| In the reality,multi-pointer instruments have been widely used in industrial and agricultural productions,such as pressure gauge and various types of equipment meter on ships and aircrafts,and so on.Many domestic water meters are still pointer type.The development of image processing technology enables machines to have vision like human beings,thus replaces manual and dangerous tasks.This makes automatic recognition of multi-pointer meters possible.The automatic recognition of the multi-pointer instruments by machine vision reduces the manpower and improves the work efficiency,but there are also a series of problems such as the inaccurate positioning of the dial meter,the interference of identifying the readings and the influence of external interference.Due to the problem of automatic recognition of pointer reading in multi-pointer instruments,we proposed a more rapid automatic pointer reading method.This method uses an existing image and template image to sift feature matching,and uses the geometric transformation to straighten existing image,then uses the color extracting to divide the pointer,according to the shape of the pointer to determine the pointer direction,then gets further readings.Experimental results show that the location and identification algorithm is simple and high accuracy,the running time is less than three seconds,and the recognition accuracy is 95%,overcome the random inclination angle of reading recognition algorithms.Taking the pointer water meter as an example,we contract this method to Hough transform pointer positioning,and they have basically the same pointer recognition accuracy but reduce halved running time.In addition this method has more general applicability,and can be applied to multiple dials.The results of this thesis can be used in research and application of automatic meter reading of multiple pointer and a reference significance for the study of complicated or dangerous instrument readings. |
PDF Full Text Request