Numeral Instrument Reading Research And Application Based On Machine Vision

Numeral instrument recognition (NIR) based on machine vision is a breach for achieving interconnection of all instruments and realizing management-control integration and improving the intelligent management level. At present, the research on numeral recognition mostly focus on handwritten numeral recognition and printing numeral recognition and license plate recognition at home, but study on numeral instrument recognition is seldom reported. So, this paper has fairly explored the recognition of LCD and LED numeral instrument. The main achievement can summarized as follows:1.A numeral instrument recognition system of image processing has beeninvented.This system is composed of image processing software platform and displaying module of LCD and LED of single chip and image sampling module. It can show some statistics which include recognition rate and refuel recognition rate and error recognition rate to proving the performance of algorithm combination of image processing.2. An algorithm combination of image processing has been proposed.1)LED recognitionColor image segmentation can transform the color image to gray image. Incising algorithm based on Gradient matrix detach the number object region from the background. Global threshold method achieves binary images. Character incising algorithm get the coordinate of LED number character. Cross-line algorithm recognizes the integer and template matching algorithm recognizes the decimal. Recognition rate in Statistic result is near 100%, and recognition time is in 70ms under CPU of AMD 1.43G.2) LCD recognitionOctree and color formula method achieve the transformation of color image to gray image. Gradient matrix incising method detach the number region from the background. Canny edge detecting get the edge of LCD 7-segment numeral character. Edge tracking compensating or open of mathematical morphology remedy the canny edge breakpoint. Seed filling get binary image. Template matching method recognizes the LCD numeral. The recognition rate exceeds 99%.3.The debugging result verified that the algorithm combination of LCD and LED recognition can meet the application demands, having the practical value.Image processing software platform based on VC++ has been perfected. Engineering file system was introduced to manage the course of image processing. Besides single step debugging and continuum debugging, the image processing platform gives real time debugging that based on single chip. View segmentation gives two image showing windows that make the debugging of image processing more intuitionistic and convenient.