JPH09218024A - Method for inspecting surface unevenness of vegetable and fruit - Google Patents

Abstract

PROBLEM TO BE SOLVED: To determine the grade of vegetables and fruits automatically by comparing the surface unevenness of a fruit being conveyed, detected from the gray image data at two points, with a preset threshold value. SOLUTION: The image of a material (vegetables and fruits) 10 being conveyed on a conveyor belt 2 at a constant speed is picked up at two points within a predetermined pickup period (e.g. 66ms) by means of a TV camera 3 and two gray images are inputted to an image processor 8. Two gray images picked up at two points are then subjected to A/D conversion and one image data is set with a match window. The gay level is interpolated for each window and a maximum position of correlation value is determined through cross correlation operation with other image. Subsequently, the distribution of height on the surface of a vegetable or fruit, i.e., the surface unevenness thereof, is determined from the parallax which is determined from the maximum position of correlation value for each match window. The surface unevenness is compared with a preset threshold value thus determining the grade of vegetables and fruits.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a fruit or vegetable surface inspection method for inspecting the surface of fruits or vegetables to determine the grade.

[0002]

2. Description of the Related Art A conventional fruit and vegetable grade discriminating apparatus based on image processing using a television camera discriminates a grade based on the contour shape of a discrimination material as a whole, such as bending, constriction, buttocks, etc. Inspection is performed visually on the partial uneven shape of the discriminating material such as open, skin vein, groove and the like.

[0003]

Since the conventional visual inspection for the partial unevenness of the discriminating material is an inspection of fine local defects on the surface of the material, it is more difficult for a human to continuously perform it for a long time because it causes more eyestrain. It is difficult and causes many inspection omissions. In addition, when a visual inspection is performed by a plurality of inspectors, the class is determined by the judgment of each inspector according to the degree of unevenness, so that there is a problem that the determination standard is not constant. On the other hand, there has been no surface inspection method that enables complete automation instead of a visual inspection on the partial surface shape of the discrimination material.

In view of the above situation, the present invention can perform the partial surface irregularity shape of the grade discrimination material without lowering the processing capacity,
An object of the present invention is to provide a method of classifying fruits and vegetables which can be completely automated in the classification of fruits and vegetables.

[0005]

[Means for Solving the Problems] After using a grayscale image at two points of fruits and vegetables moving on a conveyor such as a belt conveyor, a matching window is set in one of the images, and density interpolation processing is performed for each matching window. , The cross-correlation calculation with the other image is performed, the maximum position of the correlation value is obtained, and the difference in surface height of the fruits and vegetables obtained from the maximum position of the correlation value of each matching window is compared with the unevenness of the surface of the fruits and vegetables. The grade is discriminated by a preset threshold value.

[0006]

[Function] 2 of fruits and vegetables obtained by one TV camera
The parallax amount is obtained for each matching window by performing the cross-correlation calculation for each matching window between the images at the points, and the maximum position of the correlation value is obtained, and the unevenness degree that is the distribution of the fruit and vegetable surface height is calculated from the parallax amount. Understand and be able to distinguish grades.

[0007]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a plan perspective view of a vegetable and fruit surface unevenness inspection apparatus according to the present invention. In FIG. 1, 1 is a vegetable such as sweet potato, potato, carrot, or fruit such as apple, pear, peach, etc. A belt conveyor that conveys a material 10 (showing sweet potatoes in the figure) that serves as a grade determination factor, 2 is a conveyor belt,
3 is a television camera that images the material 10 on the conveyor belt 2 from above, 4 and 5 are television cameras that image the material 10 from the left and right sides, and 6 is an image captured by the television cameras 4 and 5. A background plate to prevent extraneous objects other than 10 from being reflected, 7 is a lighting device that illuminates the upper surface and both side surfaces of the material 10, and 8 is a TV camera 3, 4, 5
An image processing device for inspecting the surface irregularity shape of the material 10 and discriminating the grade by A / D converting the video signal from the device and subjecting the obtained digital image data to image processing, and 9 is the surface irregularity inspection device. It is a storage box for storing the devices constituting the.

FIG. 2 is a view of the surface unevenness inspection apparatus shown in FIG. 1 as viewed from the direction of arrow a. The same parts as those in FIG. 1 are designated by the same reference numerals. In the figure,
Reference numeral 7a is a fluorescent light which is turned on at a high frequency, and 7b is a light diffusion plate for uniformly illuminating the material 10.

The method for inspecting the surface irregularities of fruits and vegetables according to this embodiment having the above-mentioned structure will be described below. FIG. 3 is a processing flow of the fruit and vegetable surface unevenness inspection method of this embodiment, which will be described in the order of this processing flow.

First, the two-point image capturing will be described. FIG. 4 is an enlarged perspective view of a visual field portion (indicated by a broken line 3a in the figure) of the captured image by the television camera 3 of the surface unevenness inspection apparatus shown in FIG. While the material 10 is being conveyed at a constant speed by the belt 2, the image processing device 8 causes the analog video signal from the television camera 3 to be taken in at a constant period (for example, 66).
ms), 2 screens are taken in and sequentially A / D converted. When the image of the material 10 is captured by the camera 3 at two points A and B, an image of points A and B is shown, an image of the point A is shown in FIG. 5A, and an image of the point B is shown. 5 (b).

Next, the setting of the matching window will be described. The matching window is a divided area for associating the images of the two points, and as shown in FIG. Set a number of windows to divide the image into small pieces. In FIG. 5A, it is provided at the bottom of the image, but in the case of FIG. 5B, it is provided at the top of the image.

Next, the density interpolation in the matching window will be described with reference to FIG.
This is for associating the images at the two points with sub-pixel accuracy. The position of the image data in the matching window shown in FIG. 5A is represented by u and v coordinates, and the coordinate position is represented by (u, v), (u, v + 1), (u + 1, v),
(U + 1, v + 1), and the density values are f (u,
v), f (u, v + 1), f (u + 1, v), f (u +
1, v + 1) and the density value at the (u ₀ , v ₀ ) position after density interpolation is f (u ₀ , v ₀ ), the density value f (u ₀ ,
v ₀ ) is calculated by an equation. f (u ₀ , v ₀ ) = f (u, v) (1-α) (1-β)
+ F (u + 1, v) α (1-β) + f (u, v + 1)
(1-α) β + f (u + 1, v + 1) αβ ...

When the original image data is interpolated by dividing it into, for example, 10 parts, α and β are 0.1 and 0.1 as shown in FIG.
The density value at each point is obtained by changing the density value from 0.2 to 1. In the figure, positions at α = 0.4 and β = 0.3 are shown. FIG.
FIG. 8 shows an image when the density of the matching window 11 shown in FIG.

Next, the density interpolation image data in the matching window of the point A image shown in FIG.
The cross-correlation calculation for performing the matching scan with the density interpolation image data within a certain range (determined by the speed of the conveyor belt 2 and the image capturing period) of the point B image indicated by the broken line in (b) will be described.

As shown in FIG. 9, the density interpolated image data in the matching window of the image at the point A is g (x,
y), the density-interpolated image data within the scanning range at the point B is expressed as f
Given (x, y), the normalized cross-correlation R (x, y) is
Assuming that the center coordinate of the image data g (x, y) is (k, ι), it is calculated by the equation. ΣΣf (x, y) g (x−k, y−ι) Xy R (x, y) = ≦ 1 ΣΣf ² (x, y) ΣΣg ² (x−k, y−ι) xy Xy The distribution of the normalized cross-correlation values is shown in FIG. As a result, the skin with the highest correlation can be obtained with sub-pixel accuracy.

Next, from the position of the maximum value of the cross correlation,
A method for obtaining the height of the surface of the region divided by the matching window of the material 10 by using monocular stereoscopic vision based on the principle of triangulation will be described with reference to FIG. In the figure, when the camera position set to the height H from the conveyor belt 2 is set to O point, and the area on the surface of the material 10 moves from A point to B point after a certain image capturing time, the A point image Based on the angles α and β obtained from the maximum position of the correlation calculation between the image of B and the point B, and the actual moving distance ι obtained from the speed of the conveyor 2 and the image capture time, the height h of the area on the surface of the material 10 is calculated by You can ask more. h = H-ι / (tanα + tanβ) ……………………

If the processing for obtaining the heights of the areas divided by the matching window is performed for all matching windows of the portion where the material 10 is reflected, as shown in FIG. A surface height distribution close to the contour is obtained. Here, the height distribution closer to the actual contour can be obtained by setting the number of area divisions of the matching window more finely within a range that can be associated by the cross-correlation calculation. Next, in order to obtain the unevenness of the material surface from the height distribution, the height difference indicated by Δh in FIG. 12 may be obtained.

If the above-described processing is performed on two screens that are sequentially captured as the material 10 is conveyed, the unevenness of the entire surface of one material can be obtained. The grade is determined by comparing with a threshold value determined for the degree of unevenness such as striations, tears, and the like. Although the image of the TV camera 3 that images the material 10 from above is used in the description, the degree of unevenness is similarly obtained in the images of the TV cameras 4 and 5 that image from the left and right sides of the material 10. The grade determination can be performed by including the surface unevenness of the upper surface and the side surface of the material 10.

Also, as shown in FIG. 13, a mirror 12 is installed in place of the left and right side cameras, and the upper and side surfaces of the material can be viewed by one upper camera. The surface roughness of the material can be obtained.

[0020]

As described above, according to the present invention, by using grayscale images at two points of fruits and vegetables moving on a conveyor such as a belt conveyor, density interpolation and cross-correlation calculation of the images are performed. By determining the degree of surface irregularity of fruits and vegetables and enabling classification to be performed, it is now possible to automatically perform an irregularity inspection of the surface of fruits and vegetables, which has traditionally relied on visual inspection by inspectors, making continuous work difficult for a long time. The visual inspection is automated, there is no omission of detection, the determination result is stable, and there is a great effect that it can be realized without adding any new device to the conventional surface inspection device by image processing.

[Brief description of drawings]

FIG. 1 is a perspective plan view of an embodiment according to the present invention.

FIG. 2 is an internal view of the embodiment.

FIG. 3 is a processing flowchart of the embodiment.

FIG. 4 is an enlarged perspective view of a television camera viewing area.

FIG. 5 is a diagram showing a captured image.

FIG. 6 is a diagram for explaining density interpolation.

FIG. 7 is a diagram for explaining density interpolation data.

FIG. 8 is a diagram showing a density interpolation image.

FIG. 9 is a diagram for explaining cross-correlation.

FIG. 10 is a diagram showing distribution of cross-correlation.

FIG. 11 is a diagram for explaining monocular stereo vision.

FIG. 12 is a diagram showing a surface height distribution.

FIG. 13 is an internal arrow view of another embodiment.

[Explanation of symbols]

 1. Belt conveyor 2. Conveyor belt 3.4.5. TV camera 6. Background plate 7. Lighting device 8. Image processing device 9. Storage box 10. Material

Continuation of front page (51) Int.Cl. ⁶ Identification number Office reference number FI Technical indication location G06F 15/70 460D (72) Inventor Akio Tateishi 3-11-5 Iwamotocho, Chiyoda-ku, Tokyo Daido Electric Industry Incorporated (72) Inventor Mitsutoshi Akai 3-11-5 Iwamotocho, Chiyoda-ku, Tokyo Daido Electric Industry Co., Ltd.

Claims (1)

[Claims] 1. A transport device such as a belt conveyor for transporting fruits and vegetables, a television camera for imaging the fruits and vegetables on the transport device, and an illuminating device for illuminating the surface of the fruits and vegetables on the transport device.
In a fruit and vegetable surface inspection device including an image processing device for A / D converting a video signal from a television camera and inspecting the surface of the fruit and vegetables based on the A / D converted image data, Using the grayscale image data at two points of moving fruits and vegetables, a matching window is set in one of the images, the density interpolation processing is performed for each matching window, and the cross-correlation calculation with the other image is performed. Obtain the maximum position of the correlation value, for the difference in the height of the fruit and vegetable surface obtained from the maximum position of the correlation value of each matching window, depending on the degree of unevenness of the fruit and vegetable surface, it is possible to determine the grade by a preset threshold value. A characteristic method for inspecting the surface irregularities of fruits and vegetables.