required the characteristics of portable,real-time processing and able to recognition patterns accurately.So the author uses field programmable gate array(FPGA) as the tool to realize the system after studing the method.The theise focuses on the landmark character recognition and then analysis and improves the method of their realization. Finally, the author carries out using FPGA for hardware implemention.The main work of the paper is to study the neural network algorithm, using MATLAB language for algorithm implementation and getting the weights and thresholds of the sample landmark signs; And then,the weights and thresholds are cured to the ROM in the FPGA. Finally, in the Quartus II platform, HDL language is used in the system.First of all,BP neural network is used for pattern classification of landmark character.The landmark character to be recognized is the size of10×5binary image.And then, the author use perceptron neural network to classify the pattterns of landmark character. In perceptron neural network,the landmark character to be recognized is the size of32x25binary image. Finally, by analyzing the type of FPGA to process data and so on, this design uses the perceptron neural network as the method of hardware implementation.The character recognition for landmark system is achieved with Altera corporation Quartus Ⅱ as the development platform and DE2board.In this thesis, the mode of FPGA chip is EP2C35F672C6and the method of download is JTPA and AS.The mode of configuration chip is EPCS16U30. Finally, by debugging the system, the system is achieved in DE2board.Finally,also taking into account the image with noise sometimes,the design simulate some samples with noises to test the robustness of the network. The results show that the design of the landmark of the characters recognition system has certain fault tolerance.The system also implements real-time patterns recognition capabilities,high stability and practicality.