The Research On Reading Recognition Approaches Of Water Meter Based On Template Matching

The development of the science and technology accelerates the processing speed of the computer, which makes it possible to use the computer to intelligently identify complex images. At present, various digital image processing systems such as vehicle license plate identification, face identification and various meter detection and identification systems have been already applied to the daily life and industrial production. In particular, the intelligent identification of digital images has irreplaceable advantage over manual identification in special environment such as high temperature, high pressure, high radiation, etc. or systems connected with other digital equipment. Instead of the traditional manual meter reading, the water meter reading identification system proposed in this thesis applies to the water meter verification system and the automatic meter reading system, which improves the identification accuracy as well as the labor efficiency, which will save cost for enterprises and is therefore highly applicable.The proposed system covers the character wheel identification and pointer identification of the water meter. It adapts different demands and types of water meter to have the character wheel identification and pointer identification work simultaneously or respectively. The character wheel identification adopts the traditional template matching to identify characters:Hough conversion is used to precisely position the character wheel area that features rectangular frame. Regarding the possible half-character phenomenon, a bar template vertically arraying the11numbers of0-9-0is designed to match the area to be identified by sliding up and down, left and right. As for the pointer identification, the users will need to firstly set the template according to different types of water meter for the system to acquire parameters such as number of pointers, color of pointer, length of pointer, and relative position of pointer, which is a very easy action for the users who need only to click the center of each pointer that is to be identified and drag it to the pointer head. The users only need to set the template once for the same type of water meter and the template information can be shared. Meanwhile, this system also applies to specially shaped pointer with magnetic bottom. The users can select to set the template according to the shape of pointer. Taking advantage of the different ratios of pixel components, the system identifies the target pixel and it identifies the traditional red pointer as well as blue pointer. Twice partitioning to self-adapt binarization is adopted to cope with the uneven distribution of the gradation of image caused by unfixed light source position. After the pointer is located, the template matching method is adopted to identify the pointer.In the final experiment part,200test pictures are posted to show the strong anti-interference performance and satisfactory time complexity and detection accuracy of this system, which is therefore highly applicable.