CN106940800A - Measuring instrument Recognition of Reading method and device - Google Patents

Abstract

The present invention relates to a method and device for identifying readings of metering instruments, including acquiring a digital character wheel image collected by a camera; performing preliminary positioning on the digital domain window area in the digital character wheel image, and then removing the outside of the digital domain window area Obtain an accurate digital domain window area image of the edge interference area, and the accurate digital domain window area image contains a plurality of characters; character segmentation is performed on the accurate digital domain window area image to obtain a single character image; A machine learning method is used to identify each of the individual character images. The invention can realize the high-precision recognition of the readings of the measuring instruments, and ensure the popularization and application of the intelligent meter reading of the water and electrical measuring instruments.

Description

Measuring instrument reading recognition method and device

technical field

The invention relates to the technical field of image recognition, in particular to a method and device for recognizing readings of meters.

Background technique

Water, electricity and heat are commonly used resources in family life, but "difficulty in meter reading" has become a problem that plagues society. If the problem of "meter" reading can be completely solved, and all meters such as electric meters, water meters, gas meters, heat meters, etc. Great convenience.

At present, the construction of intelligent user management and two-way interactive platform under the smart grid framework has been gradually realized, allowing ordinary households to realize user energy management, mobile terminal power purchase, water and electricity multi-meter collection, comprehensive information service, and remote home appliance control through the smart grid etc., to comprehensively improve the level of intelligent life of the people.

However, at present, for some old communities, most of the meters used in remote areas are still mechanical meters, and the collection of mechanical meter readings is mainly realized through manual entry. However, manual entry to collect readings consumes a lot of labor, time and effort, labor intensity, and high labor costs. Moreover, because the residents are not at home, it is difficult to ensure that the instrument data such as water, electricity, heat, etc. are obtained on time.

For the above-mentioned mechanical measuring instruments, through the digital recognition technology, the readings of water, electricity, heat and other meters are automatically obtained on time and accurately, so as to avoid many problems of manual meter reading become an urgent technical problem to be solved. However, the current identification technology cannot be accurately and reliably identified, and the error rate is high. Therefore, it cannot be popularized and applied to meter counting such as water, electricity, heat, etc.

Contents of the invention

The main purpose of the present invention is to provide a method and device for identifying readings of meters in view of the above-mentioned deficiencies in the prior art.

In order to achieve the above object, on the one hand, the present invention provides a method for identifying readings of meters, including:

Obtain the digital character wheel image collected by the camera;

Preliminarily locate the digital domain window area in the digital character wheel image, and then remove the edge interference area outside the digital domain window area to obtain an accurate digital domain window area image, and the accurate digital domain window area image contains multiple characters;

performing character segmentation on the precise digital domain window area image to obtain a single character image;

A machine learning method is used to identify each of the individual character images.

Preferably, the preliminary positioning of the digital domain window area in the digital character wheel image, and then removing the edge interference area outside the digital domain window area to obtain an accurate digital domain window area image specifically includes:

Detecting the upper and lower boundary lines of the digital domain window area in the digital character wheel image by means of a gradient in the vertical direction;

Detecting the left and right boundary lines of the digital domain window area after the upper and lower boundary lines are determined by means of a gradient in the horizontal direction;

Preliminary positioning of the digital domain window area according to the determined upper, lower, left, and right boundary lines, and cutting to form an image of the digital domain window area;

The image of the digital domain window area is initially processed by global binarization and its edge interference area is eliminated by using the connected domain analysis method to obtain an accurate image of the digital domain window area.

Preferably, performing character segmentation on the precise digital domain window area image to obtain a single character image specifically includes:

Using a global binarization algorithm to perform global binarization reprocessing on the precise digital domain window area image;

Performing noise elimination on the precise digital domain window area image after global binarization and reprocessing, to remove the interference area on the precise digital domain window area image;

Carry out character segmentation to the character region in the accurate digital domain window area image by adopting the method of vertical projection, to obtain the left and right boundaries of each character, and cut out a single character image;

The cropped single character image is binarized by local binarization algorithm, and noise points and interference areas are eliminated by prior knowledge and connected domain analysis.

Preferably, the method of using machine learning to identify each of the individual character images is specifically:

Each of the individual characters is recognized by using a neural network algorithm and combining the geometric prior knowledge features of each character.

In another aspect, the present invention provides a meter reading identification device, comprising:

Obtain module, be used for obtaining the digital character wheel image that camera collects;

The positioning unit is used to initially locate the digital domain window area in the digital character wheel image, and then remove the edge interference area outside the digital domain window area to obtain an accurate digital domain window area image. The image in the digital domain window area contains multiple characters;

A character segmentation unit, configured to perform character segmentation on the precise digital domain window area image to obtain a single character image;

The recognition unit is configured to use a machine learning method to recognize each of the individual character images.

Preferably, the positioning unit specifically includes:

The first detection module is used to detect the upper and lower boundary lines of the digital domain window area in the digital character wheel image with a vertical gradient method;

The second detection module is used to detect the left and right boundary lines of the digital domain window area after the upper and lower boundary lines are determined by the method of horizontal direction gradient;

A cropping module, configured to initially locate the digital domain window area according to the determined upper, lower, left, and right boundary lines, and crop to form an image of the digital domain window area;

The initial processing module is used to perform global binarization initial processing on the digital domain window area image and remove its edge interference area by using connected domain analysis method to obtain an accurate digital domain window area image.

Preferably, the character segmentation unit specifically includes:

A global binarization module is used to perform global binarization reprocessing on the precise digital domain window area image with a global binarization algorithm;

A noise elimination module, configured to perform noise elimination on the precise digital domain window area image after global binarization and reprocessing, so as to remove the interference area on the accurate digital domain window area image;

The character segmentation module is used to perform character segmentation on the character area in the precise digital domain window area image by adopting the method of vertical projection, so as to obtain the left and right boundaries of each character, and cut out a single character image;

The local binarization module is used to binarize the cropped single character image by using a local binarization algorithm, and eliminate noise points and interference areas by using prior knowledge and connected domain analysis.

Preferably, the identification unit is specifically used for:

Each of the individual characters is recognized by using a neural network algorithm and combining the geometric prior knowledge features of each character.

According to the meter reading recognition method and device provided by the present invention, the digital domain window area in the digital character wheel image is initially positioned, and then the edge interference area outside the digital domain window area is eliminated to obtain an accurate digital domain window area image The precise digital domain window area image contains a plurality of characters; character segmentation is performed on the precise digital domain window area image to obtain a single character image; each of the single character images is processed by machine learning identify. In this way, the recognition of meter readings can be realized, and the recognition accuracy is high, which can ensure the popularization and application of water and electricity meter counts.

Description of drawings

Fig. 1 is the flowchart of an embodiment of the meter reading identification method of the present invention;

Fig. 2 is a flow chart of another embodiment of the meter reading recognition method of the present invention;

Fig. 3 is a structural schematic diagram of an embodiment of a meter reading recognition device of the present invention;

Fig. 4 is a schematic structural view of the positioning unit in the meter reading identification device of the present invention;

Fig. 5 is a structural schematic diagram of the character segmentation unit in the meter reading recognition device of the present invention.

The realization of the purpose of the present invention, functional characteristics and advantages will be further described in conjunction with the embodiments and with reference to the accompanying drawings.

detailed description

Embodiments of the present invention are described in detail below, examples of which are shown in the drawings, wherein the same or similar reference numerals designate the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the figures are exemplary only for explaining the present invention and should not be construed as limiting the present invention.

Referring to Fig. 1, Fig. 1 shows an implementation flow of a meter reading recognition method provided by an embodiment of the present invention. For ease of description, only the parts related to the embodiment of the present invention are shown. Specifically, the method for identifying readings of meters includes the following steps:

S101. Acquire a digital character wheel image captured by a camera.

The camera is generally installed at a certain distance from the front of the meter dial. After receiving the collection command, the camera can collect the digital word wheel image and send it back.

S102. Preliminarily locate the digital domain window area in the digital character wheel image, and then remove the edge interference area outside the digital domain window area to obtain an accurate digital domain window area image, the precise digital domain window area The area image contains multiple characters.

Since the image of the digital character wheel captured by the camera includes the digital domain window area and the blank interference area located on the periphery of the digital domain window area, and the digital domain window area is the position where the metering data is displayed, the metering data is generally ten characters from 0 to 9 The blank interference area that makes up the periphery of the digital domain window area is a useless area. To recognize the numbers on the character wheel, it is necessary to perform preliminary positioning on the digital domain window area in the collected digital character wheel image, and then remove the initially positioned numbers. The edge interference area of the domain window area is used to obtain an accurate image of the digital domain window area.

S103. Perform character segmentation on the precise digital domain window area image to obtain a single character image.

Specifically, the characters in the image of the digital domain window area are arranged at intervals from left to right, and character recognition is character-by-character recognition. Therefore, it is necessary to first perform character segmentation on the image of the digital domain window area to obtain a single After the character image is binarized by the local binarization algorithm, the noise interference area is removed through the connected domain analysis, and then the recognition is carried out.

S104. Recognize each of the individual character images using a machine learning method. That is, recognize and judge which number from 0 to 9 the character is.

According to the meter reading recognition method provided in this embodiment, the digital domain window area in the digital character wheel image is initially positioned, and then the edge interference area outside the digital domain window area is eliminated to obtain an accurate digital domain window area image, The precise digital domain window area image contains a plurality of characters; character segmentation is performed on the precise digital domain window area image to obtain a single character image; each of the single character images is processed by machine learning Recognition, in this way, can realize the recognition of meter readings, and the recognition accuracy is high, which can ensure the popularization and application of water, electricity, heat and other meter counts.

Referring to FIG. 2 , FIG. 2 shows another implementation flow of a meter reading recognition method provided by an embodiment of the present invention. For ease of description, only parts related to the embodiment of the present invention are shown. Specifically, the method for identifying readings of meters includes the following steps:

S201. Acquire a digital character wheel image captured by a camera.

The camera is generally installed at a certain distance from the front of the meter dial. After receiving the collection command, the camera can collect the digital word wheel image and send it back.

S202. Detect the upper and lower boundary lines of the digital domain window area in the digital character wheel image by using a vertical gradient method.

Specifically, the structure of the measuring instrument includes a dial and a counting digital wheel, the dial is provided with a window, and the counting digital wheel is just located in the window. Based on the above-mentioned structure of the meter, in the digital character wheel image collected by the camera, the corresponding upper, lower, left and right boundary lines will be formed corresponding to the window edges on the dial. Therefore, by detecting the upper, lower, and The lower boundary line can know the upper and lower boundaries of the digital domain window area.

S203. Detect the left and right boundary lines of the digital domain window area after the upper and lower boundary lines are determined by means of a gradient in the horizontal direction.

That is to say, the digital domain window area is generally roughly rectangular. After determining the upper and lower boundary lines of the digital domain window area, the left and right boundaries of the digital domain window area can be known by detecting the left and right boundary lines in the digital domain window area.

S204. Preliminarily locate the digital domain window area according to the determined upper, lower, left, and right boundary lines, and crop to form an image of the digital domain window area.

That is to say, after the upper, lower, left, and right boundaries of the digital domain window area on the image are determined, cropping is performed according to the upper, lower, left, and right boundaries to form an image of the digital domain window area.

It should be noted that since the image of the digital character wheel captured by the camera includes the digital domain window area and the blank interference area located on the periphery of the digital domain window area, and the digital domain window area is the position where the metering data is displayed, the metering data generally ranges from 0 to 9 consists of ten characters in total, and the blank interference area outside the window area of the digital domain is a useless area. Therefore, through the above steps S202 to S204, the digital domain window area image can be obtained by positioning and cropping.

S205. Perform global binarization preliminary processing on the digital domain window area image, and use a connected domain analysis method to eliminate its edge interference area, so as to obtain an accurate digital domain window area image.

When the camera collects images, there may be interference pixels and/or interference areas in the collected digital character wheel image due to factors such as brightness and environment, and the image of the digital domain window area obtained by positioning and cutting also contains some interference pixels and /or the interference area, especially the edge area of the digital domain window area image just after cropping, there may be relatively obvious interference areas, and these interference areas will affect the recognition of characters. Therefore, the global binary value of the digital domain window area image After the initial processing, the edge interference area of the binarized digital domain window area is eliminated by the connected domain analysis method, and then an accurate digital domain window area image is obtained.

S206. Using a global binarization algorithm to perform global binarization reprocessing on the precise digital domain window area image. That is to say, this step needs to perform global binarization on the accurate digital domain window area image again to form a more standard binarized image.

S207. Perform noise removal on the precise digital domain window area image after global binarization and reprocessing, so as to remove interference areas on the precise digital domain window area image. That is to say, in this step S208, the interference pixels in the image of the digital domain window area can be removed through noise processing, and the interference pixels mainly refer to pixels near the character position, while the edge interference area in step S205 refers to Pixels at the edge of the image.

S208. Perform character segmentation on the character area in the precise digital domain window area image by using a vertical projection method to obtain the left and right boundaries of each character, and cut out a single character image.

Since the characters in the image of the digital domain window area are arranged at intervals from left to right, each character corresponds to a single character image formed by segmenting the precise digital domain window area image. in the image. In this way, a single character image can be recognized and its recognition accuracy can be improved.

S209. Use a local binarization algorithm to binarize the cropped single character image, and use prior knowledge and connected domain analysis methods to eliminate noise points and interference areas. That is to say, step S209 further binarizes the single character recognition area, and further eliminates the interfering pixels and interfering areas in the single character image, so as to facilitate character recognition in subsequent steps and improve character recognition efficiency and accuracy. Rate.

S210. Recognize each of the individual character images using a machine learning method. That is, a single character image is recognized to recognize which number from 0 to 9 the character in the character image is.

That is to say, in this embodiment, after the digital domain window area image is formed, the entire digital domain window area image is firstly subjected to global binarization initial processing and noise elimination processing, and then further global binarization reprocessing and further noise elimination , and then further perform projection segmentation on the image in the digital domain window area to form a single character image, then perform local binarization and noise removal processing on the single character image, and finally, perform local binarization and noise removal processing on the single character image The image is recognized. By adopting the above-mentioned processing method, the recognition accuracy can be greatly improved, and the reliability of the transcribed data can be ensured.

It can be understood that, in some embodiments of the present invention, in step S210, the method of machine learning is used to identify each of the individual character images: using a neural network algorithm and combining the geometric prior knowledge features of each character to identify each individual character image Recognize the above individual characters.

According to the meter reading recognition method provided in this embodiment, the digital domain window area in the digital character wheel image is initially positioned, and then the edge interference area outside the digital domain window area is eliminated to obtain an accurate digital domain window area image, The precise digital domain window area image contains a plurality of characters; character segmentation is performed on the precise digital domain window area image to obtain a single character image; each of the single character images is processed by machine learning identify. In this way, the recognition of meter readings can be realized, and the recognition accuracy is high, which can ensure the popularization and application of water and electricity meter counts.

Referring to FIG. 3 , FIG. 3 shows a meter reading identification device provided by an embodiment of the present invention. For ease of description, only parts related to the embodiment of the present invention are shown. Specifically, the meter reading identification device provided by the embodiment of the present invention includes:

Obtaining module 301, for obtaining the digital word wheel image that camera collects;

The positioning unit 302 is used to initially locate the digital domain window area in the digital character wheel image, and then remove the edge interference area outside the digital domain window area to obtain an accurate digital domain window area image. The image in the digital domain window area contains multiple characters;

A character segmentation unit 303, configured to perform character segmentation on the precise digital domain window area image to obtain a single character image;

The recognition unit 304 is configured to use a machine learning method to recognize each of the single character images.

Referring to Figure 4, in one embodiment of the present invention, the positioning unit 302 specifically includes:

The first detection module 3021 is used to detect the upper and lower boundary lines of the digital domain window area in the digital character wheel image by using a vertical gradient method;

The second detection module 3022 is used to detect the left and right boundary lines of the digital domain window area after the upper and lower boundary lines are determined by the method of horizontal direction gradient;

A cropping module 3023, configured to preliminarily locate the digital domain window area according to the determined upper, lower, left, and right boundary lines, and crop to form an image of the digital domain window area;

The initial processing module 3024 is used to perform global binarization initial processing on the digital domain window area image and remove its edge interference area by means of connected domain analysis to obtain an accurate digital domain window area image.

With reference to shown in Figure 5, in one embodiment of the present invention, described character segmentation unit 303 specifically comprises:

The global binarization module 3031 is used to perform global binarization reprocessing on the precise digital domain window area image with a global binarization algorithm;

The noise elimination module 3032 is used to perform noise elimination on the precise digital domain window area image after global binarization and reprocessing, so as to remove the interference area on the precise digital domain window area image;

The character segmentation module 3033 is used to perform character segmentation on the character area in the precise digital domain window area image by using a vertical projection method, so as to obtain the left and right boundaries of each character, and cut out a single character image;

The local binarization module 3034 is used to binarize the cropped single character image by using a local binarization algorithm, and eliminate noise points and interference areas by prior knowledge and connected domain analysis.

It can be understood that, the recognition unit 304 is specifically configured to: use a neural network algorithm and combine geometric prior knowledge features of each character to recognize each of the individual characters.

It should be noted that the meter reading recognition device in the embodiment of the present invention can be used to realize all the technical solutions in the above-mentioned method embodiments, and the functions of each functional unit can be realized according to the method in the above-mentioned method embodiments. For the implementation process, reference may be made to relevant descriptions in the foregoing method embodiments, and details are not repeated here.

According to the meter reading recognition device provided by the present invention, the digital domain window area in the digital character wheel image is initially positioned, and then the edge interference area outside the digital domain window area is eliminated to obtain an accurate digital domain window area image, so The precise digital domain window area image contains a plurality of characters; the precise digital domain window area image is character-segmented to obtain a single character image; each of the single character images is recognized by a machine learning method . In this way, the recognition of meter readings can be realized, and the recognition accuracy is high, which can ensure the popularization and application of water and electricity meter counts.

It should be noted that each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. For the same and similar parts in each embodiment, refer to each other, that is, Can. For the device or system embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and for relevant parts, please refer to part of the description of the method embodiments.

It should also be noted that in this article, relational terms such as first and second etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that these entities or operations Any such actual relationship or order exists between. Furthermore, the term "comprises", "comprises" or any other variation thereof is intended to cover a non-exclusive inclusion such that a process, method, article, or apparatus comprising a set of elements includes not only those elements, but also includes elements not expressly listed. other elements of or also include elements inherent in such a process, method, article, or device. Without more limitations, an element defined by the phrase "comprising a ..." does not preclude the presence of additional identical elements in the process, method, article or apparatus comprising said element.

The steps of the methods or algorithms described in connection with the embodiments disclosed herein may be directly implemented by hardware, software modules executed by a processor, or a combination of both. Software modules can be placed in random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other Any other known storage medium.

The above description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention will not be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (8)

1. a kind of measuring instrument Recognition of Reading method, it is characterised in that including：

Obtain the digital character wheel image of camera collection；

Primary Location is carried out to the numeric field window region in the digital character wheel image, then rejects the numeral Marginal interference area outside the window region of domain is to obtain accurate numeric field window region image, the accurate number Multiple characters are included in the window region image of word domain；

Character cutting is carried out to the accurate numeric field window region image, to obtain single character picture；

Each single character picture is identified using the method for machine learning.

2. measuring instrument Recognition of Reading method according to claim 1, it is characterised in that to described Numeric field window region in digital character wheel image carries out Primary Location, then reject the numeric field window region it Outer marginal interference area is specifically included with obtaining accurate numeric field window region image：

The upper and lower of numeric field window region in the digital character wheel image is detected in the method for vertical gradient Boundary line；

The numeric field window region after up-and-down boundary line is determined with the method detection of horizontal direction gradient Left and right boundary line；

The numeric field window region is gone out according to the upper and lower, left and right boundary line Primary Location of determination, And cutting forms numeric field window region image；

The numeric field window region image is carried out to handle at the beginning of global binaryzation and using the side of connected domain analysis Method rejects its marginal interference area, obtains accurate numeric field window region image.

3. measuring instrument Recognition of Reading method according to claim 1, it is characterised in that to described Accurate numeric field window region image carries out character cutting, is specifically included with obtaining single character picture：

Global binaryzation is carried out to the accurate numeric field window region image with global Binarization methods again to locate Reason；

The accurate numeric field window region image after being reprocessed to global binaryzation carries out noise elimination, To remove the interference range on the accurate numeric field window region image；

The character zone in the accurate numeric field window region image is carried out using the method for upright projection Character segmentation, to obtain the right boundary of each character, and cuts out single character picture；

Using local binarization algorithm to cutting after single character picture carry out binaryzation, and using first The method for testing knowledge and connected domain analysis eliminates noise spot and interference range.

4. measuring instrument Recognition of Reading method according to claim 1, it is characterised in that described right Each single character picture is identified specially using the method for machine learning：

Using neural network algorithm and with reference to each character geometry priori feature to each single word Symbol is identified.

5. a kind of measuring instrument reading identification device, it is characterised in that including：

Acquisition module, the digital character wheel image for obtaining camera collection；

Positioning unit, for carrying out Primary Location to the numeric field window region in the digital character wheel image, The marginal interference area outside the numeric field window region is rejected again to obtain accurate numeric field window region figure Multiple characters are included in picture, the accurate numeric field window region image；

Character segmentation unit, for carrying out character cutting to the accurate numeric field window region image, with Obtain single character picture；

Recognition unit, for each single character picture to be identified using the method for machine learning.

6. measuring instrument reading identification device according to claim 5, it is characterised in that described fixed Bit location is specifically included：

First detection module, for detecting number in the digital character wheel image in the method for vertical gradient The upper and lower boundary line of word domain window region；

Second detection module, the institute after up-and-down boundary line is determined for the method detection with horizontal direction gradient State the left and right boundary line of numeric field window region；

Module is cut, described in going out according to the upper and lower, left and right boundary line Primary Location of determination Numeric field window region, and cutting forms numeric field window region image；

First processing module, handles and adopts at the beginning of for carrying out global binaryzation to the numeric field window region image Its marginal interference area is rejected with the method for connected domain analysis, accurate numeric field window region image is obtained.

7. measuring instrument reading identification device according to claim 5, it is characterised in that the word Symbol cutting unit is specifically included：

Global binarization block, for global Binarization methods to the accurate numeric field window region figure As carrying out global binaryzation reprocessing；

Noise cancellation module, for the accurate numeric field window region after global binaryzation reprocessing Image carries out noise elimination, to remove the interference range on the accurate numeric field window region image；

Character segmentation module, for the method using upright projection to the accurate numeric field window region figure Character zone as in carries out Character segmentation, to obtain the right boundary of each character, and cuts out single Character picture；

Local binarization module, for using local binarization algorithm to cutting after single character picture Binaryzation is carried out, and noise spot and interference range are eliminated using priori and the method for connected domain analysis.

8. measuring instrument reading identification device according to claim 5, it is characterised in that the knowledge Other unit specifically for：

Using neural network algorithm and with reference to each character geometry priori feature to each single word Symbol is identified.