Study on the nameplate information detection method based on image recognition
|School||University of Electronic Science and Technology|
|Course||Communication and Information System|
|Keywords||Image Processing Character Recognition BP neural network FeatureExtract|
Inertial navigation system (INS) is widely used because it can work independently.But the INS can’t work independently in high accuracy for a long time, because thelocation error drift goes unbounded with time. GPS can be used to correct the INS error,but it is not a kind of completely autonomous navigation, and satellite signals may beblocked by high-rise buildings. INS/Vision Integrated Navigation can be an effectivesolution to the problems of autonomous navigation and system error correction.A method of vision positioning based on the road sign as a navigation mark isproposed, which is an exploratory study of self-positioning method. Based on the imagerecognition, the problem of road sign detection in the vision system is in-depth study.Firstly, analysis and research the methods of edge detection, color similaritydetermination and corner detection, a method of road sign image detection is proposedin this paper. Experimental results show that the road sign region can be extracted out ofthe image quickly and efficiently by using the proposed method. And then a method ofthreshold setting in a corner detection process is proposed on the basis of analysis forcorner features of road sign. The experimental results show the method is useful toretain the target corners while removing most of the interference corners.Secondly, a method of skew emendation is proposed, which can overcome theproblem of road sign image deformation caused as a result of the shooting angle. Theproposed method is better than the rotation method for the reason of needing a smallamount calculation. Then, split the road sign image into single text images by improvedprojection method.Finally, Gabor features and grid features are employed to distinguish differentChinese characters on the basis of summarization and analysis of method of featureextraction. And a kind of BP artificial neural network structure for Chinese characterrecognition in the natural environment is proposed, and experiment results show that thefeatures and network structure are robust and the system has a high recognition rate.