JP3653063B2 - Agricultural product quality judgment device and agricultural product quality judgment method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an agricultural product quality determination apparatus and an agricultural product quality determination method, in which an agricultural product mask image is created from an acquired agricultural product image, and the quality of the agricultural product is determined based on the mask image.
[0002]
[Prior art]
Various automatic quality judgment devices that automatically judge and grade the quality of agricultural products have been developed. In the automatic quality determination device, first, an agricultural product is imaged, and the quality of the agricultural product is determined from the shape of the agricultural product based on the acquired image.
In general, in the quality determination apparatus, each crop is conveyed by a conveyor or the like while being placed on a tray, and is imaged by a CCD camera or the like. The reason why the crops are placed on the tray in this way is to prevent the crops from being damaged or to efficiently handle the crops in the subsequent work (for example, boxing work).
[0003]
In the following, a conventional quality judgment device will be described by taking coconut as an agricultural product.
FIGS. 14A and 14B are schematic views showing the configuration of a general tray 2 used in the insulator quality judgment apparatus. As shown in FIG. 14A, the tray 2 includes two frames 2a and a plurality of pins 2b attached so as to be spanned between the frames 2a and 2a. The insulator 1 is imaged by a CCD camera or the like when the quality is judged by the quality judging device. At this time, the insulator 1 can be imaged from various directions (particularly from below) so that the tray 1 can be imaged. 2 is configured such that the insulator 1 is placed on the pins 2b arranged at intervals.
[0004]
In addition, as the quality of agricultural products, for example, there is a thing related to the shape, but in order to determine such a shape, it is important to extract only the insulator from the captured image. As such an extraction method, there is a method in which image processing for binarizing each pixel of a captured image with a specific gradation as a boundary is performed on an image acquired by a CCD camera (hereinafter referred to as a camera image). is there.
[0005]
In this method, a color close to the insulator 1 is generally used as the color of the pin 2b. This is because, for example, if the pin 2b has a color significantly different from that of the insulator 1, the binarization process is performed on the camera image acquired from the lower side (from the side with the pin 2b). This is because the obtained image is an image in which the insulator is divided by the pin 2b.
[0006]
When a color close to the insulator is used for the color of the pin 2b as in the camera image 20 of FIG. 6, an image 21 obtained by binarizing the camera image 20 has an insulator as shown in FIG. The image looks like a fusion of pins. Therefore, this time, the image 21 is subjected to the contraction process a predetermined number of times and then subjected to the expansion process a predetermined number of times to obtain an image 22 from which the pins 2b are removed as shown in FIG. .
[0007]
The contraction process and the expansion process will be described with reference to FIGS. In FIGS. 15A to 15C, each cell represents a pixel.
The contraction process is a process for thinning an image composed of “1” pixels (pixels that are turned on and displayed in white in the drawing) by one pixel. For example, if a contraction process is performed on an image as shown in FIG. 15A, an image as shown in FIG. 15B is obtained. In contrast to the contraction process, the expansion process is a process that thickens an image composed of “1” pixels by one pixel. For example, if an expansion process is performed on an image as shown in FIG. 15B, an image as shown in FIG. 15C is obtained. If the shrinkage process and the expansion process are combined in this way, only the fine dots 41a are removed from the image 41 as can be seen by comparing the image 41 in FIG. 15A and the image 42 in FIG. 15C. It becomes possible.
[0008]
Therefore, by combining such a contraction process and an expansion process, the thin line indicating the pin can be removed from the image 100 shown in FIG. 8 as described above.
[0009]
[Problems to be solved by the invention]
However, even if the shrinkage process and the expansion process are combined as in the prior art described above, the pin cannot be completely removed at the place where the insulator and the pin overlap as shown in FIG. 8 (this place is actually Is displayed in white as a part of the mask image, but is hatched for convenience in FIG. 8). For this reason, the thickness (measurement value) of the insulator measured based on the image of FIG. 8 causes an error corresponding to the true thickness of the insulator. If the thickness of the insulator cannot be detected accurately, for example, an appropriate number of insulators cannot be boxed in the subsequent boxing.
[0010]
Further, when the camera image 20 shown in FIG. 6 is binarized according to the gradation of the pixel, if there is a light-colored portion in the insulator (here, there is a light-colored portion), this portion is As shown in FIG. 8, it is erroneously determined that it is not an insulator (the insulator is missing) (the pixel is set to “0” (turned off) and displayed black in FIG. 8).
The present invention was devised in view of such a problem, and is capable of accurately detecting the shape of a crop and determining the quality of the crop based on the detected shape of the crop with high accuracy. An object is to provide an apparatus and a quality determination method.
[0011]
[Means for Solving the Problems]
For this reason, the agricultural product quality determination apparatus according to the present invention (claim 1) includes an image acquisition unit that acquires an image of an agricultural product, and a mask image generation unit that generates a mask image of the crop from the image acquired by the image acquisition unit. And a mask image correcting means for correcting the mask image created by the mask image creating means, and a quality judging means for judging the quality of the produce based on the mask image corrected by the mask image correcting means. A contour line extracting unit configured to extract a contour line of an image representing the crop from the mask image created by the mask image creating unit; and A feature point detecting means for detecting a feature point that is bent at a larger angle than a predetermined angle toward a predetermined direction; and the angle is inverted before and after the feature point in the contour line Skirt point detecting means for detecting the skirt points of the feature points in pairs, and the contour by interpolating the two skirt points located outside the skirt points related to the feature points adjacent to each other. It is characterized by comprising contour line correcting means for correcting the line.
[0012]
In this case, when the straight line distance between the two skirt points to be interpolated is shorter than a predetermined distance, the contour correcting means performs interpolation by connecting the two skirt points with a straight line. If the tangent line at the two skirt points of the contour line is parallel to each other, the two skirt points are interpolated by connecting them with a straight line. If the tangents are not parallel to each other, it is preferable that the two skirt points are interpolated by a spline function made of a polynomial.
[0013]
When the crop is transported by the crop transport means in a state of being placed on a tray comprising a pair of frame members and a pin-shaped member spanned between the pair of frame members, the mask image creating means The binarization processing unit that binarizes the image acquired by the image acquisition unit according to the pixel value, and the contraction process is performed on the image binarized by the binarization processing unit After that, it is preferable to have a contraction / expansion processing means for performing expansion processing.
[0014]
The agricultural product quality determination method of the present invention (claim 4) acquires an image of an agricultural product, creates a mask image of the crop from the acquired image, and determines the quality of the agricultural product based on the mask image. In the quality determination method for agricultural products, a first step of extracting an outline of an image representing the agricultural product from the mask image, and a feature in which the outline is bent at a larger angle than a predetermined angle toward a predetermined direction. A second step of detecting a point, a third step of detecting two pairs of points whose angles are inverted before and after the feature point on the contour line as a skirt point of the feature point, and features adjacent to each other It is characterized by comprising a fourth step of correcting the contour line by interpolating two skirt points located outside of the skirt points related to the points.
[0015]
In this case, in the fourth step, when the linear distance between the two skirt points to be interpolated is shorter than a predetermined distance, the two skirt points are connected by a straight line, and the interpolation is performed. When the straight line distance between the two skirt points is equal to or greater than a predetermined distance, if the tangent lines at the two skirt points of the contour line are parallel to each other, the two skirt points are connected by a straight line and interpolated. If the tangents are not parallel to each other, it is preferable to interpolate the above two skirt points with a spline function consisting of a polynomial.
[0016]
When the crop is transported in a state of being placed on a tray composed of a pair of frame members and a pin-shaped member spanned between the pair of frame members, and acquiring an image of the crop being transported, The first step includes a step of binarizing the image of the crop according to a pixel value, and a step of performing an expansion process after performing a contraction process on the binarized image. It is preferable that
[0017]
DETAILED DESCRIPTION OF THE INVENTION
Embodiments of the present invention will be described below with reference to the drawings.
FIGS. 1-13 is a figure shown about the quality determination apparatus and the quality determination method of agricultural products as one Embodiment of this invention, FIG. 1 is a schematic diagram which shows the structure of a quality determination apparatus, FIG. FIG. 3B and FIG. 4A to FIG. 4C are schematic diagrams for explaining the quality judgment method, and FIG. 5 shows the quality judgment method. FIG. 6 to FIG. 13 are diagrams for explaining image processing according to the quality judgment method (FIGS. 6 to 8 are diagrams used for explaining the conventional quality judgment method. And combined use). In addition, about the part demonstrated as a prior art, the same code | symbol is attached | subjected and the description is abbreviate | omitted.
[0018]
An agricultural product quality determination apparatus 10 according to the present embodiment determines the quality of an eggplant as an agricultural product, and is configured as shown in FIG. 1, placed on a tray 2 and transported by a conveyor (a crop transport means) 3. The insulator 1 is imaged by a CCD camera (or TV camera) 11 as an image acquisition means, and the quality of the insulator 1 is determined based on the image (camera image) of the insulator 1 acquired by the CCD camera 11. It is. The insulators 1 whose quality has been determined are sorted according to their quality on the downstream side (not shown).
[0019]
In FIG. 1, one CCD camera 11 is simplified and shown above the conveyor 3, but a plurality of CCD cameras 11 are provided so that the insulator 1 on the conveyor 3 can be imaged from various angles. . Moreover, in the range imaged by the CCD camera 11, the insulator 1 on the conveyor 3 is irradiated with light from a light source (not shown).
The quality judgment device 10 includes the CCD camera 11, a mask image creation unit (mask image creation unit) 12, a mask image correction unit (mask image correction unit) 13, and a quality judgment unit (quality judgment unit) 14. Yes. Based on the camera image acquired by the CCD camera 11, a mask image obtained by extracting the shape of the crop (here, coconut) 1 is created by the mask image creating unit 12, and this mask image is corrected by the mask image correcting unit 13. It is like that. Based on the corrected mask image, the quality determining unit 14 determines the quality of the insulator 1. The quality of the insulator 1 as used herein refers generally to those used for grading, and includes dimensions such as thickness as well as those relating to the shape such as the degree of warpage.
[0020]
The mask image creation unit 12 includes a binarization processing unit (binarization processing unit) 12a and a contraction / expansion processing unit (contraction / expansion processing unit) 12b. For each pixel of the camera image, the binarization processing unit 12a sets “1” (luminance to a predetermined value) if the pixel value (darkness) is larger than a predetermined threshold, while the pixel value is predetermined. If the threshold value is equal to or less than the threshold value, binarization processing is performed to set “0” (luminance is zero). When the binarization processing is performed on the camera image 20 as shown in FIG. A binarized image 21 as shown in FIG. Since the color of the pin 2b of the tray 2 is set to a color close to the insulator 1 (here, black), in this binarized image 21, not only the insulator 1 but also the pin 2b is an image made up of “1” pixels. It will be displayed as (white image).
[0021]
The contraction / expansion processing unit 12b performs the contraction process on the binarized image 21 a predetermined number of times, and then performs the expansion process a predetermined number of times. As a result, an image showing the pin 2b (FIG. 7). A white line running in the horizontal direction in FIG. 8 is removed, and a mask image 22 as shown in FIG. 8 is used, in which a pixel forming an insulator is set to “1” and a pixel forming a background (an area other than the insulator) is set to “0”. Will be obtained.
[0022]
However, in the binarization process, if there is a light-colored portion (or white scratches or large wormholes made by a machine) in the insulator 1 as in this example, this portion is converted into the binarized image 21 and the mask image. 22 is set to “0”. In addition, the number of pixels forming the lump among the lump consisting of “1” pixels or the number of pixels having the width dimension if the lump is linear are removed by the contraction / expansion processing. The number is less than a predetermined number determined by the number of times of contraction processing (for example, a linear image having a width of 4 pixels or less can be removed by two contraction processings). For this reason, in the mask image 22, the pin 2 b is not completely removed at the portion where the insulator 1 and the pin 2 b begin to overlap (the portion shown by hatching in FIG. 8).
Therefore, the mask image 22 is corrected by the mask image correcting unit 13.
[0023]
As shown in FIG. 1, the mask image correction unit 13 includes a contour line extraction unit (contour line extraction unit) 13a, a feature point detection unit (feature point detection unit) 13b, a skirt point detection unit (base point detection unit) 13c, and An outline correction unit (outline correction means) 13d is provided.
[0024]
The contour line extraction unit 13a generates a chain code from the mask image 22 created by the mask image creation unit 11, and based on the chain code, the mask image ("1" pixels of the mask image 22 are assembled. The boundary pixel image 23 shown in FIG. 9 is created by extracting the outline of the image) (= boundary pixel: pixel adjacent to the pixel “0”).
[0025]
A chain code is used to represent a curve in a binarized image. The chain cord will be briefly described with reference to FIGS. 2 (a) and 2 (b). The squares in FIGS. 2A and 2B indicate pixels. In FIG. 2B, for convenience, the pixel “1” is indicated by “◯” and the pixel “0” is indicated by a blank.
[0026]
In the chain code, as indicated by eight arrows in FIG. 2A, code numbers 0 to 7 are assigned to each of the adjacent pixels 31 to 38 with respect to a predetermined pixel 30. On the other hand, the pixel 32 is expressed as 1 in the chain code. If the closed curve (contour line) 50 shown in FIG. 2B is expressed by a chain code counterclockwise starting from the pixel 51, “1,0, 1, 2, 2, 2, 2, 3, 4, 4, 4, 5, 6, 6, 5, 6, 7, 0, 7 ".
[0027]
The feature point detection unit 13 b extracts feature points in the outline of the boundary pixel image 23. A feature point is a pixel that bends at a larger angle than a predetermined angle in a predetermined direction (here, toward the inside of a contour line that is a closed loop).
The feature point detection unit 13b will be described with reference to FIG. In FIG. 3, for convenience, “1” pixels (pixels forming the contour line 50) are indicated by “☆”, “◯”, “●”, “★”, and “0” pixels are indicated by blanks. The feature point detection unit 13b sequentially determines the pixels forming the contour line 50, starting from a predetermined pixel as a starting point (in this case, counterclockwise). The feature points are detected from the contour line 50. Hereinafter, the direction in which feature points are detected is referred to as detection direction A.
[0028]
For example, when determining whether or not the pixel 53 indicated by “●” is a feature point, this determination is made by extending the arms L1 and L2 before and after the pixel 53 and changing the directions of the arms L1 and L2. This is done based on the quantity θ. The arms L1 and L2 are pixels that form the outline 50 from the pixel 53 to be determined, and are separated from the pixel 53 by a predetermined number (here, three, the predetermined number is also referred to as the arm length). The line segments connecting the pixels 52 and 54 respectively.
[0029]
In the contour line 50, the angle toward the inside of the contour line 50 (here, the angle tilted to the left with respect to the detection direction A because the detection direction A is counterclockwise) is negative, and the angle toward the outside of the contour line 50 (here, the detection direction) Assuming that A is counterclockwise and the angle tilted to the right with respect to the detection direction A) is positive, the feature point detection unit 13b has the angle θ negative and a predetermined angle θ. ₀ If the angle is larger than that, the pixel 53 is determined as a feature point.
[0030]
The length of the arms L1 and L2 and a predetermined angle θ ₀ Is appropriately set according to the type of contour detection target (agricultural crops). For example, when the crop is an insulator, the arm lengths L1 and L2 are set to 3, and a predetermined angle θ. ₀ It is known that the contour line can be extracted with high accuracy if each is set to 50 degrees.
When a feature point is detected by the feature point detector 13b in the image 23 shown in FIG. 9, an image 24 as shown in FIG. 10 is obtained. In the image 24, the feature points are indicated by “◯”.
[0031]
The skirt point detection unit 13c detects skirt points located before and after the feature point detected by the feature point detection unit 13b, and the skirt point is an angle of the contour line from positive to negative, or from negative to positive. This is a point (pixel) that is inverted to. Then, the contour line correction unit 13c further corrects the contour line by interpolating outer skirt points among the skirt points of the adjacent feature points.
[0032]
The function of the contour correction unit 13c will be further described with reference to FIGS. 4 (a), 4 (b), and 4 (c). 4 (a) to 4 (c), for convenience, the “1” pixel (the pixel forming the outline 60) is shown in black, and the “0” pixel is shown in white.
FIG. 4A shows a case where a part of the outline 24A is greatly recessed inward. In this case, as shown in the figure, the pixel a that bends greatly to the left with respect to the detection direction A. ₀ , B ₀ Are detected as feature points, and a feature point a ₀ Pixel a as the foot of ₁ , A ₂ Is detected and the feature point b ₀ Pixel b as the base of ₁ , B ₂ Is detected. These feature points a ₀ , B ₀ And skirt point a ₁ , A ₂ , B ₁ , B ₂ Are integrated to generate data indicating pixels to be complemented. That is, four skirt points a ₁ , A ₂ , B ₁ , B ₂ Outer hem point a ₁ , B ₂ Is the feature point a ₀ , B ₀ Are set as pixels to be interpolated. The outer skirt point among the four skirt points refers to the skirt point that is the most separated combination at the distance Y along the contour line.
[0033]
And these skirt points a ₁ , B ₂ Is _, The contour 60 is interpolated by the contour correcting unit 13c, and the contour 60 is corrected so that the line segment 61 indicated by the two-dot chain line in FIG. 4A formed by this interpolation forms part of the contour 60. Become.
FIG. 4B shows a case where a convex portion is formed on a part of the contour line 62. In this case, as shown in FIG. ₁ , C ₂ Is the feature point a ₀ , B ₀ , And these hem points c ₁ , C ₂ Is interpolated by a line segment as indicated by a two-point difference line by the outline correcting unit 13c, and the outline 62 is corrected.
[0034]
FIG. 4C shows a case where the convex portion of the contour line is slightly recessed at the center with respect to FIG. ₀ , B ₀ And its skirt point a ₁ , A ₂ , B ₁ , B ₂ Is detected and the foot point a ₁ , B ₂ Is complemented by the contour correcting unit 13c as indicated by a two-point difference line, and the contour 63 is corrected.
The interpolation of the two skirt points by the contour correcting unit 13c is such that the linear distance D between the skirt points to be complemented is a predetermined value D. ₀ Less than (D <D ₀ ), Linear interpolation is performed. On the other hand, the distance D is the predetermined value D. ₀ If this is the case (D ≧ D ₀ ), Tangent line M of contour lines 60, 62, 63 at the skirt points to be complemented ₁ , M ₂ Are parallel, they are complemented by straight lines, and the tangent line M ₁ , M ₂ If they are not parallel, they are interpolated by a spline function consisting of a polynomial such as a cubic spline equation. Here, the tangent refers to the tangent of the contour line outside the base point to be interpolated.
[0035]
When the outline correction unit 13c corrects the outline of the image 24 shown in FIG. 10, first, an image 25 as shown in FIG. 11 is obtained. In the image 25, the skirt point “Δ” is shown. Next, by interpolating the skirt points, an image 26 in which the contour line is corrected as shown in FIG. 12 is obtained.
Then, as a final process, the mask image correcting unit 13 performs a hole filling process (a process of setting all pixels in the closed region surrounded by “1” pixels to “1”) to the image 26, and then performs FIG. A mask image 27 as shown is created.
[0036]
The agricultural product quality determination apparatus according to the embodiment of the present invention is configured as described above, and the quality determination is performed by the method shown in the flowchart of FIG. 5 (the agricultural product quality determination method according to the embodiment of the present invention). Is done
That is, first, in step A10, an image 20 (see FIG. 6) of the insulator 1 conveyed by the conveyor 3 while being placed on the tray 2 is acquired by imaging with the CCD camera 1. In step A20, the camera image 20 is binarized and contracted / expanded by the mask image creation unit 12, and as a result, a mask image 22 as shown in FIG. 8 is obtained. As shown in the camera image 20, since there is a lightly colored portion (portion where the pixel value is equal to or less than the threshold value), the mask image 22 has a pixel of “0” (not a lever). In addition, a part of the pin 2b of the tray 2 is left as a refracted portion bent at a substantially right angle at the outer edge of the mask image 22.
[0037]
In step A30, a chain code is generated from the mask image 23 by the contour extraction unit 13a, and an image 23 as shown in FIG. 9 obtained by extracting the contour of the mask image of the mask image 23 is obtained. Then, a feature point is detected by the feature point detection unit 13b in step A40 with respect to the contour line, and an image 24 as shown in FIG. Further, in step A50, the skirt point corresponding to the feature point is detected by the skirt point detection unit 13c, and an image 25 as shown in FIG. 10 in which the skirt point is indicated by “Δ” is obtained.
[0038]
Then, in step A60, the feature point detected in step A40 and the skirt point detected in step A50 are integrated by the contour correcting means 13d. That is, the combination of the two skirt points outside the four skirt points corresponding to the feature points adjacent to each other is generated as a set of data between the feature points. In step A70, the contour correcting means 13d further interpolates the two skirt points forming the data set to correct the contour, and an image 26 as shown in FIG. 12 is obtained.
[0039]
By interpolating the skirts of feature points adjacent to each other in this way, a large bent portion (a portion bent at an angle larger than a predetermined angle with respect to the tangential direction of the contour) that does not exist as the contour of the insulator is obtained. The convex portions and concave portions that are included are corrected so that the concave portions are removed and the concave portions are blocked. As a result, in the mask image 23, correction is performed on a portion that is erroneously determined to be a dent because of its light color, and a part of the remaining pin 2b, and the contour of the insulator 1 is accurately defined as shown in FIG. A reproduced image can be obtained.
[0040]
This image is finally subjected to hole filling processing (processing to set all the pixels in the closed region surrounded by “1” pixels to “1”), and as a result, a mask image 27 corrected as shown in FIG. In step A80, the quality determining unit 14 determines the quality of the insulator 1 based on the mask image. Since the mask image 27 used for quality determination accurately reproduces the actual contour of the insulator 1 by the above correction, the quality of the insulator 1 can be determined with high accuracy.
[0041]
In addition, the quality determination apparatus and quality determination method of the agricultural product of this invention are not limited to embodiment mentioned above, It can change variously in the range which does not deviate from the meaning of this invention.
For example, in the above-described embodiment, an example in which the image acquisition unit that acquires an image of a crop is configured by a CCD camera (imaging unit) has been described. However, an image captured by an external device is recorded on a recording medium such as a floppy (registered trademark) disk. After the image is once recorded, the image may be read from the recording medium by a recording / reading device such as a floppy (registered trademark) disk drive. In this case, an image acquisition unit is configured with the recording / reading apparatus.
In the above embodiment, the feature point is a point that bends at a predetermined angle or more at a negative angle (that is, toward the inside of the contour line). A point that turns at a predetermined angle or more (outward) may be a feature point.
[0042]
Moreover, although the said embodiment demonstrated the example which handled the eggplant as agricultural products, this invention is applicable also to another agricultural product.
[0043]
【The invention's effect】
As described above in detail, according to the present invention, the adjacent features described above are interpolated by using, for example, a straight line or a spline function, between the skirts of the feature points adjacent to each other. A large bent portion of the contour line including the point is corrected. As a result, an extremely large bent portion that does not exist in the actual crop produced in the mask image due to misrecognition is corrected. There is an advantage that the shape of the crop can be accurately detected, and the quality of the crop can be accurately determined based on the detected shape of the crop.
[0044]
In particular, the crops are placed on a tray and transported, and the images of the crops being transported are binarized according to pixel values to create a binarized image, and the binarized image is further shrunk. After processing, when expanding the mask image to create a mask image, if there is a light-colored part in the crop, this part will be misrecognized as not being a farm product, or the pin-shaped member of the tray will be accurately removed from the mask image. Although it may not be possible, there is a possibility that an uneven portion that does not actually exist on the outer edge of the mask image may be generated.
[Brief description of the drawings]
FIG. 1 is a schematic diagram showing the configuration of an agricultural product quality determination apparatus according to an embodiment of the present invention.
FIGS. 2A and 2B are schematic diagrams for explaining contour line extraction using a chain code. FIGS.
FIG. 3 is a schematic diagram for explaining a method for judging the quality of agricultural products as one embodiment of the present invention.
FIGS. 4A to 4C are schematic diagrams for explaining a method for determining the quality of agricultural products according to an embodiment of the present invention.
FIG. 5 is a schematic flowchart for explaining a method for judging the quality of agricultural products as one embodiment of the present invention.
FIG. 6 is a schematic diagram for explaining image processing according to an agricultural product quality determination method and a conventional agricultural product quality determination method according to an embodiment of the present invention;
FIG. 7 is a schematic diagram for explaining image processing according to an agricultural product quality determination method and a conventional agricultural product quality determination method according to an embodiment of the present invention;
FIG. 8 is a schematic diagram for explaining image processing according to an agricultural product quality determination method and a conventional agricultural product quality determination method according to an embodiment of the present invention;
FIG. 9 is a schematic diagram for explaining image processing according to the agricultural product quality determination method of the embodiment of the present invention.
FIG. 10 is a schematic diagram for explaining image processing according to the agricultural product quality determination method of the embodiment of the present invention.
FIG. 11 is a schematic diagram for explaining image processing according to the agricultural product quality determination method of the embodiment of the present invention.
FIG. 12 is a schematic diagram for explaining image processing according to the agricultural product quality determination method of the embodiment of the present invention.
FIG. 13 is a schematic diagram for explaining image processing according to the agricultural product quality determination method of the embodiment of the present invention.
14A and 14B are diagrams showing a configuration of a general tray used for transporting agricultural products, in which FIG. 14A is a schematic plan view, and FIG. 14B is an xx cross-sectional view of FIG.
FIGS. 15A to 15C are schematic diagrams for explaining image processing according to a conventional method for judging quality of agricultural products. FIGS.
[Explanation of symbols]
1 Isogo (agricultural crops)
2 trays
2a frame
2b pin
3 conveyor (crop transportation means)
10 Quality judgment device
11 CCD camera (image acquisition means)
12 Mask image creation unit (mask image creation means)
12a Binarization processing unit (binarization processing means)
12b Shrinkage expansion processing unit (shrinkage expansion processing means)
13 Mask image correcting unit (mask image correcting means)
13a Contour line extraction unit (contour line extraction means)
13b Feature point detector (feature point detector)
13c Bottom point detection unit (base point detection means)
13d contour correction unit (contour correction means)
14 Quality judgment part (quality judgment means)
20 Camera images
21 Binarized image
22, 27 Mask image
23-26 border pixel image
50, 60 contour lines
30-37, 51-54 pixels
M ₁ , M ₂ Tangent
θ Angle at a predetermined position of the contour line
a ₀ , B ₀ Feature point
a ₁ , B ₁ , C _1, c ₂ Hem point

Claims (6)

Image acquisition means for acquiring images of agricultural products;
Mask image creating means for creating a mask image of the crop from the image obtained by the image obtaining means;
Mask image correcting means for correcting the mask image created by the mask image creating means;
Comprising quality judging means for judging the quality of the produce based on the mask image corrected by the mask image correcting means,
The mask image correcting means is
Contour extraction means for extracting an outline of an image representing the agricultural product from the mask image created by the mask image creation means;
A feature point detecting means for detecting a feature point that bends at a larger angle than a predetermined angle toward the predetermined direction in the contour line;
In the contour line, skirt point detection means for detecting two points in which the angle is inverted before and after the feature point as a skirt point of the feature point,
It is characterized by comprising contour correction means for correcting the contour line by interpolating two skirt points located outside among the skirt points related to feature points adjacent to each other. , Agricultural product quality judgment device. The contour line correcting means is
When the linear distance between the two skirt points to be interpolated is shorter than a predetermined distance, the two skirt points are connected by a straight line and interpolated,
If the straight line distance between the two skirt points to be interpolated is greater than or equal to a predetermined distance, the two skirt points are connected by a straight line if the tangent lines at the two skirt points of the outline are parallel to each other. 2. The agricultural product quality judging apparatus according to claim 1, wherein the two skirt points are interpolated by a spline function comprising a polynomial if the tangent lines are not parallel to each other. . The crop is transported by the crop transport means in a state of being placed on a tray composed of a pair of frame members and a pin-shaped member spanned over the pair of frame members.
The mask image creating means
Binarization processing means for binarizing the image acquired by the image acquisition means according to pixel values;
2. The image processing apparatus according to claim 1, further comprising a contraction / expansion processing unit configured to perform an expansion process after the contraction process is performed on the image binarized by the binarization processing unit. Or the quality determination apparatus of agricultural products of 2. In the quality determination method for agricultural products, an image of an agricultural product is acquired, a mask image of the crop is created from the acquired image, and the quality of the agricultural product is determined based on the mask image.
A first step of extracting an outline of an image representing the crop from the mask image;
A second step of detecting a feature point that bends at a larger angle than a predetermined angle toward a predetermined direction of the contour line in the contour line;
A third step of detecting, as a set of foot points of the feature points in pairs, points whose angles are reversed before and after the feature points in the contour line;
A fourth step of correcting the contour line by interpolating two skirt points located outside among the skirt points related to feature points adjacent to each other, A method for judging the quality of agricultural products. In the fourth step,
When the linear distance between the two skirt points to be interpolated is shorter than a predetermined distance, the two skirt points are connected by a straight line and interpolated,
If the straight line distance between the two skirt points to be interpolated is greater than or equal to a predetermined distance, the two skirt points are connected by a straight line if the tangent lines at the two skirt points of the outline are parallel to each other. 5. The method for judging the quality of agricultural products according to claim 4, wherein if the tangent lines are not parallel to each other, the two skirt points are interpolated by a spline function comprising a polynomial. The crop is transported in a state of being placed on a tray composed of a pair of frame members and a pin-shaped member spanned between the pair of frame members, and an image of the crop being transported is acquired.
The first step comprises
Binarizing the image of the crop according to the pixel value;
6. The method for judging the quality of agricultural products according to claim 4 or 5, further comprising a step of performing an expansion process after performing a contraction process on the binarized image.

Images