| As the terminal unit of power grid construction,power meters are widely used in power station field.Due to the wide variety of instruments in stations and the low efficiency of human verification,it is difficult to verify and identify a large number of instruments in real time.In order to solve this problem,this paper dose some research on automatic identification methods for power meter based on machine vision,which can automatic identify different types of instruments during the inspection process.The main works of the paper includes as following(1)During image acquisition procedure,due to the affection of image acquisition equipment and the external environment,noise is inevitably introduced to degrade the image of power meter.In this paper,a BM3D-based method is proposed to denoise the power meter image,which combines the non-local idea and the transform domain method.The propose method find the similar image blocks by using block matching method,and then implement denoising in the transform domain.The experimental results show that the method can effectively eliminate the noise of power meter image.(2)For pointer-type meters with uniform scale,this paper proposes a pointer-based meter reading recognition method based on two-dimensional code matching.Firstly high-quality instrument state image is collected in real time,and two-dimensional code positioning point information and instrument type information stored in corresponding database are acquired simultaneously by using two-dimensional code.And then the instrument image can be corrected according to the two-dimensional code positioning point information.The dial of power meter can be extracted quickly by using the priori geometric position relationship between the QR code and the meter.Finally,according to the acquired meter type information,select the corresponding meter reading recognition algorithm to quickly and accurately identify the meter reading.The experimental results show that the method can effectively improve the accuracy of pointer meter reading recognition,especially for complex background instrument images.(3)In this paper,the identifying method for liquid level meters is also investigated.Firstly,an image of the liquid level meter is acquired,which include the liquid level line and at least two sets of numbers.Then,the liquid column portion of the liquid level meter is extracted,and the entire instrument image is tilt-corrected according to the tilt angle thereof.Furthermore,the liquid column portion is re-extracted from the corrected image and the liquid level height is detected.And the two digitals close to the liquid level line are identified.Finally,the final level meter reading can be deduced based on the proportional relationship between the distance between the numbers and the height of the liquid level.The experimental results show that the method can effectively implement the identification of liquid level meter recognition.(4)Moreover,the identification method of the digital power meter is also presented.The digital meter image is acquired and the reading area of the digital meter image is extracted,and the reading area is tilt-corrected by Hough transform.In addition,the corrected reading area is normalized,and each digit of the reading area is extracted for recognition,then the recognition result is output.The experimental results show that the method can accurately extract the reading area of the digital meter and accurately identify the numbers in the area in order.Under the development environment of Visual Studio 2012 software,the OpenCV library is used to realize the automatic identification method of power meter based on machine vision,which can realize automatic recognition of pointer meter,liquid level meter and digital meter.In all,these experimental results show that the reliability of the automatic identification method of power meter studied in this paper can meet the requirements of practical applications and has certain development prospects. |
PDF Full Text Request