JPH07325880A - Watthour meter character recognition device - Google Patents

Abstract

PURPOSE:To improve the recognition rate of lacking rotation characters of a watthour meter and to accurately recognize division characters by providing a character segmentation means, a reference measure segmentation means, a measurement recognition means and a judgement means. CONSTITUTION:In a measure and character segmentation part 6, the position address, height and width of the character are detected and also the position address and height of a reference measure are detected. Character information is supplied to a character recognition part 7, a prescribed dictionary is referred to and the character is recognized. The recognized result of the character, the degree of similarity, a segmentation information address and a reference measure information address are supplied through a result output part 8 to the result input part 9 of a personal computer 4. The supplied information is supplied to a measure value calculation part 10 further and a minimum measure value is calculated in the part 10. Then, the character is judged in a similarity degree calculation part 11 and a high-order digit is judged in a high-order digit judgement part 12 further. Thus, even when a part of the character of a recognition object lacks, by recognizing the next object character, the character is accurately recognized.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a watt hour meter character recognizing device for recognizing a rotating character of a watt hour meter in order to automate verification of a watt hour meter.

[0002]

2. Description of the Related Art Conventionally, the reading value of a watt hour meter has been read by visual observation. However, if such a reading value, that is, the letter, that is, the number indicated by the watt hour meter can be automatically read, such a reading is made. It is possible to perform automatic visual reading, and improve efficiency.

For automatic reading of characters, for example, a composite similarity method (see Japanese Patent Application No. 44-88109), a matching method and the like are used. The composite similarity method and the matching method basically recognize a character based on the degree of similarity between the recognition target character and the dictionary (reference character).

Further, in the watt hour meter, particularly in the case of a single phase, the minimum unit number is indicated by a scale, and it is necessary to read this minimum scale, but conventionally, such a minimum scale value is imaged. There is no general method for processing and measuring.

[0005]

The indication value of the watt hour meter is a rotating character such as a rotating counter character. Depending on the rotation state at the time of carrying the rotating character or the shooting position, the upper side of the character is displayed. Alternatively, since it seems that a part of the lower side is missing, the character recognition rate is lowered, and there is a problem that an erroneous determination or non-reading of the character occurs.

Further, in the case of the minimum scale value, as described above, there is no general method for measuring conventionally, and in order to measure such a minimum scale value by image processing,
There is a problem that a field of view with high resolution is required and high positioning accuracy is required.

Further, there is a problem in that the similarity between the character to be recognized, which is the most missing character, and the middle point of the next candidate character is reduced, resulting in erroneous determination and non-reading.

Further, when the upper part of the rotating counter character 7 of the watt hour meter is missing due to the photographing condition, the similarity is remarkably lowered as compared with other characters, resulting in erroneous determination and unreading. Furthermore, the rotation counter character 1 has a character width of only about ⅔ of other characters, and it is difficult to appropriately cut out the character, and the cutout failure causes the similarity to be lower than other characters. There is.

The present invention has been made in view of the above,
It is an object of the present invention to provide a watt hour meter character recognizing device capable of improving the recognition rate of a rotating character lacking a watt hour meter and accurately recognizing a divided character.

[0010]

To achieve the above object, a watt hour meter character recognition device of the present invention is a watt hour meter character recognition device for recognizing a rotating character of a watt hour meter. An image pickup means for picking up a rotated character, a character cutout means for cutting out a character from an image picked up by the image pickup means, a reference scale cutout means for cutting out a reference scale from the image picked up by the image pickup means, and the character from the cutout character. Measuring and recognizing the character and extracting a plurality of candidates from the recognition result and the similarity, and a determining means for determining the character based on the similarity and the upper digit information. It is a gist to have.

In addition to the above, the watt hour meter character recognition device of the present invention is the size of the character appearing in the window frame of the character on which the minimum scale is formed based on the image picked up by the image pickup means. And a minimum scale value calculating means for calculating the minimum scale value based on the size of the character.

Further, in the watt hour meter character recognition device of the present invention, when recognizing a character divided into upper and lower parts, a plurality of candidates of character likeness of each character are measured, and the continuity of numbers is determined from the similarity of these candidate characters. The gist is to have a means for recognizing characters by paying attention to.

A watt hour meter character recognition device of the present invention is a watt hour meter character recognition device for recognizing a rotating character of a watt hour meter, and an imaging means for imaging a rotating character of a watt hour meter, and the imaging means. Cutout means for cutting out the character width and the character height from the captured image, and detecting the midpoint between the character and the character based on the character height cut out by the cutout means, and based on the position of the midpoint A divided character determination unit that determines that the image captured by the image capturing unit is a divided character, a divided character dictionary that stores the divided character, and character recognition that performs character recognition of the divided character based on the divided character dictionary. The gist is to have means.

Further, the watt hour meter character recognition device of the present invention comprises a numeral judgment means for judging that the image information picked up by the image pickup means is the numeral 1 or 7 based on the character width cut out by the cutting means. The gist of the character recognizing means is to recognize the character by using the judgment of the numeral judging means in combination.

[0015]

In the watt hour meter character recognition device of the present invention, the character and the reference scale are cut out from the image picked up by the image pickup means, the height of the character is measured to recognize the character, and a plurality of characters are recognized from the recognition result and the similarity. The candidate is taken out, and the character is judged based on the judgment based on the similarity and the information of the upper digit.

In the watt hour meter character recognition device of the present invention, in addition to the above, the size of the character appearing in the window frame of the character on which the minimum scale is formed based on the image picked up by the image pickup means is set. Judgment is made and the minimum scale value is calculated based on this size.

Further, in the watt hour meter character recognition device of the present invention, when recognizing a character divided into upper and lower parts, a plurality of candidates for character likelihood of each character are measured, and the continuity of numbers is calculated from these candidate characters and similarity. Recognize characters by focusing on.

In the watt hour meter character recognition apparatus of the present invention, the character width and the character height are cut out from the image picked up by the image pickup means, and the middle point between the characters is detected based on the cut out character height. When it is determined that the divided character is a divided character based on the position of the middle point, character recognition of the divided character is performed based on the divided character dictionary.

Further, in the watt hour meter character recognition device of the present invention, it is judged that the numeral is 1 or 7 based on the character width cut out by the cutting means, and the character is recognized by using this judgment together. .

[0020]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of a watt hour meter character recognition apparatus according to an embodiment of the present invention. The watt hour meter character recognition device shown in the figure has a CCD camera 2 for reading the number of the watt hour meter 1 for displaying the number on the rotary counter, and the image information read by the CCD camera 2 is stored in the processing device 3. It is supplied to the image processing device 5, where it is binarized and supplied to the scale / character cutout unit 6.

Then, the scale / character cutout unit 6 detects the position address, height and width of the character, and also detects the position address and height of the reference scale. Then, the character information is supplied to the character recognition unit 7. The character recognition unit 7 has a dictionary as shown in the figure, and performs character recognition by referring to the dictionary. That is, character recognition is performed by comparing the cut-out character with the dictionary. For the recognized characters, five results are extracted as candidate characters as character-likeness, and the similarity is added to each of them.

The result output unit 8 outputs the character recognition result, the degree of similarity, the cutout information address, and the reference scale information address.
It is supplied to the result input section 9 of the personal computer 4 via. The information supplied to the result input unit 9 is supplied to the scale value calculation unit 10, where the minimum scale value is calculated. then,
The similarity calculation unit 11 determines the character, and the upper digit determination unit 12 further determines the upper digit. The upper digit is a digit one digit higher, and when the upper digit is a divided character, the digit one digit below the upper digit is nine.

FIG. 2A is a flow chart showing processing from character detection to character recognition and result communication, and FIGS. 2B to 2G are explanatory views in each processing of FIG. 2A. .

As shown in FIGS. 2 (b) and 2 (c), the counter type rotating character indicated by the watt hour meter 1 is different between the case of the single phase and the case of the three phases. 2 (b) is
In the case of a single phase, the lowest digit M ₃ is indicated by the scale, that is, the minimum scale value, and this minimum scale value and M ₂ one digit above it are displayed outside the window frame. ing. On the other hand, in the case of three phases, as shown in FIG.
As shown in (c), they are all inside the window frame. In any case, the smallest scale value of the least significant digit is represented by M ₃ , the digit above it is represented by M ₂ , the digit above it is represented by M ₁ , and the digit above it is It is represented by M ₀ .

Since the process of FIG. 2A is started by an interrupt, it is checked whether or not there is an interrupt (step 110).
First, the single-phase parameters are set (step 12).
0). Then, it is checked whether it is a single phase or a three phase (step 130), and in the case of three phases,
Three-phase parameters are set (step 140). Then, a window frame as shown in FIG. 2B or 2C is detected (step 150).

Then, as shown in FIG. 2D, M ₂
Is measured (step 160). In the measurement of the ratio of M ₂ , as shown in FIG. 2D, in the case of the same figure, the lower part of the number 2 partially appears in the upper half and the upper part of the number 3 appears in the lower half. Although partially appearing, the positions of the upper end and the lower end of each of the partial numbers are measured as y ₁ , y ₂ , y ₃ , and y ₄ , respectively.

Next, at step 170, it is checked whether or not M ₂ is a divided character, and it is not divided 1
If it is recognized as a character (step 180), the minimum scale value M ₃ is detected (step 210), but if it is a divided character, the upper character is recognized and the upper character is recognized as shown in FIG. 2 (e). The lower character is recognized (step 19
0, 200), the minimum scale value M ₃ is detected (step 210). In the detection of the minimum scale value M ₃ , the positions of the scale reference line are measured as y ₅ and y ₆ as shown in FIG. 2 (f).

Then, M ₁ is detected (step 220), and it is checked whether this M ₁ is a divided character (step 230). When it is recognized as one character that is not divided (step 240), M ₀ is detected (step 270), but when it is a divided character, the upper character and the lower character are recognized. (Steps 250 and 260), M ₀ is detected (Step 270). In the detection of M ₀ , as shown in FIG. 2G, the lower part of the number partially appears in the upper half and the upper part of the number partially appears in the lower half. The respective positions of the upper end and the lower end of the are measured as y ₇ , y ₈ , y ₉ , and y ₁₀ . then,
The above results are displayed and communicated (step 280,
290).

The results measured as described above are created as data in the format shown in FIG.

4 (a), (b) and (c) respectively show the number of the watt hour meter, the name of the scale recognition, the coordinate address on the image and the size of each part.

In FIG. 4A, a four-digit number "3590" is shown in a black window frame, each number is further surrounded by a small window frame, and a right side of these numbers is surrounded by a white window frame. The numbers are shown split. The numbers surrounded by the white window frame are for single phase only. Then, on the right side of the white window frame, there is a reference line, that is, a graduation scale reference line, and the scale of the smallest digit is carved in the white window frame with respect to the reference line. The number reading on the reference line of this scale corresponds to the minimum digit M ₃ , the number in the white window frame corresponds to M ₂ , the number at the right end of the black window frame is M ₁ , and the number to the left of this number Is M ₀ . As shown in FIG. 4B, the coordinate address on the image has the coordinate origin (0, 0) at the upper left end and the coordinates (511, 511) at the lower right end. Then, the x-axis is defined to the right of the coordinate origin, and the y-axis is defined below.

The size of the window frame and the characters in the window frame are defined as shown in FIG. 4 (c). That is, the character width is W, the character height is H, the character pitch is P,
The length of the window frame is K.

Next, with reference to FIG. 5, a basic measuring method for recognizing the minimum scale M ₃ in the scale value calculating section 10 will be described.

In FIG. 5, the character pitch P indicated by the centers of the numbers 2 and 3 is divided into 10 equal parts, and the reference line is located between them, but this reference line is divided into 10 equal parts between the characters 2 and 3. It shows the position of the fourth scale from the number 2 on the scale. Therefore, the numbers of the minimum scale M ₃ indicated by the reference line is 4, that is, read and 2.4. Assuming that the length from the center of the character pitch P divided by 2 to the reference line is n and the length from the reference line to the upper end of the character pitch is m, M ₃ is M ₃ = m / P as shown in the figure. expressed. Note that P / 2 = m + n and m = P / 2
Since it is −n, it can be calculated as M ₃ = (P / 2−n) / P.

FIG. 6 shows a method of recognizing characters in the window frame of the watt hour meter. In this figure, if there is dust in the window frame, first take a profile in the Y direction in the window frame, and remove dust at the diagonal threshold level so that dust is not detected as characters in the diagonal data at that time. Strong against noise so that it is not detected.

FIG. 7 shows a method for obtaining a measurement method suitable for recognizing characters in the window frame of the watt hour meter by expanding the basic measurement formula. It should be noted that this method is for using the reference scale.

In the case of FIG. 7A, the basic measurement formula is developed as shown below.

[Number 1] _{M 3 = m / P = [} y 5 - {y 2 - (y 2 -y 1) / 2}] / P = {y 5 - (y 2 + y 1) / 2} / P Note that, When it is negative, M ₃ * = 1−M ₃ . However, when there is a shadow on the upper edge, y ₁ is y ₃ and y ₂ is y.
Calculate by substituting ₄ .

In the case of FIG. 7B, the basic measurement formula is developed as shown below. _{M 3 = m / P = {} y 5 - (y 2 -H / 2)} / P

In the case of FIG. 7C, the basic measurement formula is developed as shown below.

[Number 2] _{M 3 = m / P = {} y 5 - (y 2 -H / 2)} / P M 3 = (P / 2-n) / P = [P / 2 - {(y 3 + y 2 ) / 2−y ₅ }] / P (when P≈k) = (2y ₅ + y ₄ −y ₃ −y ₂ −y ₁ ) / 2 (y ₄ −y
₁ ) Fig. 8 shows a method for obtaining a measurement method suitable for recognizing characters in the window frame of the watt hour meter by expanding the basic measurement formula. It should be noted that this method is for the case where the standard scale is not used.

In the case of FIG. 8C, the basic measurement formula is developed as shown below. l = P / 2−u = P / 2− (P−m−K ₁ ) = K ₁ −P / 2 + m Here, if K ₁ ≈P / 2, then 1≈m, so M ₃ = m / P≈l / P = {y ₄ − (y ₃ + y ₂ ) / 2} / P Here, if K≈P,

Equation 3] _{M 3 = {y 4 - (} y 3 + y 2) / 2} / (y 4 -y 1) FIG. 9 shows a method of recognizing divided character performed by the similarity calculating unit 11 . The counter type rotating character of the watt hour meter has the appearance as shown in FIG. 9 depending on its state with respect to the window frame.

As shown in FIG. 9A, in the case of a normal single character, character recognition is performed by comparing with the dictionary based on the height width of the detected character. When there is a missing character as shown in c), when the height of the character is less than a certain value, a virtual frame as shown by a dotted line in FIGS. Set the character height and compare with the dictionary. It should be noted that a portion of the virtual character having no character information is subjected to a meshing process in order to eliminate the characteristic, and the characteristic of the portion where the character is missing is offset. Then, as shown in FIG. 9D, each character is compared with the dictionary to calculate the degree of similarity. In addition, also in FIG. 9D, a portion indicated by a dotted line is a virtual frame.

Next, the determination method will be described with reference to FIG.

The data used in this determination method includes character candidate value C _iU for the upper character, similarity m _iU , character size h _iU , character candidate value C _iD for the lower character, and degree of similarity m _iD ,
Use character size h _iD . Note that i is 1 to 5,
There are 5 candidates.

In the determination method of FIG. 10, the presence or absence of character division is first determined (step 410), which is determined from the character size. If the height of the character is greater than or equal to the value set by the parameter, normal one-character recognition is performed, and the similarity of the first candidate character is checked (step 420). In the case of divided characters, the continuity of the upper and lower character candidates is confirmed (step 430), the sum of the similarities of the consecutive characters at that time is taken, and the one with the largest sum of the similarities when added is the tentative result. (Step 440).

Then, the upper digit judging section 12 checks whether or not the upper digit is a divided character (step 45).
0), if it is a divided character, the recognition result of the character is set to "9" (step 460). When the provisional result has not been changed, it is compared with the allowable similarity by the parameter, and when it is less than the allowable value, it is unread (step 48).
0), if it is larger than the allowable value, the character having the maximum similarity is taken as the answer (step 490).

FIGS. 11A to 11D are respectively shown in FIG.
It is a figure similar to (a)-(d). In FIG. 9, character recognition is performed using the virtual frame indicated by the dotted line as described above. For this reason, the part where the character is off is regarded as a space and is correlated with the dictionary pattern. In the actual characters, the part has a certain ratio for each character, that is, the width of the virtual frame is w, the length of the upper part of the virtual frame is h ₁ , and the length of the lower part is h as shown in FIG. _{2. If} the area of the character portion in the virtual frame is a, there should be a character pattern portion and a background with a ratio of (a / w × h ₁ ). Therefore, erroneous recognition may occur depending on the magnitude of this ratio. For this reason, in FIG. 11, the blank portion of this virtual frame is subjected to white / black shading processing as shown in the drawing, thereby making the correlations (similarities) of the blank portions different for each character type equal. The evaluation is based on the degree of correlation of the valid characters in the recognition frame.

As shown in FIG. 12A, in the same case as in FIG. 9, the degree of correlation of the blank portion of the virtual frame is (w × h ₁
-A) and the correlation degree of the character portion is proportional to (w × h ₂ ). However, when the hatching process is performed, the correlation degree of the hatched portion is as shown in FIG. Is proportional to (w × h ₁ ) / 2, and the degree of correlation of the character portion is proportional to (w × h ₂ ).

FIGS. 13 (a) and 13 (b) are diagrams showing the degree of similarity of the upper and lower characters of the character divided into two with respect to the recognized characters 0 to 9 before and after the shading processing, respectively. Is. In both figures, a circle ◯ indicates the similarity after the shading process, and a triangle Δ indicates the similarity before the shading process. The numbers in circles and triangles indicate the number of the candidate having the best similarity among the five candidates. As can be seen from both figures, it is shown that the overall degree of similarity is better when the shading process is performed.

14 to 23 show divided characters 0 to 0, respectively.
9, the similarity before and after the halftone processing is 5
It is a figure which shows the data taken about one candidate.

FIG. 24 is a block diagram showing the construction of a watt hour meter character recognition apparatus according to another embodiment of the present invention. As in the embodiment shown in FIG. 1, the watt hour meter character recognition device shown in the same figure captures the number of the watt hour meter 1 which displays the number on the rotary counter with the CCD camera 2, and the captured image information. Is supplied to the image processing device 5 of the processing device 31, the image information is binarized, and supplied to the character width cutout unit 21.

The character width cutting section 21 detects the character width from the image information from the image processing device 5. The character width cutout unit 21 has a “1 detection flag” indicating that the number “1” has been detected and a “7 detection flag” indicating that the number “7” has been detected, as shown in FIG. When the character width is detected from the image information, the flag is set when the character width is 1 or 7 based on the detected character width.

A character width detecting process in the character width cutting section 21 will be described with reference to FIG.

First, the process for determining that the detected character width corresponds to the numeral 1 and setting the 1 detection flag will be described. As the character width, as shown in FIG. 25B, the reference setting character width Mh is set in advance. Then, as shown in FIG. 25C, the center of gravity M of the detected character
The value of the width W1 from the virtual character start X address (Msx) obtained by subtracting the setting character width judgment Mh / 2 from j to the right side of "1", for example, 2Mh / 3 is added to the detected address X The start window address is M1x.

Then, the detected character Y start address M
sy, set character height Yw, 1 detection window width Mh-
(M1x−Msx) = Mh / 3 is set, and this area is set as one detection window as shown by the diagonal lines in FIG. Note that W1 varies depending on the type of rotary counter of the watt hour meter 1.

When one detection window is set as described above, if there is no character information in this one detection window, the character of the detected image information is judged to be "1", and
Set the detection flag. Note that this one detection window is set like the shadow shown on the right side of the numbers in FIG.

Next, the case where the detected character width is judged to correspond to the numeral 7 and the 7 detection flag is set will be described. In the case of the numeral 7, as shown in FIG. 25 (d), there is a groove for carry, and the character width is expanded by this groove for carry. Therefore, in this case, the detected character width is As shown in FIG. 25D, the set character width Mh + the width Ms2 that exceeds the allowable value is set. Therefore, in this case, the character of the detected image information is determined to be “7”, and the 7 detection flag is set. In addition, there is no groove for the carry in other characters,
It is unique.

As described above, when the 1-detection flag or the 7-detection flag is set based on the character width detected by the character-width cutting section 21, the image information is supplied to the character-height cutting section 22. The character height is detected.

The character height detection processing in the character height cutout unit 22 is to detect the character height in the divided characters. For example, as shown in FIG. In this case, the addresses of the upper end and the lower end of the number “2” which is the upper side character are Y1 and Y2 respectively, and the addresses of the upper end and the lower end of the number “3” which are the lower side character are Y3 and Y4 respectively. Then, the midpoint between the divided characters, that is, the middle point Ys is calculated.

That is, the middle point Ys is calculated by dividing the sum of the address Y3 at the upper end of the lower character "3" and the address Y2 at the lower end of the upper character "2" by 2 and dividing it by 2. The position point Ys is calculated by the following equation. Ys = (Y3 + Y2) / 2

Next, the image information is supplied to the reference scale cut-out portion 23, where the address Yk of the center of the black rectangular reference scale line shown on the right side of FIG. 26 is detected.

That is, in FIG. 26, the address Y5 at the upper end and the address Y6 at the lower end of the reference graduation line are detected, and the address Yk at the center of the reference graduation line is calculated from this address by the following equation. Yk = (Y6 + Y5) / 2

Next, the image information is supplied to the character recognizing section 24, and character recognition is performed. The character recognition unit 24 has a normal dictionary and a division dictionary as shown in the figure. The normal dictionary and the divided dictionary are as shown in FIG.

The character recognition unit 24 is not sure whether the character of the input image information is a complete character or a divided character. Therefore, when performing character recognition, a normal dictionary is used or a divided dictionary is used. Decide whether to use. For this determination, as shown in the following equation, the address Ys of the middle point is
The absolute value of the difference obtained by subtracting the center address Yk of the reference scale line from is within a predetermined value n. | Ys−Yk | ≦ n The predetermined value n is a value of about (Y4−Y1) × 0.05, and is usually about 5.

That is, if the absolute value of the difference obtained by subtracting the address Yk of the center of the reference scale line from the address Ys of the middle point is within a predetermined value n, it is determined that the character is a divided character and the divided dictionary is used. The divided one character is recognized by.

When the absolute value of the difference obtained by subtracting the address Yk of the center of the reference scale line from the address Ys of the middle point is larger than the predetermined value n, the recognition is usually performed by the dictionary.

For the character recognized by the character recognition unit 24, a plurality of results are extracted as candidates for character-likeness, and the similarity is added to each of them.

The result of character recognition by the character recognition unit 24,
The similarity, the character position address, the reference scale address, and the character information are supplied to the result output unit 25, and are supplied from the result output unit 25 to the personal computer 4. The personal computer 4 comprehensively determines the character based on the character recognition result, the degree of similarity, the presence or absence of the 1 detection flag, the 7 detection flag, and the like.

[0069]

As described above, according to the present invention,
Characters and reference scales are cut out from the image taken by the imaging means, the height of the character is measured to recognize the character, and a plurality of candidates are extracted from the recognition result and the similarity, and the judgment based on the similarity and the information of the upper digit are performed. Character recognition is performed based on the determination of characters based on the continuity of counter-type rotating characters of the watt hour meter. It can be done accurately. Further, the cutout information can be used to calculate the minimum scale, and the character of the recognition target digit can be determined in the state of the character of the higher digit.

Further, according to the present invention, the character width and the character height are cut out from the image picked up by the image pickup means, the middle point between the characters is detected based on the cut out character height, and the middle position is detected. When it is determined that the divided character is the divided character based on the position of the position, the character recognition of the divided character is performed based on the divided character dictionary, so that the accuracy of the character recognition of the divided character can be improved.

Further, according to the present invention, it is judged that the numeral is 1 or 7 based on the cut-out character width, and the character is recognized by using this judgment together. Therefore, the recognition accuracy can be further improved. You can

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a watt hour meter character recognition device according to an embodiment of the present invention.

FIG. 2 is a flowchart showing processing from character detection to character recognition and result communication of the watt hour meter character recognition device shown in FIG.

FIG. 3 is a diagram showing a data format of a result measured by the processing of FIG.

FIG. 4 is a diagram showing numbers of a watt hour meter, names of scale recognition, coordinate addresses on an image, and sizes of respective parts.

FIG. 5 is an explanatory view showing a basic measurement method for the recognition of the minimum scale M ₃ in the scale value calculator used in the watt hour meter character recognition apparatus of FIG.

FIG. 6 is a diagram showing a method of recognizing characters in a window frame of a watt hour meter.

FIG. 7 is a diagram showing a method of developing a basic measurement formula to obtain a measurement method suitable for recognizing characters in a window frame of a watt hour meter.

FIG. 8 is a diagram showing a method for developing a basic measurement formula to obtain a measurement method suitable for recognizing characters in a window frame of a watt hour meter.

9 is an explanatory diagram showing a method of recognizing divided characters, which is performed by a similarity calculation unit used in the watt hour meter character recognition device of FIG. 1. FIG.

FIG. 10 is a flowchart showing a character determination method.

11 is an explanatory diagram showing a case where a divided character recognition method performed in a similarity calculation unit used in the watt hour meter character recognition device of FIG.

FIG. 12 is a diagram illustrating the degree of correlation when the hatching process shown in FIG. 11 is performed and when the hatching process is not performed.

FIG. 13: Recognized characters 0 to 9 of upper and lower characters of a character divided into two before and after halftone processing
It is a figure which shows the similarity degree with respect to.

FIG. 14 is a diagram showing data obtained by taking the similarity between the divided characters 0 and 1 before and after the halftone processing for each of five candidates.

FIG. 15 is a diagram showing data obtained by calculating the similarity between the divided characters 1 and 2 before and after the halftone processing for each of five candidates.

FIG. 16 is a diagram showing data obtained by taking the similarity between the divided characters 2 and 3 before and after the halftone processing for each of five candidates.

FIG. 17 is a diagram showing data obtained by taking the similarity between the divided characters 3 and 4 before and after the halftone processing for each of five candidates.

FIG. 18 is a diagram showing data obtained by taking the similarity between the divided characters 4 and 5 before and after the halftone processing for each of five candidates.

FIG. 19 is a diagram showing data of the similarity between the divided characters 5 and 6 before and after the hatching process for each of five candidates.

FIG. 20 is a diagram showing data obtained by calculating the similarity between the divided characters 6 and 7 before and after the halftone processing for each of five candidates.

FIG. 21 is a diagram showing data of the similarity between divided characters 7 and 8 before and after halftone processing for each of five candidates.

FIG. 22 is a diagram showing data obtained by calculating the similarity between the divided characters 8 and 9 before and after the halftone processing for each of five candidates.

FIG. 23 is a diagram showing data obtained by calculating the similarity before and after the halftone processing for the divided characters 9 and 0 for each of five candidates.

FIG. 24 is a block diagram showing a configuration of a watt hour meter character recognition device according to another embodiment of the present invention.

25 is an explanatory diagram showing a character width detection process in a character width cutout unit used in the watt hour meter character recognition device of FIG. 24.

FIG. 26 is an explanatory diagram showing a character height detection process in a character height cutout portion used in the watt hour meter character recognition device of FIG. 24.

FIG. 27 is a diagram showing a division dictionary and a normal dictionary provided in a character recognition unit used in the watt hour meter character recognition device of FIG. 24.

[Explanation of symbols]

 1 Watt hour meter 2 CCD camera 3 Processing device 4 Personal computer 5 Image processing device 6 Scale / character cutout part 7 Character recognition part 8 Result output part 9 Result input part 10 Scale value calculation part 11 Similarity calculation part 12 Higher digit judgment part

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Ryumi Setomoto 2-24 Harumi-cho, Fuchu-shi, Tokyo 1 Toshiba Toshiba FA System Engineering Co., Ltd. (72) Inventor Shingo Sekiguchi 1-Toshiba-cho, Fuchu, Tokyo Toshiba Fuchu factory

Claims (5)

[Claims] 1. A watt hour meter character recognition device for recognizing a rotating character of a watt hour meter, comprising: an image pickup means for picking up an image of the rotating character of the watt hour meter; and a character cutout for cutting out a character from the image picked up by the image pickup means. Means, a reference scale cutout means for cutting out a reference scale from the image picked up by the image pickup means, the height of the character is measured from the cut out character, the character is recognized, and a plurality of characters are obtained from the recognition result and the similarity. A watt hour meter character recognizing device comprising: a measurement recognizing means for extracting the candidate of ???, and a judging means for deciding the character based on the information of the higher digit and the judgment based on the similarity. 2. The size of the character appearing in the window frame of the character on which the minimum scale is formed is determined based on the image captured by the image capturing means, and the minimum scale value is determined based on the size of the character. The watt hour meter character recognition device according to claim 1, further comprising a minimum scale value calculation means for calculating. 3. When recognizing vertically divided characters,
The watt hour meter character recognition device according to claim 1, further comprising means for measuring a plurality of candidates for each character-likeness and recognizing the character by paying attention to continuity of numbers based on the similarity with these candidate characters. . 4. A watt hour meter character recognizing device for recognizing a rotating character of a watt hour meter, comprising: an image pickup means for picking up an image of the rotating character of the watt hour meter; And a middle point between the character based on the height of the character cut out by the cutting means, and the image captured by the imaging means is divided based on the position of the middle point. It is characterized by including a divided character determination means for determining that it is a character, a divided character dictionary storing divided characters, and a character recognition means for performing character recognition of the divided characters based on the divided character dictionary. Watt hour meter character recognition device. 5. The numeral recognition means for judging, based on the character width cut out by the cutting means, that the image information picked up by the image pickup means is a numeral 1 or 7, and the character recognition means has the numeral The watt hour meter character recognition device according to claim 4, wherein the character recognition is performed by using the judgment of the judgment means together.