JP5377560B2 - Image processing method - Google Patents

Abstract

Provided is an image processing method for examining the junction intervals of a plurality of flat-state corrugated fiberboard boxes that are bundled together. In a boundary detecting step, boundary positions each between a side part of a stack of corrugated fiberboard boxes and a background part both in a captured image are detected. In a concentration correcting step, a concentration correction is performed so as to correct the effect caused by a difference in condition of image capturing with respect to the side part of the stack of corrugated fiberboard boxes defined by the detected boundary positions. In a projection step, a projection is performed, with respect to the side part subjected to the concentration correction, in the direction perpendicular to the stack direction, thereby acquiring one-dimensional concentration data in the direction perpendicular to the stack direction. In a combination detecting step, a differential filter is used to filter the one-dimensional concentration data, and a combination between a trough part, which is a range in which the concentration is lower than a minimum side threshold, and a crest part, which is adjacent to the trough part and which is a range in which the concentration is higher than a maximum side threshold, is detected in the one-dimensional concentration data as filtered. Then, in a determining step, the centre position of the detected combination is acquired and this central position is determined as the central position of the junction interval sequence.

Description

  The present invention is an image processing method for specifying the position of a joining interval in order to measure the joining interval in a state where a plurality of flat corrugated cardboard boxes are stacked.

  In the box making process for producing a cardboard box from corrugated cardboard sheets, a sheet of a predetermined size including slots, ruled lines, and prints necessary for assembling as a cardboard box is folded along a fold line by a folder gluer, and is bonded to each other in advance. Adhesive glue is attached to the joints of the desired end portions, and the end portions are joined together.

  The joined corrugated cardboard sheet, that is, the flat corrugated cardboard box, is then sent to the square ring part (right angle straightening part), sent one box at a time from the gap provided at the lower rear end of this square ring part, and sent to the counter part. The count is performed here, and the bundle is shipped every predetermined number.

  When the end portions of the corrugated cardboard sheet are joined, if the joined end portions are non-parallel or the joining is insufficient, defects such as corners not being perpendicular to each other occur when assembled.

  This bonding failure can be found by inspecting the bonding interval in a flat cardboard box. The joining interval is an interval between two flaps connected to the side plates of the joined end portions. For example, when the joining interval is too wide or too narrow than a prescribed interval, the end portions are joined in parallel. You can see that it has not been done.

Inspection of the bonding interval is performed, for example, by visual inspection, but problems such as variations in inspection results occur.
In the inspection apparatus described in Patent Document 1, the flap interval can be inspected by image processing.

JP 2001-124528 A

  The inspection apparatus described in Patent Document 1 merely inspects the flap interval, and the flap interval is generally inspected in the state of the cardboard sheet before joining.

  In addition, as described above, since a predetermined number of cardboard boxes in a flat state after being joined are shipped in a bundle, the cardboard boxes in a bundled state before shipment are photographed and joined based on the photographed images. It is required to check the interval.

  By combining general-purpose image processing, for example, binarization labeling is performed on a density-corrected image, and the joining interval is detected, the corrugated plate interval is erroneously detected as the joining interval. There are problems such as erroneously detecting something other than the cardboard box that is reflected, and erroneously detecting a string that binds the flat cardboard boxes.

  An object of the present invention is to provide an image processing method capable of specifying with high accuracy the position of a joining interval included in an inspection image in order to inspect the joining interval of a plurality of flat corrugated cardboard boxes. That is.

The present invention provides a plurality of joining interval portions arranged in the stacking direction with respect to a photographed image obtained by photographing a side surface on the flap side of a corrugated cardboard stack in which a plurality of flat corrugated cardboard boxes are bundled with a binding string. An image processing method for specifying a position of a joining interval row comprising:
A boundary detection step for detecting a boundary position between a side surface portion and a background portion of the corrugated cardboard laminate in the photographed image;
A concentration correction step for performing a concentration correction process on the side surface portion of the corrugated cardboard stack specified by the detected boundary position; and
With respect to the side portion of the density correction process is performed, by performing the projection shadow processing in the stacking direction, and a projection step of obtaining a one-dimensional density data in the direction perpendicular to the stacking direction,
Performs differential processing on the 1-dimensional density data, the density in one dimension density data differential processing adjacent to the valley portion and one valley in the range to decrease the concentration one range to increase A combination detection step for detecting a combination with a mountain portion;
Obtains the center position of the detected combination, an image processing method characterized by the in heart position including a determining step of determining a center position of said joint spacing sequence.

  According to the present invention, in the combination detection step, an average value of the density in the one-dimensional density data is calculated, and a density value larger than the calculated average value in the one-dimensional density data is replaced with the average value. After that, the combination is detected.

In the present invention, after the combination detection step, before the determination step,
The method further includes a determination step of determining whether or not the side surface portion includes a binding string portion for binding a plurality of cardboard boxes based on the one-dimensional density data.

In addition, the present invention provides the following determination step:
A straight line parallel to the stacking direction including the determined center position is obtained, and a band-like region extending parallel to the straight line is set on both sides of the straight line, and the side surface portion is set with respect to the set belt-like region. The method further includes detecting a boundary position with the background portion.

According to the present invention, in the boundary detection step, the boundary position between the side surface portion and the background portion of the cardboard box stack in the photographed image is detected, and in the density correction step, the cardboard box stack specified by the detected boundary position. Density correction processing is performed on the side part of the body. The projection step, with respect to the side portion of the density correction process is performed, by performing the projection shadow processing in the stacking direction to obtain a one-dimensional density data in the direction perpendicular to the stacking direction.

In combination detection step, it performs a differential processing on the 1-dimensional density data, the density in one dimension density data differential processing adjacent to the valley portion and one valley in the range of reduction, the concentration is increased Upon detection of a combination of one of the peak portions in the range, at decision step obtains a center position of the detected combination, to determine the heart position as a central position of the joint spacing column therein.
Thereby, the position of the joining interval included in the inspection image can be specified with high accuracy.

According to the invention, in the combination detection step, an average value of the density in the one-dimensional density data is calculated, and a density value larger than the calculated average value in the one-dimensional density data is set as the average value. After the replacement, the combination is detected.
Thereby, when noise etc. mix, the influence can be made smaller.

  Further, according to the present invention, after the combination detection step and before the determination step, the side surface portion includes a binding string portion for bundling a plurality of cardboard boxes based on the one-dimensional density data. Determine whether or not.

  Thereby, when the binding string is included in the inspection image, the influence can be reduced, and erroneous detection of the joining interval row can be prevented.

Further, according to the present invention, a straight line parallel to the stacking direction including the determined center position is obtained, and on both sides of the straight line, a belt-like region extending in parallel to the straight line is set, and the set belt-like region is set. On the other hand, a boundary position between the side surface portion and the background portion is detected.
Thereby, the boundary position between the side surface portion and the background portion can be detected with higher accuracy.

1 is a perspective view showing a cardboard box laminate 10. It is a figure which shows an example of the image which image | photographed the side surface of the cardboard box laminated body. It is a flowchart which shows the image processing method of this invention. It is a flowchart which shows the procedure of cardboard edge detection (rough position). It is a flowchart which shows the procedure of a joining space | interval row | line | column detection. It is a flowchart which shows the procedure of cardboard edge detection (detailed position). It is a flowchart which shows the procedure of a string mask process.

  The present invention is an image processing method for specifying the position of a joining interval included in an image in order to inspect the joining interval of a plurality of flat corrugated cardboard boxes.

  In the image processing method of the present invention, the boundary position between the bundled flat cardboard boxes and the background is obtained in the photographed image, and density correction is performed on the inside of the boundary, that is, the image of the cardboard box from which the background is excluded. , Select a candidate region that will contain the splice interval. A portion corresponding to the binding string is removed from the selected candidates, and an area including the joining interval is determined.

FIG. 1 is a perspective view showing a cardboard box laminate 10.
In the present invention, when inspecting a cardboard box laminate 10 in which a plurality of flat cardboard boxes 1 are bundled, the side surface of the laminate is parallel to the direction perpendicular to the stacking direction and the depth direction of the cardboard box. Shoot from any direction.

FIG. 2 is a diagram illustrating an example of an image obtained by photographing the side surface of the cardboard box laminate 10.
The image obtained by photographing includes a plurality of joining intervals (hereinafter referred to as “joining interval sequence”) 2 arranged in the stacking direction, and an area including the joining interval sequence 2 is determined from the captured image.

  After determining the region including the bonding interval row 2, the bonding interval included in the region is measured, and a good or defective cardboard box is determined based on the measurement result.

The procedure of a specific image processing method after image capturing is as follows.
FIG. 3 is a flowchart showing the image processing method of the present invention.

The image processing method of the present invention is performed in the following procedure.
Step S1: Corrugated cardboard edge detection (approximate position)
Step S2: Density correction Step S3: Joining interval row detection Step S4: Cardboard edge detection (detailed position)
Step S5: String detection determination Step S6: String mask processing

<Step S1: Cardboard edge detection (approximate position)>
In step S1, in order to separate the cardboard box laminate portion and the background portion in the side-shot image of the cardboard box laminate 10, the boundary between the cardboard box laminate and the background (hereinafter also referred to as “cardboard end”) is used. To detect. First, in this step, the approximate position of the cardboard end is detected.

FIG. 4 is a flowchart showing a procedure of cardboard edge detection (approximate position).
As shown in FIG. 2, in the present embodiment, the image to be processed has the stacking direction of the cardboard box laminate 10 as the horizontal direction (X-axis direction), and the direction orthogonal to the vertical direction (Y-axis direction). And In such a photographed image, the boundary portion between the cardboard box laminate 10 and the background exists along the vertical direction. And since the arrangement direction of the joining space | interval row | line | column 2 is parallel to the lamination direction of the cardboard box laminated body 10, it exists along a horizontal direction.

  First, in step A1, vertical differentiation is performed over the entire image to be processed. As the differentiation process, a conventionally known process can be performed. For example, a filter process using a differentiation filter may be performed. As the differential filter, for example, a Sobel filter can be used.

  In step A2, vertical projection processing is performed on the differentiated data. The projection processing here is a projection addition of pixel data in the vertical direction. By performing projection processing in the vertical direction, one-dimensional pixel data arranged in the horizontal direction can be obtained.

  In step A3, smoothing processing is performed on the obtained one-dimensional pixel data. As the smoothing process, a conventionally known process can be performed. For example, an arithmetic average of pixel data included in a predetermined range is calculated, and the arithmetic average value is used as pixel data of representative pixels included in the range. The pixel data of the representative pixel obtained while shifting the predetermined range with respect to the one-dimensional pixel data is obtained as smoothed one-dimensional pixel data.

  The predetermined range for calculating the arithmetic average may be, for example, ± 10 pixels in the X-axis direction of the representative pixel.

In step A4, a threshold value is determined.
A density histogram in which pixel data (in this case, density value) is a class and the number of pixels is a frequency for 1/3 of each pixel on both ends (left and right sides) of the smoothed one-dimensional pixel data. And determine the threshold using discriminant analysis.

  In step A5, peak positions on the left and right sides are obtained in the smoothed one-dimensional pixel data. Since the left and right peak positions correspond to the boundary between the cardboard box image and the background image, they can be detected as cardboard edges on both sides.

<Step S2: Density Correction>
The density correction target image is subjected to density correction based on the two cardboard edges detected in step S1 with a central area sandwiched between the cardboard edges as a target. The density correction can be performed by a conventionally known process. For example, the density correction can be performed by calculating an average density for each pixel block having a predetermined size, or by fitting using a curve obtained by the method of least squares.

<Step S3: Joining interval row detection>
In step S3, a joining interval row is detected.

FIG. 5 is a flowchart showing a procedure for detecting a joining interval row.
In step B1, horizontal projection processing is performed on the pixel data whose density has been corrected. The projection processing here is a projection addition of pixel data in the horizontal direction. By performing projection processing in the horizontal direction, one-dimensional pixel data arranged in the vertical direction can be obtained. Further, an average value of the pixel data of the one-dimensional pixel data arranged in the vertical direction is calculated in advance.

  In step B2, smoothing processing is performed on the obtained one-dimensional pixel data. As the smoothing process, a conventionally known process can be performed. For example, an arithmetic average of pixel data included in a predetermined range is calculated, and the arithmetic average value is used as pixel data of representative pixels included in the range. The pixel data of the representative pixel obtained while shifting the predetermined range with respect to the one-dimensional pixel data is obtained as smoothed one-dimensional pixel data.

  The predetermined range for calculating the arithmetic average may be, for example, ± 10 pixels in the Y-axis direction of the representative pixel.

  In step B3, the maximum value and the minimum value of the density value are determined in the smoothed one-dimensional pixel data in the vertical direction.

  In step B4, high cut processing is performed once on the smoothed one-dimensional pixel data in the vertical direction. The high cut processing is processing for lowering the density value of one-dimensional pixel data having a specific high density value. For example, the high cut processing is performed using the average value calculated in step B1. Specifically, in the one-dimensional pixel data in the vertical direction, all density values larger than the average value are replaced with the average value. Accordingly, the one-dimensional pixel data in the vertical direction after the high cut is the average value obtained in step B1.

  In step B5, differentiation processing (for example, differentiation width = 40) is performed on the one-dimensional pixel data subjected to the high cut processing. The differentiation process can be a conventionally known process. For example, a filtering process using a differentiation filter may be performed.

  In step B6, a projection threshold value is calculated. There are two projection threshold values: a maximum value side threshold value and a minimum value side threshold value. The maximum value side threshold value is calculated by, for example, a value that is 40% of the maximum value calculated in step B3, that is, maximum value × 0.4. The minimum value side threshold value is calculated by, for example, a value that is 40% of the minimum value calculated in step B3, that is, the minimum value × 0.4.

  In step B7, threshold processing is performed on the differentiated one-dimensional pixel data, and pixel portions exceeding the maximum value side threshold value are defined as peak portions, and pixel portions smaller than the minimum value side threshold value are defined as valley portions.

  A combination (pair) of a valley portion and a mountain portion adjacent to the valley portion is searched. In addition, a case where the distance between the valley portion and the mountain portion adjacent to the valley portion exceeds 100 pixels is not regarded as a pair.

  In many cases, a pair of a valley portion and a crest portion adjacent to the valley portion indicates a junction interval sequence, and when a plurality of such pairs are found, a region including the pair is defined as a candidate region for the junction interval sequence. To do.

  In step B8, the binding string is checked. For example, the contraction process is performed on the candidate area to determine whether or not it is a binding string. If the number of white pixels in the region after contraction is equal to or greater than a predetermined number, the candidate region is determined to be a portion where a binding string has been detected, and is excluded from the candidate region.

  In step B9, the remaining candidate region is a region including the joining interval row, and the midpoint position of the candidate region is set as the Y coordinate of the center line passing through the center of the joining interval row. When there are a plurality of remaining candidate areas, for example, one candidate area closer to the center in the Y-axis direction is determined.

  In step B10, a pixel region having a predetermined number of pixels is set in the vertical direction with the Y coordinate of the center line as the center, and the set region is determined as a region including a joining interval row.

<Step S4: Cardboard Edge Detection (Detailed Position)>
In step S4, the detailed position of the cardboard end is detected. Based on the Y coordinate determined in step B9, a straight line parallel to the X-axis (hereinafter referred to as “column center line”) is obtained, and two strip regions are set up and down across the column center line. The band-like area is an area where the number of pixels in the vertical direction is the same, and the horizontal direction extends to both ends of the image. These four belt-like regions are arranged adjacent to each other without overlapping each other.

FIG. 6 is a flowchart showing a procedure of cardboard edge detection (detailed position).
First, in step C1, vertical differentiation is performed over the entire first belt-like region. As the differentiation process, a conventionally known process can be performed. For example, a filter process using a differentiation filter may be performed. As the differential filter, for example, a Sobel filter can be used.

  In step C2, vertical projection processing is performed on the differentiated data. The projection processing here is a projection addition of pixel data in the vertical direction. By performing projection processing in the vertical direction, one-dimensional pixel data arranged in the horizontal direction can be obtained.

In step C3, a threshold value is determined.
Creates a density histogram with the density value as the class and the number of pixels as the frequency for each one-third pixel on both ends (left and right) in the X-axis direction for the projected one-dimensional pixel data, and discriminant analysis The threshold is determined using the method.

  In step C4, the rising position of the density value is detected from the left and right ends toward the center of the projected one-dimensional pixel data. Specifically, threshold processing is performed on the density value of each pixel from the right end to the center of the one-dimensional pixel data, and the position of the pixel that matches the threshold is set as the rising position. The same applies when going from the left end to the center.

  In step C5, a position where the decrease in density value ends from the detected rising position toward the left and right ends is detected. Specifically, when the density value of each pixel continues to decrease from the right-side rising position of the one-dimensional pixel data toward the right end, the decrease continues, that is, the position where the density value does not change It is detected as the position of the cardboard end. The same applies to the case where the left side rises toward the left end.

  In step C6, it is determined whether or not the position of the corrugated cardboard edge has been detected for all the belt-like regions. If all the belt-like regions have been detected, the process proceeds to step C7. Returning, the area to be detected is changed to another belt-like area.

  In step C7, the positions of the cardboard edges detected in all the belt-like regions are compared, and the outermost one is determined as the cardboard edge position. Since the position of the cardboard end is detected in each of the four belt-like regions, four candidate positions for the left cardboard end and four candidate positions for the right cardboard end are detected.

  From the four candidate positions on the left cardboard end, the leftmost one is determined as the left cardboard end position, and from the four candidate positions on the right cardboard end, the outermost rightmost one is determined as the right Determine the position of the cardboard edge.

<Step S5: String detection determination>
In step S5, it is determined whether or not a binding string is detected in the handling image. If it is detected, the string mask process of step S6 is performed, and if not detected, the process is terminated.

  Whether or not a binding string has been detected is determined in step B3 of step S3 if a binding string has been detected in a part of the candidate area.

<Step S6: String mask processing>
In step S6, when a binding string is detected, a masking process is performed on the binding string portion in the target image.

FIG. 7 is a flowchart showing a string mask processing procedure.
In step D1, binarization processing is performed on the target image, and in step D2, the binarized image is compressed to ¼. In step D3, contraction processing and expansion processing are performed on the compressed image. For example, it contracts 5 times in the Y-axis direction, contracts 15 times in the X-axis direction, expands 10 times in the X and Y directions, and expands 50 times in the X direction.

  In step D4, a labeling process is performed, and in step D5, a bound string area is specified from the label. In step D6, the specified binding string area is masked.

1 Cardboard box 2 Joining interval row 10 Cardboard box laminate

Claims (4)

Bonding interval composed of a plurality of bonding interval portions arranged in the stacking direction with respect to a photographed image obtained by photographing the side surface on the flap side of a corrugated cardboard stack in which a plurality of flat cardboard boxes are bundled with a binding string An image processing method for specifying a column position,
A boundary detection step for detecting a boundary position between a side surface portion and a background portion of the corrugated cardboard laminate in the photographed image;
A concentration correction step for performing a concentration correction process on the side surface portion of the corrugated cardboard stack specified by the detected boundary position; and
With respect to the side portion of the density correction process is performed, by performing the projection shadow processing in the stacking direction, and a projection step of obtaining a one-dimensional density data in the direction perpendicular to the stacking direction,
Performs differential processing on the 1-dimensional density data, the density in one dimension density data differential processing adjacent to the valley portion and one valley in the range to decrease the concentration one range to increase A combination detection step for detecting a combination with a mountain portion;
Obtains the center position of the detected combination, an image processing method characterized by the in heart position including a determining step of determining a center position of said joint spacing sequence.   In the combination detection step, an average value of the density in the one-dimensional density data is calculated, and in the one-dimensional density data, a density value larger than the calculated average value is replaced with the average value, and then the combination is determined. The image processing method according to claim 1, wherein detection is performed. After the combination detection step and before the determination step,
3. The method according to claim 1, further comprising a determination step of determining whether or not the side surface portion includes a binding string portion that bundles a plurality of cardboard boxes based on the one-dimensional density data. Image processing method. After the determining step,
A straight line parallel to the stacking direction including the determined center position is obtained, and a band-like region extending parallel to the straight line is set on both sides of the straight line, and the side surface portion is set with respect to the set belt-like region. The image processing method according to claim 1, further comprising a step of detecting a boundary position with the background portion.