The Research Of Watermeter Detection System Based On Automatic Recognition Technology

The existing Water meter detecting system has the function that it can detect more than one Water meters in a time with the help of electric device. But it is a fact that, it is inefficient and great labor intensity detection method when control the machine manually or recognize the indication by eyes. Obviously,it doesn't match the reality that computer applications have been deeply used. And more importantly, the image recognition technology which based on the computer vision technology has also been widely used, in the fields of Pointer water meter recognition, using different methods which is accurate and efficient to combine the computer vision and automatic recognition, has been a hot topic of the scholars' research. And this paper presents a new scheme that recognizing the Pointer water meter automatically based on the SVM classifier and the HOG feature, besides, according to this automatic recognition, the paper also does some research on the automatic water meter detecting system.The first is the Pointer dial detection, according to the related knowledge about the technology and methods of image recognition on water meter dial. We propose the method which use the HOG feature and SVM classifier to detect the water meter dial. We extract the HOG feature of the Pointer dial image to be the positive samples, and also extract the bad pictures' HOG feature to form the negative samples. Then, we input these two kinds of samples into the SVM classifier to get the Pointer dial images from the water meter dial.The second is the Pointer image recognition. Before we recognize the indication, it is necessary to correct the image in right position during the automatic recognition. So according to the location of the Pointer dial, we use the Symmetry locate the circle center of the water meter dial image; and in the part of Pointer recognition, we should do the pretreatment, we use the method of Histogram equalization to weaken the light and color effect of the black Pointer image, and to the red one, we use the method of color analysis to get the target image by using the red color to minus the black color. After that, the otsu method is used to get the Pointer binary image. Furtherly using the Hough transform to detect the line which match the Pointer, we can calculate the indication by the rotation of this line.The last is the design of the detect process. When we apply this technology to the actual production, we should take all circumstances into consideration, including that the Pointer is seriously affected by the bubble or some dirty things to read. To solve the problem, according to the contrast function of the standard measurement water meter, this paper uses the auxiliary recognition method to help recognize the indication automatically in the circumstance that the Pointer is badly affected by the noise, moreover, the detection system will prompt the inspector to recheck the indication manually. At the same time, to adjust the detecting process of the manufacturers', this paper also discusses the functions of the and design the database of the information system which is to replace the method of giving the result statistics by manual computation.