JP2021060204A - Inspection device - Google Patents

Abstract

To provide an inspection device making a burden to the device small and having good workability when collecting learning data, executing machine learning of a learning model, and executing an inspection using the learning model.SOLUTION: The inspection device comprises: an inspection unit 110 which includes an electromagnetic wave irradiation part 111 for irradiating an inspection target, a transmission amount detection part 113 for detecting a transmission amount of the inspection target, a storage part 114 for storing the distribution data of the transmission amount detected by the transmission amount detection part 113, and an abnormality detection part 115 for detecting a predetermined abnormality in the inspection target based on the distribution data of the transmission amount of the inspection target; a CPU 121 which executes control of the inspection unit 110; and a GPU 122 which executes processing of generating a learning model for detecting a predetermined abnormality in the inspection target under the control of CPU 121 by machine learning using the distribution data of the transmission amount of the inspection target accumulated in the storage part 114 as learning data, in which the abnormality detection part 115 detects the predetermined abnormality in the inspection target by inputting the distribution data of the transmission amount of the inspection target into the learning model.SELECTED DRAWING: Figure 1

Description

The present invention relates to an inspection device that irradiates an object to be inspected with an electromagnetic wave and inspects a foreign substance or the like based on the amount of transmission.

When the inspection target is irradiated with electromagnetic waves such as X-rays, ultraviolet rays, visible rays, infrared rays, or microwaves and the transmitted amount is detected, the strength of the transmitted amount due to the presence or absence of foreign matter in the inspection object and the material of the foreign matter is determined. A distribution is obtained. By generating a two-dimensional image in which this intensity distribution is expressed by the color tone of each pixel (tone of color tone, shade, intensity, etc.), the situation inside the inspection object that cannot be seen from the outside is visualized. be able to.

As described above, the strength of the permeation amount depends on the presence or absence of foreign matter in the inspection object, the material of the foreign matter, etc., but depending on the combination of the foreign matter and the non-foreign matter, the difference in the permeation amount between the two is very small and is generated. It may be difficult to distinguish between the two from the two-dimensional image.

As a solution to such a problem, a method using a machine-learned learning model has recently been proposed. For example, a machine is used to collect a large amount of data such as an image showing the distribution of the permeation amount of an inspection object including a foreign substance to be detected, and use this as learning data to detect the foreign substance contained in the inspection object. Generated by learning. By inputting data such as an image showing the distribution of the transmission amount of the inspection object into this learning model, it becomes possible to distinguish foreign matter from non-foreign matter and detect it with high accuracy.

For example, by executing machine learning using an X-ray inspection image of a product in which a plurality of articles are overlapped as a learning image and X-ray inspection of the product using the feature amount obtained thereby, the products overlap each other. Patent Document 1 discloses an inspection device capable of suppressing a decrease in inspection accuracy of a product containing a plurality of articles.

In addition, a method has also been proposed in which the learning data collected by the inspection device is machine-learned by another computer device, and the learning model obtained is applied to the inspection device again to perform the inspection.

Japanese Patent No. 6537008

In the inspection device described in Patent Document 1, a CPU (Central Processing Unit) is responsible for both collection of learning data, execution control of inspection, and machine learning processing. Therefore, when the machine learning process is executed by the CPU, the process requires power and time, and during that time, other processes cannot be performed or the processing speed is significantly reduced.

In the method of executing machine learning on another computer device, it is troublesome to output the learning data collected by the inspection device once and input it to another computer device, or input the learning model obtained by machine learning to the inspection device. It is necessary to change the configuration of the inspection device so that the learning model input from the outside can be applied in the inspection. In addition, workability is not good because operations are required for each of the inspection device and the computer device.

An object of the present invention is to provide an inspection apparatus having a small load on the apparatus and good workability in collecting learning data, machine learning of a learning model, and performing an inspection using the learning model.

In the inspection device of the present invention, an electromagnetic wave irradiation unit that generates an electromagnetic wave and irradiates the inspection object, a transmission amount detection unit that detects the distribution of the transmission amount of the electromagnetic wave transmitted through the inspection object, and a transmission amount detection unit detect the electromagnetic wave. Control of an inspection unit having a storage unit for accumulating transmission amount distribution data and an abnormality detection unit for detecting a predetermined abnormality in the inspection object based on the transmission amount distribution data of the inspection object, and an inspection unit. An inspection device including a CPU for detecting a predetermined abnormality in the inspection object by using the distribution data of the permeation amount of the inspection object having a predetermined abnormality accumulated in the storage unit as learning data. It is further equipped with a GPU (Graphics Processing Unit) that executes the process of generating the learning model by machine learning under the control of the CPU, and the abnormality detection unit inputs the distribution data of the permeation amount of the inspection object into the learning model. , Detects a predetermined abnormality in the inspection object.

The learning model outputs an inferred value indicating the possibility of a predetermined abnormality by inputting the distribution data of the permeation amount of the inspection object, and the abnormality detection unit detects the presence or absence of a predetermined abnormality based on the inferred value. May be configured to determine.

The inspection unit and the CPU may form a main body, and a GPU provided outside the main body may be connected to the main body via a connection interface further provided on the main body.

The inspection device of the present invention includes a single GPU in addition to the CPU, and the GPU is responsible for machine learning processing. Since a single GPU can execute machine learning processing at high speed and with a low load, the heat generation time can be shortened and the heat generation amount can also be reduced. Therefore, the load on each part of the device including the CPU can be reduced.

In addition, since the machine learning processing function is implemented in the inspection device, all operations related to the collection of learning data, machine learning of the learning model, and the execution of the inspection using the learning model can be performed in the inspection device. Good workability.

It is a figure which shows the functional structure example of the inspection apparatus 100. It is a figure which shows the physical structure example of the inspection part 110. It is a figure which shows the functional structure example of the inspection apparatus 200.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. In the following description and drawings, the same functional parts are designated by the same reference numerals, and the functional parts once described will be omitted or described to the extent necessary.

FIG. 1 is a diagram showing a functional configuration example of the inspection device 100 of the present invention. The inspection device 100 includes an inspection unit 110, a CPU 121, a GPU 122, a display unit 131, and an input unit 141. Further, the inspection unit 110 includes an electromagnetic wave irradiation unit 111, a transport unit 112, a transmission amount detection unit 113, a storage unit 114, and an abnormality detection unit 115. FIG. 2 shows an example of the physical configuration of the inspection unit 110.

The electromagnetic wave irradiation unit 111 irradiates an inspection object W mounted on the transport unit 112 and transported in the Y-axis direction in FIG. 2 with electromagnetic waves such as X-rays, ultraviolet rays, visible rays, and infrared rays. It is the source.

The transport unit 112 is an arbitrary transport mechanism that transports the inspection object W placed on the transport surface forming the XY plane in the three-dimensional Cartesian coordinate system at a predetermined speed in the Y-axis direction. It is desirable that the transport unit 112 has high electromagnetic wave transmission so that the electromagnetic wave transmitted through the inspection object W reaches the transmission amount detection unit 113 without being attenuated as much as possible.

The transmission amount detection unit 113 detects and outputs the transmission amount of the electromagnetic wave transmitted through the inspection object W. The method of configuring the transmission amount detection unit 113 is arbitrary, and may be configured as, for example, a line sensor composed of a plurality of detection elements arranged in the X-axis direction. In this case, the amount of electromagnetic waves transmitted from the inspection target object W is repeatedly detected and output at a cycle corresponding to the transport speed by the transport unit 112 until the entire inspection target object W has passed. As a result, information on the amount of transmission detected by each detection element linearly in the X-axis direction is obtained in each cycle, and information on each cycle while the inspection object W passes over the line sensor is collected. It is possible to obtain the distribution of the permeation amount for the entire inspection object W.

The storage unit 114 stores the distribution data of the permeation amount in the inspection object W detected by the permeation amount detection unit 113. When machine learning of the learning model is performed, a plurality of distribution data of the permeation amount of the inspection target W including a predetermined foreign substance are accumulated prior to the machine learning.

Further, the storage unit 114 stores in advance a program in which various functions realized by executing the program by the CPU 121 are described.

For the storage unit 114, for example, in addition to a storage medium such as an HDD or a flash memory, a non-volatile memory, a volatile memory, or the like can be adopted. A part or all of the storage unit 114 may be provided outside the inspection device 100 by a cloud storage or the like connected via a communication unit provided in the inspection device 100.

The abnormality detection unit 115 detects a predetermined abnormality in the inspection target W based on the distribution data of the permeation amount of the inspection target W.

Specifically, for example, the magnitude of the transmission amount of each detection element detected linearly in the X-axis direction by the transmission amount detection unit 113 is expressed by the color tone of each pixel (tone such as lightness / darkness / shade / intensity of color). By generating the linear images formed for each period and arranging them in chronological order in the Y-axis direction, a transparent image of the entire inspection object W is generated.

Then, the generated transparent image is input to a learned learning model that has undergone machine learning to detect a predetermined abnormality in the inspection target object W, thereby detecting the predetermined abnormality in the inspection target object.

Predetermined abnormalities include, for example, when the inspection target W is a package, the presence of foreign matter inside, an abnormality in the shape of the contents, biting of the contents into the seal portion, and the like, as well as the inspection target W. Abnormal shape of itself, cracks, chips, etc. can be mentioned.

The abnormality may be determined, for example, by inputting an image to the learning model so that an inferred value indicating the possibility of an abnormality is output, and the output inferred value is compared with a predetermined threshold value. More specifically, for example, an inference value is defined in the range of 0 to 1 in which the case where the possibility of abnormality is extremely low is 0 and the case where the possibility of abnormality is extremely high is 1, and the output inference value is When it is a certain value (for example, 0.6) or more, it may be determined that there is an abnormality.

The abnormality detection unit 115 collates the generated transparent image with the image requirements for detecting a predetermined abnormality (for example, the color tone of each pixel, the mode of change in color tone between pixels, etc.), if necessary. Alternatively, it may be configured so that a predetermined abnormality can be detected.

The abnormality detection unit 115 may be realized as hardware as a single functional unit, or is realized by storing a program in which the above functions are described in the storage unit 114 in advance and executing the CPU 121. You may.

The CPU 121 reads various programs for functionally operating the inspection unit 110 from the storage unit 114 and executes them to realize the functional operations described in the programs.

The GPU 122 uses the distribution data of the transmission amount of the inspection object W having a predetermined abnormality accumulated in the storage unit 114 as learning data, and the difference in the color tone of each pixel and the mode of the change in the color tone between the pixels. Machine learning is performed based on the difference between the two, and a learning model for detecting a predetermined abnormality in the inspection object W is generated. This function is realized under the control of the CPU 121. Specifically, the CPU 121 reads a program in which the control content for realizing the function is described from the storage unit 114 and executes the control to control the GPU 122, whereby the function is realized.

The display unit 131 is a display means for displaying an input interface for inputting various instructions in the inspection device 100, an inspection status, a foreign matter detection result, and the like under the control of the CPU 121. The display unit 131 may be built in the main body of the inspection device 100 or may be externally attached.

The input unit 141 is an input means such as a pointing device and a keyboard for the device user to input information as needed. A touch panel display may be adopted as the display unit 131, and this may be used as the input unit 141.

In the inspection device 100 having the above configuration, for example, the device is started as follows to execute inspection and the like.

When the inspection device 100 is started, a program for displaying an input interface in which items to be operated in the device in a list or step by step is displayed in the display unit 131 is stored in the storage unit 114 in advance. As a result, when the inspection device 100 is started, the program is read from the storage unit 114 and executed by the CPU 121, and the input interface is displayed on the display unit 131.

The display program of the input interface is, for example, a program in which a function corresponding to the item stored in advance in the storage unit 114 is described by selecting and inputting from the input unit 141 to a button or a check box provided for each item. Is read and is described so as to be executed by the CPU 121 or the GPU 122 according to the item.

Specifically, when the selected item is the collection of learning data, the detection of foreign matter using the learning model, or the like, the program corresponding to the item read from the storage unit 114 is executed by the CPU 121. It is described as follows. Further, when the selected item is the generation of a learning model by machine learning processing from the collected learning data, the program corresponding to the item read from the storage unit 114 is executed by the GPU 122. Describe it.

As a result, the machine learning process can be performed by the GPU, and the other processes can be performed by the CPU.

In the inspection device 100 of the present invention described above, a single GPU 122 is provided in addition to the CPU 121, and the GPU 122 is responsible for the machine learning process. Since a single GPU can execute machine learning processing at high speed and with a low load, the heat generation time can be shortened and the heat generation amount can also be reduced. Therefore, the load on each part of the device including the CPU can be reduced.

Further, since the inspection device 100 is equipped with a machine learning processing function, the inspection device can perform all operations related to the collection of learning data, the machine learning of the learning model, and the execution of the inspection using the learning model. It can be done and workability is good.

The present invention is not limited to the above embodiments. The above-described embodiment is an example, and any object having substantially the same configuration as the technical idea described in the claims of the present invention and exhibiting the same effect and effect is the present invention. It is included in the technical scope of the invention. That is, it can be appropriately changed within the scope of the technical idea expressed in the present invention, and the form in which such a change or improvement is added is also included in the technical scope of the present invention.

For example, the GPU 122 does not necessarily have to be built in the inspection device 100. Specifically, for example, as shown in FIG. 3, a connection interface 151 (for example, Thunderbolt (registered trademark)) for pulling out a bus (for example, PCI-Express (registered trademark)) in the inspection device 100 to the outside of the device instead of GPU 122 (for example, Thunderbolt (registered trademark)). ) 3 etc.) may be configured, and the inspection device 200 may be configured by combining with an external GPU 122 that can be attached to and detached from the connection interface 151.

According to the inspection device 200, even if a defect occurs in the GPU 122, it can be easily repaired or replaced with a normal product. In addition, with the GPU 122 as an optional item, it is possible to take measures such as additional sales and lending as necessary.

Further, in the case of the inspection device 100 having the GPU in the inspection device, it is possible to suppress the occurrence of problems due to heat generation as compared with the case where the CPU is responsible for the learning function. Air cooling is required to suppress the temperature rise. However, for example, when the inspection device is used in an environment where powder is scattered, such as a food factory, there is a concern that the inspection device may be damaged due to dust due to the intake of outside air during air cooling. According to the inspection device 200, the connection interface 151 is usually closed with a cap or the like to keep the main body 101 in a sealed state, and the GPU 122 is connected only during learning. Therefore, dust can enter the inspection device. It can be prevented as much as possible.

100, 200 ... Inspection device 110 ... Inspection unit 111 ... Electromagnetic wave irradiation unit 112 ... Transport unit 113 ... Transmission amount detection unit 114 ... Storage unit 115 ... Abnormality detection unit 121 ... CPU
122 ... GPU
131 ... Display unit 141 ... Input unit 151 ... Connection interface W ... Inspection object

Claims (3)

An electromagnetic wave irradiation unit that generates an electromagnetic wave and irradiates the inspection object, a transmission amount detection unit that detects the distribution of the transmission amount of the electromagnetic wave that has passed through the inspection object, and the inspection object detected by the transmission amount detection unit. A storage unit that stores the distribution data of the permeation amount in the above, and an abnormality detection unit that detects a predetermined abnormality in the inspection object based on the distribution data of the permeation amount of the inspection object.
Inspection department equipped with
The CPU that executes the control of the inspection unit and
It is an inspection device equipped with
Using the distribution data of the permeation amount of the inspection object having the predetermined abnormality accumulated in the storage unit as learning data, a learning model for detecting the predetermined abnormality in the inspection object is generated by machine learning. A GPU that executes the processing to be performed under the control of the CPU is further provided.
The abnormality detection unit is an inspection device that detects the predetermined abnormality in the inspection target by inputting distribution data of the permeation amount of the inspection target into the learning model. The learning model outputs an inferred value indicating the possibility of the predetermined abnormality by inputting the distribution data of the permeation amount of the inspection object.
The inspection device according to claim 1, wherein the abnormality detection unit determines the presence or absence of the predetermined abnormality based on the inferred value. The claim is characterized in that the inspection unit and the CPU constitute a main body, and the GPU provided outside the main body is connected to the main body via a connection interface further provided in the main body. The inspection device according to 1 or 2.

Images