Dissertation > Excellent graduate degree dissertation topics show

Research of the Robot Target Recognition System in 360°Environment

Author: WangJian
Tutor: GuoLanShen
School: Hebei University of Technology
Course: Mechanical Engineering
Keywords: Image Recognition Image Stitching BP neural network Simulation MATLAB LabVIEW
CLC: TP391.41
Type: Master's thesis
Year: 2011
Downloads: 25
Quote: 0
Read: Download Dissertation

Abstract

The image is widely used in the engineering work and scientific research , image recognition is very necessary . Traditional as one of the core technologies of production and life automation , modernization and intelligent image recognition technology , identified a range of limitations has not meet the many aspects need to identify the problem . Abandoned in the subject image recognition technology to improve through the use of expensive wide-angle image capture devices from computer algorithms to improve the selected image stitching technology to achieve the expansion of the scope of the visual , coupled with the high degree of non- linear mapping ability of BP neural network technology would be more able to improve the reliability of the identification . This method is a great supporting role in the choice of round composite legs climbing obstacle robot obstacle recognition and more impaired manner . This article is based on some common objects to simulate obstructions as the target image information for research . After learning of a series of domestic and foreign advanced technology and research , according to the actual needs , build an image recognition system . Innovation is mainly reflected in : First image stitching technology to expand the visual range of the image acquisition , preprocessing of the target image ; This process is guaranteed to the acquisition extend , but also increase the clarity of the image , thereby increasing the subsequent target image accurate identification reliability . Followed by the mosaic image multi- feature extraction using BP neural network the super nonlinear mapping ability to build the mapping between the multi-feature and obstructions category of the target image , the final obstacle category pattern recognition . In the full analysis of the various parameters of the BP neural network , citing LM (Levenberg-Marquardt) a classic optimization algorithm to replace the traditional steepest gradient descent method as the BP network training function . As can be seen from the final test process , the improved algorithm has greatly improved the speed and accuracy of convergence . Finally, the joint simulation of MATLAB and LabVIEW , building a complete 360 ° environmental objectives to identify the man-machine system , which give full play to the advantage of short LabVIEW simplified procedures and the development cycle , as well as MATLAB a strong matrix operations , image processing and numerical analysis functions.

Related Dissertations

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device
© 2012 www.DissertationTopic.Net Mobile