| Dynamic text recognition technology is an important research hotspot in the field of the pattern recognition,it is widely used in finance,transportation,medical,security,petroleum and petrochemical industries.In the process of industrial production,many enterprises in order to classify different types of workpiece effectively,the product label or character encoding on the surface of the workpieces were directly printed as the identification information,especially some rotary workpieces,such as bearings,oil drill pipe,bottle and so on.Therefore,it is important to research and design a set of dynamic text recognition system that effective identifies the characters of the rotary workpieces.In this paper,the recognition process of characters on oil drill pipe is studied.The research contents of this paper are mainly the following aspects:In the process of image acquisition,according to the factors of the factory environment,the appropriate light source and image acquisition cards are selected.At the same time,according to the basic characteristics of oil drill pipe,the rotating device is designed and the reasonable rotation speed is calculated.Then according to the different characteristics of the hardware performance and the images,the appropriate camera is selected.Finally it is reasonable to design the hardware system.The recognition processing of text images mainly include five key parts: image preprocessing,target location,character segmentation,character feature extraction and character recognition.Firstly,the collected image images have been preprocessed,the image preprocessing work include: converting color image to gray image,using the adaptive median filtering algorithm to reduce noise,enhancing the uneven illumination image.In the image positioning,this paper adapted the Sobel edge detection,the value of the two operations,the mathematical morphological operation and the matrix operation to simply locate the text image,then according to the characters of the characteristics,achieve precise text locating.Then segment the single character of text,in this paper,firstly we use the Fourier Transformation to detect inclination angle of the rotating plate and correct it effectively,and then use the connected region method and vertical projection method to segment the correction text.Single character segmentation,13 point feature extraction and template matching method have been used in this paper for the first time on the single character recognition,and then according to the similarity of characters,SVM classifier is used to recognize two times for the error character.The recognition system has been programmed in VS2008 compile environment,using the C++ programming language.It was combined with the OpenCV2.1 computer vision library.Each part of the character recognition system has been designed and the new algorithm has been got into the system,at the same time,recognition system has achieved the good results. |
