Research On Technology Of Instrument Value Identification Based On Embedded Machine Vision

With the development of science and technology,the degree of automation in all fields have become increasingly demanding.Due to the feather of good anti-electromagnetic interference ability and good reliability,the pointer instrument is still widely used in the industry.The pointer instrument is read by manual method,that is not easy to manage and automatically control,the disadvantage that human is easily the visual fatigue to cause misreading,buries the hidden danger for the production process.Therefore,the research on the use of machine vision automatic reading of the pointer instrument has important practical significance.This paper takes one voltmeter pointer instrument as reading object,and mainly researches the recognition method of pointer instrument.The process of the pointer meters reading system includes meter image acquisition,image preprocessing and segmentation,the meter pointer extraction and the value identification of the meter.The preprocessing and segmentation of meter image is an important step in the recognition process,for the result of the preproccess directly affecting the subsequent extraction of pointer.For the reason this paper deeply studies the image enhancement technology,the filter technique and the threshold segmentation technique and edge detection.This paper compare the pointer instrument image segmentation effect between Otsu with Niblack algorithm.Considering disadvantage of Canny losing edge for blur image,taking bilateral filter in place of gauss filter to improveit,and achieved better effect of edge detection.In the aspect of pointer position,this paper analyzes the shortcomings of subtraction to extract pointer,For the problem of large amount of calculation in the pointer detection with Hough transform,this paper firstly uses the connected domain analysis techniques to remove the connected domain that has nothing to do with the pointer and scale line,and then extracting pointer skeleton with Zhang fast thinning algorithm,improving the speed of the pointer transform detection on the premise of guarantee accuracy.In the identification of the pointer,this paper split zero-scale and full-scale line in thin image,and obtain the value of pointer instrument based on angle method.Finally,In order to verify the feasibility and practicability of recognition algorithm,this paper designs the embedded value identification system of the pointer meters.The system uses S5PV210 processor as the core,acquires the instrument image through the USB camera,implements instrument image processing and reading by the upper application program in the Linux operating system.The reading test shows that the value identification system has high precision and real-time performance,which will be practical.