X-ray inspection apparatus and X-ray inspection method

Abstract

To provide an X-ray baggage inspection for reducing discomfort which a visitor feels.SOLUTION: An X-ray inspection device performs: holding an X-ray image of baggage of a visitor, a model for recognizing a category of an article included in the X-ray image, article attribute information indicating a category of a special article, a category of the special article, and event information indicating contents of an event for providing profit to the visitor; recognizing the category of the special article included in the X-ray image based on the X-ray image of the baggage and the model; identifying the contents of the event corresponding to the category of the recognized special article from the event information; and outputting data based on the contents of the identified first-kind event.SELECTED DRAWING: Figure 2

Description

【Technical Field】 【0001】 The present invention relates to an X-ray inspection apparatus and an X-ray inspection method. 【Background Art】 【0002】 In baggage inspection at airports or large-scale event venues, baggage inspection using X-ray images of articles is often performed. As a technique for automatically discriminating the type of an article based on an X-ray image, there is Japanese Unexamined Patent Application Publication No. 2021-120639 (Patent Document 1). 【0003】 This publication describes that "the alert output timing control device holds an analysis target image and time information indicating the time until an alert is output, outputs the analysis target image to the display device, inputs the analysis target image to the model, and when it is determined that the photographed object of the analysis target image includes an alert target object, acquires the time from the time information, and controls so that an alert is output at the timing when the acquired time has elapsed from a predetermined timing after the analysis target image is output." (See the abstract). 【Prior Art Documents】 【Patent Documents】 【0004】 【Patent Document 1】 Japanese Unexamined Patent Application Publication No. 2021-120639 【Summary of the Invention】 【Problems to be Solved by the Invention】 【0005】 The X-ray baggage inspection in the technique described in Patent Document 1 is performed for ensuring the safety of the facility, and the discomfort and stress given to visitors are not considered. Therefore, one aspect of the present invention provides an X-ray baggage inspection that reduces the discomfort given to visitors. 【Means for Solving the Problems】 【0006】 To solve the above-mentioned problems, one aspect of the present invention employs the following configuration. The X-ray inspection apparatus comprises a processor and a memory, the memory holding an X-ray image of a visitor's luggage to the facility, a model for recognizing the category of articles contained in the X-ray image, article attribute information indicating the category of special articles that are not prohibited from being brought into the facility, the category of the special article, and event information indicating the content of a Type 1 event that provides benefits to the visitor. The processor recognizes the category of the special article contained in the X-ray image based on the X-ray image of the luggage and the model, identifies the content of a Type 1 event corresponding to the recognized category of special article from the event information, and outputs data based on the content of the identified Type 1 event. [Effects of the Invention] 【0007】 According to one aspect of the present invention, it is possible to provide X-ray baggage inspection that reduces discomfort caused to visitors. 【0008】 Other issues, configurations, and effects not mentioned above will be clarified by the following description of the embodiments. [Brief explanation of the drawing] 【0009】 [Figure 1] This is a perspective view showing an example of the X-ray image processing system in Example 1. [Figure 2] This is a block diagram showing an example configuration of the X-ray imaging apparatus in Example 1. [Figure 3] This figure shows an example of the data structure of item attribute data in Example 1. [Figure 4] This figure shows an example of the data structure of event data in Example 1. [Figure 5] This flowchart shows an example of the pre-registration process in Example 1. [Figure 6A] This figure shows an example of a pattern embedded in an article in Example 1. [Figure 6B] This figure shows an example of a pattern embedded in an article in Example 1. [Figure 7] This flowchart shows an example of the item recognition process in Example 1. [Figure 8] This figure shows an example of the prohibited item confirmation screen in Example 1. [Figure 9] This figure shows an example of a screen that simultaneously displays an event guiding users on how to handle prohibited items and an entertainment event in Example 1. [Modes for carrying out the invention] 【0010】 Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In this embodiment, the same components will be denoted by the same reference numerals in principle, and repeated descriptions will be omitted. It should be noted that this embodiment is merely one example for realizing the present invention and does not limit the technical scope of the present invention. 【0011】 Unlike airports and public facilities, baggage checks are sometimes conducted at recreational facilities such as theme parks and tourist attractions. However, baggage checks at these facilities can disrupt the atmosphere they create, and because visitors become preoccupied with the fact that they are being checked, it can diminish their enjoyment of the facility. 【0012】 Thus, conventional baggage checks are conducted with the safety of visitors and facilities in mind, but they may result in waiting times for visitors or increase their stress due to being inspected. Furthermore, conventional baggage checks have forced a choice between improving visitor convenience and comfort, or reducing the accuracy of baggage checks or decreasing the types of items that trigger warnings during baggage checks, thereby compromising security. 【0013】 Therefore, in this embodiment, we aim to realize a baggage inspection that visitors can undergo comfortably and enjoyably without compromising safety. [Examples] 【0014】 FIG. 1 is a perspective view showing an example of an X-ray image processing system. The X-ray image processing system discriminates the category of an article photographed in the X-ray image based on the X-ray image. Note that the "category" of an article is the type of article used in the processing of the X-ray image processing system (for example, a notebook computer, a tablet, a water tube, a knife, etc.). For example, when the discrimination based on the X-ray image of an article is correctly performed by the X-ray image processing system, the "category" of the article, which is the discrimination result, matches the original type of the article (the type in the user's operation). 【0015】 <Configuration of X-ray inspection apparatus> The X-ray image processing system includes, for example, an X-ray inspection apparatus 100, a display apparatus 5, and a plurality of rollers R1. The X-ray inspection apparatus 100 is an apparatus that performs X-ray inspection of an article, and is used, for example, for inspecting the carry-on luggage of visitors to facilities such as airports, large-scale event venues, data centers, factories, research institutions, and public institutions. The X-ray inspection apparatus 100 includes, for example, an X-ray image processing apparatus 1, an apparatus main body 2, two display apparatuses 3, and an input apparatus 4. 【0016】 The apparatus main body 2 includes, for example, a transport mechanism that transports luggage on a roller conveyor, an irradiation mechanism that irradiates the luggage with X-rays, and an X-ray imaging mechanism that measures the amount of X-ray transmission. The apparatus main body 2 is provided with a rectangular hole H1 for allowing the luggage transported by the transport mechanism to pass through. The roller R1 is provided on the upstream side in the transport direction of the apparatus main body 2 in order to smoothly move the luggage toward the hole H1. 【0017】 The X-ray imaging mechanism in the apparatus main body 2 includes, for example, two types of X-ray sensors that measure the amount of X-ray transmission, and one X-ray sensor detects the amount of X-ray transmission of higher energy than the other X-ray sensor. As the X-ray sensor, for example, a backscattering type material determination sensor is used, but it is not limited thereto. Note that two types of X-ray sensors may be installed at each of a plurality of locations (side surfaces and ceiling surfaces) in the area where the luggage passes through the apparatus main body 2. Thereby, the apparatus main body 2 can perform X-ray imaging from a plurality of different directions. 【0018】 The X-ray image processing device 1 shown in Figure 1 determines the category of items contained in a package based on the X-ray image of the package. The X-ray image processing device 1 is connected to the main unit 2 via wiring, and is also connected to the display device 3, input device 4, and display device 5. X-ray transmission data and other information are output to the X-ray image processing device 1 from two types of X-ray sensors, which are the imaging mechanism of the main unit 2. 【0019】 The X-ray image processing device 1 determines the material of an object pixel by pixel in an X-ray image based on the difference between high-energy X-ray transmission data acquired by one of two types of X-ray sensors and low-energy X-ray transmission data acquired by the other. The X-ray image processing device 1 generates a color or grayscale image that visualizes the material and density of the object as an X-ray image, and determines the category of the object based on this X-ray image. 【0020】 If the X-ray image processing device 1 determines that an item falls under a specified prohibited item category, it outputs a predetermined signal to the device body 2 indicating that a prohibited item has been detected. In this case, an indicator lamp on the device body 2 lights up, or a predetermined alert sound is emitted from a speaker on the device body 2. This allows the user (inspector, etc.) to understand that the item being inspected is a specified prohibited item. 【0021】 The two display devices 3 shown in Figure 1 display X-ray images acquired by the X-ray image processing device 1 from the main unit 2. This allows users (inspectors, etc.) to visually confirm not only the X-ray image of the item but also the result of identifying the item's category. The two display devices 3 may each display X-ray images taken from different directions, or they may each display different types of images (for example, a color image and a grayscale image). For security reasons, it is undesirable for visitors (who are both being inspected and bringing in their belongings) to see the images displayed on these two display devices 3. Therefore, it is desirable that the display devices 3 be hidden behind a screen or similar so that only inspectors can see them. 【0022】 The input device 4 is a device for receiving inputs from a user (such as an inspector) using, for example, a keyboard or a mouse. Note that the configuration of the X-ray inspection apparatus 100 is not limited to the example in FIG. 1. For example, the X-ray image processing apparatus 1 may be built into the apparatus main body 2. 【0023】 The display device 5 is used to display an event generated according to the category of an article determined from the X-ray image of the imaging result of the apparatus main body 2, or to display a synthesized image in which the article included in the X-ray image is replaced with an animation, an illustration, or the like. 【0024】 Since the content displayed on the display device 5 is for visitors, the display device 5 is installed at a position where visitors can view it. Note that since the display by the display device 5 is for notifying visitors, instead of the display by the display device 5, a notification such as an audio guide by a speaker may be performed. 【0025】 <Configuration of the X-ray Image Processing Apparatus 1> FIG. 2 is a block diagram showing a configuration example of the X-ray image processing apparatus 1. The X-ray image processing apparatus 1 is configured by, for example, a computer including a CPU (Central Processing Unit) 11, a memory 12, a display interface 13, an input interface 14, a communication device 15, and an auxiliary storage device 16. 【0026】 The CPU 11 includes, for example, a processor and executes a program stored in the memory 12. The memory 12 includes a ROM (Read Only Memory), which is a non-volatile storage element, and a RAM (Random Access Memory), which is a volatile storage element. The ROM stores invariant programs (such as a BIOS (Basic Input / Output System)). The RAM is a high-speed and volatile storage element such as a DRAM (Dynamic Random Access Memory), and temporarily stores a program executed by the CPU 11 and data used during the execution of the program. 【0027】 The display interface 13 is an interface device for outputting the program execution results to the display device 3 and the display device 5, etc., in a format that can be visually confirmed by the operator. The input interface 14 is an interface device for receiving input from the operator via the input device 4 and input of X-ray transmission amount from the main unit 2 of the device. The input interface 14 may include, for example, a dedicated interface for the X-ray image processing device 1, or it may include a general-purpose screen input terminal such as a VGA (Video Graphics Array) terminal. 【0028】 The communication device 15 is a network interface that controls communication with other devices based on a predetermined protocol. The communication device 15 may also include a serial interface such as USB (Universal Serial Bus). 【0029】 The auxiliary storage device 16 is a non-volatile, high-capacity storage device such as an HDD (Hard Disk Drive) or flash memory. The auxiliary storage device 16 stores, for example, the OS (Operating System: not shown), as well as various programs executed by the CPU 11 and data used during the execution of each program. In other words, these programs are read from the auxiliary storage device 16, loaded into memory 12, and executed by the CPU 11. 【0030】 Some or all of the program executed by the CPU 11 may be provided to the X-ray image processing device 1 via a network from a removable media (such as a CD-ROM or flash memory) or an external computer equipped with a non-temporary storage device, and stored in a non-volatile auxiliary storage device 16, which is also a non-temporary storage device. For this reason, the X-ray image processing device 1 may have an interface for reading data from the removable media. 【0031】 The X-ray image processing device 1 is a computer system that is configured on a single physical computer or on multiple logically or physically configured computers, and may operate in separate threads on the same computer, or may operate on a virtual computer built on multiple physical computer resources. 【0032】 The CPU 11 includes, for example, functional units: an X-ray image acquisition unit 111, an object recognition unit 112, an object learning unit 113, an event generation unit 114, a screen display unit 115, and an object registration unit 116. 【0033】 The X-ray image acquisition unit 111 receives X-ray image data from the main unit 2 via the input interface 14 and acquires (generates) an X-ray image from the acquired data. An example of how the X-ray image acquisition unit 111 acquires X-ray image data will be described below. 【0034】 For example, when predetermined RAW data is input to the X-ray image processing device 1 from the two types of X-ray sensors provided in the main body 2 of the device, the X-ray image acquisition unit 111 classifies each pixel (the material captured in it) of the X-ray image into one of four types: metal, inorganic material, organic material, or other, based on the difference between the amount of high-energy X-ray transmission and the amount of low-energy X-ray transmission. 【0035】 The X-ray image acquisition unit 111 generates a predetermined color image in which, for example, the material of an object is shown in a predetermined color and the amount of X-ray transmission (density of the object) is shown by the intensity of the color. Alternatively, the X-ray image acquisition unit 111 may acquire data captured from the display screen of the display device 3. 【0036】 Furthermore, if multiple packages are loaded into the main unit 2 in succession, the X-ray image acquisition unit 111 may divide the X-ray image data for each package being photographed using, for example, one of the following three division methods. 【0037】 In the first division method, the X-ray image acquisition unit 111 identifies the portion where the X-ray transmission amount is below a predetermined value as the cargo area. The first division method is used when RAW data in a predetermined format is input to the X-ray image processing device 1 from a VGA output terminal or the like provided on the main body of the device 2. 【0038】 In the second division method, the X-ray image acquisition unit 111 divides the X-ray image for each package based on the integrated value of the amount of X-ray transmission (integrated value at a predetermined line or predetermined time on the X-ray image). In the third division method, the main body of the device 2 is equipped with a package detection sensor, and the X-ray image acquisition unit 111 acquires from the main body of the device 2 the time when the package detection sensor detected a package and the time when it stopped detecting a package, and divides the X-ray image for each package based on the acquired times. 【0039】 The object recognition unit 112 recognizes objects on an X-ray image on a pixel-by-pixel basis, for example, based on deep learning segmentation technology using AI (Artificial Intelligence). The object recognition unit 112 inputs the X-ray image into the object recognition model 161, which will be described later, to determine the category of objects contained in the X-ray image and to identify the position (coordinates) of those objects on the X-ray image. 【0040】 Furthermore, the item recognition unit 112 inputs the X-ray image into the item recognition model 161 to calculate the confidence level of the item category classification result. The confidence level is an indicator of how much the item category classification result can be trusted. The higher the confidence level, the more likely it is that the item category classification result is correct. 【0041】 The object recognition unit 112 may output multiple categories for an object contained in an X-ray image. In this case, for example, the item recognition unit 112 may select the category with the highest confidence level among the multiple categories as the category of the object. Alternatively, for example, the object recognition unit 112 may select a predetermined number of categories as the category of the object in order of increasing confidence level, or it may select all categories with a confidence level equal to or greater than a predetermined value as the category of the object. 【0042】 Furthermore, the object recognition unit 112 may appropriately use OSS (Open Source Sortware) libraries such as Fully Convolutional Instance-aware Semantic Segmentation or Mask R-CNN in the process of recognizing objects in the X-ray image. 【0043】 The object learning unit 113 generates and updates an object recognition model 161 by learning the training data 163, for example, based on AI-based deep learning segmentation technology. The event generation unit 114 determines that the recognition result of an object on the X-ray image satisfies the conditions for an event to occur, and then displays the event details on the display device 5 or notifies the user by voice. 【0044】 The screen display unit 115 displays information for inspectors on the display device 3 and information for visitors on the display device 5. The screen display unit 115 displays information necessary for inspection, such as X-ray images and recognition results from the item recognition unit 112, on the display device 3, and displays luggage information and event details that visitors can enjoy, generated based on the X-ray images, on the display device 5. The item registration unit 116 registers items that may be related to the conditions for the occurrence of an event. 【0045】 For example, the CPU 11 functions as an X-ray image acquisition unit 111 by operating according to an X-ray image acquisition program loaded into memory 12, and functions as an object recognition unit 112 by operating according to an object recognition program loaded into memory 12. The relationship between programs and functional units is similar for other functional units included in the CPU 11. 【0046】 Furthermore, some or all of the functions performed by the functional units included in the CPU 11 may be implemented by hardware such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field-Programmable Gate Array). 【0047】 The auxiliary storage device 16 holds, for example, an object recognition model 161, X-ray image storage data 162, training data 163, object attribute data 164, and event data 165. When an X-ray image is input, the object recognition model 161 outputs the category of the object that is presumed to be depicted in the X-ray image, the location in the X-ray image where the object of that category is presumed to be depicted, and the confidence level of the object category. The object recognition model 161 includes predetermined parameters based on deep learning or the like. 【0048】 The X-ray image storage data 162 stores, for example, X-ray images taken in the object recognition process described later, for a predetermined period of time. It is also preferable that the X-ray image storage data 162 includes, for example, the date and time the X-ray images were taken. 【0049】 Training data 163 is training data for generating the object recognition model 161, and includes, for example, annotated X-ray images, where the annotation indicates the category of the object contained in the X-ray image and the position of the object of that category within the X-ray image. Object attribute data 164 indicates information about objects that may be subject to event occurrence. Event data 165 indicates information about events. 【0050】 Furthermore, some or all of the information stored in the auxiliary storage device 16 may be stored in the memory 12 or in a database connected to the device. Also, in this embodiment, the information used by the X-ray image processing system may be represented in any data structure, regardless of the data structure. In this embodiment, the information is represented in a table format, but for example, a data structure appropriately selected from a list, database, or queue can store the information. 【0051】 Figure 3 shows an example of the data structure of item attribute data 164. The item attribute data 164 includes, for example, an item ID field 1641, an item category field 1642, a recognition target type field 1643, a first threshold field 1644, and a second threshold field 1645. 【0052】 The item ID field 1641 holds an ID that identifies the item. The item category field 1642 holds information indicating the name of the item's category. The recognition target type field 1643 holds information indicating whether the item is a special item or a prohibited item. A special item is an item that is not prohibited from being brought into the facility where the X-ray inspection device 100 is installed, and which may be related to the conditions for the occurrence of an event. A prohibited item is an item that is prohibited from being brought into the facility where the X-ray inspection device 100 is installed. 【0053】 The first threshold column 1644 holds the first threshold for the confidence level assigned to the recognition result of the item. The second threshold column 1645 holds the second threshold (a value greater than the first threshold) for the confidence level assigned to the recognition result of the item. In the example in Figure 3, multiple thresholds (first and second thresholds) are set for special items, and one threshold (first threshold) is set for prohibited items. However, only one threshold may be set for some or all special items, or multiple thresholds may be set for some or all prohibited items. 【0054】 The first and second thresholds are confidence thresholds for determining whether an item is included in an X-ray image; in other words, they are confidence thresholds for deciding whether to reject or accept the recognition result. When an X-ray image is input to the item recognition model 161, if the confidence level of an item category is equal to or greater than the first threshold, it can be determined that the item is definitely included in the X-ray image. If the confidence level is equal to or greater than the second threshold (but less than the first threshold), it can be determined that the item may be included in the X-ray image, although it cannot be said with certainty that the item is definitely included. 【0055】 Furthermore, for prohibited items, only a second threshold may be set; that is, the recognition result that an X-ray image contains a prohibited item may be adopted only when a confidence level is output that confirms the presence of the prohibited item. This can reduce the frequency of detecting prohibited items in X-ray images that do not contain them. 【0056】 Figure 4 shows an example of the data structure of event data 165. Event data 165 includes, for example, event content data 1650 and lucky item data 1657. Event content data 1650 includes, for example, an event name field 1651, a target item field 1652, an occurrence condition field 1653, an occurrence flag field 1654, an event type field 1655, and an event content field 1656. 【0057】 The event name field 1651 holds information indicating the name of the event. The target item field 1652 holds information indicating the item related to the conditions for the occurrence of the event (either the item ID or the lucky item ID described later). In the example in Figure 4, the target item field 1652 stores information indicating one item, but it may also store information indicating multiple items. 【0058】 The occurrence conditions field 1653 holds information indicating the conditions under which an event occurs. Examples of occurrence conditions include the detection of one item of one type, or the detection of multiple items of one type. Furthermore, if information indicating items from multiple categories is registered in the target item field 1652, examples of occurrence conditions include the detection of at least some of those items, the detection of all of those items, or the detection of items from each of those categories combined by AND or OR conditions. 【0059】 The occurrence flag field 1654 holds an occurrence flag indicating whether the event will occur. The event occurs when the occurrence conditions are met and the occurrence flag is set to "1" (i.e., if the occurrence flag is "0", the corresponding event will not occur even if the occurrence conditions are met). Note that for events where the target item is a prohibited item, only an occurrence flag of "1" may be set. 【0060】 The event type field 1655 holds information indicating the type of event. For prohibited items, an event called "Prohibited Item Handling Guidance Event" is defined in the event content data 1650, and for special items, an event data 165 defines an entertainment event (an example of a Type 1 event) and a special item confirmation event (an example of a Type 2 event). In other words, the event type field 1655 corresponding to a prohibited item stores "Prohibited Item Handling," which indicates an event called "Prohibited Item Handling Guidance Event," and the event type field 1655 corresponding to a prohibited item stores "Special Item Handling," which indicates a special item confirmation event, and "Entertainment," which indicates an entertainment event. 【0061】 A prohibited items guidance event is, for example, an event that guides attendees on how to handle prohibited items included in their luggage. A special items confirmation event is, for example, an event that asks attendees to confirm whether their luggage contains any special items. An entertainment event is an event that is related to special items and also provides attendees with benefits. The benefits provided by an entertainment event include not only monetary and material benefits such as coupons and gifts, but also benefits that provide mental enjoyment to attendees. 【0062】 The event details field 1656 holds information indicating the content of the event. When an event occurs, for example, the message written in the event details field 1656 is displayed on the display device 5 or output from the speaker. However, if the event details field 1656<> is written, when an event occurs, for example, the image of the character shown inside the <> is displayed on the display device 5 along with the message written outside the <>, or the message written outside the <> is output from the speaker in the voice of the character shown inside the <>. 【0063】 The lucky item data 1657 includes, for example, a lucky item ID column 1658 indicating a lucky item ID for identifying a lucky item, and an item ID column 1659 indicating an item ID of an item that is a lucky item. Note that only the item IDs of special items can be stored in the item ID column 1659. That is, although special items can become lucky items, prohibited items cannot become lucky items. 【0064】 [[ID==3]] <Processing by the X-ray image processing apparatus 1> Hereinafter, the processing executed by the X-ray image processing apparatus 1 will be described. The processing executed by the X-ray image processing apparatus 1 includes a pre-registration process and an item recognition process. 【0065】 <Pre-registration process> FIG. 5 is a flowchart showing an example of the pre-registration process. In the pre-registration process, the item learning unit 113 learns the appearance of the item on the X-ray image based on the learning data 163 including a large number of X-ray images in which the items of the registration target category are shown, and updates the item recognition model 161. Hereinafter, the details of the pre-registration process will be described. 【0066】 The X-ray image acquisition unit 111 selects the category of the item to be registered and acquires a plurality of X-ray images in which the items of the category are photographed (S501). Specifically, for example, the X-ray image acquisition unit 111 acquires the category name of the item to be registered and information indicating whether the item to be registered is a prohibited item or a special item via the input of the user to the input device 4 (the user in the pre-registration process is, for example, an inspector or an administrator of the X-ray image processing apparatus 1, etc.), and acquires an X-ray image from the X-ray RAW data of the item photographed received from the apparatus main body 2. 【0067】 Furthermore, even if X-ray images are taken of items in the same category, the appearance of the item may differ depending on the shooting direction and the overlapping of items. Therefore, the main unit 2 takes multiple X-ray images (for example, about 100) of the item with different shooting directions and overlapping arrangements with other items. Additionally, since a certain number of X-ray images are required for training the item recognition model 161, the X-ray image acquisition unit 111 acquires multiple X-ray images of items in the category to be registered. Furthermore, items in the category to be registered may have specific patterns embedded in them (for example, patterns with shapes not typically embedded in general items), and in this case, an X-ray image of the specific pattern may be taken as the X-ray image of the item. 【0068】 Figures 6A and 6B show examples of patterns embedded in articles. Pattern 610 is, for example, a metal plate, and pattern 620 is, for example, a three-dimensional metal object. In particular, by embedding such patterns in articles that are difficult to capture in X-ray images, the article recognition model 161 can recognize those articles. Furthermore, by preparing multiple patterns with different shapes, it becomes possible to distinguish between multiple articles with different patterns embedded in them and generate events accordingly. 【0069】 Returning to the explanation of Figure 5, the item recognition unit 112 performs a false detection determination using the existing item recognition model 161 (i.e., the model has not been trained on items in the category to be registered) and the X-ray images of items in the category to be registered acquired in step S501 (S502). 【0070】 Specifically, for example, the item recognition unit 112 calculates the percentage (hereinafter also referred to as the false positive rate) in which, when multiple X-ray images of items in the category to be registered, acquired in step S501, are input to the existing item recognition model 161, the item recognition unit determines that items from other categories (for example, categories already registered in the item recognition model 161, or categories of prohibited items or special items already registered in the item attribute data 164) are included. 【0071】 In calculating the false positive rate, the confidence threshold for determining whether items from other categories are included may be the first threshold for each category indicated by the item attribute data 164, the second threshold for each category indicated by the item attribute data 164, or a predetermined threshold regardless of category. 【0072】 For example, if an item to be registered looks similar in appearance in an X-ray image to an item already registered in the item recognition model 161, such as a prohibited item or a special item, adding the category of the item to be registered to the item recognition model 161 would increase the likelihood that the X-ray image of the item to be registered would be mistakenly detected as another prohibited item or special item when input into the item recognition model 161, thus increasing the likelihood of an inappropriate event occurring. Since it is undesirable for such an item to be registered (especially as a special item), the false detection determination in step S502 is performed. 【0073】 Furthermore, if the X-ray image acquired in step S501 includes an item other than the item to be registered, it is desirable to provide the item recognition unit 112 with location information of the item to be registered within the X-ray image (for example, through input from the user via the input device 4) so ​​that the item recognition unit 112 can recognize which item in the X-ray image is the item subject to false detection in step S502. 【0074】 The item recognition unit 112 determines whether there is a problem with the result of the false detection determination (S503). Specifically, for example, the item recognition unit 112 determines that there is no problem with the result of the false detection determination if the false detection rate is less than a predetermined value, and determines that there is a problem with the result of the false detection determination if the false detection rate is equal to or greater than the predetermined value. 【0075】 If the item recognition unit 112 determines that there is a problem with the result of the false detection determination (S503: NO), it returns to step S501 (i.e., the category of the item to be registered is re-selected). However, if the item recognition unit 112 determines that there is a problem with the result of the false detection determination, it may terminate the pre-registration process without returning to step S501. 【0076】 If the item recognition unit 112 determines that there is no problem in the false detection determination (S503: YES), the item learning unit 113 includes the X-ray image acquired in step S501 in the learning data 163 and performs retraining of the item recognition model 161 to add the category of the item to be registered as a new category to the item recognition model 161 based on the learning data 163 (S504). In this embodiment, the item recognition model 161 can recognize categories for both prohibited items and special items, but separate item recognition models 161 capable of recognizing categories for prohibited items and items capable of recognizing categories for special items may be provided. 【0077】 Furthermore, since retraining the object recognition model 161 takes a considerable amount of time, it is desirable for the object learning unit 113 to perform retraining in batch processing during times when object recognition processing is not being performed, such as at night. 【0078】 Furthermore, relearning requires an X-ray image annotated with respect to the registered items (the annotation indicates the category of the registered items included in the X-ray image and the position of the registered items within the X-ray image). For example, in step S501, annotation of the registered items in the X-ray image may be performed by user input via the input device 4, or in step S501, an X-ray image of a tray containing only the registered items may be acquired, and in step S504, the item learning unit 113 may perform automatic annotation on the X-ray image using methods such as background subtraction or interactive segmentation. 【0079】 Furthermore, the item learning unit 113 may generate annotations by combining images of the item to be registered, extracted from X-ray RAW data taken from several directions of the item to be registered, with a pre-prepared background image. For example, this can be done by the method described in the patent document (Japanese Patent Application Publication No. 2021-026742). 【0080】 The item recognition unit 112 performs an overdetection determination using the item recognition model 161 retrained in step S504 and the X-ray images contained in the X-ray image storage data 162 (S505). Specifically, for example, the item recognition unit 112 calculates the percentage (hereinafter also called the overdetection percentage) in which each X-ray image contained in the X-ray image storage data 162 is determined to be included in the category of the item to be registered when input into the retrained item recognition model 161. Note that at the time of calculating the overdetection percentage, the first and second thresholds corresponding to the item to be registered have not been determined, so a predetermined threshold is used as a confidence threshold for determining whether the item to be registered is included. 【0081】 The item recognition unit 112 determines whether there is a problem with the result of the overdetection determination (S506). Specifically, for example, the item recognition unit 112 determines that there is no problem with the result of the overdetection determination if the overdetection rate is less than a predetermined value, and determines that there is a problem with the result of the overdetection determination if the overdetection rate is equal to or greater than the predetermined value. 【0082】 For example, if a newly released keychain is a registered item, and if information indicating the release date of the keychain is obtained in step S501, then, assuming that the keychain did not exist before that release date, the item recognition unit 112 may, in calculating the false detection rate in step S505, only include images from before the release date among the X-ray image storage data 162 as input to the item recognition model 161. 【0083】 Furthermore, if the registered items include items whose release date is unknown or items for which it is difficult to specify a release date due to their long history, the item recognition unit 112 may, for example, obtain the percentage of images that do not include the registered item from the X-ray images input to the item recognition model 161 in step S505 via input to the input device 4, and determine in step S506 whether the difference obtained by subtracting this percentage from the calculated over-detection percentage is greater than or equal to a predetermined value. In other words, the item recognition unit 112 may determine that there is no problem with the over-detection determination result if the difference is less than the predetermined value, and may determine that there is a problem with the over-detection determination result if the difference is greater than or equal to the predetermined value. 【0084】 The execution of the false detection determination in step S505 reduces the probability that an event corresponding to the registered item will be falsely generated in the item recognition process, even if an X-ray image that does not contain the item in question is input. 【0085】 If the item recognition unit 112 determines that there is a problem with the result of the over-detection judgment (S506: NO), it discards the retrained item recognition model 161 and returns to the state before retraining (S509), and returns to step S501 (i.e., the category of the item to be registered is re-selected). It is desirable to temporarily save the item recognition model 161 before retraining is performed in step S504 in order to process step S509. Alternatively, after executing the process in step S509, the item recognition unit 112 may terminate the pre-registration process without returning to step S501. 【0086】 Furthermore, if the item recognition unit 112 determines that there is a problem with the result of the over-detection determination, it may, for example, randomly display on the display device 3 some of the X-ray images used in the over-detection determination in step S505 that were determined to contain the item to be registered, and allow the user to confirm this. At this time, the item recognition unit 112 may also receive confirmation results from the user via the input device 4 indicating whether there is a problem with the result of the over-detection determination. If the confirmation results indicate that there is a problem with the result of the over-detection determination, the item recognition unit 112 may proceed to step S509, and if the confirmation results indicate that there is no problem with the result of the over-detection determination, it may proceed to step S507, which will be described later. 【0087】 If the item recognition unit 112 determines that there is no problem with the result of the over-detection judgment (S506: YES), the item learning unit 113 adjusts the confidence threshold (S507). Specifically, for example, the item recognition unit 112 randomly samples some of the X-ray images acquired in step S501 as evaluation images (it is desirable that these evaluation images are not used for learning in step S506), generates an item recognition model that recognizes the category of the item to be registered using the evaluation images, and uses the generated model and the evaluation images to obtain, for example, a confidence threshold of 95% or more as the first threshold (confidence that the item to be registered is definitely included in the X-ray image), and a confidence threshold of 50% or more as the second threshold (confidence that the item to be registered is not definitely included in the X-ray image, but may be included). 【0088】 The item recognition unit 112 may acquire the first threshold and the second threshold according to the user input via the input device 4. Furthermore, if the item to be registered is a prohibited item, it is desirable to set only the first threshold, as it is undesirable for detection failures to occur in the item recognition process described later. 【0089】 Next, the item registration unit 116 registers the item in the item attribute data 164, and the event generation unit 114 registers an event related to the registered item in the event data 165 (S508), and the pre-registration process ends. 【0090】 In step S508, specifically, for example, the item registration unit 116 generates an item ID and registers the item ID, the category of the item to be registered, the type of item to be recognized, the first threshold, and the second threshold in the item attribute data 164. The event generation unit 114 also receives input of the event name, event type (however, if the item to be registered is a prohibited item, only "Prohibited Item Response" can be selected, and if the item to be registered is a special item, "Special Item Confirmation" and "Enjoyment" can be selected), the item ID of the item to be registered, the event occurrence conditions, and the event content via the input device 4 and registers the input information in the event content data 1650. 【0091】 Furthermore, in step S508, the event generation unit 114 may register not only information about events related to the registered item, but also information about events related to lucky items. Specifically, for example, the event generation unit 114 receives input of the event name, event type (however, since prohibited items cannot be lucky items, "Special Item Confirmation" or "Fun" can be selected as the event type) via the input device 4, and registers the input information in the event content data 1650. 【0092】 Furthermore, in order to make it difficult for visitors to guess which items are lucky items, it is preferable that the setting of lucky item data 1657 (i.e., the setting of which items correspond to lucky items) is not performed during the pre-registration process, but rather is performed, for example, in the process of step S701 described later in the item recognition process (for example, a process performed at the start of business each day at the facility). 【0093】 Furthermore, if the item to be registered is a prohibited item, it is desirable that information regarding the prohibited item be reliably registered in the item recognition model 161, item attribute data 164, and event data 165. Therefore, it is desirable to transition from step S501 to step S504, and from step S505 to step S504 to step S507. 【0094】 <Item Recognition Processing> Figure 7 is a flowchart showing an example of the item recognition process. The event generation unit 114 sets the event data 165 (S701). Specifically, for example, the event generation unit 114 sets the lucky item in the lucky item data 1657, sets the value of the occurrence flag in the event content data 1650, and updates other values ​​in the event data 165, etc., via input from a user (e.g., an inspector or the administrator of the X-ray image processing device 1) to the input device 4. For example, by executing the process in step S701 at the start of business each day at the facility, events that may occur on that day and lucky items are set. Note that for events of the event type "Prohibited Item Response", only "1" can be set as the occurrence flag (i.e., the event will be executed as long as the occurrence conditions are met). 【0095】 The X-ray image acquisition unit 111 acquires an X-ray image of the package from the X-ray RAW data input from the main unit 2 of the device (S702). The item recognition unit 112 performs item recognition on the X-ray image acquired in step S701 (S703). Specifically, for example, the item recognition unit 112 inputs the X-ray image acquired in step S701 into the item recognition model 161 and outputs the category of the item contained in the X-ray image, the confidence level corresponding to the item, and the position of the item on the X-ray image as the item recognition result. 【0096】 The item recognition unit 112 determines, based on the results of item recognition in step S703, whether a prohibited item is included in the X-ray image (whether the conditions for an event related to a prohibited item are met) (S704). 【0097】 Specifically, for example, the item recognition unit 112 determines that a prohibited item is included in the X-ray image if, in the item recognition result in step S703, the category of the item corresponding to the prohibited item indicated by the item attribute data 164 is output, there is an event in the event content data 1650 where the occurrence flag corresponding to that category is "1", and the confidence level corresponding to the item output in step S703 is equal to or greater than the first threshold corresponding to the category of the item indicated by the item attribute data 164. Otherwise, it determines that a prohibited item is not included in the X-ray image. 【0098】 If the item recognition unit 112 determines that no prohibited items are included in the X-ray image (S704: NO), it proceeds to step S707, which will be described later. If the item recognition unit 112 determines that a prohibited item is included in the X-ray image (S704: YES), the screen display unit 115 displays a prohibited item confirmation screen on the display device 5 (i.e., the screen for visitors) (S705). 【0099】 Figure 8 shows an example of a prohibited items confirmation screen. The prohibited items confirmation screen displayed on the display device 5 shows, for example, the X-ray image acquired in step S702. The prohibited items confirmation screen also displays a frame 800 indicating the location (area) where the prohibited items determined to be included in step S703 are displayed, and a message 810 indicating instructions for the visitor. 【0100】 Message 810 displays a message to prompt the visitor to confirm whether there are any prohibited items in the X-ray image, such as, "Is this a plastic bottle? If it is, please touch within the frame. If there is no plastic bottle, please wait 10 seconds." (Assume that the display device 5 is a touch panel that also functions as an input device 4). When the screen display unit 115 receives feedback from the visitor indicating that there are no prohibited items in the X-ray image (in the example in Figure 8, if 10 seconds have passed without the frame 800 being touched), it notifies the inspector, for example, by displaying information indicating the feedback on the display device 3. In this case, the inspector reconfirms the X-ray image displayed on the display device 3 or the display device 5, and, at the inspector's discretion, speaks to the visitor or checks the contents of the visitor's luggage as necessary. 【0101】 In this way, by displaying a screen for checking prohibited items, visitors themselves can check for prohibited items, reducing the workload of inspectors while allowing visitors to enjoy the process of checking for prohibited items. 【0102】 Returning to the explanation of Figure 7, the screen display unit 115, when it receives feedback from a visitor indicating that a prohibited item is included in the X-ray image (in the example in Figure 8, when frame 800 is touched), executes a prohibited item handling guidance event corresponding to the item in the event content data 1650 (S706). 【0103】 In events providing guidance on how to handle prohibited items, audio guides prompting participants to take appropriate action, such as disposing of the prohibited items or storing them in lockers, are output from speakers or displayed on the display device 5. 【0104】 Next, the item recognition unit 112 determines, based on the results of item recognition in step S703, whether a special item is included in the X-ray image (whether the conditions for the occurrence of an event related to a special item are met) (S707). 【0105】 Specifically, for example, the item recognition unit 112 determines that a special item is included in the X-ray image if, in the item recognition result in step S703, the category of the item corresponding to the special item indicated by the item attribute data 164 is output, there is an event in the event content data 1650 with an occurrence flag of "1" corresponding to that category, and the confidence level corresponding to the item output in step S703 is equal to or greater than the first threshold corresponding to the category of the item indicated by the item attribute data 164. Otherwise, it determines that a special item is not included in the X-ray image. 【0106】 If the item recognition unit 112 determines that no special item is included in the X-ray image (S707: NO), it proceeds to step S711, which will be described later. If the item recognition unit 112 determines that a special item is included in the X-ray image, but its reliability is low (i.e., it determines that the reliability corresponding to the special item is above the first threshold but below the second threshold) (S707: YES (low reliability)), the screen display unit 115 executes a special item confirmation event corresponding to the item in the event content data 1650 (S708). 【0107】 If the confidence level for a special item indicated by the item recognition result is above the first threshold but below the second threshold, there is a significant possibility that the special item is not included in the X-ray image (or that another item has been mistakenly identified as the special item). Therefore, by triggering a special item confirmation event, inspectors can confirm whether a special item is included in the luggage in an entertaining way for visitors. 【0108】 In the special item verification event, a message such as "Do you have a character A keychain?" is displayed on the display device 5 to confirm whether the special item is included in the package. For example, the user inputs feedback to the message (i.e., feedback indicating whether the special item is included in the package) via the display device 5 (for example, the display device 5 is a touch panel). 【0109】 The screen display unit 115 confirms whether the special item is included in the package according to the feedback result (S709). If the screen display unit 115 receives feedback indicating that the special item is not included in the package (S709: NO), it proceeds to step S711. 【0110】 If the screen display unit 115 receives feedback indicating that the special item is included in the luggage (S709: YES), it executes an entertainment event corresponding to the item in the event content data 1650 (S710). In the entertainment event, a message associated with the special item that is intended to entertain the visitor is displayed on the display device 5 or output from the speaker. In addition, during the entertainment event, a benefit associated with the special item (for example, a two-dimensional code indicating a coupon) may be displayed on the display device 5. Through such an entertainment event, the user can enjoy undergoing the X-ray inspection of their luggage. 【0111】 In addition, there is a possibility that visitors may enter false feedback during the special item verification event in step S708. However, even if false feedback is entered, it will only affect whether or not the fun event occurs, and there will be no major operational risks. Therefore, inspectors are not required to speak to visitors or check their luggage based on the feedback results. 【0112】 Furthermore, if the item recognition unit 112 determines that a special item is included in the X-ray image and that its reliability is high (i.e., that the reliability corresponding to the special item is determined to be at or above the second threshold) (S707: YES (high reliability)), there is no need to confirm with the visitor that the special item is included in the luggage, and the process proceeds to step S710. 【0113】 In the example shown in Figure 7, the event providing guidance on how to deal with prohibited items and the entertainment event are displayed at different times, but it is also acceptable to display both the event providing guidance on how to deal with prohibited items and the entertainment event on the same screen. 【0114】 Figure 9 shows an example of a screen that simultaneously displays an event guiding users on how to handle prohibited items and an entertainment event. In the example in Figure 9, items detected in the X-ray image are replaced with illustrations. Specifically, for example, prohibited items are replaced with illustrations of bad characters or illustrations that convey a negative image, and special items are replaced with illustrations of characters related to the special item or illustrations of the special item itself. 【0115】 In the example shown in Figure 9, the X-ray image 900 recognizes a prohibited item (a knife 901), special items (binoculars 902, a spray bottle 903, and a dry cell battery 905), and a special item that is also a lucky charm (a wristwatch 904). 【0116】 At this time, the display device 5 actually displays the prohibited item confirmation screen 910, and instead of the knife 901, binoculars 902, spray bottle 903, wristwatch 904, and dry cell battery 905, the prohibited item confirmation screen 910 displays illustrations of a scorpion 911, binoculars 912, spray bottle 913, wristwatch 914, and dry cell battery 915, which give a bad image. 【0117】 Furthermore, at the same time these illustrations are displayed, audio indicating the content of the ongoing event may be output from the speaker. The illustrations corresponding to prohibited items and special items are registered in the item attribute data 164, for example, before the item recognition process shown in Figure 7 begins. Instead of displaying the X-ray images directly on the display device 5, replacing prohibited items and special items with illustrations, as shown in Figure 9, allows visitors to enjoy the inspection more. 【0118】 Furthermore, such illustrations may be used on the prohibited items confirmation screen and special items confirmation events, as well as when prohibited items handling guidance events and fun events occur separately. 【0119】 Returning to the explanation of Figure 7, the X-ray image acquisition unit 111 determines whether the X-ray inspection is complete (S711). The X-ray image acquisition unit 111 may determine that the X-ray inspection is complete when a predetermined time, such as the closing time of business, arrives, or it may determine that the X-ray inspection is complete when it receives an X-ray inspection completion instruction via the input device 4. 【0120】 If the X-ray image acquisition unit 111 determines that the X-ray inspection is complete (S711: YES), it terminates the item recognition process. If the X-ray image acquisition unit 111 determines that the X-ray inspection is not complete (S711: NO), it returns to step S701. 【0121】 Furthermore, steps S708 and S709 may be omitted. In other words, the special item confirmation event may be prevented from occurring. In this case, the special item confirmation event does not need to be defined in the event content data 1650, nor does the second threshold need to be set. Also, for some special items, the processing in steps S708 and S709 may be omitted. 【0122】 In steps S704 and S707, if prohibited items or special items are included in the X-ray image (the conditions for occurrence are met) and the occurrence flag is "1", the corresponding event is guaranteed to occur. However, for example, the probability of occurrence for each event may be predetermined, and if the conditions for occurrence are met and the occurrence flag is "1", the event may occur according to the probability of occurrence for the corresponding event. 【0123】 However, the probability of events corresponding to special items and prohibited items can be greater than 0 and less than or equal to 1, but it is desirable that the probability of events corresponding to prohibited items always be 1. On the other hand, for events corresponding to special items, since no major operational problems occur even if they do not occur, the probability of occurrence may be set to less than 1. 【0124】 In this way, by setting a probability of occurrence, it becomes more difficult for visitors to guess which items are part of the event. In particular, for entertainment events that offer benefits to visitors, such as issuing coupons, setting a probability of occurrence of less than 1 can reduce the operating costs of the facility. 【0125】 Furthermore, if, for example, multiple X-ray image processing systems are installed in a facility, by setting different event content data 1650 (e.g., the items targeted by the event) and lucky item data 1657 for each X-ray image processing system, it becomes more difficult for visitors to guess which items are the items targeted by the event. 【0126】 Furthermore, for example, by setting a higher probability of events corresponding to certain special items in only some of the multiple X-ray image processing systems, it is possible to create lanes in X-ray inspections where events corresponding to special items are more likely to occur and lanes where events corresponding to special items are less likely to occur. In particular, in X-ray image processing systems that correspond to certain lanes that are difficult for visitors to access, such as lanes at the edge of a facility, setting a higher probability of events corresponding to special items than in other lanes can help distribute congestion among visitors seeking X-ray inspections. 【0127】 In summary, the X-ray image processing device 1 of this embodiment, by performing the aforementioned events corresponding to prohibited items and special items, allows visitors to enjoy X-ray inspections, which tend to give a negative impression, and also reduces the workload of inspectors. 【0128】 Furthermore, the X-ray image processing device 1 of this embodiment can increase sales of souvenirs and other items related to special items within the facility by conducting entertainment events related to the special items, thereby improving visitors' impressions of the special items and allowing visitors to receive benefits or other perks corresponding to the special items. 【0129】 Furthermore, the X-ray image processing device 1 of this embodiment can display a prohibited item confirmation screen or execute a special item confirmation event, thereby confirming the X-ray image inspection results (whether prohibited items or special items are in the luggage) through interaction between the visitor and the X-ray image processing system, and ultimately reducing the workload of inspectors while entertaining visitors. 【0130】 It should be noted that the present invention is not limited to the embodiments described above, and various modifications are included. For example, the embodiments described above are described in detail to make the present invention easier to understand, and are not necessarily limited to those having all the configurations described. It is also possible to replace parts of the configuration of one embodiment with the configuration of another embodiment, and it is also possible to add configurations from other embodiments to the configuration of one embodiment. Furthermore, it is possible to add, delete, or replace parts of the configuration of each embodiment with other configurations. 【0131】 Furthermore, each of the above configurations, functions, processing units, and processing means may be implemented in hardware, either partially or entirely, by designing them as integrated circuits, for example. Alternatively, each of the above configurations and functions may be implemented in software by having the processor interpret and execute programs that implement each function. Information such as programs, tables, and files that implement each function can be stored in memory, a recording device such as a hard disk or SSD (Solid State Drive), or a recording medium such as an IC card, SD card, or DVD. 【0132】 Furthermore, the control lines and information lines shown are those deemed necessary for explanatory purposes, and not all control lines and information lines are necessarily shown in the actual product. In reality, it is safe to assume that almost all components are interconnected. [Explanation of symbols] 【0133】 1 X-ray image processing device, 3 Display device, 4 Input device, 5 Display device, 11 CPU, 12 Memory, 13 Display interface, 14 Input interface, 15 Communication device, 16 Auxiliary storage device, 111 X-ray image acquisition unit, 112 Item recognition unit, 113 Item learning unit, 114 Event generation unit, 115 Screen display unit, 116 Item registration unit, 161 Item recognition model, 162 X-ray image storage data, 163 Learning data, 164 Item attribute data, 165 Event data

Claims

[Claim 1] X-ray inspection device, Equipped with a processor and memory, The aforementioned memory is X-ray images of visitors' belongings to the facility, A model that recognizes the category of items contained in an X-ray image from that X-ray image, Item attribute information indicating the category of special items, The system maintains event information indicating the category of the special item, the content of the Type 1 event that provides benefits to the visitors, and the conditions under which the Type 1 event occurs. The aforementioned special items are items that are not prohibited from being brought into the facility related to the aforementioned conditions for occurrence, The aforementioned processor, Based on the X-ray image of the package and the model, the category of the special item included in the X-ray image is recognized. The content of the first type event corresponding to the category of the special item recognized above is identified from the event information, An X-ray inspection device that outputs data based on the content of the aforementioned identified Type 1 event. [Claim 2] An X-ray inspection apparatus according to claim 1, The aforementioned model outputs a confidence level for the category of items recognized as being included in the X-ray image, The aforementioned item attribute information indicates, for each category of the special item, a first threshold for the reliability and a second threshold higher than the first threshold. The event information indicates the category of the special item and the content of a Type 2 event to allow the visitor to confirm whether the special item is included in the package. The aforementioned processor, The confidence level for the category of the recognized special item is obtained based on the X-ray image of the package and the model. If the acquired confidence level is equal to or greater than the second threshold, The content of the first type event corresponding to the category of the special item recognized above is identified from the event information, Output data based on the content of the aforementioned identified Type 1 event. If the acquired confidence level is greater than or equal to the first threshold and less than the second threshold, The content of the Type 2 event corresponding to the category of special item recognized above is identified from the event information, An X-ray inspection device that outputs data based on the content of the aforementioned identified Type 2 event. [Claim 3] An X-ray inspection apparatus according to claim 2, The aforementioned processor, If, after outputting data based on the content of the aforementioned Type 2 event, a confirmation result is obtained indicating that the special item is included in the package, The content of the first type event corresponding to the category of the special item recognized above is identified from the event information, An X-ray inspection device that outputs data based on the content of the aforementioned identified Type 1 event. [Claim 4] An X-ray inspection apparatus according to claim 2, An X-ray inspection apparatus in which the first threshold and the second threshold for each category are determined based on a recall rate derived from a plurality of evaluation X-ray images containing articles of that category, and an evaluation model that recognizes the category contained in the X-ray images from the X-ray images. [Claim 5] An X-ray inspection apparatus according to claim 1, The aforementioned memory is It stores multiple X-ray images of registered special articles, each being an X-ray image of a registered special article, and information indicating the category of the registered special article. The aforementioned processor, Each of the multiple X-ray images to be registered is input into the model, and a false detection determination is performed to calculate the percentage of cases where a category different from the category of the special item to be registered is recognized. An X-ray inspection device that determines, based on the result of the false detection determination, whether the model performs learning to enable it to recognize the category of special items to be registered. [Claim 6] An X-ray inspection apparatus according to claim 5, Each of the aforementioned memories holds a plurality of past X-ray images, each of which is an image of an item in a category that the model can output. The aforementioned processor, Based on the aforementioned model and the plurality of X-ray images to be registered, a retrained model capable of recognizing the category of the special item to be registered is generated. The above-mentioned multiple past X-ray images are input into the retraining model to perform a false positive detection, which calculates the percentage of times the category of the special item to be registered is recognized. Based on the results of the false positive detection, a decision is made to either discard the retrained model or update the model using the retrained model. If it is decided to update the aforementioned model, The category of the special item to be registered is registered in the item attribute information. The content of the Type 1 event corresponding to the category of the special item to be registered is obtained, An X-ray inspection device that registers the category of the special item to be registered and the content of the acquired Type 1 event in the event information. [Claim 7] An X-ray inspection apparatus according to claim 1, The event information indicates the probability of the occurrence of the first type event, An X-ray inspection apparatus, wherein the processor identifies the content of a first-class event corresponding to the category of the recognized special item from the event information according to the probability of occurrence. [Claim 8] An X-ray inspection apparatus according to any one of claims 1 to 7, An X-ray inspection device characterized by detecting prohibited items that are not allowed to be brought into the aforementioned facility. [Claim 9] A program for causing an X-ray inspection device to perform X-ray inspection processing, The aforementioned X-ray inspection apparatus has a processor and memory, The aforementioned memory is X-ray images of visitors' belongings to the facility, A model that recognizes the category of items contained in an X-ray image from that X-ray image, Item attribute information indicating the category of special items, The system maintains event information indicating the category of the special item, the content of the Type 1 event that provides benefits to the visitors, and the conditions under which the Type 1 event occurs. The aforementioned special items are items that are not prohibited from being brought into the facility related to the aforementioned conditions for occurrence, The aforementioned program, A process to recognize the category of the special item included in the X-ray image based on the X-ray image of the package and the model, A process to identify the content of a Type 1 event corresponding to the category of the recognized special item from the event information, A program that causes the processor to perform a process to output data based on the content of the identified Type 1 event. [Claim 10] The program according to claim 9, The aforementioned model outputs a confidence level for the category of items recognized as being included in the X-ray image, The aforementioned item attribute information indicates, for each category of the special item, a first threshold for the reliability and a second threshold higher than the first threshold. The event information indicates the category of the special item and the content of a Type 2 event to allow the visitor to confirm whether the special item is included in the package. The aforementioned program, The processor is instructed to perform a process to obtain the confidence level for the category of the recognized special item based on the X-ray image of the package and the model. If the confidence level obtained by the aforementioned processor is equal to or greater than the second threshold, A process to identify the content of a Type 1 event corresponding to the category of the recognized special item from the event information, The processor is instructed to perform a process that outputs data based on the content of the identified Type 1 event. If the confidence level obtained by the processor is greater than or equal to the first threshold and less than the second threshold, A process to identify the content of a Type 2 event corresponding to the category of the recognized special item from the event information, A program that causes the processor to perform a process to output data based on the content of the identified Type 2 event. [Claim 11] The program according to claim 10, If the processor outputs data based on the content of the second type event and then obtains a confirmation result indicating that the special item is included in the package, A process to identify the content of a Type 1 event corresponding to the category of the recognized special item from the event information, A program that causes the processor to perform a process to output data based on the content of the identified Type 1 event. [Claim 12] The program according to claim 10, A program in which the first threshold and the second threshold for each category are determined based on a recall rate derived from a plurality of evaluation X-ray images containing items of that category, and an evaluation model that recognizes the category contained in the X-ray images. [Claim 13] The program according to claim 9, The aforementioned memory is It stores multiple X-ray images of registered special articles, each being an X-ray image of a registered special article, and information indicating the category of the registered special article. The aforementioned program, The process involves inputting each of the multiple X-ray images to be registered into the model and calculating the percentage of cases where a category different from the category of the special item to be registered is recognized, thereby determining whether a false positive occurred. A program that causes the processor to perform a process to determine, based on the result of the false detection determination, whether the model performs learning to enable it to recognize the category of the special item to be registered. [Claim 14] The program according to claim 13, Each of the aforementioned memories holds a plurality of past X-ray images, each of which is an image of an item in a category that the model can output. The aforementioned program, A process to generate a retrained model capable of recognizing the category of the special item to be registered, based on the aforementioned model and the plurality of X-ray images to be registered. The above-mentioned multiple past X-ray images are input into the retraining model to calculate the percentage of times the category of the special item to be registered is recognized, and a false positive detection is performed. The processor is instructed to perform a process to determine whether to discard the retrained model or update the model using the retrained model, based on the result of the over-detection determination. If the processor decides to update the model, A process to register the category of the special item to be registered in the item attribute information, A process to obtain the content of a Type 1 event corresponding to the category of the special item to be registered, A program that causes the processor to perform a process of registering the category of the special item to be registered and the content of the acquired Type 1 event in the event information. [Claim 15] The program according to claim 9, The event information indicates the probability of the occurrence of the first type event, The program causes the processor to perform a process to identify the content of a Type 1 event corresponding to the category of the recognized special item from the event information, according to the probability of occurrence.

Images