CN113191351B

CN113191351B - Reading identification method and device of digital electric meter and model training method and device

Info

Publication number: CN113191351B
Application number: CN202110138431.4A
Authority: CN
Inventors: 沈美丽; 刘尊年; 高新亚; 张晓双
Original assignee: Qingdao University of Technology
Current assignee: Qingdao University of Technology
Priority date: 2021-02-01
Filing date: 2021-02-01
Publication date: 2022-04-01
Anticipated expiration: 2041-02-01
Also published as: CN113191351A

Abstract

The invention discloses a reading identification method and device of a digital ammeter and a model training method and device, and belongs to the technical field of image identification. The method comprises the following steps: extracting an interested area from the acquired digital electric meter image; determining at least one target region in the region of interest; determining a first rectangular area based on the at least one target area; cutting the first rectangular area to obtain a plurality of second rectangular areas with equal widths, wherein the width of each second rectangular area is determined based on the width of the first rectangular area and the width of each number in the readings; and inputting the images corresponding to the plurality of second rectangular areas into the reading recognition model to obtain the reading output by the reading recognition model. The method can reduce the cost of acquiring the readings of the electric meters, improve the efficiency and the accuracy of acquiring the readings of the electric meters, and can be used for identifying the readings of the digital electric meters.

Description

Reading identification method and device of digital electric meter and model training method and device

Technical Field

The invention relates to the technical field of image recognition, in particular to a method and a device for identifying the number of a digital ammeter and a method and a device for training a model.

Background

The digital electric meter records the electric quantity service condition of a user by acquiring pulse signals and automatically records and saves data. It is currently common to manually record readings on a digital electric meter.

However, since the number of digital electric meters is usually large, the amount of data collected manually is large, which results in high cost, low efficiency and easy error of data.

Disclosure of Invention

The invention provides a method and a device for identifying readings of a digital electric meter and a method and a device for training a model, which can reduce the cost of acquiring the readings of the electric meter and improve the efficiency and the accuracy of acquiring the readings of the electric meter, and the technical scheme is as follows:

in a first aspect, a method for identifying the number of a digital electric meter is provided, the method comprising:

extracting an area of interest from the acquired digital meter image, the area of interest including a digital meter screen, the digital meter image including opposing upper and lower edges, and opposing left and right edges;

determining at least one target region in the region of interest, the target region comprising a number in the representation of the digital electricity meter, the at least one target region being at a minimum upper spacing from the upper edge and a maximum upper spacing; a minimum spacing from the left edge is a minimum left spacing and a maximum spacing is a maximum left spacing;

determining a first rectangular area based on the at least one target area, wherein the first rectangular area comprises an upper side length and a lower side length which are opposite, and a left side length and a right side length which are opposite, the interval between the upper edge and the upper side length is smaller than or equal to the minimum upper interval, and the interval between the upper edge and the lower side length is larger than or equal to the maximum upper interval; the distance between the left edge and the left side length is smaller than or equal to the minimum left distance, and the distance between the left edge and the right side length is larger than or equal to the maximum left distance;

cutting the first rectangular area to obtain a plurality of second rectangular areas with equal widths, wherein the widths of the second rectangular areas are determined based on the width of the first rectangular area and the width of each number in the readings;

and inputting the images corresponding to the plurality of second rectangular areas into a reading recognition model to obtain the reading output by the reading recognition model.

Optionally, the target region is a rectangular region, and the determining at least one target region in the region of interest includes:

performing contour processing on the region of interest to obtain a plurality of third rectangular regions, wherein the contour processing comprises edge detection processing and/or expansion processing;

determining a third rectangular area meeting a target condition from the plurality of third rectangular areas as an initial target area;

wherein the target condition comprises at least one of: the area is within a target area range, the length is within a target length range, the width is within a target width range, the interval between the upper side length of the third rectangular region and the upper edge is within a first interval range, and the interval between the lower side length of the third rectangular region and the upper edge is within a second interval range;

a target area is determined based on the initial target area.

Optionally, the determining a target region based on the initial target region includes:

changing a width of the initial target area to a target width;

for any initial target area with the length larger than the first length, changing the length of any initial target area into the first length by moving the right length of any initial target area;

for any initial target area having a length less than the first length, changing the length of the any initial target area to the first length by moving the left length of the any initial target area.

Optionally, the method further comprises:

when the minimum interval between the initial target area and the left edge is larger than a first interval, moving the side length corresponding to the minimum interval to the left by a first distance;

and when the maximum interval between the initial target area and the left edge is larger than a second interval, moving the side length corresponding to the maximum interval to the left by a second distance.

Optionally, before the first rectangular region is cut to obtain a plurality of second rectangular regions with equal widths, the method further includes:

when the distance between the left side length of the first rectangular area and the left edge is larger than a third distance, moving the left side length of the first rectangular area to the left by a third distance;

and when the interval between the right side length of the first rectangular area and the left edge is smaller than a fourth interval, moving the right side length of the first rectangular area to the right by a fourth distance.

In a second aspect, a model training method is provided, the method comprising:

extracting a region of interest from the acquired digital meter sample image, the region of interest including a digital meter screen, the digital meter sample image including opposing upper and lower edges, and opposing left and right edges;

and training the recognition model by using the images corresponding to the plurality of second rectangular areas to obtain the reading recognition model.

a target area is determined based on the initial target area.

changing a width of the initial target area to a target width;

Optionally, the method further comprises:

In a third aspect, there is provided a reading identification apparatus for a digital electric meter, the apparatus comprising:

the extraction module is used for extracting an interested area from the acquired digital electric meter image, wherein the interested area comprises a digital electric meter screen, and the digital electric meter image comprises an upper edge, a lower edge, a left edge and a right edge which are opposite;

a first determining module for determining at least one target area in the region of interest, the target area including a number of the digital electric meter, the minimum interval between the at least one target area and the upper edge being a minimum upper interval, and the maximum interval being a maximum upper interval; a minimum spacing from the left edge is a minimum left spacing and a maximum spacing is a maximum left spacing;

a second determining module, configured to determine a first rectangular region based on the at least one target region, where the first rectangular region includes an upper side length and a lower side length that are opposite to each other, and a left side length and a right side length that are opposite to each other, a distance between the upper edge and the upper side length is smaller than or equal to the minimum upper distance, and a distance between the upper edge and the lower side length is larger than or equal to the maximum upper distance; the distance between the left edge and the left side length is smaller than or equal to the minimum left distance, and the distance between the left edge and the right side length is larger than or equal to the maximum left distance;

the cutting module is used for cutting the first rectangular area to obtain a plurality of second rectangular areas with equal widths, and the width of each second rectangular area is determined based on the width of the first rectangular area and the width of each number in the readings;

and the input module is used for inputting the images corresponding to the plurality of second rectangular areas into the reading recognition model to obtain the reading output by the reading recognition model.

Optionally, the target area is a rectangular area, and the first determining module includes:

the processing unit is used for carrying out contour processing on the region of interest to obtain a plurality of third rectangular regions, and the contour processing comprises edge detection processing and/or expansion processing;

a first determining unit configured to determine, as an initial target region, a third rectangular region that satisfies a target condition among the plurality of third rectangular regions;

a second determining unit for determining a target area based on the initial target area.

Optionally, the second determining unit is configured to:

changing a width of the initial target area to a target width;

Optionally, the second determining unit is further configured to:

Optionally, the device for identifying the number of the digital electric meter further includes:

the first moving module is used for moving the left side length of the first rectangular area to the left by a third distance when the distance between the left side length of the first rectangular area and the left edge is larger than a third distance;

and the second moving module is used for moving the right side length of the first rectangular area to the right by a fourth distance when the interval between the right side length of the first rectangular area and the left edge is smaller than a fourth interval.

In a fourth aspect, there is provided a model training apparatus, the apparatus comprising:

the extraction module is used for extracting an interested area from the acquired digital electric meter sample image, wherein the interested area comprises a digital electric meter screen, and the digital electric meter sample image comprises an upper edge, a lower edge, a left edge and a right edge which are opposite;

and the training module is used for training the recognition model by using the images corresponding to the plurality of second rectangular areas to obtain the reading recognition model.

Optionally, the second determining unit is configured to:

changing a width of the initial target area to a target width;

Optionally, the second determining unit is further configured to:

In a fifth aspect, there is provided a reading identification apparatus for a digital electric meter, comprising:

a processor;

a memory for storing executable instructions of the processor;

wherein the processor is configured to execute the instructions stored in the memory to implement the method of identifying a reading of a digital electricity meter of any of the first aspect of the present invention.

In a sixth aspect, there is provided a model training apparatus comprising:

a processor;

a memory for storing executable instructions of the processor;

wherein the processor is configured to execute instructions stored in the memory to implement the model training method of any of the second aspect.

In a seventh aspect, there is provided a computer storage medium having stored therein instructions that, when executed on a processing component, cause the processing component to perform the method of identifying a reading of a digital electricity meter according to any one of the first aspect.

In an eighth aspect, there is provided a computer storage medium having stored therein instructions that, when run on a processing component, cause the processing component to perform the model training method of any of the second aspects.

The technical scheme provided by the embodiment of the invention can have the following beneficial effects:

according to the method for identifying the number of the digital electric meter, the number of the digital electric meter is identified, at least one target area in the area of interest is determined after the area of interest is extracted from the acquired digital electric meter image, the first rectangular area is determined based on the at least one target area, the first rectangular area is cut to obtain a plurality of second rectangular areas with equal widths, the images corresponding to the second rectangular areas are input into the number identification model, and the number output by the number identification model is obtained.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention, as claimed.

Drawings

Fig. 1 is a flowchart of a method for identifying a number of a digital electric meter according to an embodiment of the present invention;

FIG. 2 is a flow chart of another method for identifying the number of a digital electric meter according to an embodiment of the present invention;

fig. 3 is a schematic diagram illustrating a digital electric meter image after graying according to an embodiment of the present invention;

fig. 4 is a schematic diagram illustrating the binarized image shown in fig. 3 according to an embodiment of the present invention;

FIG. 5 is a schematic diagram illustrating an example of determining an outer rectangular frame for the image shown in FIG. 4 according to an embodiment of the present invention;

FIG. 6 is a schematic diagram illustrating a reduced outer rectangular frame according to an embodiment of the present invention;

fig. 7 is a schematic diagram of the image shown in fig. 6 after edge detection processing according to an embodiment of the present invention;

FIG. 8 is a schematic diagram illustrating the image shown in FIG. 7 after being dilated according to an embodiment of the present invention;

FIG. 9 is a schematic diagram of a third rectangular area according to an embodiment of the present invention;

FIG. 10 is a schematic diagram of another third rectangular area provided in the embodiment of the present invention;

fig. 11 is a schematic diagram of a process for determining a target area based on an initial target area according to an embodiment of the present invention;

FIG. 12 is a schematic diagram of another process for determining a target area based on an initial target area according to an embodiment of the present invention;

FIG. 13 is a diagram illustrating a plurality of second rectangular areas according to an embodiment of the present invention;

FIG. 14 is a flowchart of a model training method according to an embodiment of the present invention;

FIG. 15 is a flow chart of another model training method provided by embodiments of the present invention;

FIG. 16 is a schematic diagram of a training sample image according to an embodiment of the present invention;

fig. 17 is a block diagram of a device for identifying the number of a digital electric meter according to an embodiment of the present invention;

FIG. 18 is a block diagram of a first determination module provided by embodiments of the present invention;

fig. 19 is a block diagram of an indication number recognition apparatus for a digital electric meter according to an embodiment of the present invention;

FIG. 20 is a block diagram of a model training apparatus according to an embodiment of the present invention;

FIG. 21 is a block diagram of a first determination module provided by embodiments of the present invention;

FIG. 22 is a block diagram of another model training apparatus provided in an embodiment of the present invention;

fig. 23 is a schematic structural diagram of a device for identifying a number of a digital electric meter according to an embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention clearer, the present invention will be described in further detail with reference to the accompanying drawings, and it is apparent that the described embodiments are only a part of the embodiments of the present invention, not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

The embodiment of the invention provides a method for identifying the number of a digital electric meter, which can be applied to first computer equipment, for example, a Graphics Processing Unit (GPU) in the first computer equipment. For example, referring to fig. 1, fig. 1 is a flowchart of a method for identifying a number of a digital electric meter according to an embodiment of the present invention, where the method includes:

step 101, extracting a region of interest from an acquired digital meter image, the region of interest including a digital meter screen, the digital meter image including opposite upper and lower edges and opposite left and right edges.

Step 102, determining at least one target area in the region of interest, wherein the target area comprises numbers in the display, the minimum interval between the at least one target area and the upper edge is the minimum upper interval, and the maximum interval is the maximum upper interval; the minimum spacing from the left edge is the minimum left spacing and the maximum spacing is the maximum left spacing.

103, determining a first rectangular area based on at least one target area, wherein the first rectangular area comprises an upper side length and a lower side length which are opposite, and a left side length and a right side length which are opposite, the interval between the upper edge and the upper side length is smaller than or equal to the minimum upper interval, and the interval between the upper edge and the lower side length is larger than or equal to the maximum upper interval; the left edge is spaced from the left edge by less than or equal to the minimum left spacing and spaced from the right edge by greater than or equal to the maximum left spacing.

And 104, cutting the first rectangular area to obtain a plurality of second rectangular areas with equal widths, wherein the width of each second rectangular area is determined based on the width of the first rectangular area and the width of each number in the readings.

And 105, inputting the images corresponding to the plurality of second rectangular areas into the reading recognition model to obtain the reading output by the reading recognition model.

In summary, according to the method for identifying the number of the digital electric meter provided by the embodiment of the invention, after the region of interest is extracted from the acquired digital electric meter image, at least one target region in the region of interest is determined, the first rectangular region is determined based on the at least one target region, the first rectangular region is cut to obtain a plurality of second rectangular regions with equal widths, and the images corresponding to the plurality of second rectangular regions are input into the number identification model to obtain the number output by the number identification model.

Referring to fig. 2, fig. 2 is a flowchart of another method for identifying a number of a digital electric meter according to an embodiment of the present invention, where the method may be applied to a first computer device, and as shown in fig. 2, the method may include:

step 201, extracting a region of interest from the acquired digital meter image, wherein the region of interest comprises a digital meter screen, and the digital meter image comprises an upper edge and a lower edge which are opposite, and a left edge and a right edge which are opposite.

The digital electricity meter image may be acquired by an image acquisition device. For example, it may be captured by a camera. The digital electric meter image is a color image and generally comprises a complete digital electric meter screen, the readings are positioned on the right side of the digital electric meter screen, the frame of the digital electric meter screen is obvious, and the brightness is obviously different from the background. The number may include three numbers or six numbers, and each number may be any one of "0" to "9".

Optionally, image processing may be performed on the digital electric meter image, then an external rectangular frame of the screen in the digital electric meter image is determined, and then an area surrounded by the external rectangular frame is determined as the region of interest, and the image processing may include graying and/or binarization.

Exemplarily, it is assumed that the image processing includes graying and binarization. Referring to fig. 3 to 5, fig. 3 is a schematic diagram illustrating a digital electric meter image after graying, fig. 4 is a schematic diagram illustrating the image shown in fig. 3 after binarization, and fig. 5 is a schematic diagram illustrating the image shown in fig. 4 after determining an external rectangular frame. Because the acquired digital electric meter image is a color image, the digital electric meter image is subjected to gray level and binarization in sequence so as to extract an interested area more accurately subsequently and identify the readings more accurately. After the digital electric meter image is subjected to graying and binarization in sequence, the edges of the digital electric meter screen and the external area are obvious, so that an edge rectangular frame can be determined. When a plurality of edge rectangular boxes are determined, the rectangular box with the largest area is determined as the outer rectangular box, so that the complete readings can be ensured to be included in the region of interest.

Optionally, the length and width of the outer rectangular frame may be scaled down to reduce the content included in the region of interest, other than the indication, for facilitating subsequent processing. And the scaling down ensures that each number in the index has a similar size. For example, referring to fig. 6, fig. 6 is a schematic diagram of the reduced external rectangular frame provided by the embodiment of the present invention, and compared with fig. 5, the content included in the region of interest in fig. 6 is less than the number of the readings, and the number of the readings is more obvious.

In fig. 3 to 6, the edge a1 is an upper edge, the edge a2 is a lower edge, the edge a3 is a left edge, and the edge a4 is a right edge.

Step 202, performing contour processing on the region of interest to obtain a plurality of third rectangular regions, where the contour processing includes edge detection processing and/or dilation processing.

Alternatively, the region of interest may be grayed before the contour processing.

Referring to fig. 7 and 8, taking contour processing including edge detection and dilation processing as an example, fig. 7 is a schematic diagram of the image shown in fig. 6 after edge detection processing according to an embodiment of the present invention, and fig. 8 is a schematic diagram of the image shown in fig. 7 after dilation processing according to an embodiment of the present invention. As shown in fig. 7, the positions of the respective digits in the index can be clearly observed after the edge detection processing. But the outline segments of the numbers are relatively random and discontinuous, which may result in inaccuracies in the subsequently obtained third rectangular area, e.g. a third rectangular area comprising the upper half of a number. As shown in fig. 8, after the expansion processing, the contour line segments can be made continuous, and the accuracy of subsequently determining the third rectangular area is ensured.

For example, referring to fig. 9, fig. 9 is a schematic diagram of a third rectangular area according to an embodiment of the present invention, and an area framed by each white rectangular frame in fig. 9 is the third rectangular area. As shown in fig. 9, of the plurality of third rectangular regions, there are a third rectangular region including the indication number and a third rectangular region including the other contents in the region of interest than the indication number. In fig. 9, an upper side length, a lower side length, a left side length, and a right side length of a third rectangular region are illustrated as an example of one third rectangular region. The side length of b1 is the upper side length of the third rectangular area, the side length of b2 is the lower side length of the third rectangular area, the side length of b3 is the left side length of the third rectangular area, and the side length of b4 is the right side length of the third rectangular area.

And step 203, determining a third rectangular area meeting the target condition from the plurality of third rectangular areas as an initial target area.

Optionally, the target condition may include at least one of: the area is within the range of the target area, the length is within the range of the target length, the width is within the range of the target width, the interval between the upper side length and the upper edge of the third rectangular region is within the range of the first interval, and the interval between the lower side length and the upper edge of the third rectangular region is within the range of the second interval.

The first and second ranges of intervals may be characterized in terms of coordinates. Optionally, a coordinate system is established with the upper left corner point of the digital electric meter image as an origin, the upper edge as a horizontal axis (i.e., x-axis) and the left edge as a vertical axis (i.e., y-axis), the horizontal axis is oriented to the right and the vertical axis is oriented to the bottom. At this time, the interval between the upper side length and the upper edge of the third rectangular region is the ordinate of the upper side length of the third rectangular region, and the interval between the lower side length and the upper edge of the third rectangular region is the ordinate of the lower side length of the third rectangular region. The first interval range may be a first ordinate range and the second interval range may be a second ordinate range. Of course, the way of establishing the coordinate system is only an exemplary one, and is not limited thereto.

As shown in fig. 9, the determined third rectangular areas may have at least one of the following problems: the third rectangular area comprises other characters except the readings in the interested area, and the other characters can cause interference on the subsequent recognition result; there is a third rectangular area comprising the entire region of interest; there is a third rectangular area comprising a plurality of consecutive digits in the display; there is a third rectangular area that includes both the numbers and other characters in the display.

For a third rectangular region including only characters other than the display, the area thereof is generally small. For a third rectangular region comprising the entire region of interest, its area is typically larger. And for a third rectangular region comprising only the numbers in the display, the area is generally between the two. Therefore, the third rectangular region having a smaller or larger area among the plurality of third rectangular regions can be excluded by setting the target area range.

Illustratively, the area of one number (except 1) in the index is generally 1200, the area of number 1 is generally 500, the total area of two consecutive numbers is between 2500-. This enables the third rectangular area including the readings to be retained while the third rectangular area not including the readings is removed.

Further, after excluding a part of the third rectangular region by the target area range, there may be a third rectangular region excluding the number in the index. For example, referring to fig. 10, fig. 10 is a schematic diagram of another third rectangular area provided in the embodiment of the present invention, and an area framed by each white rectangular frame in fig. 10 is the third rectangular area. As shown in fig. 10, the third rectangular region not including the number is generally larger in length, larger in width, or significantly different in position from the third rectangular region including the number. It is therefore possible to further exclude the third rectangular region not including the number by setting the target length range, the target width range, the first interval range, or the second interval range.

Illustratively, one number is typically 55 in width, so the target width range may be set to [0, 55 ]. The total length of two consecutive numbers is typically about 55, so the target length range can be set to 0, 65. In the region of interest, the area where the index is located is generally spaced apart from the upper edge (i.e., ordinate) by more than 5 and less than 37, so that the first interval range (i.e., the first ordinate range) may be set to [5, 37] and the second interval range (i.e., the second ordinate range) may be set to [5, 37 ].

Step 204, determining a target area based on the initial target area, wherein the target area comprises numbers in the readings, the minimum interval between the target area and the upper edge is the minimum upper interval, and the maximum interval is the maximum upper interval; the minimum spacing from the left edge is the minimum left spacing and the maximum spacing is the maximum left spacing.

After the coordinate system is established in the digital electric meter image in the manner described above, the target area includes a plurality of coordinate points. At this time, the coordinate point corresponding to the minimum upper interval is the minimum ordinate, the coordinate point corresponding to the maximum upper interval is the maximum ordinate, the coordinate point corresponding to the minimum left interval is the minimum abscissa, and the coordinate point corresponding to the maximum left interval is the maximum abscissa.

The target area may be a rectangular area, and the length and width of each initial target area may be uniform. Illustratively, the width of the initial target area may be changed to a target width so as to make the width of each initial target area uniform, for example, the target width may be 50.

When the lengths of the initial target areas are unified, it is necessary to ensure that the obtained target areas include complete numbers. Optionally, for any initial target area with a length greater than the first length, the length of any initial target area is changed to the first length by moving the right length of any initial target area. For any initial target region having a length less than the first length, the length of any initial target region is changed to the first length by moving the left length of the any initial target region. The width of two consecutive digits in the index is typically greater than 30, and the width of one digit (excluding 1) is about 25. In the case where 1 is not included in the index, the first length may be set to 25. This ensures that the resulting target area includes only digits and that the initial target area including consecutive digits can be changed to a target area including only the leftmost digit of consecutive digits.

Exemplarily, referring to fig. 11, fig. 11 is a schematic diagram illustrating a process of determining a target area based on an initial target area according to an embodiment of the present invention, where fig. 11 shows an image 11a and an image 11b, where neither of the indicators includes 1, an area framed by a white rectangle in fig. 11 is the initial target area, and an area framed by a white rectangle in fig. 11b is the target area. After unifying the length and width of the initial target area in fig. 11a, the target area in fig. 11b is obtained, the initial target area in fig. 11a includes other characters except numbers, and the target area in fig. 11b does not include other characters.

Assuming that the coordinate system is established in the digital electric meter image in the manner described above, since the length of the number 1 is smaller than the lengths of the other numbers, in this step 204, if the number 1 is located at the leftmost position in the index, the minimum interval between the initial target area and the left edge (i.e. the minimum abscissa in the initial target area) will be larger; if the number 1 is located at the rightmost digit of the display, the maximum separation of the initial target area from the left edge (i.e., the maximum abscissa in the initial target area) is larger. It is therefore possible to detect the minimum and maximum intervals of the initial target area from the left edge, and then to make an adjustment of the initial target area based on the detection result.

Optionally, when the minimum interval between the initial target region and the left edge is greater than the first interval, the side length corresponding to the minimum interval is moved to the left by a first distance. And when the maximum interval between the initial target area and the left edge is larger than the second interval, moving the side length corresponding to the maximum interval to the left by a second distance. Illustratively, the second distance may be 25. For example, referring to fig. 12, fig. 12 is a schematic diagram of another process for determining a target area based on an initial target area according to an embodiment of the present invention, where fig. 12 shows an image 12a and an image 12b with a number 1 located at the leftmost digit of the number, an area framed by a white rectangle in the image 12a is the initial target area, and an area framed by a white rectangle in fig. 12b is the target area. The minimum distance between the initial target region and the left edge in fig. 12a is larger, and the left length (i.e. the side length corresponding to the minimum distance) of the initial target region including the number 1 is moved to the left by a first distance, so that the target region including 1 shown in fig. 12b is obtained.

Alternatively, the foregoing process is described in terms of coordinates, where a first abscissa and a second abscissa may be determined. And when the minimum abscissa of the initial target area is larger than the first abscissa, moving the side length corresponding to the minimum abscissa leftwards by a first distance. And when the maximum abscissa of the initial target area is larger than the second abscissa, moving the side length corresponding to the maximum abscissa by a second distance to the left.

Step 205, determining a first rectangular area based on at least one target area, where the first rectangular area includes an upper side length and a lower side length which are opposite to each other, and a left side length and a right side length which are opposite to each other, a distance between the upper edge and the upper side length is less than or equal to a minimum upper distance, and a distance between the upper edge and the lower side length is greater than or equal to a maximum upper distance; the left edge is spaced from the left edge by less than or equal to the minimum left spacing and spaced from the right edge by greater than or equal to the maximum left spacing.

Taking the coordinate system established in the digital electric meter image as an example, the target area comprises a plurality of coordinate points, the interval between the upper edge and the upper side length of the first rectangular area is less than or equal to the minimum upper interval, and the ordinate corresponding to the upper side length of the first rectangular area is less than or equal to the minimum ordinate of the plurality of coordinate points; the interval between the upper edge and the lower side length of the first rectangular area is greater than or equal to the maximum upper interval, and the ordinate of the lower side length of the first rectangular area is equal to or greater than the maximum ordinate of the plurality of coordinate points; the distance between the left edge and the left side length of the first rectangular area is smaller than or equal to the minimum left distance, and the abscissa, which is equivalent to the left side length of the first rectangular area, is smaller than or equal to the minimum abscissa of the plurality of coordinate points; the distance between the left edge and the right side of the first rectangular area is greater than or equal to the maximum left distance, and the abscissa corresponding to the right side of the first rectangular area is greater than or equal to the maximum abscissa among the plurality of coordinate points.

The target regions are rectangular regions, and in step 205, the left side length and the right side length of each target region may be sorted, and the left side length with the minimum interval is taken as the left side length of the first rectangular region to be generated. And taking the right side length with the largest interval as the right side length of the first rectangular area to be generated.

In another way, describing the foregoing process in terms of coordinates, in this step 205, the abscissas (i.e. x coordinates) of the respective target areas may be sorted, the smallest abscissa is taken as the abscissa of the left side length of the first rectangular area to be generated, and the largest abscissa is taken as the abscissa of the right side length of the first rectangular area to be generated.

Further, the upper side length and the lower side length of each target region and the interval between the upper edge may be sorted, the upper side length with the minimum interval is used as the upper side length of the first rectangular region to be generated, and the lower side length with the maximum interval is used as the lower side length of the first rectangular region to be generated.

In another way, describing the foregoing process in terms of coordinates, the ordinate (i.e., y coordinate) of each target region may be sorted, the smallest ordinate is used as the ordinate of the upper side length of the first rectangular region to be generated, and the largest ordinate is used as the ordinate of the lower side length of the first rectangular region to be generated.

There may be missed detection in the foregoing process, for example, the rightmost digit of the number is not detected due to interference of other characters. Therefore, the number in the readings is not in the target area and is not in the first rectangular area, and the images obtained by cutting the first rectangular area subsequently lack the number which is missed to be detected, so that the reading identification result is wrong. Therefore, it is required to detect whether the interval between the left side length and the left edge of the first rectangular region is greater than the third interval, and whether the interval between the right side length and the left edge of the first rectangular region is less than the fourth interval. And selects to execute the subsequent step 206 and/or step 207 according to the detection result. When the interval between the left side length and the left edge of the first rectangular region is less than or equal to the third interval and the interval between the right side length and the left edge is greater than or equal to the fourth interval, the

subsequent steps

206 and 207 are not required to be performed, and the subsequent step 208 is directly performed.

In another way, describing the foregoing process in terms of coordinates, it may be detected whether the abscissa of the left side of the first rectangular region is greater than the third abscissa, and whether the abscissa of the right side of the first rectangular region is less than the fourth abscissa. And selects to execute the subsequent step 206 and/or step 207 according to the detection result. When the abscissa of the left side length of the first rectangular region is less than or equal to the third abscissa and the abscissa of the right side length is greater than or equal to the fourth abscissa, the

subsequent steps

And step 206, when the distance between the left side length and the left edge of the first rectangular area is larger than the third distance, moving the left side length of the first rectangular area to the left by a third distance.

Describing step 206 in terms of coordinates, when the abscissa of the left side length of the first rectangular area is greater than the third abscissa, the left side length of the first rectangular area is moved to the left by a third distance. Illustratively, the third distance may be 30.

And step 207, when the interval between the right side and the left edge of the first rectangular area is smaller than the fourth interval, moving the right side of the first rectangular area to the right by a fourth distance.

Describing step 207 in terms of coordinates, when the abscissa of the right side of the first rectangular area is less than the fourth abscissa, the right side of the first rectangular area is moved to the right by a fourth distance. Illustratively, the fourth distance may be 30.

And 208, cutting the first rectangular area to obtain a plurality of second rectangular areas with equal widths.

The width base of the second rectangular regionThe width of each digit in the first rectangular area and the display is determined. Alternatively, the width of the second rectangle may be the ratio of the width of the first rectangular area to the width of each number. The width of the first rectangular area is the absolute value of the difference between the interval of the right side length and the left edge of the first rectangular area and the interval of the left side length and the left edge of the first rectangular area. Describing in a coordinate angle, the width of the first rectangular area is an abscissa x of the right side length of the first rectangular area₁Abscissa x of left side length of first rectangular region₂Difference x of₁－x₂. Illustratively, the width of each number may be 25, and the width of the second rectangular region is (x)₁－x₂）/25。

After determining the width of the second rectangular region to be generated, vertical line segments perpendicular to the horizontal axis (x-axis) are drawn based on the width, and the interval between every two adjacent vertical line segments is the width. For example, referring to fig. 13, fig. 13 is a schematic diagram of a plurality of second rectangular areas provided by the embodiment of the present invention, and an area framed by a plurality of equal white rectangles in fig. 13 is the second rectangular area. Fig. 13 shows three second rectangular regions including the numeral 1, the numeral 3, and the numeral 6, respectively.

And step 209, inputting the images corresponding to the plurality of second rectangular areas into the reading recognition model to obtain the reading output by the reading recognition model.

Alternatively, the index recognition model may be trained on a K-nearest neighbor (KNN) model based on a KNN classification algorithm. The core idea of the KNN classification algorithm is as follows: if most of the K nearest neighbors of a sample in the feature space belong to a certain class, then the sample also belongs to this class and has the characteristics of the sample on this class. The KNN classification algorithm only determines the class of the sample to be classified according to the class of the nearest sample or a plurality of samples in the determination of classification decision.

After any image corresponding to the second rectangular area is input into the KNN scale recognition model, the KNN scale recognition model searches for an image nearest to the input image, and numbers in the image nearest to the input image are used as the result of the input image. For example, the KNN registration recognition model avoids matching problems between objects by calculating inter-object distances, which may include euclidean distances or manhattan distances, as an indicator of non-similarity between objects. Meanwhile, the KNN index recognition model makes a decision according to the dominant category in the k objects, rather than making a single object category decision.

In the above embodiment, the first rectangular area determined in step 205 is taken as an example, and the first rectangular area is cut subsequently. In another implementation, only the left side length and the right side length may be determined in step 205. And then detecting whether the interval between the left side length and the left edge is larger than a third interval or not, and detecting whether the interval between the right side length and the left edge is smaller than a fourth interval or not (namely detecting whether the abscissa of the left side length is larger than a third abscissa and detecting whether the abscissa of the right side length is smaller than a fourth abscissa or not). And adjusting the left length, adjusting the right length and/or not adjusting any length according to the detection result.

And then, firstly, determining the width of a second rectangular area to be generated according to the width of the left side length and the right side length, drawing a vertical line segment perpendicular to a horizontal axis (x axis) based on the width, wherein the interval between every two adjacent vertical line segments is the width. And then sorting the upper side length, the lower side length and the interval of the upper edge of each target area, and taking the upper side length with the minimum interval as the upper side length of the second rectangular area to be generated. And taking the lower side length with the largest interval as the lower side length of the second rectangular area to be generated.

Describing the process in terms of coordinates, sorting the vertical coordinates (i.e. y coordinates) of the target areas, and plotting the minimum vertical coordinate as the vertical coordinate of the upper side length of the second rectangular area to be generated to obtain the upper side length. And sorting the sum of the vertical coordinate of the upper left corner of each target area and the width of the target area, and drawing the lower side length by taking the maximum sum as the vertical coordinate of the lower side length of the second rectangular area to be generated. The manner of determining the upper side length and the lower side length of the second rectangular region to be generated in the embodiment of the present invention is only an exemplary illustration, and is not limited to this.

It should be noted that, the sequence of the steps of the method for identifying the number of a digital electric meter according to the embodiment of the present invention may be adjusted appropriately, and the steps may be increased or decreased according to the circumstances.

In summary, in the method for identifying the number of the digital electric meter according to the embodiment of the present invention, after the region of interest is extracted from the acquired digital electric meter image, the region of interest is subjected to contour processing to obtain a plurality of third rectangular regions, a third rectangular region that meets the target condition among the plurality of third rectangular regions is determined as the initial target region, and then the target region is determined based on the initial target region and the first rectangular region is determined based on the target region. Later cut first rectangle region and obtain a plurality of second rectangle regions that the width equals, and with the image input registration identification model that a plurality of second rectangle regions correspond, obtain the registration of this registration identification model output, the registration that can automatic identification digital ammeter in this process, need not artifical manual record, the cost of gathering the ammeter registration has been reduced and the efficiency of gathering the ammeter registration has been improved, and the image that the second rectangle region corresponds can accurate frame select the registration, thereby the accuracy of the registration of discernment digital ammeter has been improved.

In addition, when the interval between the left side length and the left edge of the first rectangular area is larger than the third interval, the left side length of the first rectangular area is moved to the left by a third distance, and when the interval between the right side length and the left edge of the first rectangular area is smaller than the fourth interval, the right side length of the first rectangular area is moved to the right by a fourth distance, so that the condition that the reading identification result is inaccurate due to digital omission can be avoided, and the reading accuracy of the digital electric meter is further improved.

The above describes the method for identifying the number of the digital electric meter, and when the method is executed, the identification model needs to be trained in advance to obtain the number identification model in the foregoing embodiment, and the training process of the identification model is described below.

The embodiment of the invention provides a model training method which can be applied to second computer equipment. The second computer device may be the same as the first computer device performing the method for recognizing the number of readings of the digital electric meter, or may be different from the first computer device performing the method for recognizing the number of readings of the digital electric meter. When the second computer device is different from the first computer device executing the indication recognition method of the digital electric meter, the second computer device may transmit the trained indication recognition model to the first computer device executing the indication recognition method of the digital electric meter. Exemplarily, fig. 14 is a flowchart of a model training method provided in an embodiment of the present invention, and referring to fig. 14, the method may include:

step 301, extracting a region of interest from the acquired digital meter sample image, wherein the region of interest includes a digital meter screen, and the digital meter sample image includes opposite upper and lower edges and opposite left and right edges.

Step 302, determining at least one target area in the region of interest, wherein the target area comprises numbers in the readings, the minimum interval between the at least one target area and the upper edge is the minimum upper interval, and the maximum interval is the maximum upper interval; the minimum spacing from the left edge is the minimum left spacing and the maximum spacing is the maximum left spacing.

Step 303, determining a first rectangular area based on at least one target area, where the first rectangular area includes an upper side length and a lower side length which are opposite to each other, and a left side length and a right side length which are opposite to each other, a distance between the upper edge and the upper side length is less than or equal to a minimum upper distance, and a distance between the upper edge and the lower side length is greater than or equal to a maximum upper distance; the left edge is spaced from the left edge by less than or equal to the minimum left spacing and spaced from the right edge by greater than or equal to the maximum left spacing.

And step 304, cutting the first rectangular area to obtain a plurality of second rectangular areas with equal widths, wherein the width of each second rectangular area is determined based on the width of the first rectangular area and the width of each number in the readings.

And 305, training the recognition model by using the images corresponding to the plurality of second rectangular areas to obtain a reading recognition model.

In summary, in the model training method provided in the embodiment of the present invention, after the region of interest is extracted from the acquired digital electric meter image, at least one target region in the region of interest is determined, then a first rectangular region is determined based on the at least one target region, the first rectangular region is cut to obtain a plurality of second rectangular regions with equal widths, and the recognition model is trained by using the images corresponding to the plurality of second rectangular regions to obtain the indication recognition model. The image corresponding to the second rectangular area can accurately frame the display number, so that the accuracy of the display number being modeled is improved. The registration identification model can output registration based on the input image corresponding to the second rectangular area, so that manual registration of registration is not needed, the cost of collecting electric registration is reduced, and the efficiency and the accuracy of collecting electric registration are improved.

Referring to fig. 15, fig. 15 is a flowchart of another model training method according to an embodiment of the present invention, where the method may be applied to a second computer device, and as shown in fig. 15, the method may include:

step 401, extracting a region of interest from the acquired digital meter sample image, where the region of interest includes a digital meter screen, and the digital meter sample image includes opposite upper and lower edges and opposite left and right edges.

And 402, performing contour processing on the region of interest to obtain a plurality of third rectangular regions, wherein the contour processing comprises edge detection processing and/or expansion processing.

And step 403, determining a third rectangular area meeting the target condition from the plurality of third rectangular areas as an initial target area.

Step 404, determining a target area based on the initial target area, wherein the target area comprises numbers in the readings, the minimum interval between the target area and the upper edge is the minimum upper interval, and the maximum interval is the maximum upper interval; the minimum spacing from the left edge is the minimum left spacing and the maximum spacing is the maximum left spacing.

Step 405, determining a first rectangular area based on at least one target area, wherein the first rectangular area comprises an upper side length and a lower side length which are opposite, and a left side length and a right side length which are opposite, the interval between the upper edge and the upper side length is smaller than or equal to the minimum upper interval, and the interval between the upper edge and the lower side length is larger than or equal to the maximum upper interval; the left edge is spaced from the left edge by less than or equal to the minimum left spacing and spaced from the right edge by greater than or equal to the maximum left spacing.

And 406, when the distance between the left side length and the left edge of the first rectangular area is greater than the third distance, moving the left side length of the first rectangular area to the left by a third distance.

And 407, when the interval between the right side length and the left edge of the first rectangular area is smaller than the fourth interval, moving the right side length of the first rectangular area to the right by a fourth distance.

And 408, cutting the first rectangular area to obtain a plurality of second rectangular areas with equal widths.

And 409, training the recognition model by using the images corresponding to the plurality of second rectangular areas to obtain a reading recognition model.

Optionally, the images corresponding to the plurality of second rectangular regions may be divided into 10 types of images according to 0 to 9, and each type of image is used to train the recognition model, so as to obtain the reading recognition model. For example, referring to fig. 16, fig. 16 is a schematic diagram of a training sample image according to an embodiment of the present invention, where the training sample image includes 10 types of images with numbers 0 to 9 in a one-to-one correspondence.

In this embodiment, reference may be made to the aforementioned step 201 to step 208 in the steps 401 to 408, which are not described herein again in this embodiment of the present invention.

In summary, in the model training method provided in the embodiment of the present invention, after the region of interest is extracted from the acquired digital electric meter image, the region of interest is subjected to contour processing to obtain a plurality of third rectangular regions, a third rectangular region that meets the target condition among the plurality of third rectangular regions is determined as the initial target region, and then the target region is determined based on the initial target region and the first rectangular region is determined based on the target region. And then cutting the first rectangular area to obtain a plurality of second rectangular areas with equal width, and training the recognition model by using the images corresponding to the second rectangular areas to obtain the reading recognition model. The image corresponding to the second rectangular area can accurately frame the registration number, so that the accuracy of the registration number recognition model is improved. The registration identification model can output registration based on the input image corresponding to the second rectangular area, so that manual registration of registration is not needed, the cost of collecting electric registration is reduced, and the efficiency and the accuracy of collecting electric registration are improved.

Optionally, in the above embodiment, the second computer device executing the model training method is taken as an example for explanation. In one example, different steps in the model training method may be performed by different modules. The different modules may be located in one device or in different devices. The embodiment of the invention does not limit the device for executing the model training method.

In this embodiment of the present invention, the first computer device or the second computer device may be a host, a server, or the like, which is not limited in this embodiment of the present invention.

It should be noted that, the order of the steps of the model training method provided in the embodiment of the present invention may be appropriately adjusted, and the steps may be increased or decreased according to the circumstances, and any method that can be easily conceived by those skilled in the art within the technical scope of the present invention should be covered within the protection scope of the present invention.

The method for identifying the number of the digital electric meter and the method for training the model provided by the embodiment of the invention are described in detail with reference to fig. 1 to 16, and the device for identifying the number of the digital electric meter and the device for training the model provided by the embodiment of the invention are described with reference to fig. 17 to 22.

Referring to fig. 17, fig. 17 is a block diagram of a device for identifying a number of a digital electric meter according to an embodiment of the present invention, where the device 50 includes:

an extracting module 501, configured to extract an area of interest from the acquired digital meter image, where the area of interest includes a digital meter screen, and the digital meter image includes an upper edge and a lower edge which are opposite to each other, and a left edge and a right edge which are opposite to each other;

a first determining module 502 for determining at least one target region in the region of interest, the target region comprising a number in the indication, the minimum spacing of the at least one target region from the upper edge being a minimum upper spacing, the maximum spacing being a maximum upper spacing; the minimum interval from the left edge is a minimum left interval, and the maximum interval is a maximum left interval;

a second determining module 503, configured to determine a first rectangular region based on the at least one target region, where the first rectangular region includes an upper side length and a lower side length which are opposite to each other, and a left side length and a right side length which are opposite to each other, a distance between the upper edge and the upper side length is smaller than or equal to a minimum upper distance, and a distance between the upper edge and the lower side length is greater than or equal to a maximum upper distance; the distance between the left edge and the left side length is less than or equal to the minimum left distance, and the distance between the left edge and the right side length is greater than or equal to the maximum left distance;

a cutting module 504, configured to cut the first rectangular region to obtain a plurality of second rectangular regions having equal widths, where the width of each second rectangular region is determined based on the width of the first rectangular region and the width of each number in the readings;

and an input module 505, configured to input the images corresponding to the plurality of second rectangular regions into the registration identification model, so as to obtain a registration output by the registration identification model.

To sum up, the device for identifying the number of readings of the digital electric meter according to the embodiment of the present invention extracts the region of interest from the acquired image of the digital electric meter through the extraction module, determines at least one target region in the region of interest through the first determination module, determines the first rectangular region based on the at least one target region through the second determination module, cuts the first rectangular region through the cutting module to obtain a plurality of second rectangular regions having the same width, inputs the images corresponding to the plurality of second rectangular regions into the number of readings identification model through the input module to obtain the number of readings output by the number of readings identification model, can automatically identify the number of the digital electric meter during the process, does not need to manually record the number of readings, reduces the cost for acquiring the number of the electric meter and improves the efficiency for acquiring the number of the electric meter, and the image corresponding to the second rectangular region can accurately frame the number of readings, thereby improving the accuracy of identifying the readings of the digital electric meter.

Optionally, the target area is a rectangular area, please refer to fig. 18, and fig. 18 is a block diagram of a first determining module according to an embodiment of the present invention, where the first determining module 502 includes:

the processing unit 5021 is used for carrying out contour processing on the region of interest to obtain a plurality of third rectangular regions, wherein the contour processing comprises edge detection processing and/or expansion processing;

a first determining unit 5022, configured to determine a third rectangular region that meets a target condition among the plurality of third rectangular regions as an initial target region;

wherein the target conditions include at least one of: the area is within the range of the target area, the length is within the range of the target length, the width is within the range of the target width, the interval between the upper side length and the upper edge of the third rectangular region is within the range of the first interval, and the interval between the lower side length and the upper edge of the third rectangular region is within the range of the second interval;

a second determining unit 5023, configured to determine the target area based on the initial target area.

Optionally, the second determining unit 5023 is configured to:

changing the width of the initial target area to a target width;

for any initial target region having a length less than the first length, the length of any initial target region is changed to the first length by moving the left length of any initial target region.

Optionally, the second determining unit 5023 is further configured to:

when the minimum interval between the initial target area and the left edge is larger than the first interval, moving the side length corresponding to the minimum interval leftwards by a first distance;

and when the maximum interval between the initial target area and the left edge is larger than the second interval, moving the side length corresponding to the maximum interval to the left by a second distance.

Optionally, referring to fig. 19, fig. 19 is a block diagram of another device for identifying the number of a digital electric meter according to an embodiment of the present invention, and on the basis of fig. 17, the device 50 further includes:

a first moving module 506, configured to move the left length of the first rectangular region to the left by a third distance when the distance between the left side length and the left edge of the first rectangular region is greater than the third distance;

the second moving module 507 is configured to move the right side of the first rectangular area to the right by a fourth distance when the distance between the right side and the left edge of the first rectangular area is smaller than the fourth distance.

In summary, in the device for identifying the number of the digital electric meter according to the embodiment of the present invention, after the region of interest is extracted from the acquired digital electric meter image by the extraction module, the processing unit performs contour processing on the region of interest to obtain a plurality of third rectangular regions, the first determination unit determines, as the initial target region, a third rectangular region that satisfies the target condition among the plurality of third rectangular regions, the second determination unit determines the target region based on the initial target region, and the second determination module determines the first rectangular region based on the target region. Later cut first rectangle region through the cutting module and obtain a plurality of second rectangle regions that the width equals, and through the image input registration recognition model that input module corresponds a plurality of second rectangle regions, obtain the registration of this registration recognition model output, the registration of digital ammeter can be discerned automatically to this in-process, need not artifical manual record, the cost of gathering the ammeter registration has been reduced and the efficiency of gathering the ammeter registration has been improved, and the image that the second rectangle region corresponds can accurate frame selection registration, thereby the accuracy of the registration of discernment digital ammeter has been improved.

In addition, when the interval between the left side length and the left edge of the first rectangular area is larger than the third interval, the first moving module moves the left side length of the first rectangular area to the left by a third distance, and when the interval between the right side length and the left edge of the first rectangular area is smaller than the fourth interval, the second moving module moves the right side length of the first rectangular area to the right by a fourth distance, so that the condition that the reading identification result is inaccurate due to digital omission can be avoided, and the reading accuracy of the identification digital electric meter is further improved.

It can be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working process of the above-described device for identifying a number of a digital electric meter may refer to the corresponding process in the foregoing method embodiment, and further description of the embodiment of the present invention is omitted here.

Referring to fig. 20, fig. 20 is a block diagram of a model training apparatus according to an embodiment of the present invention, the apparatus 60 includes:

an extracting module 601, configured to extract an area of interest from the acquired digital meter image, where the area of interest includes a digital meter screen, and the digital meter image includes opposite upper and lower edges, and opposite left and right edges;

a first determining module 602 for determining at least one target region in the region of interest, the target region including a number in the indication, the at least one target region having a minimum upper spacing from the upper edge and a maximum upper spacing; the minimum interval from the left edge is a minimum left interval, and the maximum interval is a maximum left interval;

a second determining module 603, configured to determine a first rectangular region based on the at least one target region, where the first rectangular region includes an upper side length and a lower side length which are opposite to each other, and a left side length and a right side length which are opposite to each other, a distance between the upper edge and the upper side length is smaller than or equal to a minimum upper distance, and a distance between the upper edge and the lower side length is greater than or equal to a maximum upper distance; the distance between the left edge and the left side length is less than or equal to the minimum left distance, and the distance between the left edge and the right side length is greater than or equal to the maximum left distance;

a cutting module 604, configured to cut the first rectangular region to obtain a plurality of second rectangular regions having equal widths, where the width of each second rectangular region is determined based on the width of the first rectangular region and the width of each number in the readings;

the training module 605 is configured to train the recognition model by using the images corresponding to the plurality of second rectangular regions to obtain a reading recognition model.

In summary, in the model training device provided in the embodiment of the present invention, after the extraction module extracts the region of interest from the acquired digital electric meter image, the first determination module determines at least one target region in the region of interest, the second determination module determines the first rectangular region based on the at least one target region, the cutting module cuts the first rectangular region to obtain a plurality of second rectangular regions with equal widths, and the training module trains the recognition model by using the images corresponding to the plurality of second rectangular regions to obtain the indication recognition model. The image corresponding to the second rectangular area can accurately frame the display number, so that the accuracy of the display number being modeled is improved. The registration identification model can output registration based on the input image corresponding to the second rectangular area, so that manual registration of registration is not needed, the cost of collecting electric registration is reduced, and the efficiency and the accuracy of collecting electric registration are improved.

Optionally, the target area is a rectangular area, please refer to fig. 21, and fig. 21 is a block diagram of a first determining module according to an embodiment of the present invention, where the first determining module 602 includes:

a processing unit 6021, configured to perform contour processing on the region of interest to obtain a plurality of third rectangular regions, where the contour processing includes edge detection processing and/or dilation processing;

a first determining unit 6022 configured to determine a third rectangular region satisfying the target condition among the plurality of third rectangular regions as an initial target region;

a second determination unit 6023 for determining the target region based on the initial target region.

Optionally, the second determining unit 6023 is configured to:

changing the width of the initial target area to a target width;

Optionally, the second determining unit 6023 is further configured to:

Optionally, referring to fig. 22, fig. 22 is a block diagram of another model training apparatus according to an embodiment of the present invention, and on the basis of fig. 20, the apparatus 60 further includes:

a first moving module 606, configured to move the left length of the first rectangular region to the left by a third distance when the distance between the left side length and the left edge of the first rectangular region is greater than the third distance;

the second moving module 607 is configured to move the right side length of the first rectangular area to the right by a fourth distance when the distance between the right side length and the left edge of the first rectangular area is less than the fourth distance.

In summary, in the model training apparatus provided in the embodiment of the present invention, after the region of interest is extracted from the acquired digital electric meter image by the extraction device, the processing unit performs contour processing on the region of interest to obtain a plurality of third rectangular regions, the first determination unit determines, as the initial target region, a third rectangular region that satisfies the target condition among the plurality of third rectangular regions, then the second determination unit determines the target region based on the initial target region, and the second determination module determines the first rectangular region based on the target region. And then cutting the first rectangular area through a cutting module to obtain a plurality of second rectangular areas with equal widths, and training the recognition model through a training module by utilizing images corresponding to the plurality of second rectangular areas to obtain a reading recognition model. The image corresponding to the second rectangular area can accurately frame the registration number, so that the accuracy of the registration number recognition model is improved. The registration identification model can output registration based on the input image corresponding to the second rectangular area, so that manual registration of registration is not needed, the cost of collecting electric registration is reduced, and the efficiency and the accuracy of collecting electric registration are improved.

It can be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working process of the model training apparatus described above may refer to the corresponding process in the foregoing method embodiment, and details of the embodiment of the present invention are not described herein again.

The embodiment of the invention provides a number identification device of a digital electric meter, which comprises: a processor; a memory for storing executable instructions of the processor; the processor is configured to execute the instructions stored in the memory to implement the method for identifying the number of the digital electric meter according to any one of the embodiments of the present invention.

For example, referring to fig. 23, fig. 23 is a schematic structural diagram of a number identification device of a digital electric meter according to an embodiment of the present invention, and as shown in fig. 23, the number identification device 70 of the digital electric meter includes: a memory 701 and a processor 702. The memory 701 is used for storing a program, and the processor 702 is used for executing the program stored in the memory 701, so as to implement any one of the methods for identifying the number of the digital electric meter provided in the embodiments of the present application.

Optionally, as shown in fig. 23, the reading identification device 70 of the digital electric meter may further include at least one communication interface 703 and at least one communication bus 704. The memory 701, processor 702, and communication interface 703 are communicatively connected via a communication bus 704.

The embodiment of the invention provides a model training device, which comprises: a processor; a memory for storing executable instructions of the processor; wherein the processor is configured to execute instructions stored in the memory to implement any of the model training methods described in embodiments of the present invention. Fig. 23 may be referred to for a structure of the model training apparatus, and details are not described herein in the embodiment of the present invention.

The embodiment of the invention provides a computer storage medium, wherein the storage medium is stored with instructions, and when the instructions are run on a processing assembly, the processing assembly is enabled to execute the method for identifying the number of the digital electric meter.

An embodiment of the present invention provides a computer storage medium, where instructions are stored in the storage medium, and when the instructions are executed on a processing component, the processing component is caused to execute any one of the model training methods according to the embodiments of the present invention.

The above-described embodiments may be implemented by software, hardware, firmware, or any combination thereof. When implemented in software, may be embodied in the form of a computer program product comprising computer instructions for causing a computer to perform the method of any one of the embodiments of the present invention.

The computer may comprise a general purpose computer or a network of computers, among others. The computer stores computer instructions by its storage medium or retrieves computer instructions from another storage medium. The storage medium may be any available medium that can be accessed by a computer or may comprise one or more data storage devices such as an integrated server and data center. The available media may be magnetic media (e.g., floppy disks, hard disks, tapes), optical media, or semiconductor media (e.g., solid state drives), among others.

In the embodiments of the present invention, "first", "second", "third", and "fourth", etc. are used for descriptive purposes only and are not to be construed as indicating or implying relative importance. "at least one" means one or more, "a plurality" means two or more, "and/or" is merely an associative relationship describing an associated object, meaning that three relationships may exist, e.g., a and/or B, may mean: a exists alone, A and B exist simultaneously, and B exists alone. In addition, the character "/" herein generally indicates that the former and latter related objects are in an "or" relationship. Unless explicitly defined otherwise.

Other embodiments of the invention will be apparent to those skilled in the art from consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the invention following, in general, the principles of the invention and including such departures from the present disclosure as come within known or customary practice within the art to which the invention pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the invention being indicated by the following claims.

It will be understood that the invention is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the invention is limited only by the appended claims.

Claims

1. A reading identification method for a digital electric meter, the method comprising:

2. The method of claim 1, wherein the target region is a rectangular region, and wherein determining at least one of the regions of interest comprises:

a target area is determined based on the initial target area.

3. The method of claim 2, wherein determining a target region based on the initial target region comprises:

changing a width of the initial target area to a target width;

4. The method of claim 3, further comprising:

5. The method of claim 1, wherein before the cutting the first rectangular area into a plurality of second rectangular areas of equal width, the method further comprises:

6. A method of model training, the method comprising:

7. An apparatus for identifying a reading of a digital electric meter, said apparatus comprising:

8. A model training apparatus, the apparatus comprising:

9. An indication recognition device for a digital electric meter, comprising:

a processor;

a memory for storing executable instructions of the processor;

wherein the processor is configured to execute the instructions stored in the memory to implement the method of identifying a reading of a digital electric meter of any of claims 1 to 5.

10. A model training apparatus, comprising:

a processor;

a memory for storing executable instructions of the processor;

wherein the processor is configured to execute instructions stored in the memory to implement the model training method of claim 6.