JP7407189B2 - Egg-checking equipment, egg-checking programs, and egg-checking methods - Google Patents

Description

The present invention relates to an egg checking device, an egg checking program, and an egg checking method.

Conventionally, a fertilized egg is irradiated with light to capture an image of the inside of the egg, an inspection area is extracted from the captured image, and blood vessel information within the inspection area is measured. A technique for automatically determining normal eggs based on the total length of blood vessels is known (see Patent Document 1).

Patent No. 3998184

With conventional technology, it was possible to determine whether the eggs to be inspected were normal or defective eggs, but if the eggs were defective, the problem was that it was not possible to identify the cause of the defect with high accuracy. there were. In recent years, there has been a need to improve the breeding conditions for chickens and eggs and vaccine production conditions by verifying the causes of defects, and there has been a need for technology that can identify the causes of defects with high accuracy.

An object of the present invention is to provide an egg checking device, an egg checking program, and an egg checking method that can identify the cause of defective eggs with high accuracy.

The egg checking device according to the present invention includes a first determining means for determining whether the egg has a first defective factor using a first trained model created in advance using an image of the egg as training data; Determining whether the egg has a second defective factor different from the first defective factor using a second trained model different from the first trained model, which is created in advance using images as training data. a second discriminating means for capturing an image of the egg to be inspected by irradiating light onto the egg to be inspected, and a control means, and the control means is configured to: By causing the first determining means and the second determining means to determine the defective factor of the egg to be inspected, it is determined whether or not the egg to be inspected has a defective factor, and if the egg to be inspected has a defective factor; The cause of the defect is determined.
In the egg checking device, the control means causes the first determining means to determine whether the egg to be examined has the first defective factor based on the image of the egg to be examined, and If the eggs do not have the first defective factor, the second determining means may be configured to determine whether the egg to be inspected has the second defective factor.
In the egg checking device, the control means causes the first determining means to determine whether the egg to be examined has the first defective factor based on the image of the egg to be examined, and Regardless of whether or not the egg to be inspected has the first defective factor, the second determining means may be configured to determine whether the egg to be inspected has the second defective factor.
In the egg checking device, the first determining means determines a failure factor based on the state of blood vessels of the egg as the first failure factor, and the second determining means determines a failure factor based on the state of blood vessels of the egg as the second failure factor. The defect factor can be determined based on any one of the state of the chamber, the state of growth, and the presence or absence of cracks.
In the egg checking device, a third trained model different from the first trained model and the second trained model, which is created in advance using an image of the egg as training data, is used to detect the first defective cause and the second trained model. It may be configured to further include a third determining means for determining whether or not there is a third failure factor different from the second failure factor.
In the egg checking device, using a fourth trained model different from the first trained model, the second trained model, and the third trained model, which is created in advance using an egg image as teacher data, It may be configured to further include fourth determining means for determining whether a fourth failure factor different from the first failure factor, the second failure factor, and the third failure factor is present.
In the egg checking device, the first defective factor, the second defective factor, the third defective factor, and the fourth defective factor are respectively a blood vessel condition, an air chamber condition, and a growth condition. , and can be configured to be a failure factor based on the presence or absence of cracks.
The egg checking device may be configured to further include a fifth discriminating means for discriminating normal eggs using a fifth trained model created in advance using an egg image as training data.
The egg checking device may be configured such that the eggs to be examined are eggs to be used in vaccine production.
The above-mentioned egg checking device may be configured to have a function of totaling the frequency of occurrence of each defect factor.
The egg checking program according to the present invention acquires an image of the egg to be examined by irradiating the egg to be examined with light on a computer, and uses a first trained model created in advance using the image of the egg as training data. , determine whether the egg to be inspected has the first defective factor, and use a second trained model different from the first trained model, which is created in advance using an image of the egg as training data, to determine whether the egg to be inspected has the first defective factor; By determining whether the egg has a second defective factor different from the first defective factor, it is determined whether the egg to be inspected has a defective factor, and the egg to be inspected has a defective factor. If so, a process for determining the cause of the defect is executed.
The egg checking method according to the present invention uses a computer to irradiate the egg to be examined with light to obtain an image of the egg to be examined, and to use a first trained model created in advance using the image of the egg as training data. to determine whether the egg to be inspected has the first defective factor, and using a second trained model different from the first trained model, which was created in advance using an image of the egg as teacher data, By determining whether the egg to be inspected has a second defective factor different from the first defective factor, it is determined whether the egg to be inspected has a defective factor; If so, determine the cause of the defect.

According to the present invention, instead of determining the defective factors of eggs to be inspected at once, a first trained model and a second trained model different from the first trained model are separately used to By determining the defect factor of eggs, it is not only possible to determine whether the egg to be inspected has a defect factor, but also to identify the defect factor with high accuracy when the egg to be inspected has a defect factor.

FIG. 2 is a schematic diagram illustrating the structure of a normal fertilized egg that is more than ten days old. FIG. 3 is a diagram showing an example of a captured image of a normal egg that is more than ten days old. It is a figure which shows an example of the captured image of the defective egg (excluding the egg with a crack) of more than 10 days old. FIG. 2 is a diagram showing an example of a captured image of a defective egg (egg with cracks) that is more than 10 days old. It is a block diagram of the egg checking device concerning this embodiment. It is a flowchart for explaining the egg checking method concerning a 1st embodiment. It is a figure for explaining the egg checking method concerning a 1st embodiment. It is a flow chart for explaining the egg checking method concerning a 2nd embodiment. It is a figure for explaining the egg checking method concerning a 2nd embodiment.

Embodiments of the present invention will be described below based on the drawings. The present invention relates to an egg-checking device, an egg-checking program, and an egg-checking method for testing chicken eggs (fertilized eggs) used in the production of vaccines and the like. The eggs to be tested in the present invention are fertilized eggs of chickens and the like, and the egg surface may have any color such as white or brown. It does not matter the type of virus such as influenza or the type of medicine or other injected material, whether or not it was injected, but it does not matter whether it is injected or not, as it grows, the fetal blood vessels are widely distributed inside the egg. The technical significance of the present invention is great in the case of destructive testing.

《First embodiment》
FIG. 1 is a diagram illustrating the general structure of a normal fertilized egg of more than 10 days old. The external structure of an embryonated chicken egg is covered by a shell called the eggshell. There is an eggshell membrane just inside the eggshell membrane, and the eggshell and eggshell membrane are used to exchange oxygen with the inside. There are large blood vessels just inside the eggshell. For example, the internal structure of an embryonated chicken egg has an air chamber, which is a layer of air, at the tip of the egg, and a fetus wrapped in amniotic membrane containing amniotic fluid in the center. Between the fetus and the air chamber is the chorioallantoic cavity, which is surrounded by the chorioallantoic membrane. In the case of a normally growing fertile egg that is more than 10 days old, the chorioallantoic cavity has a certain size, and blood vessels are widely distributed in the chorioallantoic membrane.

FIG. 2 shows an example of a captured image of a normal egg that is more than 10 days old. In the example shown in FIG. 2, by irradiating only the egg with light in a dark room, it is possible to image the egg in a state where the inside of the egg is transparent (the same applies to FIG. 3). In the case of a normally growing fertile egg that is more than 10 days old, the chorioallantoic cavity is of a certain size and the fetus is located near the center of the egg. The color contrast between the chamber and the chorioallantoic cavity is clearly different, there is no bleeding within the shell membrane, the chorioallantoic membrane, and the chorioallantoic cavity, and there is an air chamber of a certain size in the upper part of the egg. characteristics can be seen.

On the other hand, in the case of defective eggs, the following characteristics are observed for each factor of defect. Note that FIGS. 3 and 4 are diagrams showing examples of captured images of defective eggs taken using the same method as in FIG. 2.
(1) Infertile eggs As shown in FIG. 3(A), infertile eggs have no blood vessels and no fetuses. Therefore, a characteristic can be seen that the difference in hue becomes smaller (contrast becomes lower) as a whole.
(2) Discontinued eggs (dead eggs)
As shown in FIG. 3(B), aborted eggs have a significantly narrower and fewer blood vessels than normal eggs. In addition, the blood vessels distributed are extremely thin and pale in color, the difference in color near the border between the air chamber and the chorioallantoic space may be small, and bleeding is observed within the shell membrane, on the chorioallantoic space, and in the chorioallantoic space. There are some characteristics that can be seen.
(3) Underdeveloped eggs As shown in Figure 3 (C), underdeveloped eggs have characteristics such as the distribution of blood vessels on the chorioallantoic membrane being extremely narrow and few, and the blood vessels distributed being extremely thin and pale in color. .
(4) Eggs with defective air chambers As shown in Figure 3 (D), eggs with defective air chambers have deformed air chambers (air chambers are tilted), air chambers and chorion. Characteristics can be seen, such as the boundary with the membrane not forming a smooth curve.
(5) Upside-down eggs Upside-down eggs are eggs that are placed upside down (with the blunt end (tip on the air chamber side) and the sharp end facing the opposite direction) when placing the eggs to be inspected on a special tray. As shown in FIG. 3(E), the characteristic is that no air chambers are visible.
(6) Eggs with cracks Eggs with cracks have cracks, which are represented by lines with high brightness on the imaging screen, as shown in FIG. Characteristics include a significantly large chamber.

As described above, there are multiple causes of defective eggs, and the egg checking device 1 according to the present embodiment also aims to identify the defective factor with high accuracy when the egg to be inspected is a defective egg. . Therefore, in this embodiment, as shown in FIGS. 2 to 4, experts are asked to actually check the captured images of eggs to be used for vaccine production, and check whether the eggs are normal or not (see (1) above). ) to (6), the eggs are judged to be defective: unfertilized eggs, discontinued eggs, poorly developed eggs, eggs with poor air chambers, upside down eggs, or cracked eggs. The results were labeled. In addition, a trained model was constructed by subjecting the image data labeled in this way to machine learning using deep learning using a multilayered deep neural network as a model. Then, the egg checking device 1 uses the constructed learned model to determine from the image of the egg to be examined whether the egg to be examined is a normal egg or a defective egg. In such cases, the cause of the failure will also be identified. This makes it possible, for example, to feed back defective factors and their frequency of occurrence to egg and vaccine producers, which can be used to improve breeding conditions for chickens and eggs and vaccine production conditions. Furthermore, in the egg checking device 1 according to the present embodiment, as will be described later, by using a plurality of trained models in multiple stages, it is possible to solve the problem of defective eggs, which is difficult to solve with a single trained model. can be identified with high accuracy at a practical level.

FIG. 5 is a configuration diagram of the egg checking device 1 according to this embodiment. As shown in FIG. 5, the egg checking device 1 according to the present embodiment includes an imaging device 10, an illumination device 20, a discrimination device 30, and a transport device 40. Each device will be explained below.

The imaging device 10 is a camera for capturing an image of an egg to be inspected, and includes, for example, a color CCD camera that images the egg to be inspected and outputs color image data of the egg to be inspected. Furthermore, the imaging device 10 is not limited to a color CCD camera, and may be any known camera that captures color images, such as a color CMOS camera.

Further, the illumination device 20 is a light for irradiating light to the egg to be inspected, and as shown in FIG. 5, the optical axis L1 of the imaging device 10 and the light of the illumination device 20 It is arranged at a position intersecting with the axis L2. Note that the lighting device 20 may be, for example, an LED light, although it is not particularly limited.

When imaging an egg to be inspected, as shown in FIG. 5, the egg to be inspected is placed on a support stand 42, and in order to prevent disturbances during imaging, the support stand 42 is raised and lowered, and the egg to be inspected is imaged in a light-shielding manner. 50. An imaging device 10 and an illumination device 20 are arranged within the imaging section 50. The illumination device 20 is placed in a tube 21 for illumination, and the supporting platform 42 is raised until the tip of the tube 21 comes into contact with the egg to be inspected, and the egg to be inspected is placed on the side of the air chamber by the illumination device 20. More lighting. The tip of the tube 21 for housing the illumination device 20 is made of a soft material and has a bellows structure to prevent illumination light from leaking to the outside even if the tube has a different egg shape or size. ing. When the egg to be inspected is illuminated by the illumination device 20 in the imaging unit 50, the egg to be inspected is imaged by the imaging device 10. By illuminating only the egg to be inspected with the lighting device 20 in a dark room, the imaging device 10 can image the inside of the egg to be inspected. In addition, as shown in FIG. 2, the imaging device 10 captures an image of the egg to be examined from the side, thereby detecting the air chambers, the inside of the shell membrane, the chorioallantoic membrane, the chorioallantoic cavity, and the state of the fetus. can be imaged so that they can be observed all at once. Note that the operations of the imaging device 10 , the illumination device 20 , and the support stand 42 are controlled by the discrimination device 30 , and the captured image of the egg to be inspected captured by the imaging device 10 is transmitted to the discrimination device 30 .

The discrimination device 30 acquires a captured image of the egg to be inspected from the imaging device 10, and determines whether the egg to be inspected is a normal egg or a defective egg based on the captured image of the egg to be inspected. Furthermore, when the egg to be inspected is a defective egg, the discrimination device 30 identifies the cause of the defect. As shown in FIG. 5, the discrimination device 30 includes a storage device 31 and a processing device 32.

The storage device 31 stores in advance a plurality of trained models using a plurality of captured images of eggs as training data. Specifically, the storage device 31 stores four different learned models M1 to M4 for determining the cause of defective eggs. The trained models M1 to M4 stored in the storage device 31 will be explained below.

The trained model M1 is constructed by machine learning using deep learning using the first learning data group classified based on the state of egg blood vessels as training data in order to distinguish between "unfertilized eggs" and "abandoned eggs". This is the model that was created. The first learning data group consists of a group of images in which eggs have been previously determined by experts to be infertile eggs due to reasons such as the absence of blood vessels, and are labeled as "unfertilized eggs," and a group of images that have a small distribution of blood vessels, short blood vessels, and eggs that are labeled as "infertile eggs." The images include a group of images in which eggs are determined to be discontinued eggs due to thinness or other reasons and are labeled as "discontinued eggs," and a group of images in which the eggs are determined to be neither infertile eggs nor discontinued eggs and are labeled as "first normal eggs." By using the trained model M1, it is possible to determine from the image of the egg to be inspected whether the egg to be inspected is an "unfertilized egg", a "discontinued egg", or a first normal egg including a normal egg. can be determined.

The trained model M2 is machine learned by deep learning using the second learning data group classified based on the state of the air chambers as training data in order to distinguish between "upside down eggs" and "eggs with poor air chambers." This is a model built with. The second learning data group consists of a group of images in which an egg was previously identified as an upside-down egg by an expert because no air chambers were observed and was labeled as an "upside-down egg," and a group of images in which the shape of the air chamber was deformed. One group of images was determined to be an egg with a defective air chamber and was labeled as an ``egg with a defective air chamber,'' and the other was a group of images in which an egg was determined to be neither an upside-down egg nor an egg with a defective air chamber and was labeled as a ``secondary normal egg.'' This includes a group of images. By using the trained model M2, from the image of the egg to be inspected, it is possible to determine whether the egg to be inspected is an "upside down egg", "an egg with poor air chamber", or a second normal egg including a normal egg. It is possible to determine whether there is

The trained model M3 is a model constructed by performing machine learning using deep learning using the third learning data group classified based on the developmental state as a feature as training data in order to identify "underdeveloped eggs". . The third learning data group has been previously determined by experts to be poorly developed eggs due to reasons such as the distribution of blood vessels on the chorioallantoic membrane being extremely narrow and few, and the blood vessels distributed being extremely thin and pale in color. The images include a group of images labeled as "defective eggs" and a group of images labeled as "third normal eggs" that were determined not to be poorly developed eggs. By using the trained model M3, it is possible to determine from the image of the egg to be inspected whether the egg to be inspected is a "underdeveloped egg" or a third normal egg including a normal egg.

The trained model M4 is a model constructed by performing machine learning using deep learning using the fourth learning data group classified based on the presence or absence of cracks as training data in order to identify "cracked eggs". . The fourth learning data group consists of images of eggs that have been determined by experts to have cracks by direct visual inspection. This includes a group of images labeled as "egg with cracks" due to the presence of a crack or a significantly large air chamber as shown in , and a group of images labeled as "fourth normal egg" other than those. By using the learned model M4, it is possible to determine from the image of the egg to be inspected whether the egg to be inspected is a "cracked egg" or a fourth normal egg including a normal egg.

Note that the construction of the learned models M1 to M4 may be performed in advance by the processing device 32, or the learned models M1 to M4 constructed by an external device may be acquired and stored in the storage device 31. . The algorithm of the machine learned model is not particularly limited, such as support vector machine, logistic regression, neural network, etc., but as mentioned above, it is a neural network, especially a deep neural network with three or more layers. Among deep neural networks, it is preferable to use a convolutional neural network suitable for image recognition.

The processing device 32 executes the egg checking program stored in the memory, and uses the captured image of the egg to be tested obtained from the imaging device 10 and the learned models M1 to M4 stored in the storage device 31. It is determined whether the eggs to be inspected are normal eggs or defective eggs. Furthermore, if the egg to be inspected is a defective egg, the processing device 32 also identifies the cause of the defect. Furthermore, the processing device 32 also has a function of compiling the defect factors of defective eggs and their frequency of occurrence, and outputting the results to a display (not shown) or the like. Note that the details of the method for determining the eggs to be inspected by the processing device 32 will be described later.

The transport device 40 transports the eggs to be inspected to the imaging section 50 and also transports the eggs to be inspected from the imaging section 50. Specifically, the transport device 40 drives a conveyor and transports a plurality of eggs to be inspected placed on a special tray 41 to a position of a support base 42 located below the imaging section 50. Further, the transport device 40 raises the support stand 42 to transfer the eggs to be inspected from the special tray 41 to the support stand 42, and then further raises the support stand 42 to transfer the eggs to be inspected to the inside of the imaging section 50. lift. When the imaging unit 50 finishes imaging the eggs to be inspected, the transport device 40 lowers the support base 42 and returns the imaged eggs to be inspected to the dedicated tray 41 . Then, the transport device 40 moves the conveyor to transport the next egg to be tested to the position of the support stand 42 . In this way, the eggs to be inspected are sequentially transported to the imaging section 50, and images are taken by the imaging device 10. In addition, the conveyance device 40 can also eliminate defective eggs based on the discrimination result of the eggs to be inspected by the discrimination device 30.

Next, the egg checking process according to the first embodiment will be described based on FIGS. 6 and 7. FIG. 6 is a flowchart showing the egg checking process according to the first embodiment, and FIG. 7 is a diagram for explaining the egg checking process according to the first embodiment.

In step S101, the imaging device 10 images the egg to be inspected. In this embodiment, the egg to be inspected is placed in the imaging section 50, which is a dark room, and only the egg to be inspected is illuminated by the illumination device 20, so that the imaging device 10 captures an image while seeing through the inside of the egg to be inspected. can do. Image data of the egg to be inspected captured by the imaging device 10 is transmitted to the processing device 32 . Then, in step S102, the processing device 32 acquires image data of the egg to be inspected captured by the imaging device 10.

In step S103, the processing device 32 performs a process of determining whether the egg to be inspected is an infertile egg or a discontinued egg. Specifically, the processing device 32 executes the egg checking program stored in the memory, and uses the image data of the egg to be examined acquired in step S102 and the trained model M1 to determine whether the egg to be examined is an infertile egg. It is determined whether the eggs are present, discontinued eggs, or other first normal eggs. Then, the process proceeds to step S104, and if it is determined that the egg to be inspected is an infertile egg or a discontinued egg as a result of the determination in step S103, the process proceeds to step S112, and the egg to be inspected is determined to be a defective egg. Then, in the subsequent step S113, the cause of the defect in the eggs to be inspected (infertile eggs or discontinued eggs) is output.

On the other hand, if it is determined in step S104 that the egg to be inspected is the first normal egg, the process proceeds to step S105. In step S105, the processing device 32 performs a process of determining whether the egg to be inspected is an upside-down egg or an egg with poor air chambers. Specifically, the processing device 32 executes the egg checking program stored in the memory, and uses the image data of the egg to be examined acquired in step S102 and the learned model M2 to determine whether the egg to be examined is an upside-down egg. It is determined whether the egg is a defective egg, an egg with a defective air chamber, or a second normal egg. Then, the process proceeds to step S106, and if the result of the determination in step S105 is that the egg to be inspected is determined to be an upside-down egg or an egg with a defective air chamber, the process proceeds to step S112, and the egg to be inspected is determined to be a defective egg. Ru. Then, in the subsequent step S113, the cause of the defect in the egg to be inspected (upside down egg or egg with defective air chamber) is output.

On the other hand, if it is determined in step S106 that the egg to be inspected is the second normal egg, the process proceeds to step S107. In step S107, the processing device 32 performs a process of determining whether or not the egg to be inspected is a poorly developed egg. Specifically, the processing device 32 executes the egg checking program stored in the memory, and uses the image data of the egg to be examined acquired in step S102 and the trained model M3 to determine whether the egg to be examined is underdeveloped. It is determined whether it is an egg or a third normal egg. Then, the process proceeds to step S108, and if it is determined that the egg to be inspected is a poorly developed egg as a result of the determination in step S107, the process proceeds to step S112, where the egg to be inspected is determined to be a poorly developed egg. Then, in the subsequent step S113, the cause of the defect (underdeveloped egg) of the egg to be inspected is output.

On the other hand, if it is determined in step S108 that the egg to be inspected is the third normal egg, the process proceeds to step S109. In step S109, the processing device 32 performs a process of determining whether the egg to be inspected is a cracked egg. Specifically, the processing device 32 executes the egg inspection program stored in the memory, and uses the image data of the egg to be inspected acquired in step S102 and the learned model M4 to determine whether the egg to be inspected has cracks. It is determined whether it is an egg or a fourth normal egg. Then, the process proceeds to step S110, and if the result of the determination in step S109 is that the egg to be inspected is determined to be an egg with cracks, the process proceeds to step S112, where the egg to be inspected is determined to be an egg with cracks. Then, in the subsequent step S113, the cause of the defect in the egg to be inspected (cracked egg) is output.

On the other hand, if it is determined in step S110 that the egg to be inspected is the fourth normal egg, the process proceeds to step S111. In step S111, the processing device 32 determines that the egg to be inspected is a normal egg. This ends the egg checking process according to this embodiment.

In addition, in the egg checking process shown in FIG. 6, a process for determining whether the egg to be examined is an infertile egg or an abandoned egg (step S103), a process for determining whether the egg to be examined is an upside-down egg or an egg with a defective air chamber. (Step S105), the process of determining whether the egg to be inspected is a poorly developed egg (Step S107), and the process of determining whether the egg to be inspected is an egg with cracks (Step S110) are exemplified. did. However, the execution order of these processes is not limited to the example shown in FIG. 6, and the order may be changed as appropriate.

As described above, the egg checking device 1 according to the present embodiment uses image data of eggs labeled with the result of determining whether the egg is a normal egg or a defective egg, and the cause of the defect if the egg is a defective egg. By using a trained model built by machine learning using deep learning using a multi-layered deep neural network model as training data, it is possible to determine whether the eggs to be inspected are normal eggs or defective eggs. In addition to being able to determine if the egg is defective, it is also possible to identify the cause of the defect if the egg is defective.

Further, in the egg checking device 1 according to the present embodiment, by using the plurality of trained models M1 to M4 in multiple stages, it is possible to determine the defect factor of defective eggs with higher accuracy. In particular, in the production of vaccines, it is necessary to distinguish between "unfertilized eggs," "aborted eggs," "upside down eggs," and "normal eggs" with high accuracy. With this method, "infertile eggs", "aborted eggs", "upside down eggs", and "normal eggs" can be distinguished with high accuracy. For example, a model was constructed using training data that labeled "unfertilized eggs," "abandoned eggs," "egg with poor air chambers," "upside down eggs," "underdeveloped eggs," "cracked eggs," and "normal eggs." When the eggs to be inspected are classified using only a single trained model, the results are: "unfertilized eggs", "abandoned eggs", "poorly ventilated eggs", "upside down eggs", "underdeveloped eggs", "cracked eggs" Only about 90% discrimination accuracy was obtained for all classifications (factors) of ``egg'' and ``normal egg.'' On the other hand, in the egg checking process according to the present embodiment, when the eggs to be examined are discriminated using a plurality of trained models M1 to M4 in multiple stages, the results are "unfertilized eggs", "abandoned eggs", The researchers were able to identify almost 100% (100% when testing on just under 90,000 eggs) of ``upside-down eggs,'' which are extremely important to prevent contamination in vaccine production. Furthermore, "normal eggs" could be identified with an accuracy of 99% or more.

The egg checking device 1 according to the present embodiment also has a function to aggregate the frequency of occurrence of each defective factor, and feeds back the identified defective factors and their frequency of occurrence to egg and vaccine production managers. The results can be used to improve egg rearing conditions and vaccine production conditions. For example, if the cause of defects is a high proportion of poorly developed eggs, the production manager can take measures such as changing the settings (temperature and humidity) of the incubator. Furthermore, if the percentage of eggs with cracks is high as a cause of failure, the production manager can take measures such as changing the feed mix or inspecting the transport equipment. In this way, the operation of the production process can be improved depending on the frequency of occurrence of defective factors. In addition, various conditions in incubation work (e.g., rearing conditions such as temperature and humidity, feed composition, parent chicken history, incubator settings such as temperature and humidity), various conditions in virus production work (e.g., type of virus, Establish the optimal production conditions for fertile eggs for vaccine production by associating various conditions during egg checking (e.g. incubator settings such as temperature and humidity) and the proportion of each defective factor that occurs. You can also expect to do so.

《Second embodiment》
Next, a second embodiment of the present invention will be described. The egg checking device according to the second embodiment has the same configuration as the egg checking device 1 according to the first embodiment, except for what is explained below. perform an action.

In the second embodiment, the storage device 31 stores seven trained models M11 to M17. The seven learned models M11 to M17 stored in the storage device 31 will be described below.

The learned model M11 is a model constructed by performing machine learning by deep learning using a learning data group classified based on the state of blood vessels of eggs as training data in order to determine "abandoned eggs." The learning data group includes a group of images that have been previously determined to be discontinued eggs by an expert due to reasons such as poor blood vessel distribution, short blood vessels, thin blood vessels, etc., and are labeled as "abandoned eggs." By using the trained model M11, the processing device 32 can determine from the image of the egg to be inspected whether or not the egg to be inspected is a "discontinued egg."

The trained model M12 is a model constructed by performing machine learning by deep learning using a learning data group classified based on the state of the blood vessels of the egg as training data in order to determine whether it is an "unfertilized egg". The learning data group includes a group of images that have been previously determined by an expert to be infertile eggs due to reasons such as the absence of blood vessels in the eggs, and have been labeled as "infertile eggs." By using the trained model M12, the processing device 32 can determine whether the egg to be inspected is an "infertile egg" from the image of the egg to be inspected.

The learned model M13 is a model constructed by performing machine learning using deep learning using a group of learning data classified based on the state of the air chambers as training data in order to determine "egg with defective air chambers". . The learning data group includes a group of images that have been previously determined by an expert to be eggs with poor air chambers due to deformed shape of the air chambers, and are labeled as "eggs with poor air chambers." By using the trained model M13, the processing device 32 can determine from the image of the egg to be inspected whether or not the egg to be inspected is an "egg with poor air chambers."

The trained model M14 is a model constructed by performing machine learning using deep learning, using a group of learning data classified based on the state of the air chamber as a feature, as training data, in order to identify an "upside down egg." The learning data group includes a group of images that have been previously determined by an expert to be an upside-down egg because no air chambers are observed in the egg, and have been labeled as "upside-down eggs." By using the learned model M14, the processing device 32 can determine whether the egg to be inspected is an "upside down egg" from the image of the egg to be inspected.

The trained model M15 is a model constructed by performing machine learning using deep learning using a learning data group classified based on the developmental state as a feature as training data in order to identify "underdeveloped eggs." The learning data group has been previously determined by experts to be underdeveloped eggs due to reasons such as the distribution of blood vessels on the chorioallantoic membrane being extremely narrow and few, and the blood vessels distributed being extremely thin and pale in color. Contains images labeled as "Eggs". By using the learned model M15, the processing device 32 can determine from the image of the egg to be inspected whether or not the egg to be inspected is a "slowly developed egg."

The learned model M16 is a model constructed by machine learning using deep learning using a learning data group classified based on the presence or absence of cracks as a feature, in order to identify "cracked eggs." The learning data group is based on images of eggs that have been determined by experts to have cracks due to direct visual inspection, and a high-intensity linear pattern on the captured image. Contains images labeled as "cracked eggs" due to visible cracks, significantly larger air chambers, etc. By using the learned model M16, the processing device 32 can determine from the image of the egg to be inspected whether or not the egg to be inspected is a "cracked egg."

The learned model M17 is a model constructed by machine learning using deep learning using a group of learning data classified to identify "normal eggs" as training data. The learning data group is a captured image of an egg that has been determined in advance by an expert to be normal with no defective factors in the egg through direct visual inspection, and includes a group of images labeled as "normal eggs." By using the trained model M17, the processing device 32 can determine whether the egg to be inspected is a "normal egg" from the image of the egg to be inspected.

Next, the egg checking process according to the second embodiment will be described based on FIGS. 8 and 9. FIG. 8 is a flowchart for explaining egg checking processing according to the second embodiment. Moreover, FIG. 9 is a diagram for explaining the egg checking process according to the second embodiment.

As shown in FIG. 8, in steps S201 and S202, similarly to steps S101 and S102 of the first embodiment, the imaging device 10 images the egg to be inspected (step S201), and the inspection object imaged in step S201 Image data of the target egg is acquired (step S202).

In step S203, the processing device 32 performs a process of determining whether the egg to be inspected is a "discontinued egg." Specifically, the processing device 32 executes the egg checking program stored in the memory, and uses the image data of the egg to be examined acquired in step S202 and the trained model M11 to determine whether the egg to be examined is a "cancelled egg". ”. Then, the processing device 32 stores the determination result as to whether the egg is a "cancelled egg" in the storage device 31, and proceeds to the next step S204.

In steps S204 to S209, similarly to step S203, the processing device 32 determines whether the eggs to be inspected are "unfertilized eggs", "eggs with poor air chambers", "upside down eggs", "underdeveloped eggs", "eggs with cracks", and Processing is performed to determine whether each egg is a "normal egg". That is, the processing device 32 executes the egg checking program stored in the memory, and determines whether the egg to be examined is an "infertile egg" or It is determined whether the egg is an "egg with poor air chamber", "an upside-down egg", "an egg with poor growth", "an egg with cracks", or a "normal egg", and the results of each determination are stored in the storage device 31. In the second embodiment, even if the egg to be inspected has a defective factor, as shown in steps S203 to S209 of FIG. 8 and FIG. It is determined whether the eggs are "defective eggs," "upside down eggs," "underdeveloped eggs," "cracked eggs," or "normal eggs." Then, in step S210, the processing device 32 outputs the results determined in each of the determination processes in steps S203 to S209.

As described above, in the second embodiment, if an egg to be inspected has any defective factor, it is not immediately determined to be a defective egg and the defective factor is determined, but as shown in FIG. By determining a plurality of defective factors in a target egg in parallel, it is possible to determine a plurality of defective factors, such as "upside down egg" and "cracked egg," for example, in one test target egg. In addition, in the second embodiment, not only the trained models M11 to M16 using images of eggs with defective factors as training data, but also the trained model M17 using images of "normal eggs" as training data are used. , it is possible to improve the accuracy of discrimination from "normal eggs". For example, in the first embodiment, an egg to be inspected for which no defective factor has been determined is determined as a "normal egg," but in the second embodiment, a defective factor is not determined and the egg is determined as a "normal egg." By identifying the tested eggs as "normal eggs," it is possible to improve the accuracy of identifying "normal eggs," and also to identify eggs that have not been identified as defective and are not determined to be "normal eggs." This makes it possible for the operator to recognize that a new defect factor has occurred and to visually classify it.

Although preferred embodiments of the present invention have been described above, the technical scope of the present invention is not limited to the description of the above embodiments. Various changes and improvements can be made to the embodiments described above, and forms with such changes and improvements are also included within the technical scope of the present invention.

For example, in the embodiment described above, the configuration is illustrated in which the egg to be inspected is imaged from only one direction by the imaging device 10, and the egg to be inspected is discriminated using the captured image taken from the direction, but this configuration is limited. For example, by rotating the support stand 42 of the egg to be examined, images of the egg to be examined are taken from a plurality of directions (for example, four directions by rotating the support stand 42 by 90 degrees), and images are taken from a plurality of directions. The configuration can be such that the eggs to be inspected are determined using the captured image. In this case, the accuracy of identifying eggs to be inspected can be further improved.

Further, in the above-described first embodiment, a configuration was exemplified in which eggs to be inspected are determined using four trained models, but the number of trained models may be two or more; for example, three trained models, or Eggs to be inspected can be identified using five or more trained models. For example, when manufacturing a vaccine, it is important to distinguish between three defective factors: "unfertilized eggs," "discontinued eggs," and "upside down eggs." It is possible to adopt a configuration in which two trained models, a trained model M1 for determining an "upside down egg" and a trained model M2 for determining an "upside down egg", are used to determine the egg to be inspected. In addition, in this embodiment, as shown in FIGS. 4(F) and 4(G), there is a group of image data that has a line with high brightness on the imaging screen and can be determined as having a crack, and a group of image data that has a large air chamber with a crack. We have exemplified a configuration in which an "egg with a crack" is determined using the learned model M4, which uses a group of image data that can be determined as training data for an "egg with a crack", but for example, as shown in FIG. After first determining "cracked eggs" using a trained model using as training data only image data groups that have lines with high brightness on the imaging screen and can be determined to have cracks, as shown in Fig. 4 (G) As shown in the figure, a configuration may be adopted in which an "egg with cracks" is determined using a trained model that uses as training data a group of image data in which the air chambers are significantly large and can be determined as having cracks. In this case, five trained models are used to discriminate eggs to be inspected, and cracked eggs can be discriminated with higher accuracy.

Furthermore, in the embodiment described above, a configuration is illustrated in which a single illumination device 20 is provided and light is irradiated from the air chamber side of the egg to be inspected. It is also possible to have a configuration in which the eggs to be tested are inspected by irradiating light from below the eggs to be inspected. In this case, it becomes easier to check the inside of the egg, and especially when inspecting cracked eggs, it becomes possible to detect cracks in a wide range.

Furthermore, in the above-described embodiment, the configuration is exemplified to determine "abandoned eggs", "unfertilized eggs", "poor air chamber eggs", "upside down eggs", "underdeveloped eggs", and "cracked eggs" as defective factors. However, the defective factors are not limited to the above, and can be added/changed as appropriate. For example, it is also possible to use a trained model for determining eggs with bleeding accumulated at the air chamber boundaries to determine such defective factors.

1... Egg checking device 10... Imaging device 20... Illumination device 30... Discrimination device 31... Storage device 32... Processing device 40... Transport device 41... Exclusive tray 42... Support stand 50... Imaging unit

Claims (9)

An egg checking device for fertilized eggs used for vaccine production,
A first discrimination that determines whether the egg to be inspected is an infertile egg or a discontinued egg using a first trained model created in advance using a first learning data group classified based on the state of blood vessels of the egg as teacher data. means and
Determine whether the egg to be inspected is an upside-down egg or an egg with a defective air chamber, using a second trained model created in advance using a second learning data group classified based on the state of the air chamber of the egg as teaching data. a second determining means for
A third discriminating means for discriminating whether the egg to be inspected is an underdeveloped egg, using a third trained model created in advance using a third learning data group classified based on the state of egg development as teacher data. and,
A fourth determining means for determining whether the egg to be inspected is an egg with cracks, using a fourth learned model created in advance using a fourth learning data group classified based on the presence or absence of cracks in the egg as teacher data. and,
an imaging means for capturing an image of the egg to be inspected by irradiating the egg to be inspected;
control means;
The control means executes the first to fourth discrimination means based on the image of the egg to be inspected, thereby determining whether the egg to be inspected is a normal egg or a defective egg. An egg checking device that identifies the cause of defective eggs. An egg checking device for fertilized eggs used for vaccine production,
Determine whether the egg to be inspected is a discontinued egg using a first trained model created in advance using a learning data group classified based on the state of the blood vessels of the egg as a feature for determining discontinued eggs as training data. a first determining means;
A second trained model that determines whether the egg to be inspected is an infertile egg using a second trained model created in advance using a learning data group classified based on the state of the blood vessels of the egg as a feature for determining infertility as training data. Discrimination means;
Using a third trained model created in advance using a group of learning data classified based on air chamber conditions as characteristics for determining eggs with air chamber defects as training data, determine whether eggs to be inspected are eggs with air chamber defects. a third discriminating means for discriminating;
A fourth trained model is used to determine whether the egg to be inspected is an upside down egg, using a fourth trained model created in advance using a training data group classified based on the state of the air chamber as a feature for determining an upside down egg. 4 discrimination means;
Determine whether the egg to be inspected is a poorly developed egg by using a fifth trained model created in advance using a learning data group classified based on the state of the air chamber as a feature for determining a poorly developed egg as training data. a fifth determining means for
a sixth discriminating means for discriminating whether the egg to be inspected is an egg with cracks, using a sixth trained model created in advance using a learning data group classified based on the presence or absence of cracks in the egg as teacher data;
a seventh discriminating means for discriminating whether the egg to be inspected is a normal egg using a seventh trained model created in advance using a learning data group classified as teacher data for discriminating normal eggs;
an imaging means for capturing an image of the egg to be inspected by irradiating the egg to be inspected;
control means;
The control means executes the first to seventh discrimination means based on the image of the egg to be inspected, thereby determining whether the egg to be inspected is a normal egg or a defective egg. An egg checking device that identifies the cause of defective eggs. The egg checking device according to claim 1 or 2, wherein the control means is capable of determining a plurality of defective factors in one egg to be inspected. Any one of claims 1 to 3, further comprising a support stand on which the egg to be inspected is placed, and by rotating the support stand, the egg to be inspected can be imaged from a plurality of directions by the imaging means. The egg checking device described. a dark room in which the egg to be inspected is placed during imaging by the imaging means; a cylindrical member that abuts the egg to be inspected from above; and a cylindrical member that irradiates light from above onto the egg to be inspected from within the cylindrical member. The egg checking device according to any one of claims 1 to 4, comprising a first lighting device and a second lighting device that irradiates light to the egg to be inspected from below. The egg checking device according to claim 5, further comprising a lifting device that moves up and down the support base in order to place the egg to be tested in the dark room. The egg checking device according to any one of claims 1 to 6, wherein the control device has a function of totalizing the frequency of occurrence of each defect factor. to the computer,
Irradiating light to an egg to be inspected, which is a fertilized egg used for vaccine production, to obtain an image of the egg to be inspected;
Using a first trained model created in advance using a first learning data group classified based on the state of the blood vessels of the egg as teacher data, determine whether the egg to be inspected is an infertile egg or a discontinued egg;
Using a second trained model created in advance using a second learning data group classified based on the state of the air chamber of the egg as teacher data, determine whether the egg to be inspected is an upside-down egg or an egg with a defective air chamber. determine,
Determining whether the egg to be inspected is a poorly developed egg using a third learned model created in advance using a third learning data group classified based on the state of egg development as teacher data as teacher data;
By using a fourth trained model created in advance using a fourth learning data group classified based on the presence or absence of cracks in the egg as teacher data, it is determined whether the egg to be inspected is an egg with cracks. ,
An egg checking program that determines whether the egg to be inspected is a normal egg or a defective egg, and executes a process of identifying a defect factor of the defective egg. using a computer,
Irradiating light to an egg to be inspected, which is a fertilized egg used for vaccine production, to obtain an image of the egg to be inspected;
Using a first trained model created in advance using a first learning data group classified based on the state of the blood vessels of the egg as teacher data, determine whether the egg to be inspected is an infertile egg or a discontinued egg;
Using a second trained model created in advance using a second learning data group classified based on the state of the air chamber of the egg as teacher data, determine whether the egg to be inspected is an upside-down egg or an egg with a defective air chamber. determine,
Determining whether the egg to be inspected is a poorly developed egg using a third learned model created in advance using a third learning data group classified based on the state of egg development as teacher data;
By using a fourth learned model created in advance using a fourth learning data group classified based on the presence or absence of cracks in the egg as teacher data, it is determined whether the egg to be inspected is an egg with cracks. ,
An egg checking method for determining whether the eggs to be inspected are normal eggs or defective eggs, and identifying a defective factor of the defective eggs.