JP2023023312A - Program, method and information processing system for estimating resource value of electronic apparatus or substrate - Google Patents

Abstract

To provide a program, a method, and an information processing system capable of highly accurately estimating a resource value of an electronic apparatus or a substrate included therein.SOLUTION: An estimation method comprises steps of: (A) identifying a manufacturer of an electronic apparatus; (B) identifying, on the basis of the identified manufacturer, one of more type of one or more components to be focused on among components mounted on a substrate included in the electronic apparatus; (C) extracting areas of components belonging to the one or more types of one or more components to be focused on from an image of the substrate; and (D) estimating, on the basis of a total area of areas of the extracted components and the manufacturer, a resource value of the electronic apparatus or the substrate.SELECTED DRAWING: Figure 4

Description

The present invention relates to technology for estimating the resource value of electronic equipment or substrates.

Some techniques already exist for non-destructively estimating the resource value of discarded electronic devices. For example, Patent Document 1 describes the content contained in the substrate for each resource (that is, element) such as gold, silver, copper, platinum, etc., the total area on the substrate for each type of electronic component, and the corresponding coefficient It is described that it can be calculated by the sum of products with If the content of each element is known, the resource value can be calculated from the unit price of the element. However, none of the prior arts focuses on differences between electronic device manufacturers. In other words, the relationship between the total area on the board and the resource value for each type of electronic component varies greatly depending on the manufacturer of the electronic equipment. Value cannot be estimated.

JP 2015-222511 A JP 2015-232449 A JP 2016-218979 A JP 2017-106792 A JP 2017-133850 A JP 2020-190881 A JP 2015-222510 A

Accordingly, an object of the present invention, according to one aspect, is to provide a technique that enables more accurate estimation of the resource value of an electronic device or a substrate included therein.

The estimating method according to the present invention includes (A) the step of identifying the manufacturer of the electronic device; (C) an extraction step of extracting an area of a component belonging to the type of the component of interest from the image of the substrate; (D) based on the total area of the extracted component area and the manufacturer , and estimating the resource value of the electronic device or substrate.

According to one aspect, it becomes possible to more accurately estimate the resource value of an electronic device or a substrate included therein.

FIG. 1 is a diagram showing the relationship between the Au value per electronic device model (one substrate) and the total resource value. FIG. 2 is a diagram showing the relationship between the Au concentration of each component and the ratio (distribution ratio) to the amount of Au contained in the substrate when the component is mounted. FIG. 3A is a diagram showing the relationship between the sum of the upper surface areas of ICs and the Au value. FIG. 3B is a diagram showing the relationship between the sum of the top surface areas of the connectors and the Au value. FIG. 3C is a diagram showing the relationship between the sum of the upper surface areas of the IC+connector and the Au value. FIG. 4 is a diagram showing the configuration of the information processing apparatus according to this embodiment. FIG. 5 is a diagram showing an example of a type selection table. FIG. 6A is a diagram showing a processing flow according to this embodiment. FIG. 6B is a diagram showing a processing flow according to this embodiment. FIG. 7 is a diagram for explaining a specific example of IC extraction processing. FIG. 8 is a diagram illustrating a functional configuration example of a computer that is an information processing apparatus.

[Premise of the embodiment]
Electronic devices such as home appliances include substrates (generally printed circuit boards) on which a plurality of components are mounted. Mounted parts include memories, processors and other ICs (Integrated Circuits), transistors and diodes, various capacitors, coils and transformers, crystal oscillators, connectors and the like. Such substrates contain various elements. For example, the inventor of the present application collected 26 models of discarded digital cameras (5 manufacturers, manufactured from 2000 to 2014). They were dismantled manually to take out a plurality of printed circuit boards, pulverized with a small cutter mill, roasted in an air atmosphere (450° C., 4 hours), and then subjected to fluorescent X-ray analysis. In addition, pressure acid decomposition was performed for trace element analysis, and ICP emission spectrometry and mass spectrometry were performed. The weight of each element obtained after the elemental analysis was multiplied by the elemental price as of April 30, 2020) to calculate the price of the substrate included in one model of discarded digital camera. (Au: 6411 yen/g, Ag: 56.8 yen/g, Cu: 0.56 yen/g, Ta: 16.4 yen/g, Pd: 7207 yen, etc.)

FIG. 1 shows the relationship between the value of Au (Au element price×weight) per model of electronic device (1 to 5 substrates included) and the total resource value. The total resource value varied between ¥30 and ¥492, with an average value of ¥187. The plots are distributed almost linearly (the square value of the correlation coefficient R (contribution rate) R ² =0.990), and a strong positive correlation is recognized between the value of Au and the total resource value. Since the slope of the regression line is 1.11, even if the ratio of Au values in any value substrate is uniformly estimated to be about 90.0%, almost no value error occurs as a whole. , is understood. That is, there is no problem in terms of accuracy even if attention is paid only to Au.

In addition, the inventor of the present application peeled off all the electronic parts mounted on 18 models of discarded digital cameras separately obtained for verification from the substrate, measured the weight ratio for each part type, and performed the above elemental analysis. FIG. 2 shows the result of calculating the Au concentration of each component and the proportion of Au contained in the substrate when the component was mounted as a distribution ratio from the results of the elemental analysis. Parts with a high weight ratio are aluminum electrolytic capacitors (19.3%), ports (9.3%), IC/memory (8.6%), etc., and board parts without parts (glass epoxy resin board + flexible board ) accounts for 33.8%. The components with the highest Au concentration were IC/memory (8960ppm), transistor/diode (5500ppm), and crystal oscillator (4000ppm) in that order. was overwhelmingly high (66.0%), followed by connectors (7.52%). Considering the distribution rate of only the parts excluding the board part, the two types of IC/memory and connector exceeded 80%. Thus, if attention is paid only to Au, attention is paid to ICs (including memories and processors) and connectors, which is sufficient in terms of accuracy.

Furthermore, the inventor of the present application believes that the Au value depends not only on the number of ICs and connectors but also on their size. predicted that the value could be estimated. 3A to 3C show the sum of the upper surface areas of the IC, connector, and IC+connector mounted on the board included in each of 26 models of discarded digital cameras manufactured by 5 manufacturers, and the Au value per model. indicate the relationship respectively. Since the number of ICs and connectors with a long side size of less than 2 mm is small, and it is thought that their proportion in the total top surface area is small, ICs and connectors with a length of 2 mm or more were targeted.

From FIGS. 3A to 3C, there is a positive correlation between the sum of the top surface areas (also called total areas) of ICs and connectors and the Au value, indicating that the larger the sum of areas, the higher the Au value. Specifically, as shown in FIG ^. 3A, the correlation with the top surface area of the IC is as follows. In the case of , it was shown to be around 0.6 to some extent. On the other hand, the IC upper surface area data of maker E remained in a narrow range, and no correlation was found.

Also, as shown in FIG. 3B, the correlation with the top surface area of the connector is high for all except the case of manufacturer E, with the contribution rate of manufacturer B being 0.96, followed by manufacturers A, C and D being 0.85 to 0. It was about 0.9. On the other hand, although there was data in a relatively wide range, only maker E did not show a clear positive correlation. Furthermore, as shown in FIG. 3C, regarding the correlation with the top surface area of the IC + connector, although the contribution rate was slightly lower in the case of manufacturer C (0.67), manufacturers A, B, and D were 0.94 or more. It was found that a very strong positive correlation of On the other hand, in the case of manufacturer E, a rather negative correlation was suggested.

As described above, the inventor of the present application has obtained a new finding that the types of components to be focused on (IC/connector/IC+connector) differ depending on the manufacturer. In particular, if there is a maker E that exhibits a negative correlation, it is impossible to estimate the resource value with high accuracy without the parameter of the maker. Below, the configuration for calculating the resource value will be described in detail based on the above assumptions.

[Configuration example of the embodiment]
As shown in FIG. 4, information processing apparatus 1000 that executes processing according to the present embodiment includes image acquisition unit 102, image storage unit 104, manufacturer identification unit 106, type selection unit 108, type selection table It includes a storage unit 110 , an IC extraction unit 112 , a connector extraction unit 114 , a value calculation unit 116 , a value calculation database (DB) 118 , an output unit 120 and a processing result storage unit 122 .

The image acquisition unit 102 acquires an image of the substrate from, for example, a digital camera, and stores it in the image storage unit 104 . The manufacturer identification unit 106 identifies the manufacturer of the electronic device including the board using various techniques. For example, the method described in Japanese Patent Laid-Open No. 2019-81971 may be adopted, the feature amount is extracted from the image of the electronic device or substrate to identify the manufacturer name or manufacturer identifier, or the manufacturer name or manufacturer identifier is provided to the user. may be input. Furthermore, it may be specified using a learned model that has undergone machine learning so as to output the manufacturer identifier from the image of the electronic device or circuit board.

The type selection unit 108 selects the type of component to be focused on from the type selection table storage unit 110 based on the manufacturer of the electronic device. The type selection table storage unit 110 stores, for example, a table as shown in FIG. In the table, component types to be focused on are registered in association with manufacturer names. From the results shown in FIGS. 3A to 3C, focusing on the contribution rate, the component type "IC" is registered in association with manufacturer A, and the component type "IC+connector" is registered in association with manufacturer B. The part type "connector" is associated with the manufacturer C, the part type "IC+connector" is registered with the manufacturer D, and the part type "IC+connector" is associated with the manufacturer E. is registered. Manufacturers having the same characteristics may be grouped, or general-purpose data may be registered when the manufacturer is unknown.

Based on the selection result of the type selection unit 108, if the type of the component to be focused on includes an IC, the IC extraction unit 112 extracts the area of the IC from the board image stored in the image storage unit 104. is extracted to calculate the total area (also referred to as the total area) of the region. Based on the selection result of the type selection unit 108, the connector extraction unit extracts the region of the connector from the image of the board stored in the image storage unit 104 when the type of the component to be focused on includes a connector. Extract and calculate the total area of the region.

The value calculation unit 116 calculates the specified manufacturer name or manufacturer identifier, the IC extraction unit 112 , the connector extraction unit 114 , and the resource value at the time of disposal, and stores them in the processing result storage unit 122 . The value calculation DB 118 stores, for example, mathematical formulas representing the relationship between the total area and the Au value as shown in FIGS. 3A to 3C, and the relationship between the Au value and the substrate resource value as shown in FIG. Formulas are stored for each manufacturer. However, it is also possible to store formulas obtained by integrating these formulas and data such as tables corresponding to the formulas. Further, when calculating not only the resource value of the substrate but also the resource value of the entire electronic device, a mathematical formula or the like for calculating the resource value of the entire electronic device from the resource value of the substrate may be stored. Since it is known that the resource value of the substrate is a predetermined ratio (for example, 90%) of the resource value of the entire electronic device, it is possible to estimate the resource value of the entire electronic device from the resource value of the substrate. .

The output unit 120 outputs the resource value stored in the processing result storage unit 122 to the display device or printing device of the information processing device 1000, other computers connected to the network, and the like.

Next, processing contents of the information processing apparatus 1000 will be described in detail with reference to FIGS. 6A to 7. FIG. Note that the processing of FIGS. 6A and 6B is performed for each substrate.
First, the image acquisition unit 102 acquires an image of the substrate to be processed taken by a digital camera or the like, and stores it in the image storage unit 104 (FIG. 6A: step S1). The information processing apparatus 1000 may be connected to a digital camera and acquire images from the digital camera, or may acquire images via a computer or the like connected to another network.

Next, the manufacturer identification unit 106 identifies the manufacturer of the electronic device including the substrate using various techniques as described above (step S3). For example, by preparing a file in which data of manufacturer names or manufacturer identifiers of electronic devices are registered in the order of substrates whose resource values are to be estimated, each time the processing of FIGS. 6A and 6B is executed from the file, It may be in the form of reading the name or manufacturer identifier.

The type selection unit 108 selects the type of the part to be focused on by referring to the type selection table of the specified manufacturer (step S5). If a type selection table as shown in FIG. 5 is held, IC, IC+connector (both), or connector can be identified as the type of component to be focused on from the manufacturer name or manufacturer identifier.

Then, if the type of the component to be focused on is an IC or both (IC+connector) (step S7: Yes route), the IC extraction unit 112 extracts the IC from the image stored in the image storage unit 104. An extraction process is executed (step S9). The IC extraction process may be performed by image analysis, or may be performed by a trained model machine-learned so as to extract the IC area from the image of the substrate.

When extracting the IC region by image analysis, the following processing is performed. That is, color binarization based on the HSL (Hue/Saturation/Luminance) component distribution of the IC characteristic color is performed. For example, binarization is performed by setting 1 in the ranges of hue: 20 to 200, saturation: 0 to 70, and brightness: 25 to 95, and setting the others to 0. For example, when processing an image of a substrate as shown in the upper left end of FIG. 7, color binarization is performed to obtain the second image from the upper left. Most of the ICs enclosed by dotted lines are extracted, but many parts other than ICs are also included. Next, average filtering is performed to replace pixel values in a pixel region of a predetermined size (eg, 7×7 pixels) with their average values. In the example of FIG. 7, if the second image from the left in the upper row is subjected to averaging filtering, the third image from the left in the upper row is obtained. As a result, the outline of the IC can be clarified, and noise generated at terminals other than the IC can be removed.

Next, a filling process (Filling) is executed for a minute area included inside the area of the IC at the present stage. In the third image from the top left in FIG. 7, the IC name printed on the surface of the IC can be seen, but when the hole-filling process is executed, the IC name etc. are buried in the rightmost image in the top row of FIG. 7 and cannot be seen. there is In addition, a second averaging filtering is performed to remove fine areas (grains) outside the IC. When the second averaging filtering for replacing the pixel values in a pixel area of a predetermined size (for example, 9×9 pixels) with their average value is performed on the image on the right end of the upper row in FIG. An image is obtained.

Filtering is then performed based on features of the shape (including size) of the IC to be detected. Specifically, a long side size range (for example, 2-5 mm, 5-10 mm, 10-15 mm, 15 mm-30 mm) is set, and long side size filtering is performed to remove the rest. When long side size filtering is performed on the leftmost image in the lower row of FIG. 7, the second image from the left in the lower row of FIG. 7 is obtained. Furthermore, aspect ratio filtering is performed by specifying a range of aspect ratios (eg, 1 to 3.5) and removing the rest. By performing aspect ratio filtering on the second image from the bottom left in FIG. 7, the third image from the bottom left in FIG. 7 is obtained. In addition, a range (for example, 0.6-1) of density (the ratio of the area of the extraction region to its circumscribing rectangle) is specified, and density filtering is performed to remove the rest. By performing compactness filtering on the third image from the left in the bottom row of FIG. 7, the rightmost image in the bottom row of FIG. 7 is obtained. Such processing allows areas such as overly large areas, high aspect ratio areas, low density cable sections, etc., to be removed.

Calculate the total area of the regions thus extracted. Note that the numerical values described above are merely an example, and are set based on the substrate to be processed.

Furthermore, in the case of a trained model machine-learned to extract the IC region from the substrate image, for example, learning obtained by performing supervised learning processing on the residual network (Residual Network) Use a ready-made model. More specifically, we use the well-known residual network ResNet50. In addition, when 2367 IC images on the printed circuit board of a digital camera were learned and IC detection was performed on 263 unlearned circuit board images, the detection rate was 97.3%. The accuracy rate was 90.1%. It should be noted that other types of neural networks can be used instead of residual networks.

Returning to the description of FIG. 6A, after step S9, if the types of components to be focused on are not both (IC+connector) (step S11: No route), the process proceeds to FIG. 6B via terminal A. On the other hand, if the type of the component to be focused on is neither IC nor both (IC+connector) (step S7: No route), or if the type of the component to be focused on is both (IC+connector) (step S11: Yes root), the connector extraction unit 114 executes connector extraction processing on the image stored in the image storage unit 104 (step S13). After that, the process shifts to the process of FIG. 6B via terminal A. FIG.

The basic structure of the IC extraction process is the same for the connector extraction process. That is, color binarization, averaging filtering, hole-filling processing, second averaging filtering, long-side size filtering, aspect ratio filtering, and density filtering based on the HSL component distribution of the characteristic colors are performed with appropriate parameters for the connector. Configure and run. Since the colors of connectors are various, the connector extraction process may be executed for each color. Furthermore, if the same color has different sizes, the connector extraction process may be executed for each color and size.

Furthermore, in the case of a trained model that has been machine-learned to extract the connector region from the board image, for example, the learning obtained by performing supervised learning processing on the residual network (Residual Network) Use a ready-made model. More specifically, we use the well-known residual network ResNet50. After learning 1,350 connector images of the printed circuit board of a digital camera, IC detection was performed on 150 untrained images of the circuit board. The detection rate was 95.9%, and the detected IC was correct The rate was 97.3%. It should be noted that other types of neural networks can be used instead of residual networks.

Proceeding to the process of FIG. 6B, the value calculation unit 116 totals the areas of the regions extracted by the IC extraction unit 112 and the connector extraction unit 114 that executed the extraction processing (step S15). Then, the value calculation unit 116 reads the data for calculating the resource value corresponding to the specified manufacturer from the value calculation DB 118, and calculates the resource value corresponding to the total area based on the data for calculating the resource value. , is stored in the processing result storage unit 122 (step S17). As described above, the data for calculating the resource value may be a mathematical formula for calculating the resource value of the substrate, or may be a mathematical formula for calculating the resource value of the electronic device as a whole. There is also

Then, the output unit 120 outputs the resource value data stored in the processing result storage unit 122 to an output device such as a display device (step S19). By doing so, it becomes possible to accurately estimate the resource value of the board or the electronic device as a whole by reflecting the tendency of the manufacturer.

Although the embodiments of the present invention have been described above, the present invention is not limited to these. For example, the above embodiment may be implemented by removing any element. As for the processing flow, the order may be changed or multiple steps may be executed in parallel as long as the processing result does not change. Also, the functional block configuration of the information processing apparatus shown in FIG. 4 is merely an example, and may differ from the program module configuration, file configuration, and the like.

In some cases, the functions of the information processing apparatus 1000 are not realized by one information processing apparatus, but the functions of the information processing apparatus 1000 are realized by cooperation of a plurality of information processing apparatuses. In either case, the information processing apparatus 1000 may be called an information processing system.

The information processing apparatus 1000 described above is, for example, a computer apparatus, and as shown in FIG. A display control unit 2507 connected to a device 2509, a drive device 2513 for a removable disk 2511, an input device 2515, a communication control unit 2517 for connecting to a network, and peripheral devices (image sensor 100, attenuation mechanism 300, cooling mechanism 400 etc.) are connected via a bus 2519 to a peripheral device connection unit 2521 for connection. Note that the HDD may be a storage device such as a solid state drive (SSD). An operating system (OS) and an application program for performing processing in the embodiments of the present invention are stored in the HDD 2505 and read from the HDD 2505 to the memory 2501 when executed by the CPU 2503. . The CPU 2503 controls the display control unit 2507, the communication control unit 2517, and the drive device 2513 according to the processing content of the application program to perform predetermined operations. In addition, data in the middle of processing is mainly stored in the memory 2501, but may be stored in the HDD 2505 as well. For example, an application program for executing the processing described above is stored in a computer-readable removable disk 2511 and distributed, and installed from the drive device 2513 to the HDD 2505 . It may be installed in the HDD 2505 via a network such as the Internet and the communication control unit 2517 . Such a computer device implements the various functions described above through organic cooperation between hardware such as the CPU 2503 and memory 2501 described above and programs such as the OS and application programs. .

The embodiments described above are summarized as follows.

The estimation method according to the present embodiment includes (A) the step of identifying the manufacturer of the electronic device, (C) an extraction step of extracting an area of a component belonging to the type of the component of interest from the image of the substrate; (D) a total area of the extracted component area and the specified and estimating the resource value of the electronic device or substrate based on the manufacturer.

As described above, the types of components to be focused on differ depending on the manufacturer, and the relationship between the total area of the regions on the substrate and the resource value also differs. Improves estimation accuracy.

Further, in the step of specifying the type of the part to be focused on as described above, the specified manufacturer is associated with the specified manufacturer from the data storage unit that stores the correspondence relationship between the manufacturer and one or more types of the part to be focused on. A step of identifying one or more types of the part of interest may also be included. Preparing the correspondence relationship in advance enables more accurate resource value estimation.

Furthermore, there are cases in which the type of the component to be noted above is any one of an IC (Integrated Circuit), a connector, and an IC and a connector. Based on the price of Au, it is preferable to focus on the IC and the connector, but if the price fluctuates, other elements may be focused. In such a case, the IC and the connector may not be focused on. It is possible.

The above-described extraction step includes color binarization based on the HSL (Hue/Saturation/Luminance) component distribution of the characteristic color of the extracted part, averaging filtering, and the feature of the part shape for each type of part to be focused. It may also include a combination of predetermined filtering based on. When performing extraction by image analysis, appropriate extraction can be performed by performing such processing after setting appropriate parameters.

Note that the predetermined filtering based on the feature of the part shape may include filtering based on the long side size, filtering based on the aspect ratio, and filtering based on the density. With these, the parts to be extracted are narrowed down.

Furthermore, the extraction step described above may be performed using a learned model that has been machine-learned so as to specify a region of a part belonging to the type of the part for each type of the part of interest. . For example, it may be a trained model such as a residual network trained to extract a particular type of part from an image.

A program can be created to cause a computer to perform the methods described above, and the program can be stored in various storage media.

1000 information processing device 102 image acquisition unit 104 image storage unit 106 manufacturer identification unit 108 type selection unit 110 type selection table storage unit 112 IC extraction unit 114 connector extraction unit 116 value calculation unit 118 value calculation DB
120 output unit 122 processing result storage unit

Claims (8)

identifying the manufacturer of the electronic device;
identifying a type of a component to be focused on among components mounted on a board included in the electronic device based on the identified manufacturer;
an extracting step of extracting an area of a component belonging to the type of the component to be focused on from the image of the substrate;
estimating a resource value of the electronic device or the substrate based on the extracted total area of the regions of the components and the manufacturer;
A program that causes a computer to run The step of identifying the type of the part to be focused on includes:
identifying one or more types of the target part associated with the identified manufacturer from a data storage unit that stores a correspondence relationship between the manufacturer and one or more types of the target part. A program according to claim 1. 3. The program according to claim 1, wherein the type of the part to be focused on is any one of an IC (Integrated Circuit), a connector, and an IC and a connector. The extraction step comprises:
A combination of color binarization based on the HSL (Hue/Saturation/Luminance) component distribution of the characteristic color of the extracted part, averaging filtering, and predetermined filtering based on the features of the part shape is performed for each type of the part to be focused. 4. The program according to any one of claims 1 to 3, comprising: 5. The program according to claim 4, wherein the predetermined filtering based on part shape features includes filtering based on long side size, filtering based on aspect ratio, and filtering based on density. The extraction step comprises:
4. The program according to any one of claims 1 to 3, wherein for each type of the target part, the program is performed using a machine-learned learned model so as to specify a region of the part belonging to the type of the part. identifying the manufacturer of the electronic device;
identifying a type of a component to be focused on among components mounted on a board included in the electronic device based on the identified manufacturer;
an extracting step of extracting an area of a component belonging to the type of the component to be focused on from the image of the substrate;
estimating a resource value of the electronic device or the substrate based on the extracted total area of the regions of the components and the manufacturer;
and a computer-implemented estimation method. means for identifying the manufacturer of the electronic device;
means for identifying a type of a component to be focused on among components mounted on a substrate included in the electronic device based on the identified manufacturer;
extracting means for extracting an area of a component belonging to the type of the component to be focused on from the image of the substrate;
means for estimating a resource value of the electronic device or the board based on the extracted total area of the regions of the component and the manufacturer;
An information processing system having

Images