JPH03229374A - Image processor - Google Patents

Abstract

PURPOSE:To accurately separate an object to be checked from other objects and to process the images at the low cost by changing the increase/decrease value of the illumination light corresponding to each segment via a fuzzy inference and based on each image pickup result of plural segments forming a check surface. CONSTITUTION:A check surface 5 which is photographed by a TV camera 1 is divided into the areas (segments) S1 - S16. The light sources LED1 - LED16 forming an illumination device 2 are set opposite to the areas S1 - S16 respectively. The quantity of illumination light of each light source is decided via the fuzzy inferences which are carried out based on the image pickup results of the areas S1 - S16 respectively. Then the fuzzy rule of the fuzzy inference is properly set based on the experiment rule obtained through an experiment so as to decide whether an object to be checked is included or not in each of areas S1 - S16. Thus only the quantity of light corresponding to a relevant check subject is increased. Thus it is possible to accurately separate the check subject from other object and to reduce the cost of an image processing system.

Description

DETAILED DESCRIPTION OF THE INVENTION (a) Industrial Application Field This invention illuminates an object to be inspected passing through an inspection surface on a production line in a factory with light from a light source to make an image of the object to be inspected. The present invention also relates to an image processing device that performs processing for determining an object to be inspected using the imaging results.

(b) Conventional Technology In an inspection line in which an image processing device is used, an object to be inspected and other objects may be simultaneously included in the inspection surface imaged by an imaging means included in the image processing device. For example, in determining the grade of grain, straw, rice husk, etc. are imaged at the same time as the grain, which is the object to be inspected. In order to accurately inspect only the object to be inspected in such a case, it is necessary to limit the image processing to only the object to be inspected. Therefore, there is a method of providing a detection means for detecting the relative position of the object to be inspected with respect to the inspection surface, and synchronizing the operation of the imaging means for taking an image of the inspection surface based on the detection result of the detection means with the passage of the object to be inspected, or A method is used in which the object to be inspected is positioned at a specific position on the imaging screen by using a mechanism such as an XY child table to keep the relative position of the imaging means and the object to be constant. However, in order to specify the position of the object to be inspected in the imaged screen, a complicated mechanism and detection means are required, which increases the cost due to the complexity of the apparatus.

For this reason, the captured images are binarized and labeled, and each label image obtained is characterized as being like an object to be inspected.
The determination was made using tHi, and image processing was performed using a labeled image containing the inspection object selected based on this. In many cases, the area value is used as the feature for determining the resemblance to the object to be inspected.

(C1 Problem to be Solved by the Invention) However, with the above-mentioned conventional technology, there is a problem that labeling processing in image processing requires a longer time than area measurement processing, and cannot be applied to inspection systems that require high-speed processing. In addition, in order to perform this labeling process at high speed, a dedicated LSI was required, leading to an increase in cost.Furthermore, if the object to be inspected is in contact with other objects, However, there was a problem in that these could not be separated and accurate inspection could not be performed.

This invention accurately separates the object to be inspected from other objects by changing the increase Kl of the illumination light corresponding to each segment using fuzzy inference based on the imaging results of each of the plurality of segments constituting the inspection surface. make it possible,
An object of the present invention is to provide an image processing device that can inspect an object to be inspected using image processing with a relatively simple configuration and at low cost. An image processing device that images an object to be inspected using an imaging means, which includes: a plurality of light sources corresponding to each of a plurality of segments constituting an inspection surface; The present invention is characterized in that it includes fuzzy inference means that determines the increase or decrease of illumination light for each segment according to a rule, and light rA drive means that drives each light source based on the inference result of the fuzzy inference means.

(e) Effect In this invention, the inspection surface on which the object to be inspected is placed is divided into a plurality of segments, and each segment is irradiated with light from its corresponding light source. The amount of illumination light of each of the plurality of light sources is determined by fuzzy inference performed based on the imaging results of each segment by the imaging means. Therefore, by appropriately setting fuzzy rules in fuzzy inference according to empirical rules obtained through experiments, it is possible to determine whether or not an object to be inspected is located in each segment, and to determine whether or not the object to be inspected is located in each segment. By increasing only the amount of light, it is possible to accurately divide the object to be inspected into two parts. Let me explain using an example.

FIG. 1 is an external view showing the configuration of a grade inspection device to which an image processing device according to an embodiment of the present invention is applied. Grain, which is an object to be inspected and whose grade is determined according to the size of its external shape, is conveyed on a conveyor 3. A predetermined inspection surface 5 is imaged on this conhair 3 by a television camera 1 serving as an imaging means. Further, an illumination device 2 is provided above the inspection surface 5. As shown in FIG. 2, this lighting device 2 is composed of 16 light sources LED1 to LED16.

The television camera 1 images the inspection surface 5 illuminated by the illumination device 2 as shown in FIG.
Transfer to. The grains conveyed on the belt conveyor 3 generally include straw and rice husks, and the image taken by the television camera 1 also includes the grains 1), straw and rice husks 12, as shown in FIG. Here, only the grains 1) are the objects to be inspected by the inspection device, and the straw and rice husk 12 are not the objects to be inspected.

The inspection surface 5 photographed by the television camera 1 is divided into regions (segments) 31 to S16, as shown in FIG. Light #L constituting the lighting device 2 shown in FIG.
ED1 to LED 16 face each of the regions 31 to 316, and when the light sources LEDI to LED 16 are uniformly driven, uniform illuminance can be obtained over the entire surface of the inspection surface 5. Note that the upper surface of the belt conveyor 3, which is the background in the screen imaged by the television camera 1, is sufficiently dark compared to the grains 1). The image processing device 4 measures the area of only the grain 1) located within the inspection surface 5, and determines the grade of the grain according to its size.

In addition, the television camera 1 images the inspection surface 5 at time intervals of about 10-odd milliseconds, and the moving speed of the belt conveyor 3 is determined so that the imaged screen is almost stopped at this imaging interval. ing.

FIG. 5 is a block diagram of a control section of the image processing apparatus.

CPIJ'21 is connected to lma0M22 and R via the internal bus.
AM23, G RAM 24 and Ink Hue IS 25
．． 26 are connected. An image captured by camera j is input to CP1, 321 via ink face 25.

The image captured by the camera 1 is binarized pixel by pixel by the binarization circuit 27, and this binarized screen is stored in the G-RAM.
24. Further, an LED driver 29 is connected to the CPU 21 via an interface 26.

This LED driver 29 is a light source L that constitutes the lighting device 2.
A voltage according to control data output from the CPU 21 is applied to each of ED1 to LED16. CPtJ21
executes processing operations according to a program written in advance in the ROM 22. Further, the ROM 22 stores fuzzy rules and membership functions related to fuzzy inference, which will be described later. Further, in the RAM 23, data input and output during processing operations of the CPU 21, which will be described later, is temporarily stored in each memory area. For example, as shown in FIG. 6, memory area MAI of the RAM 23 is assigned to counter i, which will be described later, and memory area
The detection areas A1 to A16 of each region constituting the inspection surface 5 are stored in the memory areas M A 2 to M A 17.
18 to MA33 respectively store fuzzy outputs Y1 to Y16 which are inference results of fuzzy inference described later.

FIG. 7 is a flowchart showing the processing procedure of the control section of the image processing apparatus.

At a predetermined inspection timing, the CPU 21 turns on the light source LED.
I~ Drive the LED 16 uniformly (I), TV camera 1
The image of the inspection surface taken by G-RA is converted into binary data.
Import into M24 (n2). Thereafter, the contents of the counter i assigned to the memory area MAL of the RAM 23 are set to 1 (n3), and the area Ai of the region Si on the inspection surface 5 specified by the contents of the counter i is measured (n4).

Here, the counter i assigned to the memory area MAI of the RAM 23 specifies the area on the inspection surface 5.

The area Ai of the portion corresponding to the area Si in n4 is measured by counting 81 the number of white pixels in the range corresponding to the area Si in the binarized image stored in the G-RAM 24. The measurement result of this area A i is stored in memory areas MA 2 to MA 17 of the RAM 23 . The measurement result of the area Ai is compared with the reference value α (
n5). When the area AiO value is less than the reference value α, the areas of four neighboring areas of the area St are measured. As shown in FIG. 8, these four neighboring areas are areas located on the upper, lower, left and right sides of the area Si. The CPU 21 uses the area Ai of this region Si and the areas of the four neighboring regions of the region Si as input variables, and uses the fuzzy rule shown in FIG.
) Fuzzy inference is performed according to the Mempage-Knob function shown in Figures.

In the fuzzy inference of n7, for example, the fuzzy rule Xi = S xz = S then y = NL is executed. In the above formula, X is the area Ai of the region Si, and Xz is the four neighboring regions S (i 4), S (i −1), S,
The area values of (f+1) and S (i+4) are substituted. Therefore, four types of equations with different x2 are obtained for the fuzzy inference of one region Si.

In each equation, the labels of X and Xz, that is, the area of each region, are obtained from the membership function shown in FIG. The inference results obtained by each of the above rules are logically summed using the centroid method in the membership function shown in FIG. 1) to obtain the increase/decrease Yi of the light source LEDi.

The increase/decrease MYi in the amount of illumination light of the light source LEDi obtained by the above fuzzy inference is stored in the memory area MA 1 of the RAM 23.
8 to MA 33 (n8). After that, the contents of the counter i are incremented (n9), and the processes 04 to n9 are performed for all areas (, n, l' O).

Note that if the white area Ai within the region is equal to or greater than the reference value α at n5, the process directly proceeds to n9. Furthermore, for example, in the region S1 shown in FIG. 4, among the four neighboring regions, the region S (
i-4), there is no area corresponding to 5(i-1). In this case, it is assumed that there is no output of Yi in the above rule. When the increase/decrease in the amount of illumination light is obtained for the entire area as described above, the memory area MA of the RAM 23 is
The fuzzy outputs Y1 to Y16 stored in each of MA18 to MA33 are read out and output to the L'ED driver 29. The LED driver 29 increases or decreases the light amount of the light sources LED 1 to LED according to this increase or decrease. In this state, a determination image is captured by the television camera 1 (n13), the number of white pixels is counted in the obtained image, the area of the grain 1) is measured, and the grade is determined (n14).

With the above processing, according to the present invention, if the area of the white part in each of the regions 81 to S16 constituting the inspection surface 5 is not sufficient, the area is determined based on the area of the white part in the four neighboring regions of that region. It is determined whether the image is of the grain 1) which is the object to be inspected or it is not the object to be inspected, and based on this judgment, if the image is the object to be inspected, the amount of light from the light source corresponding to that area is determined. If the area is not to be inspected, the amount of light from the light source corresponding to that area is decreased.

After adjusting the light intensity of the light source in this way, the inspection surface 5 is imaged again and the area of the white part in the obtained image is measured, thereby making it possible to accurately measure the area occupied only by the grains 1).

In this embodiment, a program related to fuzzy inference is stored in the ROM 22, and the CPU 21 executes this process. It may also be input to an analog fuzzy processor. Furthermore, a general-purpose ROM in which a program related to the processing procedure of fuzzy inference is written can also be used.

(g) Effects of the Invention According to this invention, if a white part with a minute area appears in any of the plurality of areas constituting the inspection surface, it is possible to determine whether it is part of the object to be inspected or not to be inspected. By using fuzzy inference to determine whether the object is an object, and increasing or decreasing the light intensity of the light source corresponding to that area according to the inference result, only the object to be inspected remains in the captured image of the inspection surface, and objects other than the object to be inspected appear. You can avoid it. As a result, even when an object other than the object to be inspected is in contact with the object to be inspected, only the object to be inspected can be accurately searched, and accurate image processing can be performed at all times. Moreover, since fuzzy inference is used to determine whether the object is part of the object to be inspected or not, the processing time is extremely faster than conventional labeling processing, making it suitable for high-speed inspection lines. It has the advantage of being fully compatible. Furthermore, when processing fuzzy inference, a general-purpose fuzzy processor or general-purpose ROM can be used, which has the advantage that a high-speed and accurate image processing system can be constructed at low cost.

[Brief explanation of drawings]

FIG. 1 is an external view of an inspection device to which an image processing device according to an embodiment of the present invention is applied, FIG. 2 is a diagram showing how a light source is installed in a lighting device provided in the image processing device, and FIG. 3 is a diagram showing an image captured by the same image processing device;
5 is a block diagram of the control unit of the image processing device, and FIG.
The figure is a memory map of the main part of the RAM that constitutes a part of the control unit, Figure 7 is a flowchart showing the processing procedure of the control unit, and Figure 8 is the four neighboring areas for one area during the processing operation of the control unit. FIG. 9 is a diagram showing the fuzzy rules related to fuzzy inference executed during the processing operation of the control section, FIG. 10 and FIG. 1) are executed during the processing operation of the control section. FIG. 3 is a diagram showing membership functions related to fuzzy inference. 1-television camera (imaging means), 2-illumination device, 4-image processing device, 5-inspection surface, 1)-inspection object, 81-516-area (segment), LEDI-LED16-light source.

Claims (1)

[Claims] (1) In an image processing device that images an object to be inspected on an inspection surface using an imaging means, a plurality of light sources corresponding to each of a plurality of segments constituting the inspection surface and a An image characterized by being provided with fuzzy inference means that determines the increase or decrease of illumination light for each segment according to predetermined fuzzy rules, and light source drive means that drives each light source based on the inference result of the fuzzy inference means. Processing equipment.