Design And Implementation Of Industrial OCR Recognition System Based On Machine Vision

With the continuous expansion of modern industrial production scale,industrial production has entered the era of big data.In the daily industrial production process,every operation link will produce many production data values that need to be recorded,which is used as a standard to measure whether the product is qualified.For many factories with older machines,their equipment does not have the corresponding data interface.Most of the time,they must rely on manual identification and manual copying to complete the recording of production data.Such a highly repetitive and boring work is undoubtedly a great waste of human resources,which greatly reduces the production efficiency of the factory.In order to solve this problem,there is an urgent need for an OCR system that can independently complete the image recognition of industrial information.Although the application of OCR recognition technology has been extended to many fields such as office,transportation,finance,etc.,its application in the field of industrial production is still insufficient.In this thesis,researches on a type of industrial information image are carried out,combined with its characteristics and innovatively applied OCR recognition technology to the field of industrial production,and an OCR automatic recognition system for industrial information images is designed and implemented.The specific work content is as follows:(1)Preprocessing the industrial information image.In addition to some basic preprocessing operations,in order to obtain a more ideal processing effect,the factors that may affect image quality in the industrial production environment are also fully considered.For industrial information images with uniform light distribution,the gray level distribution is improved through illumination compensation.The tilted image is processed based on the traditional Hough transform improvement method,which successfully corrects the tilt angle while shortening the processing time.(2)After analyzing and comparing the differences between industrial information images and ordinary text images,according to the characteristics of industrial information images,a new method that combines SSD algorithm,color space conversion and edge detection is designed to achieve precise positioning of the image information area and precise segmentation of basic recognition units.(3)According to the characteristics of different types of characters,different methods are used for recognition to improve the accuracy of recognition.Through the self-training of the font library,the recognition accuracy of label text characters has been improved.After using different methods to test the characters of the digital instrument,finally decided to use the threading method for recognition,which further reduces the time consumption while improving the recognition accuracy.(4)Carry out system testing,randomly select the same number of pictures from different data sets to compare and test the system performance,make corresponding evaluations on the overall performance of the system according to the test results,and improve the existing problems and deficiencies.