Design And Implement Of Label Character Recognition System On Industrial Production Line

Recently, the label has been widely used in all kinds of electrical products ofindustrial production line. But artificial recognitions are easily to be fatigue and makemistakes, it is necessary to develop a set of label character recognition system applied toautomatically identify characters on labels in production line.This paper is aimed to research and develop a label character recognition system,for air condition line. This recognition system should be able to recognize threeparameters on two kinds of air conditioning label model，the mark and certificate, whichcontained the air conditioning type, manufacturing number, date of manufacturing,eliminating some characters form line interference, and comparing their consistency.According to the requirement, this paper mainly finished the following work:(1) General design for the whole label character recognition system. the generalstructure of the hardware system design and equipment selection, software system of thewhole design, function analysis and the key techniques analysis of software algorithmwere completed.(2) Studying and designing character segmentation and feature extractionalgorithm for label character. Character segmentation is the basis of the characterrecognition system. After tilt correction, a step-by-step localization method is used tocharacter zone positioning, segmenting single character after positioning. Positioningcharacter zone from the table frame line accurately is the difficulty of charactersegmentation, which this paper use a kind of table frame line removal algorithm, andgood results have been achieved. Feature extraction: it is the key of characterrecognition system. This paper respectively extract hole characteristic of character;Mixed characteristics of character, including statistical characteristics, structurecharacteristics and the texture characteristics which based on Gabor transform;Character strokes features of character including the end of the character, intersection,pen section, inflection point, fine strokes of character and profile characteristics ofcharacters which are keep out.(3) Designing three layers classifier for character recognition and classifying. The first layer classifier based on Euclidean distance, locate characteristics on charactercoarse classification with character hole characteristics, the characters is divided intoseveral subclasses; The second layer classifier further subdivide several subclasses withthe BP neural network according to mixed characteristics of character; The third layerclassifier, with the template matching algorithm, identify characters that can’t beclassified at previous two identify according to the character strokes and contourfeatures. Compared with single classifier, the design to some extent, improves the labelcharacter identification accuracy of character recognition system.(4) According to the design of the label character recognition system, adoptingVC++for software programming, and finally implementing the design, including thehardware system platform, software algorithm and the realization of the function of thesoftware realization.After finishing design and implement the system, the system is tested, results showthat the system’s functions and real-time performance can meet the requirements ofproduction line. its effect has been agreed by customer.