Print media processing system and print media processing method

Abstract

To provide a printing medium processing system capable of appropriately classifying mediums such as printing mediums.SOLUTION: In a printing medium processing system, a medium identifying unit obtains spectroscopic information from at least five points of printing results containing each of the three primary colors of paper white, black, and printing color material of plural types of printing mediums, generates in advance a learned model by machine learning with these as learning data, identifies the types of the printing mediums by using these, identifies the types of the printing object mediums by applying the spectroscopic information obtained from the at least five points of printing results for the printing object mediums to be printed to the learned model, and performs processing to the printing object medium according to the identified type of the printing object medium. When the printing medium is a new printing medium which is not yet learned, learning is performed by transition learning, and the learned model is updated.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] The present disclosure relates to techniques for identifying media type for processing of print media, such as printing or reproduction. [Background technology]

[0002] In printing using ink, the type of printing medium (hereinafter referred to as media) such as paper onto which the ink is ejected is identified and printing conditions are adjusted. For this reason, it is necessary to identify the type of media. The image processing device described in Patent Document 1 below measures color values ​​as the image quality of the printout, calculates the color gamut characteristics of the media based on the measured values, identifies media with similar image quality characteristics, and selects the optimal profile. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2010-252053 A Summary of the Invention [Problem to be solved by the invention]

[0004] However, in the technology described in Patent Document 1, in order to classify image quality characteristics by parameters (e.g., color values) that are predefined, it is necessary to hold parameters of media having characteristics that are sufficiently different to distinguish the image quality characteristics as preset values, and there are cases in which media cannot be properly classified. Also, even if an attempt is made to add a new classification, there are cases in which a difference in color values ​​sufficient to distinguish the image quality characteristics cannot be obtained, and it can be difficult to arbitrarily add a classification. [Means for solving the problem]

[0005] The present disclosure can be realized in the following forms or application examples.

[0006] A first aspect of the present disclosure is a print media processing system for identifying print media. The print media processing system includes a model storage unit that stores a trained model generated in advance by machine learning using at least five pieces of spectral information obtained from the three primary colors and black printing areas of a printing color material and the paper white area of ​​the print media as training data for a plurality of types of print media, and an identification unit that applies the at least five pieces of spectral information obtained from the print target medium to be printed on to the trained model to identify the type of the print target medium.

[0007] A second aspect of the present disclosure is a print medium processing method. This print medium processing method stores a trained model generated in advance by machine learning using at least five pieces of spectral information obtained from the three primary colors and black printing areas of the printing color materials and the paper white area of ​​the print medium as training data for a plurality of types of print media, applies the at least five pieces of spectral information obtained from the print target medium to be printed on to the trained model, identifies the type of the print target medium, and performs processing on the print target medium according to the identified type of print target medium. [Brief description of the drawings]

[0008] [Figure 1] 1 is a schematic configuration diagram showing a print media learning device and a print media processing system according to an embodiment; [Diagram 2] FIG. 4 is an explanatory diagram illustrating an example of a patch used when acquiring spectral information. [Diagram 3] FIG. 1 is a schematic diagram illustrating the configuration of a print processing system including a printer. [Figure 4] 11 is a flowchart showing an example of a process in the print media learning device. [Diagram 5] An explanatory diagram showing an overview of supervised machine learning. [Figure 6] 11 is a flowchart showing an example of a media identification process in the print media processing system. [Figure 7] 11 is a flowchart outlining processing of unknown media. [Figure 8]FIG. 1 is an explanatory diagram illustrating transfer learning. [Figure 9] FIG. 11 is an explanatory diagram showing an example of a dialog box for inputting new print medium processing information. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] A. First embodiment: (1) Overall configuration: FIG. 1 is a block diagram showing a print media processing system 40 according to an embodiment together with a print media learning device 20. Prior to the operation of the print media processing system 40, the print media learning device 20 generates a media identifier 30 for identifying the type of print media. Once the media identifier 30 is generated, the media identifier 30 functions as a part of the print media processing system 40. Of course, the function of generating a trained model 33 by performing machine learning in advance and the function of identifying media using the trained model 33 may be completely separated as a device configuration. In that case, it is sufficient that the media identifier 30 including the trained model 33 generated by the print media learning device 20 can be introduced into the print media processing system 40. The media identifier 30 may be introduced using a storage medium such as a memory, or may be introduced via a network such as the Internet or a dedicated line.

[0010] Print media learning device 20 includes a spectral information acquisition unit 21 that acquires spectral information of various print media (hereinafter also referred to as media), a media data input unit 22 that inputs various data related to the media, a media identifier generation unit 25, and a supervised machine learning machine 31. Print media learning device 20 acquires spectral information of multiple types of media A, B, ...N by spectral information acquisition unit 21 and accumulates it as spectral information 26.

[0011] FIG. 2 is an explanatory diagram illustrating an example of a patch used when acquiring the spectral information of each medium. In this embodiment, a predetermined size area is actually printed on the printing medium P with ink of cyan C=100, magenta M=100, yellow Y=100, and black K=100, and the spectral information of each area is acquired together with the paper white area W. The spectral information may be the spectral reflectance or the spectral transmittance for each wavelength, or may be both. Furthermore, it is not limited to the spectral reflectance, etc., and may be an image captured by a spectral camera (spectrum camera) capable of capturing an image of the printed printing medium P in a predetermined wavelength range (band) including at least a part of visible light. The image captured by the spectral camera is obtained as a collection of monochrome images captured for each wavelength band.

[0012] In addition to the spectral information 26, media data for each print medium is input by the media data input unit 22. The media data includes a media identifier 32 that identifies the type of print medium, and processing information 27. The media identifier 32 is an identifier for identifying the medium for which the spectral information was input. The processing information 27 includes various data required when printing using that medium, such as an LUT or ICC profile for color conversion required when printing on the print medium, and a duty limit value indicating the upper limit of the amount of ink applied.

[0013] The supervised machine learning machine 31 uses the spectral reflectance data and the media identifier 32 as training data to learn a model for identifying the print medium, using the spectral information 26 for multiple types of print media A, B, ...N. The trained model 33 generated as a result of the learning, together with the spectral information 26 and processing information 27, forms the media identifier 30. The operation and configuration of the media identifier generation unit 25 that performs supervised machine learning will be described later.

[0014] The trained model 33 thus generated serves as a media identifier 30 that uses spectral information as input to identify the type of media in a print medium processing system 40, which will be described later. Once the trained model 33 has been trained and the media identifier 30 has been generated, the media identifier 30 constitutes the print medium processing system 40 together with a computer (hereinafter abbreviated as PC) 50 that instructs printing, a raster image processor (hereinafter abbreviated simply as RIP) 60, and a printer 70. When the PC 50 outputs data such as an image to be printed by the printer 70, the RIP 60 inputs this data and develops it into a raster image that can be printed by the printer 70. At this time, the RIP 60 receives processing information corresponding to the type of print medium P to be printed from the media identifier 30, and generates a raster image using this processing information. That is, in this embodiment, the raster image output by the RIP 60 is configured as dot data indicating the on / off state of each ink installed in the printer 70. Of course, the RIP 60 may store processing information corresponding to existing print media P in advance, receive information on the type of print media P from the media identifier 30, and generate dot data.

[0015] A detailed configuration of the print medium processing system 40 is shown in FIG. 3. The media identifier 30 includes the spectral information acquisition unit 21, the media data input unit 22, the spectral information 26, the processing information 27, and the trained model 33, as well as the identification unit 47, the transfer learning unit 48, and the output unit 49. When the spectral information acquisition unit 21 acquires the spectral information of the print medium to be printed (hereinafter, referred to as the print target medium) p1, the identification unit 47 uses the input spectral information to identify the print medium using the trained model 33. The identification unit 47 determines whether the print target medium p1 is a known or unknown type of print medium from the acquired spectral information, and if it is determined to be a known type, also identifies the type of print medium. In response to this, the media identifier 30 outputs the processing information 27 according to the corresponding type of print medium to the RIP 60 via the output unit 49. If the print target medium p1 is unknown, the transfer learning unit 48 performs transfer learning using the existing trained model 33 to generate a new transfer trained model 33a that has been trained for the unknown print target medium p1. The original trained model 33 is replaced with the transfer trained model 33a.

[0016] As shown in FIG. 3, the printer 70 includes a processor 71 that controls the entire printer 70, including a print control unit 72 that controls printing, a communication unit 73 that communicates with the RIP 60, a print head 77 that ejects ink droplets to print on a print medium P, a head transport device 75 that transports the print head 77 together with a carriage 80 on which the print head 77 is mounted, and a paper transport device 76 that transports the print medium P. The carriage 80 is equipped with ink cartridges 81 that supply ink to nozzles for each color provided on the print head 77. In this embodiment, the ink cartridges 81 each contain ink of cyan C, magenta M, and yellow Y, which is one combination of the three primary colors for printing, and black (real black) K. Note that the printer 70 may be equipped with ink of a color other than the three primary colors, such as special color inks such as red, blue, and green, and light color inks such as light magenta and light cyan, and perform printing.

[0017] In this embodiment, the printing medium P is rolled paper wound into a roll, and when this roll paper is loaded into the printer 70, the cyan C, magenta M, yellow Y, and black K areas shown in Figure 2 are printed on a portion of the roll paper, which is the printing target medium p1. This printed printing target medium p1, together with the paper white area W, is used as a patch for acquiring spectral information.

[0018] In this embodiment, the printer 70 is a so-called inkjet serial printer, but it may be a printer other than the inkjet type, for example a so-called page printer such as a laser printer. Or it may be a thermal dye sublimation printer. The printing color material is ink in an inkjet printer and toner in a laser printer. The three primary colors of the printing color material are typically cyan, magenta, and yellow as described above, but are not limited to this, and as long as the colors contained in the image to be printed by the printer can be expressed, any combination of color materials may be used.

[0019] In this embodiment, a spectroscopic reflectometer is used as the spectroscopic information acquisition unit 21. The spectroscopic information acquisition unit 21, which is a spectroscopic reflectometer, is provided with an LED as a light-emitting unit that irradiates light in the wavelength range of visible light, and a sensor as a light-receiving unit that detects the intensity of light reflected from the LED by the printing medium P in a predetermined wavelength range, for example, every 10 nm. Note that, in addition to the spectral reflectance, the spectral transmittance, which is the rate at which light passes through the printing medium P, may also be used as the spectroscopic information. In this case, it is sufficient to provide an LED for light emission and a sensor for light reception at a position that sandwiches the printing medium P.

[0020] (2) Configuration of supervised machine learning machine 31: In this embodiment, the supervised machine learning machine 31 performs machine learning using a training model with the spectral reflectance data of various print media P and an identifier indicating the type of print medium P corresponding to the spectral reflectance data as supervised data. This utilizes the fact that the characteristics of the spectral reflectance differ depending on the type of print medium P. In the case of white paper that does not use a fluorescent whitening agent, there tends to be little change in reflectance in each wavelength range. On the other hand, in the case of white paper that uses a fluorescent whitening agent, there is a tendency for the reflectance to increase in the short wavelength range. In this way, the spectral reflectance of the print medium P includes information such as color information of the surface of the print medium P and the presence or absence of a fluorescent whitening agent.

[0021] 4 is a flowchart showing the process in which the supervised machine learning device 31 performs machine learning to generate a trained model. Before starting this process, training data is collected for multiple types of print media P, in which the spectral reflectance data is associated with a media identifier 32 that identifies the type of print medium P. The process of generating the trained model is performed based on this training data.

[0022] When the process starts, a process of acquiring teacher data is performed first (step S100). As described above, the teacher data is collected in advance prior to the process. In this embodiment, it is assumed that the teacher data, which is the spectral reflectance data, is collected for 14 types of media from media A to media N and stored in memory. Of course, the number of types of media may be more or less. After the teacher data is acquired, machine learning processing using the training model (step S110) is then performed until generalization is completed (step S120). The teacher data is given to the training model and machine learning is performed, thereby generating a trained model 33, which is finally saved as the trained model 33. The saved trained model 33 is incorporated into the media identifier 30 and used to identify the type of print medium P.

[0023] A training model for machine learning can be defined in various ways. FIG. 5 is a diagram showing an example of a model used in this embodiment. In the figure, each layer of the total n layers by CNN is shown as a rectangular parallelepiped L1 to Ln-1, and the nodes of a normal neural network are shown as white circles. In this embodiment, a CNN (Convolutional Neural Network) is used for the data of the spectral reflectance, but other models such as various neural networks such as a capsule network type and a vector neural network type may be used. As an example, a configuration using a vector neural network shown in Japanese Patent Application No. 2020-94205, which is another application by the applicant of the present application, can be mentioned. In this vector neural network type model, the output of the intermediate layer can be used. Note that vector neural networks that can use the output of the intermediate layer are publicly known.

[0024] The first layer L1 is provided with a plurality of nodes that input reflectance for each fixed wavelength. In this embodiment, the spectral reflectance for each fixed wavelength indicated by the spectral reflectance data is input as input data to each node of the first layer L1, which is the input layer, and final output data related to the spectral reflectance data is output from the final output layer Ln-1. The reflectance is a value between 0 and 1. Instead of the spectral reflectance data, it is also possible to use transmittance for each fixed wavelength. It is also possible to use both. When both the spectral reflectance and the spectral transmittance are used, two training models are provided, and intermediate output data of the spectral reflectance and intermediate output data of the spectral transmittance are output from the final layers of each model, and the two are combined to form the overall final output layer Ln-1.

[0025] The output of each node in the first layer L1 is connected to a node in the next second layer with a predetermined weighting. This is the same for the second layer L2 and onwards, up to the n-2th layer. By repeating the process of correcting the weighting between each node in each layer using training data, learning is completed and a trained model 33 is generated. This point will be described in detail below.

[0026] In learning, the supervised machine learning machine 31 inputs one of the teacher data into the training model and calculates the output value of the final output layer Ln-1. At this time, it is determined in advance which node of the final output layer Ln-1 corresponds to which type of media. For example, if the print medium P is medium A, learning is performed so that node A outputs a value close to value 1. One teacher data is given to the training model, and the supervised machine learning machine 31 evaluates the values ​​of each node A to N of the output final output layer Ln-1 using a loss function. In other words, if the teacher data is for media A, for example, a state in which node A has a value of 1 and other nodes B to N have a value of 0 is set as the target state, and the difference from the actual output (learning error) is specified using the loss function. Then, the supervised machine learning machine 31 corrects the weighting between each node. After the correction, the learning error is calculated again using the loss function, the difference (differential value) from the previously calculated learning error is calculated, whether the learning error has been reduced by the weighting is determined, and the weighting is further corrected.

[0027] The loss function for determining the learning error can be various functions, for example, cross entropy error. The process of calculating the loss function as described above is performed on all or part of the data included in the training data, and the loss function for one learning session is expressed by the average or sum of the data. Once the learning error for one learning session is obtained, the supervised machine learning device 31 corrects the weighting using a predetermined optimization algorithm, for example, stochastic gradient descent.

[0028] When the weighting of all the training data has been corrected a predetermined number of times in this manner, the supervised machine learning device 31 determines whether generalization of the training model is complete (step S120). That is, the supervised machine learning device 31 obtains the degree of agreement between the input and output of the training model using the training data, and determines that generalization is complete when the degree of agreement is equal to or greater than a predetermined threshold.

[0029] If it is not determined that generalization is complete, the machine learning process (step S110) is repeated. That is, a process of further modifying the weights is performed, and the calculation and evaluation of the learning error using the error function is repeated. Note that the supervised machine learning device 31 may perform other processes, such as verifying the validity of hyperparameters, in addition to the supervised machine learning of the type of printing medium P.

[0030] Each node in the final output layer Ln-1 corresponds to the type of print medium P to be identified (media A, media B,...media N). Using the training data, the sum of the output values ​​of each node in the final output layer Ln-1 is normalized to 1 when one training data is given. It is possible to estimate that the type of print medium P corresponding to the node with the largest output value in the final output layer Ln-1 is the type of print medium P from which each training data input to the input layer L1 was obtained. Ideally, when each training data is given, the output of one node in the final output layer Ln-1 is value 1 and the others are value 0, and it is desirable to be able to identify the print medium P, but the output of each node in a multi-layered neural network with multiple types of outputs is usually in the range greater than value 0 and less than value 1. For this reason, the maximum value of the output values ​​of each node in the final output layer Ln-1 is treated as a parameter indicating the possibility that the print medium P is the media corresponding to the node.

[0031] In this way, in the supervised machine learning device 31, the training model is given the supervised data to learn, and a trained model 33 is generated. For the types of print media P for which supervised machine learning is performed, processing information required for printing is prepared for each type of print media P that has been learned. This processing information is input from the media data input unit 22 for each type of print media P. Since the processing information 27 differs depending on the printer 70, it may be difficult to store the processing information 27 for all types of print media P on the printer 70 side. For this reason, it is useful to store the corresponding processing information 27 on the cloud side for some types of print media P and read it as necessary. Of course, a dedicated site for storing the processing information 27 may be provided on the cloud, and the processing information 27 may be prepared there.

[0032] Such processing information 27 includes: [1] A color conversion LUT that converts image data into the amount of printing color material, in this case, ink, [2] ICC profile, [3] Heating control information, which is the conditions for heating the printing medium after printing; [4] Media transport information, which is the conditions for transporting the print media; [5] Duty limit value indicating the upper limit of the amount of ink that can be applied to the printing medium. [6] Information regarding the susceptibility of print media to cockling; It is not necessary to prepare all of these as the processing information 27, and only the necessary processing information can be determined according to the type and functions of the printer 70 to be used. Since the color conversion LUT and ICC profile have a significant effect on the color development characteristics of the printed matter, it is desirable to include these two.

[0033] Therefore, if the media identifier 30 identifies the print medium p1 to be printed on as a known medium, it outputs processing information required for printing the print medium p1 to the RIP 60. On the other hand, if the media identifier 30 identifies the print medium p1 to be printed on as an unknown medium, it updates the trained model 33 by transfer learning performed by the transfer learning unit 48. For this transfer learning, the media identifier 30 only needs to store the trained model 33 and the spectral information of the unknown print medium p1, and does not need to store the spectral information, which is the data of the spectral reflectance of each of the trained types of media A to N. If the transfer learning unit 48 determines that the print medium p1 to be printed on is not a known medium, it performs transfer learning by adding a new media identifier X to the spectral information 26p1 of the print medium p1, and updates the trained model 33 by the trained transfer learned model 33a. The actual transfer learning will be described later.

[0034] (3) Media Identification Processing: An overview of the media identification process performed by the media identifier 30 prior to printing by the printer 70 is shown in Fig. 6. Before the start of the illustrated processing routine, the media identifier 30 stores the trained model 33 and processing information 27 corresponding to multiple types of printing media P. When the processing in Fig. 6 is started, first, a process is performed in which the patch shown in Fig. 2 is placed on the printing medium P in the spectral information acquisition unit 21 (step S200).

[0035] Next, a process of acquiring the spectral reflectance of the printing medium P is performed using the spectral information acquisition unit 21 (step S210). The wavelength of the irradiated light is divided into a predetermined wavelength range, here every 10 nm, and the spectral reflectance is acquired as a set of data on the reflectance in the wavelength range. In FIG. 6, the solid line J1 shows an example of the spectral reflectance of a normal printing paper. In contrast, the printing medium P to which a fluorescent whitening agent is applied has a higher reflectance on the short wavelength side than normal printing paper, as shown by the solid line J2. The spectral reflectance may be measured by the spectral information acquisition unit 21 provided in the media identifier 30, or may be measured using a dedicated measuring device and acquired by the media identifier 30 via communication or the like.

[0036] Next, the media identifier 30 performs a media identification process to identify the type of print medium P using the spectral reflectance data acquired in step S210 and the trained model 33 (step S220). The media identifier 30 inputs the acquired spectral reflectance data to the trained model 33 and obtains an output value (see FIG. 5). The media identifier 30 acquires an identifier and parameters representing the type of print medium P based on the output values ​​of the intermediate and final layers of the trained model 33 (step S230).

[0037] Next, based on the acquired parameters, it is determined whether the print medium P to be printed on is a known medium (step S240). Of course, the parameters obtained from the output values ​​of the intermediate layer, the final output layer Ln-1, etc. and the type (identifier) ​​of the medium may be displayed, and the user may be asked to confirm whether it is OK to determine that the medium is a known medium. In this case, it is up to the user to determine whether the medium is a known medium.

[0038] If it is determined that the medium is a known medium, the known media printing condition setting process (step S300) is executed, and if it is determined that the medium is not a known medium, the unknown media printing condition setting process (step S400) is executed. In the known media printing condition setting process (step S300), the processing information 27 corresponding to the printing medium P to be printed is identified and read out. On the other hand, in the unknown media printing condition setting process (step S400), the transfer learning described later is performed, and then the processing information 27 suitable for the printing target medium p1 is identified and prepared. In either printing condition setting process, the processing information prepared in each process is output to the RIP 60 (step S250). The RIP 60, which has received this processing information, converts the image data input from the PC 50 into a raster image, here dot data indicating the on / off of dots by each color ink, using the processing information, and outputs it to the printer 70 to perform printing.

[0039] (4) Unknown Media Processing: The unknown media printing condition processing routine (step S400) performed by the media identifier 30 will now be described. When it is determined that the printing target medium p1 is an unknown medium, the media identifier 30 first prompts the user to input media data including the new media name, and performs processing to acquire the input media data (step S410). The media data includes a media identifier indicating the type of new media, the name of the media, processing information when using that media, and the like. Note that the processing information may be entirely input by the user using the media data input unit 22, but it may also be possible to use processing information 27 for existing printing medium P with reference to parameters output by the media identifier 30, or to use a portion of the information after modifying it.

[0040] When the media data of the new medium is acquired according to the user's input, a process of setting the trained model 33 is then performed (step S420). This is a process of setting the trained model 33 that has been used to identify the print medium P up to this point for transfer learning. Next, transfer learning is performed using the spectral information 26p1 when it is determined that the medium is an unknown medium and the acquired media identifier 32 (step S430). In transfer learning, the weighting between layers of the already trained model is used as is, a new layer is further provided as the final layer, and only the weighting between this layer is learned. This is because in machine learning such as CNN, the features to be identified are gradually structured from lower-order spectral features as the layers approach the final output layer from the layers closer to the input. The trained model 33 is attempting to output a result of identifying the printing medium P using an input of spectral information, whether for a known medium or an unknown medium, so as shown diagrammatically in Figure 8, a new layer Ln is added to the output of the output layer Ln-1 of the existing trained model 33, and the weighting between the layers is adjusted to train the model to select a new medium X.

[0041] In this way, transfer learning is performed for the new media, and a transfer-learned model 33a with a new layer Ln added is generated. The output X of this model is associated with a media identifier indicating the new media, and processing information such as a color conversion LUT and ink duty when printing on that media (step S440). Such processing information may be selected using a dialog box 91 illustrated in FIG. 9, which is prepared for similar media. That is, the media data input unit 22 displays a dialog box 91 for inputting processing information, and the user is prompted to select using this dialog box 91. In the illustrated example, a color conversion LUT and an ICC profile are set as processing information. In this case, when the user selects the "Browse" button 92, a color conversion LUT and an ICC profile prepared in advance are read from a database 93 that stores them, and multiple candidates are displayed. When the user selects from multiple candidates, the selected file is displayed in a box 94. When the color development characteristics are unknown, pressing the "Browse" button displays media showing similar color development characteristics. When determining that the medium is unknown and registering it as a new medium, the media identifier 30 performs the above-mentioned transfer learning, generates a new transfer-learned model 33a, and associates the selected processing information, such as the LUT and ICC profile, with the new medium that can now be identified. Note that in FIG. 9, the transfer learning unit 48 is shown as being on the cloud. The media identifier 30 instructs the transfer learning unit 48 on the cloud server to perform transfer learning via the network, and the transfer learning unit 48 returns the results. After making these selections, if the "Register" button is pressed, the trained model 33 is updated to the trained model 33a generated by transfer learning, and the selected processing information is associated with the new print medium X.

[0042] When the processing information prepared for a similar print medium is not used as is, the processing information for a print medium with a high similarity in spectral information may be displayed as default processing information, which may then be used or modified, registered as processing information for a new medium, and associated with a media identifier.The trained model of the media identifier 30 is then replaced with the new trained model 33a generated by transfer learning, and the media identifier 30 is updated (step S450). The processing of the unknown medium is completed through the above processing, and the routine ends by exiting to "NEXT".

[0043] According to the print media learning device 20 and print media processing system 40 described above, for multiple types of print media P, spectral reflectance is obtained using a batch (FIG. 2) having a total of five color regions, including four regions printed with the three primary colors (cyan C, magenta M, yellow Y) used for printing by the printer 70 and black K ink, and paper white W, and the type of print medium is identified using the results of supervised machine learning previously performed based on this. Therefore, there is no need to use special parameters that can distinguish image quality characteristics, and the type of print medium P can be identified using a simple method.

[0044] Moreover, if the type of print medium p1 to be printed on is one that has not been learned by machine learning before and the type of print medium p1 cannot be identified, the characteristics of the new print medium can be learned by transfer learning, which requires fewer man-hours than fine tuning, and the trained model 33 can be updated to one that can identify the new print medium. Since a new trained model is generated by transfer learning, there is no need to store the spectral information used during the initial learning.

[0045] When printing on the identified print medium in this manner, processing information such as a color conversion LUT or ICC profile associated with the identified print medium is used, making it possible to print in a manner suited to the color development characteristics of the print medium, thereby improving the color reproduction and image quality of the printed material.

[0046] In the above embodiment, the data output by the PC 50 is converted into a raster image using the RIP 60, but the conversion into a raster image may be performed by a printer driver of the PC 50 without using the RIP 60. The printer 70 itself may also be provided with color conversion and binarization functions, and may also incorporate a media identifier 30 that identifies the type of print medium. In this way, it is possible to print simply by inserting a memory card that stores image data into the printer 70, or by sending image data to the printer 70 from a portable device such as a tablet terminal.

[0047] B. Other embodiments: (1) The print medium processing system of the present disclosure can be implemented in the following manners, for example. The first manner is as a print medium processing system for classifying print media. This print medium processing system includes a model storage unit that stores a trained model generated in advance by machine learning using at least five pieces of spectral information obtained from the printed areas of the three primary colors of the print color materials and black and the paper white portion of the print medium as training data for multiple types of print media, and an identification unit that applies the at least five pieces of spectral information obtained from the print target medium to be printed to the trained model to identify the type of the print target medium. In this way, for multiple types of print media P, spectral information obtained from at least five print results including the paper white, black, and the three primary colors of the print color materials of the print medium is obtained, and the type of print medium can be identified using the results of supervised machine learning performed in advance based on this. Therefore, there is no need to use special parameters that can distinguish image quality characteristics, and the type of print medium can be identified by a simple method.

[0048] Here, the print medium processing system may be implemented as a system including a printer that prints on print media, or as a processing system that does not involve printing, as long as it can identify the type of print medium. For example, the system may be applied to a system that recycles used paper or a system that transports print media.

[0049] (2) In such a configuration, the three primary colors of the printing color materials may be cyan and magenta. In this way, most printers that perform multi-color printing using inks can easily perform printing required for identification. The three primary colors of the printing color materials do not need to be limited to the commonly used cyan, magenta, and yellow, and any combination of the three primary colors used in printing is acceptable. The combination of the three primary colors does not need to cover the entire visible range. The spectral information only needs to include these five points, and may include points of colors other than the above-mentioned five points.

[0050] (3) In such a configuration, the spectral information used as the learning data may include at least one of the spectral reflectance or spectral transmittance obtained by measuring the color of the printed result, or the spectral image obtained by capturing the printed result. In this way, the spectral information can be easily acquired. The spectral reflectance or the spectral transmittance may be used alone, or both may be used. When both are used, two training models are provided during machine learning, and intermediate output data of the spectral reflectance and intermediate output data of the spectral transmittance are output from the final layer of each model trained with the spectral reflectance or the spectral transmittance, respectively, and both are combined to form the overall final output layer. Similarly, the spectral image may be handled alone or in combination with other spectral information. Each piece of spectral information may be acquired for a wavelength range that includes at least a portion of the wavelength region of visible light. This wavelength range may include the shorter wavelength side than visible light. This is because a fluorescent whitening agent may be added to the printing medium, and when a fluorescent whitening agent is added, characteristics may occur in the spectral information on the shorter wavelength side of visible light.

[0051] (4) In this configuration, processing information used for printing using the print medium is [1] Color conversion LUT that converts image data into the amount of printing color materials; [2] ICC profile, [3] Heating control information, which is the conditions for heating the print medium after printing; [4] Media transport information, which is the conditions for transporting the print media; [5] a duty limit value indicating an upper limit of the amount of ink that can be applied to the printing medium when the printing color material is printing ink; [6] information regarding the susceptibility of the print medium to cockling; and a processing unit that, when the type of the print target medium to be printed is identified by the identification unit, reads out the associated processing information according to the type of the print target medium and uses it for processing the print target medium. In this way, it is possible to use the result of identifying the type of print medium to print under conditions appropriate for that print medium.

[0052] Among the above processing information, the heating control information is information on the temperature and time of a heating process that may be performed after printing. Heating may affect not only the coloring material but also the printing medium, and may affect color development. The medium transport information is the conditions for transporting the printing medium, and is information on the transport speed allowed for the medium after printing and the tension applied to the printing medium during transport. Cockling is the waviness of the printing medium that occurs in the printing medium due to the absorption of ink when printing is performed with ink. The tendency for this cockling to occur ultimately leads to the occurrence of color unevenness and affects the image quality. Note that the processing information is not limited to the above, and other processing information may be used if the printing method or processing method is different. For example, if the processing is printing with a thermal dye sublimation printer, the heating temperature of the print head and the correction amount thereof may be treated as the processing information. If the printing is with a dot impact printer, the voltage applied to the piezo and the waveform of the applied voltage may be treated as the processing information. Also, if the processing is recycling of the printing medium, the dissolving temperature of the printed printing medium, the concentration of the solvent used, the processing time, etc. may be treated as the processing information.

[0053] (5) In such a configuration, if the identification unit is unable to identify the print medium as a learned print medium, at least one of the processing information items [1] to [6] associated with the print medium may be read from the processing information storage unit in accordance with the result of the identification and used for the processing of the print medium. In this way, the processing can be performed appropriately and easily. Here, the use of processing information includes not only the use of pre-stored processing information as is, but also the use of pre-stored processing information with a portion of the information modified.

[0054] (6) In such a configuration, the process may be a process of printing on the print target medium using at least the three primary colors and black printing color materials. In this way, printing can be performed under conditions suitable for the print medium to be printed on, and high-quality printing can be achieved. Of course, these printing color materials may be ink or toner. In the case of ink, any type may be used, such as dye ink, pigment ink, or ink used for thermal dye sublimation. The printing color materials may include color materials of colors other than the three primary colors and black.

[0055] (7) In such a configuration, when it is determined that the print medium is not a trained print medium, a transfer learning unit may be provided that performs transfer learning using the trained model, increases the output layer of the trained model, and generates a new trained model. In this way, even if an untrained print medium is used, the trained model can be updated with fewer steps, and can be adapted to new print media. In transfer learning, only the output layer may be increased, or the output layer and one or more other layers may be increased. The transfer learning unit that performs transfer learning may be provided within the identification unit, or may be provided separately from the identification unit. When provided separately, for example, the transfer learning unit may be placed on the cloud side, the transfer learning unit may perform transfer learning via a network, and the trained model generated by transfer learning may be returned to the identification unit by a method such as via a network.

[0056] (8) In such a configuration, when the transfer learning unit generates a new trained model with an increased output layer by transfer learning, the trained model stored in the model storage unit may be updated. In this way, the type of print medium can be identified using the latest trained model. The trained model may be updated every time, or may be updated at a predetermined timing, for example, at regular intervals.

[0057] (9) A second aspect of the present disclosure is a printing medium processing method. This printing medium processing method stores a trained model generated in advance by machine learning using at least five pieces of spectral information obtained from the printed areas of the three primary colors of the printing color materials and black and the paper white area of ​​the printing medium as training data for multiple types of printing media, and applies the at least five pieces of spectral information obtained from the printing target medium to be printed to the trained model to identify the type of the printing target medium, and performs processing on the printing target medium according to the identified type of printing target medium. In this way, at least five pieces of spectral information obtained from the printing results of the paper white, black, and the three primary colors of the printing color materials of the printing medium are obtained for multiple types of printing media, and the type of the printing target medium can be identified using the results of supervised machine learning performed in advance based on this. Therefore, there is no need to use special parameters that can distinguish image quality characteristics, and the type of the printing medium P can be identified by a simple method.

[0058] (10) In each of the above embodiments, a part of the configuration realized by hardware may be replaced by software. At least a part of the configuration realized by software may be realized by a discrete circuit configuration. In addition, when a part or all of the functions of the present disclosure are realized by software, the software (computer program) may be provided in a form stored in a computer-readable recording medium. The term "computer-readable recording medium" is not limited to portable recording media such as flexible disks and CD-ROMs, but also includes internal storage devices in a computer such as various RAMs and ROMs, and external storage devices fixed to a computer such as a hard disk. In other words, the term "computer-readable recording medium" has a broad meaning including any recording medium to which a data packet can be fixed, not temporarily.

[0059] The present disclosure is not limited to the above-mentioned embodiment, and can be realized in various configurations without departing from the spirit of the present disclosure. For example, the technical features in the embodiments corresponding to the technical features in each aspect described in the Summary of the Invention column can be appropriately replaced or combined to solve some or all of the above-mentioned problems or to achieve some or all of the above-mentioned effects. Furthermore, if the technical feature is not described as essential in this specification, it can be appropriately deleted.

[0060] The present disclosure is not limited to the above-mentioned embodiments, and can be realized in various configurations without departing from the spirit of the present disclosure. For example, the technical features of the embodiments corresponding to the technical features in each form described in the Summary of the Invention column can be appropriately replaced or combined to solve some or all of the above-mentioned problems or to achieve some or all of the above-mentioned effects. Furthermore, if a technical feature is not described as essential in this specification, it can be appropriately deleted. For example, a part of the configuration realized by hardware in the above-mentioned embodiments can be realized by software. [Explanation of symbols]

[0061] 20...print media learning device, 21...spectral information acquisition unit, 22...media data input unit, 25...media identifier generation unit, 26, 26p1...spectral information, 27...processing information, 30...media identifier, 31...machine learning machine, 32...media identifier, 33...trained model, 33a...transfer learned model, 40...print media processing system, 47...identification unit, 48...transfer learning unit, 49...output unit, 70...printer, 71...processor, 72...print control unit, 73...communication unit, 75...head transport device, 76...paper transport device, 77...print head, 80...carriage, 81...ink cartridge, 91...dialog box, 92...button, 93...database, 94...box

Claims

1. 1. A print media processing system for identifying a type of print media, comprising: A model storage unit stores a trained model generated in advance by machine learning using at least five pieces of spectral information obtained from the three primary colors and black printing areas of the printing color material and the paper white area of ​​the printing medium as training data for a plurality of types of printing media; an identification unit that identifies the type of the print target medium by applying the at least five pieces of spectral information obtained from the print target medium to be printed to the trained model; Processing information used for printing using the print medium, [1] A color conversion LUT that converts image data into the amount of printing color materials; [4] Media transport information, which is a condition for transporting the print medium; [6] Information regarding the susceptibility of the print medium to cockling; a processing information storage unit that stores information on at least one of the types of the print medium in advance in association with the type of the print medium; a processing unit that, when the type of the print target medium on which the printing is to be performed is identified by the identification unit, reads out the associated processing information in accordance with the type of the print target medium and uses the information in processing the print target medium; A print media processing system comprising:

2. 2. The print media handling system of claim 1, wherein the three primary colors of print colorants are cyan, magenta, and yellow.

3. The print medium processing system of claim 1 or claim 2, wherein the spectral information used as the learning data includes at least one of the spectral reflectance or spectral transmittance obtained by measuring the color of at least five print results including the three primary colors and black printed areas of the printing color material and the paper white area of ​​the print medium, or a spectral image obtained by capturing the print results.

4. A print medium processing system as described in any one of claims 1 to 3, wherein when the identification unit is able to identify the print medium as a learned print medium, at least one of the processing information [1] [4] [6] associated with the print medium is read from the processing information storage unit in accordance with the result of the identification and used for the processing of the print medium.

5. 5. The print medium processing system according to claim 1, wherein the process is a process of printing on the print target medium using at least each of the three primary colors and black print color materials.

6. The print medium processing system of any one of claims 1 to 5, further comprising a transfer learning unit that, when it is determined that the printing target medium is not a trained printing medium, performs transfer learning using the trained model, increases the output layer of the trained model, and generates a new trained model.

7. The print media processing system according to claim 6 , wherein when the transfer learning unit generates a new trained model with an increased output layer by transfer learning, the trained model stored in the model storage unit is updated.

8. 1. A method for processing print media to identify a type of print media, comprising: A trained model is stored that is generated in advance by machine learning using at least five pieces of spectral information obtained from the three primary colors and black printed areas of the printing color material and the paper white area of ​​the printing medium as training data for a plurality of types of printing media; Identifying the type of the printing medium by applying the at least five pieces of spectral information obtained from the printing medium to be printed to the trained model; Processing information used for printing using the print medium, [1] A color conversion LUT that converts image data into the amount of printing color materials; [4] Media transport information, which is a condition for transporting the print medium; [6] Information regarding the susceptibility of the print medium to cockling; and storing the information in advance in a storage unit in association with the type of the print medium for at least one of the types of the print medium; When the type of the print medium to be printed is identified, the corresponding processing information is read from the storage unit in accordance with the type of the print medium, and the read information is used to perform processing on the print medium in accordance with the identified type of the print medium. Print media processing method.

Images