| With the development and changes of modern computer communications technology anddigital instrumentation using in the field of testing industry in-depth, in order to be able toread the data needed more quickly and accurately so as to regulate the entire production andto form a virtuous cycle, you need to use the digital image processing technology,thus toidentify the important information of the instrumentation with high-speed and with automatic.This article intends to identify the multiply characters quickly, after analysis of thedevelopment and current states of genetic algorithm, Digital identification system based ongenetic algorithm was proposed.The genetic algorithm system reflects the correlation betweenthe digital identification object, and also the differences between each other.Considering therelevance of differences in design algorithms, and increase the timeliness and effectiveness ofthe system.In this thesis, the work was done as follows:1，in terms of image preprocessing, the HSV color space is closer to the human eye colorperception, images collected by the camera converted from RGB space to HSV space, andfinished image denoising through the median filter.2, in terms of rough image segement,the image is segmented aspects of map by Hcomponent of the HSV space so as to find out the approximate area of the digital symbols.And then through the horizontal and vertical scanning, find the critical point of each character,thus completing the segmentation of multiple characters, which relates to the conversion ofcolor images of space and the choice of the H component.3, in terms of accurate image segement, it is denoising after rough segement. Usinggenetic algorithms through multiple enerations of genetic crossover, mutation, and using therules of life and death by the fitness function.Identify the optimal threshold of the firstcharacter, while preserving the genetic characteristics of the characters, then completeseparation of other characters in the same image, then convert all characters to binary. Thencompleting inheritance of characteristics to the next image.4, in image recognition, the character size normalized to16*20, the BP neural network istrained and finally output the character. This part is completed by the other students.This thess is to obtain a better image segmentation results, while taking into account thereal-time system,it is conducive to industrial applications. |
