Study On Automatic Reading Method Of Digital Instrument Based On Machine Vision

With the rapid development of computer technology,machine vision and image processing are also developing and mature.Image recognition technology is also applied to all areas of production and life,the use of artificial intelligence and image automatic identification technology.You can use the machine instead of people for production activities,which can set people from some heavy and repetitive work free,make the production efficiency greatly improve and bring great convenience,such as license plate automatic identification,remote medical and mechanical automatic,semi-automatic and so on.These are all in the convenience of people's daily life.Using machine vision and image processing,you can enhance the image,enhance the visual effect of the image,you can discover the human vision,which cannot capture the characteristics of information,to achieve accurate image recognition.Digital instrument automatic identification technology is the application of this aspect.The use of automatic instrument recognition technology,we can automatically access to the instrument image for digital character recognition,get accurate instrument readings and make people from a single repetition of the reading table work freed.Not only can we liberate the labor force,but also ensure the accuracy of the reading,which is widespread in the instrument today's society is undoubtedly of great significance.This paper takes into account the digital instrument automatic identification system is embedded in the chip,will be subject to CPU processing capacity constraints.Therefore,this paper aims to choose a simple,fast and relatively high recognition rate recognition system,for this situation,do the following work:(1)In the image acquisition stage,we take into account the impact of the complex environment on the image,make the digital instrument image enhance.On the bright or too dark image we use histogram equalization of the treatment,which get better results.In the case of uneven illumination of the instrument,the Otsu algorithm is used in local threshold processing when we make the threshold processing,and good results are obtained.(2)Considering the problem of shooting angle in the image acquisition stage,the first Canny operator is used for edge detection in the case of tilt in the instrument image,and then the Hough transform is used to realize the tilt correction.At the same time,we use projection and mathematical morphology theory for area positioning of the digital instrument image.(3)This paper selects the method of vertical and horizontal projection to realize the digital character segmentation of the instrument reading area,and then normalize the digital characters.(4)In the process of digital character recognition,we choose both statistical features and structural features as the feature of digital characters,and the recognition of image is achieved by using template matching,support vector machine and BP neural network.