JP2008243103A - Image processing device, method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To precisely and automatically read an indicator value after correcting distortion in an image of a round meter. <P>SOLUTION: An image input part 11 inputs an original image of the round meter 2. A meter area extraction part 12 extracts a meter area from the original image. A projection conversion part 13 carries out projection conversion of the extracted meter area for correcting distortion of a display board area included in the meter area. A reference line decision part 15 decides a reference line. An area division part 14 divides the display board area, in which distortion is corrected, into a scale area, a digit area, and a center/indicator area. An indicator position decision part 16 decides an angle made by the center line of the indicator and the reference line of the round meter 2 as a position of the indicator. A digit recognition part 17 recognizes digits in the digit area. A digit position decision/accumulation part 18 decides the positions of digits, and based on the decided positional information of the indicator and the respective digits, calculates an indicator value shown by the indicator. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an image processing technique, and more particularly to an image processing apparatus, an image processing method, and an image processing program for processing an image of a round instrument and calculating an instruction value indicated by a pointer of the round instrument.

  Instruments such as analog meters are widely used for measuring and monitoring various systems. In recent years, automation of factories and cost reduction have been desired, and automation of inspection with a meter has been attempted. Methods for automating inspections with meters include meter replacement, meter modification, and automated visual inspection. Of these, replacement or modification of the meter has a serious problem that it is costly, visibility is deteriorated, and the system must be stopped during the replacement period. On the other hand, the automation of the visual recognition work has the advantage that it is only necessary to install a camera and an image input device and does not require the system to be stopped. Further, since the meter does not come into contact with the meter, visibility can be maintained, difficult high places and dangerous areas can be inspected, and temporary automatic monitoring can be easily performed. From this point of view, it is expected to automate the visual recognition work of analog meters by image processing, particularly round meters that occupy most of analog meters.

  FIG. 30 shows a processing flow of a general analog meter automatic reading system. A meter to be inspected is photographed with the installed camera (step S100), the image is analyzed (step S101), and an instruction value is output (step S102). The inspection result can be stored or analyzed (step S103).

  When reading the indicated value of the instrument using the image, the structure information of the instrument is required. Here, the structure information is information necessary for reading the instruction value. In a conventional system, prior information necessary for reading an instruction value is set in the system by attaching a marker or the like when the system is installed. Specifically, a reference position in image analysis is set by placing a marker on the display panel of the instrument, and information on the set reference position is used when reading an instruction value. In this way, information necessary for reading is given in advance as an initial setting. There is also a method of using a template image prepared in advance for reading an instruction value of an instrument.

  Further, when the instrument is photographed with a camera from an oblique direction, the image of the instrument photographed by the camera is distorted. If the reading of the indicated value is performed using such a distorted image, an error in reading the indicated value occurs. Therefore, it is necessary to read the indicated value after correcting the distortion generated in the image of the instrument. .

As a specific conventional technique related to image recognition, for example, the following Patent Document 1 describes a device that calculates an indication value of a pointer after approximating distortion due to an imaging angle of an analog meter by an ellipse. .
JP 2006-120133 A

  However, in the system that requires the above initial setting, the reading value of the instrument cannot be read if the prior information necessary for reading the reading value is not set. Moreover, attaching prior information such as markers to the display panels of all the instruments existing in large numbers in the factory is excessive in labor and cost. In the method of using a template image prepared in advance for reading the indicator value of the instrument, since the number of templates is limited, reading of the indicator value with higher accuracy is impossible. Therefore, there is a need for a system that automatically reads the indicated value using only an image of a round instrument photographed by a camera installed in the field without requiring prior initial settings.

  Further, since an image obtained by photographing a round instrument from an oblique direction is not exactly an ellipse, if the distortion of the image is large, depending on the conventional technique described in Patent Document 1, there is an error in reading the indication value of the pointer. growing.

  Further, when automatically reading the indicated value of the pointer, it is necessary to read the indicated value in a natural reading unit as if a human had read the indicated value by looking at the pointer. For example, when the amount of change in the numerical value (voltage value) between the scales of the meter is 5 volts, it seems natural for the human to read the indicated value of the meter in units of 1 volt. The indicated value calculated as a result of the automatic reading of the indicated value of the indicator of the instrument also needs to be an indicated value in units of 1 volt.

  The present invention solves the above-described problems of the prior art, corrects distortion generated in an image of a round instrument, and automatically reads an instruction value with high accuracy, and an image processing method And an image processing program.

  The present invention also provides an image processing apparatus, an image processing method, and an image processing program that automatically read an instruction value in a natural reading unit as if a human had read the instruction value by looking at the pointer. Objective.

  An image processing apparatus according to the present invention is an image processing apparatus for processing an image of a round instrument photographed by an image photographing means to calculate an instruction value indicated by a pointer of the round instrument, wherein the round instrument Image input means for inputting the image of the above, instrument area extraction means for extracting the area including the display panel area of the round instrument from the input image of the round instrument as the instrument area, and the extracted instrument area Projective transformation means for correcting distortion of the display panel area included in the instrument area, and a straight line drawn vertically downward from the center of the display board area corrected for the distortion is determined as a reference line A reference line determining means; a display panel area in which the distortion is corrected; a scale area which is an area where a scale exists; a number area which is an area where numbers indicating scale values are written; and the display board area The center of An area dividing means for dividing the pointer into a center / pointer area, and an angle formed by a center line of the circular instrument included in the center / pointer area and the reference line is the position of the pointer A pointer position determining means for determining, a number recognizing means for cutting out and recognizing numbers from the number area, and a straight line drawn from the center of the display board area corrected for distortion with respect to the center of gravity of each recognized number And an angle formed by the reference line is determined as the position of each numeral, numeral position determination / accumulation means for accumulating the position information of each determined numeral, position information of the determined pointer, and the accumulation Instruction value calculation means for calculating an instruction value indicated by the pointer based on the position information of each numeral.

  Preferably, the image processing apparatus of the present invention further includes a scale extraction unit that extracts a scale from the scale area obtained by dividing the display panel area by the area dividing unit, and a scale extraction result by the scale extraction unit. Based on the position information of each number stored in the number position determining / accumulating means, a straight line drawn from the center of the display panel area in which the distortion is corrected and each scale is formed, and the reference line are formed. A scale position determining / accumulating means for determining an angle and a numerical value corresponding to each scale and storing the correspondence information between the angle and the numerical value as position information of each scale, and accumulating in the scale position determining / accumulating means Reading unit determining means for determining a reading unit of the indicated value of the pointer based on the position information of each of the scales, wherein the indicated value calculating means includes the indicator position Based on the position information of the pointer determined by the determining means and the position information of each scale accumulated in the scale position determining / accumulating means, the indication indicated by the pointer using the determined reading unit of the pointer Calculate the value.

  The image processing method of the present invention is an image processing method for processing an image of a round instrument photographed by an image photographing means and calculating an instruction value indicated by a pointer of the round instrument, An image input step of inputting an image of a type instrument, an instrument region extraction step of extracting an area including a display panel area of the round instrument as an instrument area from the input image of the round instrument, and the extracted Projective transformation step for projective transformation of the instrument area to correct distortion of the display board area included in the instrument area, and a straight line drawn downward from the center of the display board area in which the distortion is corrected as a reference line A reference line determination step to determine, a display panel area in which the distortion is corrected, a scale area that is an area where a scale exists, a number area that is an area where a number indicating a numerical value of the scale is written, An area dividing step for dividing the center of the display panel area and a center / pointer area, which is a combined area of the pointer, and a center line of the pointer of the round instrument included in the center / pointer area and the reference line are formed. A pointer position determining step for determining an angle as the position of the pointer, a number recognizing step for cutting out and recognizing a number from the number area, and a center of gravity of each recognized number from the center of the display panel area in which the distortion is corrected Determining the angle between the straight line drawn with respect to the reference line and the reference line as the position of each numeral, and storing the position information of each determined numeral in the first storage means; and the determined pointer And an instruction value calculation step of calculating an instruction value indicated by the pointer based on the position information of each numeral and the position information of each numeral stored in the first storage means.

  Preferably, the image processing method of the present invention further includes a scale extraction step for extracting a scale from the scale area obtained by dividing the display board area, an extraction result of the scale, and accumulation in the first storage means. And the angle formed by the reference line and the straight line drawn with respect to each scale from the center of the display panel area where the distortion has been corrected based on the position information of each number, and the numerical value corresponding to each scale And storing the correspondence information between the angle and the numerical value in the second storage means as the position information of each scale, and based on the position information of each scale stored in the second storage means A reading unit determining step for determining a reading unit of the indicated value of the pointer, and the indicated value calculating step is stored in the position information of the determined pointer and the second storage means. Based on the position information of each scale, using a reading unit of the pointer the determined, to calculate the command value indicated by the pointer.

  The image processing program of the present invention is an image processing program for processing an image of a round instrument photographed by an image photographing means and calculating an instruction value indicated by a pointer of the round instrument, An image input process for inputting an image of the round instrument, an instrument area extraction process for extracting an area including a display panel area of the round instrument as an instrument area from the input image of the round instrument, and Projective transformation processing for projective transformation of the extracted instrument area to correct distortion of the display panel area included in the instrument area, and a straight line drawn vertically downward from the center of the display panel area in which the distortion is corrected A reference line determination process for determining as a reference line, a display board area in which the distortion is corrected, a scale area that is an area where a scale exists, and an area where numbers indicating scale values are written A region dividing process for dividing the character region and a center / pointer region that is a region combining the center of the display panel region and the pointer, a center line of the pointer of the round instrument included in the center / pointer region, and the Pointer position determination processing for determining an angle formed by a reference line as the position of the pointer, number recognition processing for recognizing by cutting out a number from the number area, and recognition from the center of the display board area in which the distortion is corrected A process of determining an angle formed by a straight line drawn with respect to the center of gravity of each number and the reference line as the position of each number, and storing the position information of each determined number in the first storage unit; Based on the determined position information of the pointer and the position information of each numeral stored in the first storage unit, an instruction value calculation process for calculating an instruction value indicated by the pointer is executed.

  Preferably, the image processing program of the present invention further includes a scale extraction process for extracting a scale from the scale area obtained by dividing the display board area, a scale extraction result, and the first storage. Based on the position information of each numeral accumulated in the means, the angle formed by the straight line drawn from the center of the display panel area corrected for each scale and the reference line, and each scale A corresponding numerical value is determined, and processing for storing the correspondence information between the angle and the numerical value in the second storage means as the position information of each scale, and the position of each scale stored in the second storage means A reading unit determination process for determining a reading unit of the indicated value of the pointer based on the information, and the indicated value calculation process includes the position information of the determined pointer and the second storage unit Based on the position information of each were stored scale, using a reading unit of the pointer the determined, to calculate the command value indicated by the pointer.

  According to the present invention, it is possible to automatically read the indicated value of the pointer with high accuracy after correcting the distortion generated in the image of the round instrument. As a result, even when the round instrument is photographed obliquely using a camera, it is possible to automatically calculate the indicated value of the pointer based on the photographed image.

  Further, the present invention extracts a scale from an image of a round instrument, accumulates the position information of each scale obtained based on the scale extraction result and the position information of each number in the display panel area, Based on the position information of each scale, the reading unit of the indication value of the pointer is determined. Then, based on the position information of the pointer and the position information of each scale, an instruction value indicated by the pointer is calculated using the reading unit of the pointer. Therefore, according to the present invention, it is possible to automatically read the instruction value in a natural reading unit as if a human read the instruction value by looking at the pointer.

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

(First embodiment)
FIG. 1 is a diagram illustrating an example of the configuration of the first exemplary embodiment of the present invention. The image processing device 1 is a processing device that processes an image of the round instrument 2 photographed by the camera 3 and calculates an instruction value indicated by a pointer of the round instrument 2.

  The image processing apparatus 1 includes an image input unit 11, an instrument region extraction unit 12, a projective conversion unit 13, a region division unit 14, a reference line determination unit 15, a pointer position determination unit 16, a number recognition unit 17, a number position determination / accumulation unit. 18, a numeric position correction unit 19, and an instruction value calculation unit 20.

  An image (original image) of the round meter 2 photographed by the camera 3 is input to the image input unit 11. The instrument area extraction unit 12 extracts an instrument area from the input original image of the round instrument 2. The instrument area is an area to be subjected to a projective transformation process, which will be described later. For example, the display panel area and the display panel of the round instrument 2 such as an area enclosed by a rectangle shown in FIG. It is an area including the area of the outer frame.

  The projective transformation unit 13 performs projective transformation on the extracted instrument area, and corrects distortion of the display panel area included in the instrument area. By this correction processing, the display panel area that is distorted in the original image is corrected to a circular shape.

  The area dividing unit 14 divides the display panel area in which the distortion is corrected into a scale area that is an area where a scale exists, a number area that is an area where numbers indicating scale values are written, and the display board area The center and the pointer area are divided into the area including the center and the pointer.

  The reference line determination unit 15 determines a straight line drawn downward in the vertical direction from the center of the circular display panel area in which the distortion is corrected as a reference line. The reference line is a line that serves as a reference for the numbers and scale positions written in the display panel area. As will be described later, the number position is an angle formed by, for example, a straight line drawn from the center of the display panel area with respect to the center of the number and a reference line. The position of the scale is an angle formed by a straight line drawn with respect to the scale from the center of the display panel area and a reference line.

  The pointer position determination unit 16 determines the angle formed by the center line of the pointer of the round instrument 2 included in the center / pointer area and the reference line as the position of the pointer.

  The number recognition unit 17 recognizes numbers existing in the number area. The numerical position determining / accumulating unit 18 determines an angle formed by a straight line drawn from the center of the display panel area where the distortion is corrected with respect to the center of gravity of each recognized numeral and a reference line as the position of each numeral, The position information of each determined number is stored.

  The numerical position correction unit 19 corrects the numerical position information stored in the numerical position determination / accumulation unit 18. Specific processing of the numeral position correction unit 19 will be described later. Note that the image processing apparatus 1 may adopt a configuration in which the numerical position correction unit 19 is omitted.

  The instruction value calculation unit 20 indicates the instruction value indicated by the pointer based on the position information of the pointer determined by the pointer position determination unit 16 and the position information of each number stored in the number position determination / accumulation unit 18. Is calculated.

  The functions of the image processing apparatus 1 and each unit described above are realized by a CPU and a program executed on the CPU. The program for realizing the present invention can be stored in a computer-readable recording medium such as a semiconductor memory, a hard disk, a CD-ROM, a DVD, or the like, provided by being recorded in these recording media, or communication. It is provided by transmission / reception using a network via an interface.

  FIG. 2 is a diagram illustrating an instruction value calculation processing flow according to the first embodiment of this invention. First, the image input unit 11 inputs an image (original image) of the round meter 2 taken by the camera 3 (step S1). Next, the instrument area extraction unit 12 extracts an instrument area from the input original image of the round instrument 2 (step S2).

  The projective transformation unit 13 performs projective transformation on the extracted instrument region, and corrects distortion of the display panel region included in the instrument region (step S3). Then, the reference line determination unit 15 determines a straight line drawn downward in the vertical direction from the center of the display panel area whose distortion has been corrected as a reference line (step S4). The area dividing unit 14 divides the display board area in which the distortion is corrected into a scale area, a numerical area, and a center / pointer area (step S5).

  The pointer position determination unit 16 determines the angle formed by the center line of the pointer of the round instrument 2 included in the center / pointer area and the reference line as the position of the pointer (step S6).

  Next, the number recognition unit 17 recognizes by cutting out a number from the number area (step S7). The numerical position determining / accumulating unit 18 determines an angle formed by a straight line drawn from the center of the display panel area where the distortion is corrected and the center of gravity of each recognized numeral and a reference line as the position of each numeral. The position information of the determined numbers is accumulated (step S8).

  Then, the instruction value calculation unit 20 calculates the instruction value indicated by the pointer based on the position information of the pointer determined in step S6 and the position information of each number accumulated in step S8 (step S9). ).

(1) Extraction of Instrument Area The instrument area extraction process in step S2 of FIG. 2 will be described. The instrument area extraction unit 12 performs binarization processing on the input original image of the round instrument 2 illustrated in FIG. 3A, for example, and is an area filled with white in FIG. A display panel area such as is detected. When the original image is an image obtained by obliquely photographing the round instrument 2 with the camera 3, as shown in FIG. 3B, the outline of the display board area is a distorted circular shape.

  Next, the instrument area extraction unit 12 extracts a rectangular area including the detected display panel area from the original image, and sets the extracted area as an instrument area. For example, an area as shown in a portion surrounded by a bold rectangle in FIG. 3C is extracted as an instrument area.

  Details of the instrument region extraction processing will be described with reference to FIGS. FIG. 4 is a diagram illustrating an example of an extraction process flow of the instrument region. First, as pre-processing, in order to clarify a contour line of a meter or the like, contour line enhancement using a second derivative (Laplacian) is performed on the input image (step S51). FIG. 5A is an example of an image after outline enhancement. Thereafter, smoothing is performed using the Median Filter so that numbers unnecessary for specifying the instrument position do not affect the binarization (step S52), and these effects are suppressed. FIG. 5B is an example of an image after smoothing.

  Then, by binarizing the preprocessed image, a binarized image as shown in FIG. 5C is obtained (step S53), and an instrument is extracted from the binarized image. Therefore, the maximum area portion is selected from the image (weighted maximum area selection) (step S54). Here, when calculating the area of the region, the closer to the center of the image, the greater the weight. This is because it is assumed that the instrument is photographed in the center of the image, and is cut out even when the area of the instrument area is smaller than the area of the surrounding area. Thereby, it is considered that the display panel area can be selected even when the size of the instrument is small to some extent. FIG. 6A shows an example of an image in which the maximum area is selected.

  The circularity is then used to determine whether this selected area is an instrument area. Here, in order to correctly measure the area of the instrument area necessary for calculating the circularity, a hole filling process is performed in which black pixels included in the selected area are previously filled with white pixels, and further, the circularity is calculated. In order to correctly measure the perimeter of the necessary instrument region, smoothing (Median Filter) is performed to reduce the unevenness of the perimeter (step S55). FIG. 6B is an example of an image in which the black pixels are filled with white pixels. FIG. 6C is an example of the smoothed image.

  Next, the area and circularity of the white pixel are measured (step S56), and if the white pixel is less than 75% of the whole and the circularity is equal to or greater than a predetermined threshold (for example, 0.5). Then, it is determined that the selected area is a circle, and the area is set as a virtual display board area of the instrument. For example, the virtual display panel area (circular area) is cut out from the area expanded by 30 pixels in the upper, lower, left, and right directions (step S57). The instrument region described above with reference to (C). If the circularity is less than the threshold, the process returns to step S53. If the circularity is calculated when the white image region extends over the entire image, the circularity may be equal to or greater than the threshold value, so the area of the white pixels measured in step S56 is 75% or more of the entire image. In case of, roundness is not calculated.

(2) Projection Conversion The projection conversion process in step S3 in FIG. 2 will be described. The projection conversion unit 13 performs Hough conversion on the image of the instrument region shown in FIG. 3C to extract a straight line including the outer frame side of the display panel of the round instrument 2. For example, the projective transformation unit 13 extracts four straight lines in the instrument region image shown in FIG.

  Here, referring to the four straight lines PP ′, QQ ′, RR ′, and SS ′ shown in FIG. 7A, the straight line PP ′ corresponds to the side of the back surface of the outer frame of the display panel. Therefore, as shown in FIG. 7B, the projective transformation unit 13 displays the four straight lines on the display board area side so that the four straight lines touch the outline of the display board area in the image of the instrument area. Move to. The four straight lines are moved so as to be in contact with the contour line of the display panel region, and the projection conversion as described below is performed, whereby the contour line of the display panel region included in the instrument region after the projection conversion is performed. Distortion can be corrected more effectively.

  Next, the projective transformation unit 13 uses the four vertices (point A, point B, point C, point D) of the quadrangle formed by the four straight lines in FIG. Projective transformation is performed on the image of the instrument region in FIG. More specifically, if the four vertices of the square after projective transformation (not shown) are point a, point b, point c, and point d, point A, point B, point C, and point D are point a, point Projective transformation is performed on the image of the instrument region so as to correspond to b, point c, and point d, respectively, thereby obtaining a square instrument region image.

  By the projective transformation process, an image of an instrument region having a square outline as shown in FIG. 7C is obtained. Further, the distortion of the contour line of the display panel area included in the instrument area detected in the extraction process of the instrument area was corrected by projective transformation, and the distortion as shown in FIG. 7C was corrected. A circular display panel region 100 is obtained.

(3) Determination of Reference Line The reference line determination process in step S4 of FIG. 2 will be described. The reference line determination unit 15 uses, for example, a straight line drawn vertically downward from the center O of the circular display board region 100 whose distortion is corrected as shown in FIG.

(4) Display Panel Area Division The display panel area division process in step S5 of FIG. 2 will be described. The display panel area includes scales, numbers, and pointers. Dividing these existing areas leads to higher processing efficiency and higher accuracy. Therefore, if the display board area is denoted by R, the area dividing unit 14 divides the display board area R into a scale area R _S , a numeric area R _N , and a center / pointer area R _C (R = R _S ∪R). _N ∪R _C ). The scale area is an area where a scale exists. The number area is an area where numbers indicating scale values are written, and is an area inside the scale area. The center / pointer area is an area where the center of the display panel and the pointer exist.

  For example, in the scale area extraction, the area dividing unit 14 performs a smoothing process, an edge extraction process, and a binarization process on the image of the display board area whose distortion has been corrected by the above-described projective transformation process, thereby performing line components. Are extracted, and the most likely circle is extracted as a circle by Hough transform. The area of the outer periphery of the extracted circle is a scale area. FIG. 8A shows an example of the scale area extracted.

Next, the area R _N ∪R _C inside the circle extracted in the scale area extraction process is divided into a numerical area and a center / pointer area. Specifically, first, an area having a predetermined radius r from the center of the circle is set as an area at the center of the display panel. Further, a threshold value process is performed on the luminance value of the pixel for the region R _N ∪R _C , and the region connected to the central region of the display panel among the regions where the luminance value of the pixel is not more than the threshold value Extract as a region. The area obtained by combining the center area of the display panel and the pointer area is the center / pointer area.

FIG. 8C shows an example of the center / pointer area. In the figure, a straight line OX drawn in the center / pointer region is the centerline of the pointer. Of the region R _N ∪R _C, the region other than the center / pointer region is a numeric region. FIG. 8B shows an example of a numeric area.

(5) Determination of the position of the pointer The processing for determining the position of the pointer in step S6 in FIG. 2 will be described. In this process, the angle formed by the pointer and the reference line is determined based on the center / pointer area. For example, a white circular portion in the center / pointer region image as shown in FIG. 9A is painted in black. For example, in a circle having a radius that is half the circle radius of the number area, the white circular portion in the center / pointer area image shown in FIG. An image of the pointer as shown in is obtained. Then, for each pixel of the pointer image, the angle with the reference line is measured, the number of pixels is voted for the angle obtained by the measurement, and a histogram as shown in FIG. 9C is obtained. The angle θ corresponding to the highest peak portion of the histogram is determined as the angle formed by the pointer and the reference line.

(6) Number Recognition and Number Position Determination The number recognition process in step S7 in FIG. 2 and the number position determination process in step S8 will be described.

  FIG. 10 is a flowchart for explaining the number recognition process and the number position determination process. First, the number recognition unit 17 cuts out a number from the number area (step S11). The extraction of numbers is performed by threshold processing. A rectangle containing the object cut out by the threshold processing is cut out as a number rectangle. FIG. 11A shows an example of a number rectangle to be cut out. FIG. 11A shows a number rectangle for the number “5” and a number rectangle for the number “0” surrounded by a solid line.

  Here, in cutting out numbers, it is desirable that each rectangle includes only one number. However, there are cases where numbers adjacent to each other come into contact with each other due to the influence of lighting, the influence of shadows, the low resolution of the camera, and the like, and a plurality of numbers are included in the same rectangle (for example, FIG. )reference). In such a case, it may affect the recognition of numbers, making subsequent understanding and reading impossible. Therefore, according to one embodiment of the present invention, the numbers that are touched are separated and then the numbers are recognized. As a condition for performing the separation process, a rectangle whose width is equal to or greater than a certain value and whose horizontal length is longer than its vertical length is selected. When the distance from the upper frame of the rectangle to the number candidate included is represented on the histogram for each column, and there is an extreme value of the histogram near the center, as shown in FIG. Separation to the left and right is performed starting from the column corresponding to. FIG. 12C and FIG. 12D each show a separated number.

  Next, the number recognition unit 17 recognizes the cut out number (Step S12). That is, the cut-out rectangular number rectangle is subjected to a number recognition process by template matching, for example. Here, since objects to be cut out include symbols such as units and numbers other than characters, numbers 0 to 9 are identified, and at the same time, other than numbers are rejected.

  Here, an example of the number recognition process will be described. Since the extracted object may include symbols other than units and numbers such as letters, numbers 0 to 9 are identified, and at the same time, other than numbers are rejected. In general, symbols and the like have a smaller rectangular size than numbers. In order to reject them, the average of the numbers is used to obtain an adjusted average (trim average), and those smaller than the average are rejected as non-numbers.

  In the number recognition, for example, the recognition target is not recognized as it is, but is classified according to the number of holes (holes inside the numbers), and then subjected to recognition processing in each category. Here, the hole is defined as a hole existing inside the numeral. For example, the number “6” has one hole because the number of holes in the number is one. Similarly, the number “8” has two holes because the number of holes in the number is two.

  Major classification means that numbers are divided in advance by the number of holes that each number has. Some numbers are prone to misrecognition because of their similar structure in recognition. For example, numbers such as “6” and “8” are likely to cause misrecognition. Therefore, the numbers are categorized in advance according to the number of holes. For example, “6” is divided into category 1 and “8” is divided into category 2. Each number is categorized as follows: That is, 1, 2, 3, 5, and 7 are classified into category 0, 0, 4, 6, and 9 are classified into category 1, and 8 is classified into category 2. After categorizing by the number of holes, it is possible to perform robust recognition against character deterioration, positional deviation, etc., "Shifting similarity method" (number recognition based on shifting similarity taking into account template displacement and inclination) Method). Each recognition target is rejected when the similarity is equal to or less than a threshold value. In addition, the number recognition process mentioned above is an example, In this invention, arbitrary number recognition processes can be used.

The number recognizing unit 17 integrates adjacent ones of the number rectangles containing the numbers recognized as numbers, and recognizes them as a plurality of digits (step S13). For example, the number rectangles are integrated when the number rectangles for the two recognized numbers satisfy the following three conditions.
Condition 1: The difference in the vertical size of the number rectangle is a certain value or less.
Condition 2: The horizontal interval of the number rectangle is a certain value or less.
Condition 3: The vertical displacement of the number rectangle is less than a certain value.

  As a result of the process of step S13, for example, in FIG. 11A, the number rectangle for the number “5” surrounded by the solid line and the number rectangle for the number “0” are shown in the portion surrounded by the dotted line. Integrated. As a result of the number recognition process for the integrated number rectangle, the number “50” as shown in FIG. 11B is recognized. When recognizing the numbers for the integrated number rectangle, for example, the position is multiplied by 10 sequentially from the left number (for example, “5”), and the next number (for example, “0”) of the number is added. Add.

  Next, the number position determination / accumulation unit 18 determines the position of the recognized number (step S14).

The process of step S14 is demonstrated using FIG. A circle shown in FIG. 13 is a boundary line between the number area and the scale area, and a point O is a point indicating the center of the display board area. A straight line OA ₀ is a reference line extending vertically downward from O. Each of V _{1 to} V ₄ is a recognized single digit or a plurality of integrated digits written in the numeric area, and corresponds to each digit in the figure. A number rectangle is shown. In the figure, a straight line OA _i (1 ≦ i ≦ 4) is a straight line passing through the center of gravity of each number V _i .

Digit position determining and storing unit 18, for each digit rectangular, clockwise direction with respect to the reference line angle a _i, determined as the position of each number V _i, as shown in FIG. 14 (A), V _i and a the correspondence information between _i, accumulates as positional information of each number V _i.

For example, the angle a _{1 with} respect to the reference line of the number rectangle for the number V ₁ is an angle formed by the straight line OA ₁ and the reference line OA ₀ in FIG. Similarly, the angle a ₂ digit rectangular with respect to the reference line for numeral V ₂ is the angle formed by the straight line OA ₂ and the reference line OA _0, the angle a ₃ for the digits rectangular reference line for numeral V ₃ linearly OA ₃ and the reference line OA ₀ and is an angle, the angle a ₄ for the digits rectangular reference line for numeral V ₄ is the angle between the straight line OA ₄ and the reference line OA _0.

  Next, the numeral position correcting unit 19 detects an error in the numeral position (step S15), and deletes the position information of the erroneous numeral. In general, there are no guarantees that all numbers will be correctly recognized because the numbers on the instrument may be affected by lighting or the pointer may overlap the number. Here, in general, the scale numbers have a monotonically increasing property. Here, by using this property, a number that is missing a part of a plurality of digits or a number that may be erroneously recognized is detected and rejected. Specifically, a recognition error is detected from a magnitude relationship with an adjacent number in a number string obtained by rearranging the number position information stored in the number position determining / accumulating unit 18 in the order of angles.

For example, the numeric position correction unit 19 reads the numeric position information stored in the numeric position determination / accumulation unit 18 and rearranges the numbers V _i in ascending order with respect to the angle a _i . For example, if a ₁ <a ₂ <a ₃ <a ₄ , for example, data after rearrangement of numbers is created as shown in FIG. Here, for example, when V ₃ <V ₂ , V ₃ is an error and the position information about the number V ₃ is deleted. FIG. 14C shows an example of the position information of each number after the position information of the wrong number is deleted.

The numeral position correcting unit 19 reads out the position information of the numbers stored in the number position determining / accumulating unit 18, and in the number string rearranged in descending order with respect to the angle a _i , an error is caused due to the magnitude relationship with the adjacent numbers. May be detected.

  According to one embodiment of the present invention, the numerical position correction unit 19 may perform error detection processing using a regression line as described below. In the above-described error detection processing of the numerical position, which focuses on the property that the numerical value of the scale increases monotonously, there are cases where an error cannot be detected. Hereinafter, a method for detecting an error in the position of one number when there are four or more relationships between numbers and angles will be described.

  For example, the second information such as the case where each number accumulated in the position information number accumulation / accumulation unit 18 is 0, 5, 100, 150 or 0, 30, 6000, 9000. When the lower digit of the number is not read (for example, when “50” is erroneously recognized as “5” and “3000” as “30”), the above-mentioned scale numbers increase monotonously. The error cannot be detected depending on the error detection method for the position of the number focusing on.

  Therefore, for example, the numeral position correcting unit 19 determines whether or not the relationship between the numeral and the angle includes an error using a regression line. Here, on the assumption that the scale is a uniform scale, it is assumed that a set of four or more numbers and angles is stored in the numerical position determining / accumulating unit 18. FIGS. 15A and 15B are graphs showing the relationship between numbers and angles when there is no error and when there is an error, respectively. FIG. 16A is a graph showing four points when there is no error and its regression line, and FIG. 16B is a graph showing four points when there is an error and its regression line. Comparing FIG. 16 (A) and FIG. 16 (B), it can be seen that there is a large difference in the square error between each point and the regression line when there is no error and when there is no error. An error is detected using the difference in square error. Hereinafter, the error detection process will be described more specifically.

In order to detect errors using a regression line, four or more pairs of numbers and angles are required. In the case of three sets, it is possible to determine whether or not there is an error, but it is not possible to detect which set is incorrect. Here, an algorithm for automatically detecting one error from four sets will be described.
(Procedure 1): The sum of the square error with the regression line is calculated for the remaining three points except for one point in the four sets of numbers and angles (herein referred to as points). For example, by removing one point from the graph as shown in FIG. 16B and using the graph shown in FIG. 17A showing the regression line determined by the remaining three points, for each of the three points on the graph, A square error with the regression line is obtained, and a total of the obtained square errors is calculated.
(Procedure 2): In the case where one other point is removed from the graph shown in FIG. 16B, the sum of the square error with the regression line is calculated in the same manner. For example, for each graph from FIG. 17 (B) to FIG. 17 (D), the sum of the square error with each regression line is calculated for three points on each graph.
(Procedure 3): The total value of each square error calculated by the above procedure 1 and procedure 2 is compared, and one point that is excluded from the graph when the total value is relatively small is regarded as an error. To detect. When there is no error, there is no big difference in the total value of each square error, and the total value of the square error is small. On the other hand, when an erroneous point is included, the total square error is only one smaller. It can be said that the point that is removed when the sum of square errors is only one is the cause of increasing the square error, and is an erroneous point.

Finally, the numeral position correcting unit 19 corrects the deviation of the numeral position by associating each numeral with the scale (step S16). In the embodiment of the present invention, for example, the angle formed by the straight line OA _i drawn from the center O shown in FIG. 13 with respect to the center of gravity of each numeral and the reference line OA ₀ is determined as the position of each numeral. An error may occur between the position of each numeral and the position of the scale corresponding to the numerical value indicated by each numeral. For example, as shown in the image of the display board area shown in FIG. 18, the tip of the arrow 102 drawn from the center O of the display board area of the round instrument 2 and passing through the center of gravity of the numeral corresponds to the numerical value indicated by the numeral. In some cases, the position of the scale 103 surrounded by a thick dotted line is shifted (an error occurs).

Therefore, the numeral position correcting unit 19 changes the angle of the straight line drawn from the center O in the direction of the center of gravity of the numeral V _i within a predetermined range (for example, a range of ± 5 °), and cuts out the scale 103 existing in the range. . As a result of cutting out the scale 103, the scale 103 surrounded by the thick dotted line is cut out.

Then, the numeral position correcting unit 19 changes the angle a _i corresponding to the numeral V _i to an angle formed by a straight line drawn with respect to the scale 103 cut out from the center O and the reference line.

(7) Calculation of Instruction Value The instruction value calculation process in step S9 of FIG. 2 will be described.

  The instruction value calculation unit 20 stores the position information of the pointer determined by the pointer position determination unit 16, that is, information on the angle formed by the center line of the pointer and the reference line, and the numerical position determination / accumulation unit 18. Based on the position information of each numeral, the indicated value indicated by the pointer of the round meter 2 is calculated.

  The indicated value is obtained as follows. As the angle formed by the center line of the pointer and the reference line increases, the indicated value also increases. Further, when the scale is equally spaced, the angle and the indicated value are proportional. From this relationship, the indication value V (θ) when the angle formed by the center line of the pointer and the reference line is the angle θ is calculated by the following equation.

θ ₁ and θ ₂ respectively represent an angle less than θ closest to θ and an angle equal to or larger than θ closest to θ among the angle information stored in the numerical position determining / accumulating unit 18.

  For example, it is assumed that numerical position information as shown in FIG. 19 is stored in the numerical position determining / accumulating unit 18. Here, when the angle θ of the pointer is, for example, 91 °, the instruction value is obtained as follows.

From the numerical position information shown in FIG. 19, the angles θ ₁ = 65 ° and θ ₂ = 141 °. Also, the number V (65) = 0 and the number V (141) = 50.

Therefore, the indication value V (91) is in accordance with Equation 1 above.
V (91) = 0 + (91-65) (50-0) / (141-65) = 17.1. . .
The indicated value “17.1...” Is calculated.

(Second Embodiment)
FIG. 20 is a diagram illustrating an example of the configuration of the second exemplary embodiment of the present invention. The image processing device 4 is a processing device that processes an image of the round instrument 2 taken by the camera 3 and calculates an instruction value indicated by a pointer of the round instrument 2.

  The image processing device 4 includes an image input unit 11, an instrument region extraction unit 12, a projection conversion unit 13, a region division unit 14, a reference line determination unit 15, a pointer position determination unit 16, a number recognition unit 17, a number position determination / accumulation unit 18, a numeric position correction unit 19, a scale extraction unit 21, a scale position determination / accumulation unit 22, a reading unit determination unit 23, and an instruction value calculation unit 24. Among the constituent elements included in the image processing apparatus 4, the same reference numerals as those of the constituent elements included in the image processing apparatus 1 shown in FIG. 1 described above are the same as the constituent elements included in the image processing apparatus 1. Omitted.

  The scale extracting unit 21 extracts a scale from the scale area obtained by dividing the display board area by the area dividing unit 14. The scale position determination / accumulation unit 22 is configured to display the display panel area in which the distortion is corrected based on the scale extraction result by the scale extraction unit 21 and the position information of each number stored in the number position determination / accumulation unit 18. An angle formed by a straight line drawn from the center with respect to each scale and a reference line and a numerical value corresponding to each scale are determined, and correspondence information between the angle and the numerical value is stored as position information of each scale.

  The reading unit determination unit 23 determines the reading unit of the pointer instruction value based on the position information of each scale accumulated in the scale position determination / accumulation unit 22.

  The instruction value calculation unit 24 is determined by the reading unit determination unit 23 based on the position information of the pointer determined by the pointer position determination unit 16 and the position information of each scale accumulated in the scale position determination / accumulation unit 22. The indicator value indicated by the pointer is calculated using the reading unit of the pointer.

  The functions of the above-described image processing apparatus 4 and each unit thereof are realized by the CPU and a program executed thereon. The program for realizing the present invention can be stored in a computer-readable recording medium such as a semiconductor memory, a hard disk, a CD-ROM, a DVD, or the like, provided by being recorded in these recording media, or communication. It is provided by transmission / reception using a network via an interface.

  FIG. 21 is a diagram illustrating an instruction value calculation processing flow according to the second embodiment of this invention. First, the image input unit 11 inputs an image (original image) of the round meter 2 taken by the camera 3 (step S21). Next, the instrument area extraction unit 12 extracts an instrument area from the input original image of the round instrument 2 (step S22).

  The projective transformation unit 13 performs projective transformation on the extracted instrument region, and corrects distortion of the display panel region included in the instrument region (step S23). Then, the reference line determination unit 15 determines a straight line drawn vertically downward from the center of the display board area whose distortion has been corrected as a reference line (step S24). The area dividing unit 14 divides the display board area in which the distortion is corrected into a scale area, a numerical area, and a center / pointer area (step S25).

  The pointer position determination unit 16 determines the angle formed by the center line of the pointer of the round instrument 2 included in the center / pointer area and the reference line as the position of the pointer (step S26).

  Next, the number recognition unit 17 recognizes by cutting out a number from the number area (step S27). The numerical position determining / accumulating unit 18 determines an angle formed by a straight line drawn from the center of the display panel area where the distortion is corrected and the center of gravity of each recognized numeral and a reference line as the position of each numeral. The position information of the determined numbers is accumulated (step S28).

  Next, the scale extraction unit 21 extracts a scale from the scale area obtained in step S25 (step S29). Then, the scale position determination / accumulation unit 22 determines each distortion from the center of the display panel area in which the distortion is corrected based on the scale extraction result in step S29 and the position information of each number accumulated in step S28. The angle formed by the straight line drawn with respect to the scale and the reference line and the numerical value corresponding to each scale are determined, and the correspondence information between the angle and the numerical value is stored as scale position information (step S30).

  Next, the reading unit determination unit 23 determines the reading unit of the pointer instruction value based on the position information of each scale accumulated in the scale position determination / accumulation unit 22 in step S30 (step S31).

  The instruction value calculation unit 24 uses the pointer reading unit determined in step S31 based on the position information of the pointer determined in step S26 and the position information of each scale accumulated in step S30. Then, the instruction value indicated by the pointer is calculated (step S32).

  The processes from step S21 to S28 described above are the same as the processes from step S1 to S8 in FIG.

(1) Scale extraction The scale extraction process in step S29 of FIG. 21 will be described. FIG. 22 is a flowchart showing details of the scale extraction process. First, the scale extraction unit 21 performs a labeling process on an image of a scale area (binary image) as shown in FIG. 23, and creates a histogram for each continuous area obtained by the labeling process (step S201). The horizontal axes of the histograms shown in FIGS. 24A and 24B indicate the angle formed by the reference line and the straight line drawn from the center of the display panel area to each pixel of the image of the scaled area that has been labeled. The axis indicates the number of black pixels (the same applies to FIGS. 25 to 27). The histogram is created by counting the number of black pixels for each angle for each label. Referring to the histograms shown in FIGS. 24A and 24B, spectra indicated by thick black lines appear at a plurality of angles.

  Specifically, FIG. 24A shows a histogram for the continuous region 1, and FIG. 24B shows a histogram for the continuous region 2. Histograms for other continuous regions are created in the same manner.

  It is desirable that the continuous area includes one scale line. However, a plurality of scale lines may be included in the same region due to the influence of the edge of the display panel, shadow, and the like. In order to remove such noise caused by shadows and divide the integrated scale line, noise removal is performed for each continuous region (step S202). Here, noise removal is performed using a predetermined threshold only when the angle range of the continuous region is equal to or greater than the predetermined range. For example, as shown in FIG. 24C, threshold processing related to the number of black pixels, such as erasing a portion where the number of black pixels is smaller than the threshold P in the histogram described above with reference to FIG. Remove. FIG. 24D shows the result of the noise removal.

  Next, the scale extraction unit 21 integrates the histograms (step S203), and obtains an integrated histogram as shown in FIG. 25A, for example. Then, the scale extraction unit 21 performs a maximum value selection process on the integrated histogram (step S204), and, for example, a histogram as illustrated in FIG. 25B obtained by the maximum value selection process is extracted as a scale line. Result B.

  Next, the scale extraction unit 21 performs noise removal using the integrated histogram (step S205). That is, in order to remove noise as a continuous region, for example, as shown in FIG. 25C, noise is removed from the integrated histogram using a predetermined threshold Q. FIG. 25D shows the result of the noise removal.

  Here, since each continuous area of the histogram extends over a plurality of angles, the angles cannot be uniquely specified. Therefore, the scale extraction unit 21 selects a maximum value from each continuous region (step S206), and sets an angle having the maximum value as a scale line candidate. For example, a maximum value selection process is performed on a portion surrounded by an ellipse in the histogram shown in FIG. 26A to obtain a histogram as shown in FIG. The obtained histogram is defined as a scale line extraction result A.

  Next, the scale extraction unit 21 measures the interval (angle) of each scale line candidate using the scale line extraction result A shown in FIG. 26 (B) described above, and as shown in FIG. 26 (C), The trim average is determined as the scale line interval (step S207).

  Next, the scale extraction unit 21 determines a scale line (step S208). For example, the scale extraction unit 21 uses the scale line extraction result B shown in FIG. 25B described above to determine the scale line based on the scale line interval determined in step S207. That is, in order to perform a search while leaving as many scale line candidates as possible, a scale line is determined using the scale line extraction result B that is not subjected to noise removal in the integrated histogram described above. For example, the scale extraction unit 21 shifts by the determined scale line interval (for example, d) using the scale line corresponding to the number whose position information is stored in the number position determination / accumulation unit 18 as a search start point. A predetermined range (for example, −d / 2 to + d / 2) from the determined position is used as a search range, and one scale line candidate in the search range is searched and determined as a scale line.

  When there are a plurality of scale line candidates in the search range, the scale line candidates close to the center position of the search range (for example, a position shifted by d from the position of the scale line candidate most recently determined as the scale line) Is selected, and the selected scale line candidate is determined as a scale line. Using the determined graduation line as a new search start point, further graduation line candidates are searched. FIG. 27A shows a search process for the above-described scale line candidates. FIG. 27B shows the determination result of the scale line. When no scale line candidate exists in the search range, the next scale line candidate is searched for using the position of the center of the search range as a temporary scale line.

  Based on the scale line determination result shown in FIG. 27B, the angle between the number of scales and the reference line and the straight line drawn from the center of the display panel area to each scale is determined. Can be recognized for each scale line.

(2) Determination of Scale Position The scale position determination process in step S30 of FIG. 21 will be described. The scale position determination / accumulation unit 22 extracts numeric position information from the numeric position determination / accumulation unit 18. The scale position determination / accumulation unit 22 includes a number associated with the smallest angle (the smallest number) and a number associated with the largest angle (the largest number) in the extracted number position information, A number corresponding to each scale is determined based on the number of scales appearing in the scale line determination result shown in FIG. For example, when the number associated with the smallest angle is 0, the number associated with the largest angle is 150, and the number of scales is 31, the scale position determination / accumulation unit 22 Divide the numerical difference (150-0) by the number of scale intervals (30) determined by the number of scales (31) to obtain the amount of change in the number between the scales, and based on the obtained amount of change in the number Determine the number corresponding to each scale. For example, 0, 5, 10,..., 145, 150 are determined as numbers corresponding to each scale.

  Then, the scale position determination / accumulation unit 22 calibrates correspondence information between the number corresponding to each determined scale and the angle corresponding to each scale recognized from the determination result of the scale line in step S208 of FIG. As position information. FIG. 28 shows an example of scale position information accumulated in the scale position determination / accumulation unit 22.

(3) Determination of the reading unit of the pointer instruction value The processing for determining the reading unit of the pointer instruction value in step S31 of FIG. 21 will be described. The reading unit determining unit 23 determines the amount of change in the number between one scale from the position information of each scale stored in the scale position determining / accumulating unit 22, and the amount of change in the number between the determined scales, Based on the reading unit calculation reference information stored in a predetermined buffer in the reading unit determination unit 23 in advance, the reading unit of the pointer instruction value is determined.

  FIG. 29 is a diagram illustrating an example of the reading unit calculation reference information. The reading unit calculation reference information includes data items such as classification and multiplication rate. The classification is a round instrument classification (instrument classification) in accordance with the amount of change in numbers between one scale. The multiplying rate is a multiplying rate by which the amount of change in numbers between one scale is multiplied.

  In the 1 system, the change amount of the number between 1 scale is 1, 10, or 100, and the result of normalizing the change amount of the number to a single digit (hereinafter, normalized result) is “1”. Instrument classification. Series 2 is an instrument classification in which the change amount of the number between 1 scale is 2, 20, or 200, and the normalized result of the change amount of the number is “2”. Series 5 is an instrument classification in which the normalized result of the numerical change amount is “5”, such as the numerical change amount between one scale being 0.5, 5, or 50.

  The reading unit determination unit 23 determines the amount of change in the number between one scale from the position information of each scale stored in the scale position determination / accumulation unit 22, and determines the amount of change in the number between the determined one scales. Normalize and get the normalization result. Then, a multiplication rate corresponding to the normalization result is specified in the reading unit calculation reference information, and a value calculated by multiplying the change amount of the number by the specified multiplication rate is determined as a reading unit of the instruction value. .

  For example, referring to the position information of the scale shown in FIG. The reading unit determination unit 23 refers to the reading unit calculation reference information shown in FIG. 29 and determines a multiplication rate “1/5” corresponding to “5”, which is a normalization result of the numerical change amount “5”. To do. Then, the reading unit determination unit 23 determines “1” calculated as a result of multiplying the change amount “5” of the number by the determined multiplication rate “1/5” as a reading unit.

  Similarly, for example, when the amount of change in the number between the graduations is “200”, the reading unit determination unit 23 refers to the reading unit calculation reference information illustrated in FIG. 29, and the amount of change in the number “200”. The multiplication rate “1/4” corresponding to “2” which is the normalization result of is determined. Then, the reading unit determination unit 23 determines “50” calculated as a result of multiplying the change amount “200” of the number by the determined multiplication rate “1/4” as a reading unit.

(4) Calculation of Pointer Instruction Value The processing for calculating the pointer instruction value in step S32 of FIG. 21 will be described. The instruction value calculation unit 24 uses the pointer position information determined in step S26 of FIG. 21 and the position information of each scale accumulated in step S30 as the pointer reading unit determined in step 31 of FIG. Use to calculate the indicated value indicated by the pointer.

  First, the instruction value calculation unit 24 performs a process similar to the calculation process of the pointer instruction value in the first embodiment described above to calculate a temporary instruction value of the pointer. A temporary instruction value V ′ (θ) when the angle formed by the center line of the pointer and the reference line is an angle θ is calculated by the following equation.

θ ₁ and θ ₂ respectively represent an angle less than θ closest to θ and an angle equal to or larger than θ closest to θ among the angle information stored in the scale position determination / accumulation unit 22.

For example, when the angle θ of the pointer is 91 °, the instruction value calculation unit 24 specifies the angles θ ₁ = 86 ° and θ ₂ = 93 ° with reference to the scale position information shown in FIG. The number V ′ (86) = 15 and the number V ′ (93) = 20.

Accordingly, the provisional instruction value V ′ (91) is obtained according to the above equation 2.
V ′ (91) = 15 + (91−86) (20−15) / (93−86) = 18.57. . . Thus, a temporary instruction value “18.57...” Is calculated.

  Next, the instruction value calculation unit 24 calculates the instruction value of the pointer based on the calculated temporary instruction value and the reading unit of the pointer. Specifically, the instruction value calculation unit 24 determines that the value up to the most significant digit of the reading unit in the temporary instruction value (here, for convenience of explanation, the value P) is larger than the value P. It is determined which one of the multiple of the reading unit closest to the value P and the multiple of the reading unit smaller than the value P and closest to the value P is closest.

  When the instruction value calculation unit 24 determines that the value P is larger than the value P and is close to a multiple of the reading unit closest to the value P, the instruction value calculation unit 24 By replacing the value P with a value that is a multiple of the reading unit that is greater than the value P and closest to the value P, a true indication value is calculated. When the instruction value calculation unit 24 determines that the value P is smaller than the value P and is close to a multiple of the reading unit closest to the value P, the instruction value calculation unit 24 The true indication value is calculated by replacing the value P with a multiple of the reading unit that is smaller than the value P and closest to the value P.

  For example, referring to the position information of the scale shown in FIG. 28, since the amount of change in the number between the scales is 5, the reading unit calculated by the reading unit determination unit 23 is “1” as described above. is there. As described above, the temporary instruction value when the pointer angle θ is 91 ° is 18.57. . . . And the value P up to the most significant digit (1 digit) of the value of the reading unit in the provisional indication value is “8.57. The multiple of the reading unit “1” that is larger than the value “8.57...” And closest to the value “8.57. The multiple of the reading unit “1” that is smaller than the value “8.57...” And closest to the value “8.57. The value “8.57...” Is a value closer to “9” than “8”.

  Therefore, the instruction value calculation unit 24 calculates the true instruction value “19” by replacing the value “8.57...” In the temporary instruction value “18.57. To do.

  Similarly, for example, when the provisional instruction value is 232.6 and the reading unit is 50, the value P is 32.6, which is larger than the value “32.6” and becomes the value “32.6”. The multiple of the reading unit “50” having the closest value is “50”. The multiple of the reading unit “50” that is smaller than the value “32.6” and closest to the value “32.6” is “0”. The value “32.6” is closer to “50” than “0”.

  Therefore, the instruction value calculation unit 24 calculates the true instruction value “250” by replacing the value “32.6” in the temporary instruction value “232.6” with the value “50”.

  Similarly, for example, when the provisional instruction value is 232.6 and the reading unit is 0.5, the value P is 0.6, which is larger than the value “0.6” and the value “0.6 The multiple of the reading unit “0.5” that is closest to “” is “1”. The multiple of the reading unit “0.5” that is smaller than the value “0.6” and closest to the value “0.6” is “0.5”. The value “0.6” is closer to “0.5” than “1.0”.

  Therefore, the instruction value calculation unit 24 replaces the value “0.6” in the temporary instruction value “232.6” with the value “0.5” to calculate the true instruction value “232.5”.

  According to the above-described true instruction value calculation processing, it is possible to automatically read the instruction value in a natural reading unit as if a human had read the instruction value by looking at the pointer.

It is a figure which shows an example of a structure of the 1st Embodiment of this invention. It is a figure which shows the instruction value calculation processing flow of the 1st Embodiment of this invention. It is a figure which shows the image of a round shape instrument's original image, a display panel area | region, and an instrument area | region. It is a figure which shows an example of the extraction process flow of an instrument area | region. It is a figure explaining the extraction process of an instrument area | region. It is a figure explaining the extraction process of an instrument area | region. It is a figure explaining a projective transformation process. It is a figure which shows the example of a scale area | region, a number area | region, and a center and a pointer area | region. It is a figure explaining the decision processing of the position of a pointer. It is a figure which shows the flowchart explaining the recognition process of a number, and the determination process of a position of a number. It is a figure which shows a number rectangle and the number recognized. It is a figure explaining the number recognition process after isolate | separating the number which contacted. It is a figure explaining the determination processing of the position of a number. It is a figure which shows the example of the positional information on the number accumulate | stored in a number position determination / accumulation | storage part. It is a graph which shows the relationship between a number and an angle. It is a graph which shows a regression line. It is a graph which shows a regression line. It is a figure explaining the correction process of the position shift of a number. It is a figure which shows the example of the positional information on a number. It is a figure which shows an example of a structure of the 2nd Embodiment of this invention. It is a figure which shows the instruction value calculation processing flow of the 2nd Embodiment of this invention. It is a flowchart which shows the detail of the scale extraction process. It is a figure explaining the extraction process of a scale. It is a figure explaining the extraction process of a scale. It is a figure explaining the extraction process of a scale. It is a figure explaining the extraction process of a scale. It is a figure explaining the extraction process of a scale. It is a figure which shows the example of the positional information on a scale. It is a figure which shows the example of reading unit calculation reference | standard information. It is a figure which shows the flow of a process of the general analog meter automatic reading system.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1, 4 Image processing apparatus 2 Round instrument 3 Camera 11 Image input part 12 Instrument area extraction part 13 Projection conversion part 14 Area division part 15 Reference line determination part 16 Pointer position determination part 17 Number recognition part 18 Number position determination / accumulation part 19 Number position correction unit 20, 24 Instruction value calculation unit 21 Scale extraction unit 22 Scale position determination / accumulation unit 23 Reading unit determination unit 100 Display panel area 102 Arrow 103 passing through the center of gravity of numbers 103 Scale

Claims (6)

An image processing apparatus for processing an image of a round instrument photographed by an image photographing means and calculating an instruction value indicated by a pointer of the round instrument,
Image input means for inputting an image of the round instrument,
From the input round instrument image, instrument area extracting means for extracting the area including the display panel area of the round instrument as an instrument area;
Projective transformation means for projectively transforming the extracted instrument area and correcting distortion of the display panel area included in the instrument area;
A reference line determining means for determining, as a reference line, a straight line drawn vertically downward from the center of the display panel area in which the distortion is corrected;
The display panel area in which the distortion has been corrected includes a scale area where a scale exists, a number area where a number indicating a numerical value of the scale is written, a center of the display board area, and a pointer. Area dividing means for dividing the area into the center and the pointer area,
Pointer position determining means for determining an angle formed by the center line of the circular instrument included in the center / pointer region and the reference line as the position of the pointer;
Number recognition means for cutting out and recognizing numbers from the number area;
An angle formed by a straight line drawn from the center of the display panel area in which the distortion is corrected with respect to the center of gravity of each recognized number and the reference line is determined as the position of each number, and each determined number Numeric position determination / accumulation means for accumulating position information of
An image processing apparatus comprising: instruction value calculation means for calculating an instruction value indicated by the pointer based on the determined position information of the pointer and the accumulated position information of each numeral. The image processing apparatus according to claim 1, further comprising:
Scale extraction means for extracting a scale from the scale area obtained by dividing the display panel area by the area dividing means;
Based on the scale extraction result by the scale extraction unit and the position information of each numeral stored in the numeral position determination / accumulation unit, each scale is measured from the center of the display panel area in which the distortion is corrected. A scale position determining / accumulating means for determining an angle formed by the drawn straight line and the reference line, and a numerical value corresponding to each scale, and storing correspondence information between the angle and the numerical value as position information of each scale; ,
Reading unit determination means for determining a reading unit of the indication value of the pointer based on position information of each scale accumulated in the scale position determination / accumulation means,
The indicator value calculation means is based on the pointer position information determined by the pointer position determination means and the position information of each scale accumulated in the scale position determination / accumulation means. An instruction value indicated by the pointer is calculated using the image processing apparatus. An image processing method for processing an image of a round instrument photographed by an image photographing means and calculating an instruction value indicated by a pointer of the round instrument,
An image input step for inputting an image of the round instrument;
From the input round instrument image, an instrument area extraction step for extracting an area including a display panel area of the round instrument as an instrument area;
A projective transformation step for projectively transforming the extracted instrument area to correct distortion of a display panel area included in the instrument area;
A reference line determining step for determining, as a reference line, a straight line drawn vertically downward from the center of the display panel area in which the distortion is corrected;
The display panel area in which the distortion has been corrected includes a scale area where a scale exists, a number area where a number indicating a numerical value of the scale is written, a center of the display board area, and a pointer. An area dividing step for dividing the area into a center area and a pointer area,
A pointer position determining step for determining an angle formed by a center line of the pointer of the round instrument included in the center / pointer region and the reference line as a position of the pointer;
A number recognition step of cutting out and recognizing numbers from the number region;
An angle formed by a straight line drawn from the center of the display panel area in which the distortion is corrected with respect to the center of gravity of each recognized number and the reference line is determined as the position of each number, and each determined number Storing the position information in the first storage means;
An instruction value calculation step of calculating an instruction value indicated by the pointer based on the determined position information of the pointer and the position information of each numeral stored in the first storage means. An image processing method. The image processing method according to claim 3, further comprising:
A scale extraction step for extracting a scale from the scale area obtained by dividing the display panel area;
Based on the scale extraction results and the position information of the numbers stored in the first storage means, straight lines drawn for the scales from the center of the display panel area where the distortion has been corrected, and Determining an angle formed by a reference line and a numerical value corresponding to each scale, and storing correspondence information between the angle and the numerical value in the second storage unit as position information of each scale;
A reading unit determining step for determining a reading unit of the indicated value of the pointer based on the position information of each scale accumulated in the second storage means;
The instruction value calculating step uses the determined pointer reading unit based on the determined position information of the pointer and the position information of each scale accumulated in the second storage means. An image processing method characterized by calculating an instruction value indicated by. An image processing program for processing an image of a round instrument photographed by an image photographing means and calculating an instruction value indicated by a pointer of the round instrument,
On the computer,
An image input process for inputting an image of the round instrument,
From the input round instrument image, an instrument area extraction process for extracting an area including a display panel area of the round instrument as an instrument area;
Projective transformation processing for projective transformation of the extracted instrument area and correcting distortion of the display panel area included in the instrument area;
A reference line determination process for determining, as a reference line, a straight line drawn vertically downward from the center of the display panel area in which the distortion is corrected;
The display panel area in which the distortion has been corrected includes a scale area where a scale exists, a number area where a number indicating a numerical value of the scale is written, a center of the display board area, and a pointer. Area division processing to divide the area into the center and the pointer area,
A pointer position determination process for determining an angle formed by the center line of the circular instrument included in the center / pointer area and the reference line as the position of the pointer;
A number recognition process for cutting out and recognizing numbers from the number area;
An angle formed by a straight line drawn from the center of the display panel area in which the distortion is corrected with respect to the center of gravity of each recognized number and the reference line is determined as the position of each number, and each determined number A process for storing the position information in the first storage means;
An instruction value calculation process for calculating an instruction value indicated by the pointer based on the determined position information of the pointer and the position information of each numeral stored in the first storage unit; A characteristic image processing program. The image processing program according to claim 5, further comprising:
On the computer,
A scale extraction process for extracting a scale from the scale area obtained by dividing the display panel area;
Based on the scale extraction results and the position information of the numbers stored in the first storage means, straight lines drawn for the scales from the center of the display panel area where the distortion has been corrected, and A process of determining an angle formed by a reference line and a numerical value corresponding to each scale, and storing correspondence information between the angle and the numerical value in second storage means as position information of each scale;
Based on the position information of each scale accumulated in the second storage means, a reading unit determination process for determining a reading unit of the indication value of the pointer is executed,
The indication value calculation process uses the determined pointer reading unit based on the determined pointer position information and the position information of each scale accumulated in the second storage means. An image processing program characterized by calculating an instruction value indicated by.

Images