Intelligent image recognition device testing tool and testing method

Abstract

The application discloses an intelligent image recognition equipment testing tool and testing method, which comprises a message queue middleware and a testing result confirmation device. The message queue middleware comprises a module for simultaneously acquiring identification results of contraband identification of multiple intelligent image recognition equipment for the same X-ray machine. The identification results from different intelligent image recognition equipment are stored in different topics. The testing result confirmation device comprises a module for subscribing to the identification results in different topics, labeling the identification results with manufacturer labels of the intelligent image recognition equipment based on different topics, selecting corresponding audit labels from optional audit labels and configuring the audit labels to the identification results with the manufacturer labels, counting the number of the audit labels, and obtaining an identification rate according to the number. With the aid of storage technology and computer human-computer interaction technology or automatic comparison technology, the testing is realized on a complete online processing, so that the loss, discontinuity and inaccuracy of archives are avoided, the testing result is reliable, and the testing labor cost is reduced.

Description

Technical Field

[0001] This invention relates to the field of testing intelligent image recognition devices, and more specifically to testing tools and methods for intelligent image recognition devices. Background Technology

[0002] See appendix Figure 1 After the package passes through the X-ray machine, the X-ray machine performs a perspective scan of the package and outputs a video stream. The role of the intelligent image recognition device is to cut the package from the video stream to obtain a package image, and then identify the items in the package image. The identified prohibited items are called the recognition results, which generally include: the name of the prohibited item and the coordinates of the prohibited item recognition box.

[0003] Different manufacturers' smart image recognition devices may have varying recognition rates for different prohibited items. To test the recognition rates of smart image recognition devices from different manufacturers, such as... Figure 1 As shown, the existing method generally involves imaging the recognition results on a monitor, then manually judging whether the recognition is correct and recording it in a form. This method requires setting up the same number of monitors as the intelligent image recognition device under test, and observing and recording them synchronously, which increases the workload of the recorders and is prone to errors. Moreover, this method can only record whether the recognition result is correct or not, and cannot supplement misidentification or omission. In addition, this recording method cannot form a record, and cannot check whether the judgment process is correct, ultimately resulting in low reliability of the recognition rate.

[0004] This work mode requires testers to observe the recognition results on multiple monitors simultaneously and to make judgments within a short period of time. This requires testers to have a high level of experience and is not friendly to new testers. Summary of the Invention

[0005] The purpose of this invention is to provide a testing tool and method for intelligent image recognition devices. The device uses storage technology to store the recognition results one by one and uses push and judgment technology to manually or automatically judge the recognition results. Finally, the judgment results are automatically counted and the recognition rate is automatically calculated. The device and method do not rely on multiple displays or experienced testers. They can classify and mark misidentified and missed recognitions when the recognition results are incorrect, so as to realize records that can be used for reverse lookup.

[0006] On the one hand, intelligent image recognition equipment testing tools include: message queue middleware and test result confirmation devices;

[0007] Message queue middleware includes:

[0008] Memory;

[0009] One or more processors; and

[0010] one or more modules residing in memory and configured for execution by the one or more processors, the one or more modules comprising:

[0011] a module for simultaneously acquiring identification results of a plurality of intelligent image recognition devices for video streams of the same X-ray machine for contraband identification;

[0012] a module for storing identification results from different intelligent image recognition devices in different topics;

[0013] the test result confirmation device comprises:

[0014] a memory;

[0015] one or more processors; and

[0016] one or more modules residing in memory and configured for execution by the one or more processors, the one or more modules comprising:

[0017] a module for subscribing to identification results in different topics, labeling the identification results with manufacturer labels of intelligent image recognition devices based on different topics, and storing the identification results with manufacturer labels in a database;

[0018] a module for acquiring identification results with manufacturer labels from the database;

[0019] a module for auditing the identification results with manufacturer labels, and selecting corresponding audit labels from optional audit labels to configure the identification results with manufacturer labels, to obtain identification results of contraband identification with audit labels and manufacturer labels;

[0020] a module for counting the number of audit labels of identification results of the same manufacturer label and the same contraband identification;

[0021] a module for obtaining an identification rate of the same manufacturer label on identification results of the same contraband identification according to the number;

[0022] The message queue middleware can be a kafka middleware or other equivalent middleware.

[0023] Preferably,

[0024] When the test result confirmation device is a human-computer interaction device, the identification result comprises: a package picture corresponding to the identification result, a contraband name in the package picture, and a bounding box coordinate of the contraband in the package picture.

[0025] The test result confirmation device further comprises:

[0026] a display,

[0027] The module for generating the identification frame line according to the identification frame coordinates of the contraband, superimposing the identification frame line on the package picture, and imaging on the display.

[0028] The module for displaying the name of the contraband in the display.

[0029] Preferably,

[0030] When the test result confirmation device is a man-machine interaction device, the optional audit label includes an optional correct audit label, an optional misidentification audit label, and an optional missed identification audit label.

[0031] The module for generating the optional correct audit label, the optional misidentification audit label, and the optional missed identification audit label;

[0032] When the current identification result with the manufacturer's label is audited, the module for selecting at least one audit label from the optional correct audit label, the optional misidentification audit label, and the optional missed identification audit label;

[0033] The module for configuring the selected audit label to the identification result with the manufacturer's label.

[0034] Preferably,

[0035] The optional misidentification audit label and the optional missed identification audit label are provided with a pull-down optional contraband name.

[0036] Preferably,

[0037] When the test result confirmation device is an automatic confirmation device,

[0038] The optional audit label includes an optional correct audit label and an optional misidentification audit label.

[0039] The module for auditing the identification result with the manufacturer's label and selecting a corresponding audit label from the optional audit label to configure the identification result with the manufacturer's label to obtain the identification result of the contraband identification with the audit label and the manufacturer's label includes:

[0040] When the current identification result with the manufacturer's label is audited, the module for comparing the preset correct contraband name in the package picture with the contraband name in the identification result of the package picture;

[0041] When the comparison result is the same, the module of selecting the correct audit label to configure the identification result of the bag with the manufacturer label is selected, and when the comparison result is different, the module of selecting the misidentification audit label to configure the identification result of the bag with the manufacturer label is selected.

[0042] In another aspect, the intelligent image recognition device testing method comprises the following steps,

[0043] S1. Packing the prohibited articles into the luggage and passing the luggage through the X-ray machine;

[0044] S2. Multiple intelligent image recognition devices simultaneously perform prohibited article identification on the video stream of the same X-ray machine and output the identification results;

[0045] S3. Through the message queue middleware, the identification results of the multiple intelligent image recognition devices performing prohibited article identification on the video stream of the same X-ray machine are simultaneously obtained, and the identification results from different intelligent image recognition devices are designated to be stored in different topics;

[0046] S4. Through the test result confirmation device, the identification results in different topics are subscribed, the manufacturer labels of the intelligent image recognition devices are added to the identification results based on the different topics, and the identification results with the manufacturer labels are stored in the database;

[0047] S5. Through the test result confirmation device, the identification results with the manufacturer labels are obtained from the database;

[0048] S6. Through the test result confirmation device, the identification results with the manufacturer labels are audited and the corresponding audit labels are selected from the optional audit labels to configure the identification results with the manufacturer labels, and the identification results of the prohibited article identification with the audit labels and the manufacturer labels are obtained;

[0049] S7. Through the test result confirmation device, the number of audit labels of the identification results of the same manufacturer label and the same prohibited article identification is counted;

[0050] S8. Through the test result confirmation device, the identification rate of the identification results of the same manufacturer label on the same prohibited article identification is obtained according to the count number.

[0051] Preferably, when the test result confirmation device is a human-computer interaction device, the identification result includes: the package picture corresponding to the identification result, the name of the prohibited article in the package picture, and the identification box coordinates of the prohibited article in the package picture; and the optional audit label includes: the optional correct audit label, the optional misidentification audit label, and the optional missed identification audit label.

[0052] S1 specifically includes: when a batch of tests are performed, multiple prohibited articles are randomly packed into multiple luggage and the luggage is passed through the X-ray machine;

[0053] S6 is specifically:

[0054] The test result confirmation device generates an identification frame line according to the identification frame coordinates of the contraband, superimposes the identification frame line on the package picture, and images on the display. The name of the contraband is displayed in the display;

[0055] The test result confirmation device generates optional correct audit labels, optional misidentification audit labels, and optional missed identification audit labels.

[0056] The test result confirmation device selects at least one audit label from the optional correct audit labels, the optional misidentification audit labels, and the optional missed identification audit labels.

[0057] The test result confirmation device configures the selected audit label to the identification result with the manufacturer's label.

[0058] Preferably,

[0059] The specific calculation process of S8 is:

[0060] The number of correct audit labels, misidentification audit labels, and missed identification audit labels is obtained respectively.

[0061] The recognition rate is calculated according to formula 1.

[0062] Formula 1: Recognition rate = number of correct audit labels / (number of correct audit labels + number of misidentification audit labels + number of missed identification audit labels) %.

[0063] Preferably,

[0064] When the test result confirmation device is an automatic confirmation device, the optional audit label includes an optional correct audit label and an optional misidentification audit label. The identification result includes the name of the contraband in the package picture.

[0065] S1 is specifically: when performing a batch of tests, multiple contrabands are loaded into a suitcase, and the names of the contrabands are recorded. The suitcase passes through the X-ray machine.

[0066] S6 is specifically:

[0067] According to the record of the name of the contraband, the names of the contrabands in the preset correct package pictures of the current batch of tests are entered into the test result confirmation device.

[0068] The test result confirmation device compares the names of the contrabands in the preset correct package pictures with the names of the contrabands in the package pictures of the identification result.

[0069] When the comparison result is the same, the test result confirmation device selects a correct audit label to configure to the identification result with the manufacturer's label.

[0070] When the comparison result is different, the test result confirmation device selects the misrecognition audit label to configure the identification result of the band manufacturer label.

[0071] Preferably,

[0072] The specific calculation process of S8 is as follows:

[0073] The number of correct audit labels and the number of misrecognition audit labels are obtained respectively;

[0074] The recognition rate is calculated according to formula 2.

[0075] Formula 2: Recognition rate = number of correct audit labels / (number of correct audit labels + number of misrecognition audit labels) %.

[0076] The beneficial effects of the present application are as follows: the device uses storage technology to store identification results one by one and uses the way of one-by-one pushing and judging technology to artificially or automatically judge the identification results, finally automatically counts the judgment results and automatically calculates the recognition rate. The device and method do not need to rely on multiple displays, do not need to rely on experienced test personnel, can realize the classification marking of misrecognition and omission when the identification result is incorrect, and can realize the record for reverse checking. BRIEF DESCRIPTION OF DRAWINGS

[0077] Figure 1 It is a structural deployment schematic diagram of the existing test method.

[0078] Figure 2 It is a structural deployment schematic diagram of the present application.

[0079] Figure 3 It is a processing flow schematic diagram of the present application.

[0080] Figure 4 It is a man-machine interaction interface schematic diagram of the test result confirmation device.

[0081] Figure 5 It is a man-machine interaction interface schematic diagram of the test result confirmation device selecting the misrecognition audit label and configuring the correct article.

[0082] Figure 6 It is a man-machine interaction interface schematic diagram of the test result confirmation device simultaneously selecting the correct audit label and the omission audit label and configuring the omission article.

[0083] Figure 7 It is a statistical chart of the recognition rate and the number of each audit label output by the test result confirmation device. DETAILED DESCRIPTION

[0084] The application will be described in further detail below with reference to the embodiments and drawings, but the embodiments of the application are not limited thereto.

[0085] Embodiment 1

[0086] As Figures 2-7 shown,

[0087] The intelligent image recognition device testing tool comprises a message queue middleware and a test result confirmation device.

[0088] The message queue middleware comprises:

[0089] a memory;

[0090] one or more processors; and

[0091] one or more modules present in the memory and configured to be executed by the one or more processors, the one or more modules comprising:

[0092] a module for simultaneously acquiring the recognition results of a plurality of intelligent image recognition devices for video streams of the same X-ray machine for contraband recognition;

[0093] a module for storing the recognition results from different intelligent image recognition devices in different topics;

[0094] It should be noted that one topic and one intelligent image recognition device are in a one-to-one correspondence, and the relationship between the two is configured in advance. The message queue middleware creates the recognition results of the intelligent image recognition device which can be subscribed.

[0095] The test result confirmation device comprises:

[0096] a memory;

[0097] one or more processors; and

[0098] one or more modules present in the memory and configured to be executed by the one or more processors, the one or more modules comprising:

[0099] a module for subscribing to the recognition results in different topics, labeling the recognition results with the manufacturer labels of the intelligent image recognition devices based on the differences between the topics, and storing the recognition results with the manufacturer labels in a database;

[0100] It needs to be explained: the test result confirmation device provided by the application first completes the subscription of obtaining a single identification result, and according to different topics subscribed, the single identification result is marked with a manufacturer label. In this way, the manufacturer label corresponding to the numerous identification results is configured, and the process is realized based on different topics. The manufacturer label is automatically configured through the subscription channel, the process of manually configuring the manufacturer label is omitted, and the friendliness to the test personnel is improved.

[0101] The module for obtaining the identification result with the manufacturer label from the database;

[0102] The module for auditing the identification result with the manufacturer label, and selecting the corresponding audit label from the optional audit label and configuring the identification result with the manufacturer label with the audit label to obtain the identification result of the contraband identification with the audit label and the manufacturer label;

[0103] The module for counting the number of audit labels of the identification result of the same manufacturer label and the same contraband identification;

[0104] The module for obtaining the identification rate of the identification result of the same contraband identification according to the number of the same manufacturer label.

[0105] Preferably, the test result confirmation device of the application is divided into a man-machine interaction device and an automatic device.

[0106] As Figure 3 and Figure 4 , Figure 5 , Figure 6 , Figure 7 As shown, when the test result confirmation device is divided into a man-machine interaction device, after randomly assembling the contraband into the luggage, the test is carried out. Since the correct contraband name in each luggage is not clear, it is necessary to rely on the identification of the image by the person, and then mark the identification result with a judgment label. Therefore, it is necessary to image the package image so that the test personnel can observe whether the identification result is correct.

[0107] In order to facilitate the test personnel to observe the identification result, when the test result confirmation device is a man-machine interaction device, the identification result includes: the package picture corresponding to the identification result, the contraband name in the package picture, and the identification frame coordinates of the contraband in the package picture.

[0108] The test result confirmation device further comprises:

[0109] a display,

[0110] a module for generating an identification frame line according to the identification frame coordinates of the contraband, and imaging on the display after superimposing the identification frame line on the package picture;

[0111] A module for displaying the contraband name in the display.

[0112] Preferably, in order to facilitate the test personnel to make artificial auditing judgment on the recognition result according to observation, the general judgment result is correct or incorrect, in order to facilitate subsequent countercheck, see the attached Figure 4 、 Figure 5 、 Figure 6 In the present application, for the incorrect case, the present application sets the misrecognition auditing label and the missed recognition auditing label, wherein it needs to be explained that: since there may be missed recognition in the correct case, when the correct auditing label is selected, the missed recognition auditing label can also be selected. Therefore, the auditing label configuration of the present application is as follows: when the test result confirmation device is a man-machine interaction device, the optional auditing label includes: the optional correct auditing label, the optional misrecognition auditing label and the optional missed recognition auditing label; the module for auditing the recognition result with the manufacturer label and selecting the corresponding auditing label from the optional auditing label to configure the recognition result with the manufacturer label to obtain the recognition result of the contraband identification with the auditing label and the manufacturer label includes: the module for generating the optional correct auditing label, the optional misrecognition auditing label and the optional missed recognition auditing label; the module for selecting at least one auditing label from the optional correct auditing label, the optional misrecognition auditing label and the optional missed recognition auditing label when auditing the current recognition result with the manufacturer label; the module for configuring the selected auditing label to the recognition result with the manufacturer label.

[0113] Preferably, in order to facilitate the formation of artificial auditing record and facilitate countercheck, when judged as misrecognition or missed recognition, the optional misrecognition auditing label and the optional missed recognition auditing label are provided with a pull-down optional contraband name. When artificial auditing, the test personnel must select the pull-down optional contraband name after selecting the above-mentioned label, and the meaning of the selected contraband name is the correct contraband name determined by artificial judgment, and the record is used to form the record of countercheck.

[0114] Preferably, the test result confirmation device of the present application is divided into and automatic device.

[0115] When the test result confirmation device is an automatic confirmation device, the optional auditing label includes: the optional correct auditing label and the optional misrecognition auditing label; the recognition result includes: the contraband name in the package picture;

[0116] The module for auditing the recognition result with the manufacturer label and selecting the corresponding auditing label from the optional auditing label to configure the recognition result with the manufacturer label to obtain the recognition result of the contraband identification with the auditing label and the manufacturer label includes:

[0117] When auditing the current recognition result with the manufacturer label, the module for comparing the preset correct contraband name in the package picture with the contraband name in the package picture of the recognition result.

[0118] When the comparison result is the same, a module of selecting a correct audit label to configure the identification result of the identification result with the manufacturer label is selected, and when the comparison result is different, a module of selecting a misidentification audit label to configure the identification result of the identification result with the manufacturer label is selected.

[0119] The test result confirmation device can realize automatic comparison, but it can only output a correct audit label or a misidentification audit label, and cannot record the missed identification.

[0120] Embodiment 2

[0121] As shown in 2- Figure 7 The intelligent image recognition device test method comprises the following steps,

[0122] S1. Put the prohibited articles into the luggage, and let the luggage pass through the X-ray machine;

[0123] S2. A plurality of intelligent image recognition devices to be tested simultaneously identify the prohibited articles based on the video stream of the same X-ray machine, and output the identification results;

[0124] S3. The identification results of the prohibited articles identified by the plurality of intelligent image recognition devices based on the video stream of the same X-ray machine are simultaneously acquired through the message queue middleware, and the identification results from different intelligent image recognition devices are designated to be stored in different topics;

[0125] S4. The identification results in the different topics are subscribed by the test result confirmation device, and the manufacturer labels of the intelligent image recognition devices are marked on the identification results based on the different topics, and the identification results with the manufacturer labels are stored in the database;

[0126] S5. The identification results with the manufacturer labels are acquired from the database by the test result confirmation device;

[0127] S6. The identification results with the manufacturer labels are audited by the test result confirmation device, and the corresponding audit label is selected from the optional audit labels to configure the identification results with the manufacturer labels, and the identification results of the prohibited article identification with the audit label and the manufacturer label are obtained;

[0128] S7. The number of audit labels of the identification results of the same manufacturer label and the same prohibited article identification is counted by the test result confirmation device.

[0129] S8. The identification rate of the identification results of the same manufacturer label to the same prohibited article identification is obtained by the test result confirmation device according to the count number.

[0130] Preferably, when the test result confirmation device is a man-machine interaction device, the identification result comprises: a package picture corresponding to the identification result, a name of the prohibited item in the package picture, and coordinates of a bounding box of the prohibited item in the package picture; and the optional audit label comprises: an optional correct audit label, an optional misidentification audit label, and an optional missed identification audit label.

[0131] S1 specifically comprises: when a batch of tests are performed, a plurality of prohibited items are randomly loaded into a plurality of luggage, and the luggage passes through an X-ray machine; this process is random, which can ensure the randomness of the test.

[0132] S6 specifically comprises:

[0133] The test result confirmation device generates a bounding box line according to the coordinates of the bounding box of the prohibited item, superimposes the bounding box line on the package picture, and then images on the display to display the name of the prohibited item in the display.

[0134] The test result confirmation device generates an optional correct audit label, an optional misidentification audit label, and an optional missed identification audit label.

[0135] The test result confirmation device selects at least one audit label from the optional correct audit label, the optional misidentification audit label, and the optional missed identification audit label.

[0136] The test result confirmation device configures the selected audit label to the identification result with the manufacturer's label.

[0137] Preferably,

[0138] The specific calculation process of S8 is:

[0139] The number of correct audit labels, misidentification audit labels, and missed identification audit labels is obtained respectively.

[0140] The identification rate is calculated according to formula 1.

[0141] Formula 1: Identification rate = number of correct audit labels / (number of correct audit labels + number of misidentification audit labels + number of missed identification audit labels) %.

[0142] Preferably,

[0143] When the test result confirmation device is an automatic confirmation device, the optional audit label comprises: an optional correct audit label and an optional misidentification audit label; and the identification result comprises: a name of a prohibited item in a package picture.

[0144] S1 is specifically: when performing a batch of tests, a plurality of contraband is loaded into a luggage, and the name of the contraband is recorded, and the luggage passes through an X-ray machine; this process is directional, can realize directional verification, extract the contraband name in the preset correct package picture, and then facilitate subsequent automatic comparison, without relying on manual review, greatly simplifying the labor cost.

[0145] S6 is specifically:

[0146] According to the record of the name of the contraband, the name of the contraband in the preset correct package picture of the current batch of tests is entered into the test result confirmation device;

[0147] The test result confirmation device compares the name of the contraband in the preset correct package picture with the name of the contraband in the package picture of the recognition result;

[0148] When the comparison result is the same, the test result confirmation device selects a correct audit label to configure the recognition result with the manufacturer's label;

[0149] When the comparison result is different, the test result confirmation device selects a misrecognition audit label to configure the recognition result with the manufacturer's label.

[0150] Preferably,

[0151] The specific calculation process of S8 is:

[0152] The number of correct audit labels, the number of misrecognition audit labels, and the number of missed recognition audit labels are obtained respectively;

[0153] The recognition rate is calculated according to formula 2;

[0154] Formula 2: Recognition rate = number of correct audit labels / (number of correct audit labels+number of misrecognition audit labels) %.

[0155] As Figure 7 shown, for the same manufacturer's label, the number of correct audit labels, the number of misrecognition audit labels, and the number of missed recognition audit labels are recorded respectively, and the recognition rate of the same manufacturer for different contraband names can be calculated by the above formula 1. In the Figure 7 , the meaning of the correct recognition times is: the number of times that the contraband recognition result is consistent with the actual result. The meaning of the misrecognition times is: the number of times that the contraband recognition result is inconsistent with the actual result. The meaning of the missed recognition times is: the number of times that the contraband is not recognized. The above three meanings correspond to the three audit labels in the present application.

[0156] The identification result of the application is subscribed from the message queue middleware, so the data is introduced into the confirmation device one by one, and the confirmation device performs artificial judgment or automatic judgment in sequence, the pressure on the test personnel is small, and in the artificial judgment, the correct prohibited article name when the misidentification or missed identification is manually configured to form the anti-search record.

[0157] By means of storage technology and computer human-computer interaction technology or automatic comparison technology, the test is realized on the complete online processing, the loss, discontinuity and inaccuracy of the archives are avoided, the test result is reliable, and the test labor cost is reduced.

[0158] It can be understood that the above embodiments are only exemplary embodiments adopted for illustrating the principles of the application / invention, and the application / invention is not limited thereto. Various modifications and improvements can be made by those skilled in the art without departing from the spirit and essence of the application / invention, and these modifications and improvements are also regarded as the protection scope of the application / invention.

Claims

1. A testing tool for intelligent image recognition devices, characterized in that, include: Message queue middleware, test result confirmation device; Message queue middleware includes: Memory; One or more processors; and One or more modules, stored in memory and configured to be executed by the one or more processors, the one or more modules comprising: A module that simultaneously acquires the identification results of contraband identification from video streams from multiple intelligent image recognition devices targeting the same X-ray machine; The recognition results from different intelligent image recognition devices are assigned to modules in different topics of the message queue middleware for storage. The test result confirmation device includes: Memory; One or more processors; and One or more modules, stored in memory and configured to be executed by the one or more processors, the one or more modules comprising: This module subscribes to the recognition results in different topics, assigns manufacturer tags to the recognition results based on the different topics, and stores the recognition results with manufacturer tags in the database. A module that retrieves recognition results with manufacturer tags from the database; This module reviews the identification results with manufacturer labels and selects the corresponding review label from the available review labels to configure for the identification results with manufacturer labels, thereby obtaining the identification results of prohibited items with review labels and manufacturer labels. A module for counting the number of verification labels that show the identification results of the same manufacturer's labels and the same prohibited items; This module calculates the recognition rate of the same prohibited item identified by the same manufacturer's label based on the number of samples collected.

2. The intelligent image recognition device testing tool according to claim 1, characterized in that, When the test result confirms that the device is a human-computer interaction device, the recognition result includes: the package image corresponding to the recognition result, the name of the prohibited items in the package image, and the coordinates of the identification box of the prohibited items in the package image; The test result confirmation device also includes: monitor, A module that generates a frame line based on the coordinates of the frame line for prohibited items, overlays the frame line onto the package image, and then displays the image on the monitor. A module that displays the names of prohibited items on the monitor.

3. The intelligent image recognition device testing tool according to claim 2, characterized in that, When the test result confirmation device is a human-computer interaction device, the optional verification labels include: optional correct verification labels, optional misidentification verification labels, and optional omission verification labels; the module that verifies the identification results with manufacturer labels and selects the corresponding verification label from the optional verification labels to configure for the identification results with manufacturer labels, and obtains the identification results of prohibited items with verification labels and manufacturer labels, includes: A module for generating optional correct review tags, optional misidentified review tags, and optional missed review tags; When reviewing the current identification results with manufacturer tags, the module selects at least one review tag from the available correct review tags, the available misidentified review tags, and the available missed review tags; Configure the selected audit label to the module that displays the recognition results with the manufacturer label.

4. The intelligent image recognition device testing tool according to claim 3, characterized in that, The optional misidentification review label and optional omission review label are equipped with a drop-down list of prohibited product names.

5. The intelligent image recognition device testing tool according to claim 1, characterized in that, When the test result confirmation device is an automatic confirmation device, The optional verification labels include: optional correct verification labels and optional misidentified verification labels; the identification results include: the names of prohibited items in the package image; The module that reviews the identification results with manufacturer labels and selects the corresponding review label from the available review labels to configure for the identification results with manufacturer labels, and obtains the identification results of prohibited items with review labels and manufacturer labels, includes: This module compares the names of prohibited items in the pre-set, correct package images with the names of prohibited items in the identified package images when reviewing the current recognition results with manufacturer labels. When the comparison results are the same, the module with the correct verification label is selected for the recognition result with the manufacturer label. When the comparison results are different, the module with the incorrect verification label is selected for the recognition result with the manufacturer label.

6. A testing method for intelligent image recognition devices, characterized in that, Includes the following steps, S1. Pack the contraband into a suitcase and pass the suitcase through the X-ray machine; S2. Multiple intelligent image recognition devices to be tested simultaneously identify contraband from the video stream of the same X-ray machine and output the recognition results respectively; S3. Simultaneously obtain the identification results of contraband identification from multiple intelligent image recognition devices targeting the same X-ray machine's video stream through message queue middleware, and store the identification results from different intelligent image recognition devices in different topics; S4. Confirm the recognition results of the device by subscribing to different topics through test results, and label the recognition results with the manufacturer's label of the intelligent image recognition device based on the different topics, and store the recognition results with manufacturer's label in the database; S5. Confirm the identification results with manufacturer labels obtained from the database by the test results; S6. The device verifies the identification results with manufacturer labels by verifying the test results and selects the corresponding verification label from the available verification labels to configure the identification results with manufacturer labels, thereby obtaining the identification results of prohibited items with verification labels and manufacturer labels. S7. Confirm the number of labels verified by the test results confirmation device for the identification results of the same manufacturer's labels and the same prohibited items; S8. Confirm the recognition rate of the device for identifying the same prohibited item using the same manufacturer's label based on the number of samples obtained through the test results.

7. The testing method for an intelligent image recognition device according to claim 6, characterized in that, When the test result confirmation device is a human-computer interaction device, the recognition result includes: the package image corresponding to the recognition result, the name of the prohibited items in the package image, and the coordinates of the identification box of the prohibited items in the package image; the optional review labels include: optional correct review labels, optional misidentified review labels, and optional missed review labels; S1 specifically refers to: during a batch of testing, multiple prohibited items are randomly packed into multiple suitcases and the suitcases are passed through an X-ray machine; S6 specifically refers to: The test results confirm that the device generates a frame line based on the coordinates of the prohibited item's frame, overlays the frame line onto the package image, and then displays the image on the monitor, showing the name of the prohibited item. The test results confirm the module that generates optional correct audit labels, optional misidentified audit labels, and optional missed audit labels. The test results confirm the module that selects at least one audit label from the optional correct audit label, optional misidentified audit label, and optional missed audit label. The device confirms the selected audit label by configuring it onto the recognition result with the manufacturer label based on the test results.

8. The testing method for an intelligent image recognition device according to claim 7, characterized in that, The specific calculation process for S8 is as follows: Count the number of correctly identified review tags, incorrectly identified review tags, and missed review tags respectively; The recognition rate is calculated according to Formula 1; Formula 1: Recognition rate = Number of correctly recognized labels / (Number of correctly recognized labels + Number of misrecognized labels + Number of missed labels)%.

9. A testing method for an intelligent image recognition device according to claim 6, characterized in that, When the test result confirmation device is an automatic confirmation device, the optional verification labels include: optional correct verification labels and optional misidentified verification labels; the identification results include: the names of prohibited items in the package image; S1 specifically refers to: when performing a batch of tests, packing multiple prohibited items into one suitcase, recording the names of the prohibited items, and then passing the suitcase through an X-ray machine; S6 specifically refers to: Based on the record of prohibited items, the prohibited item names in the correct package images preset for the current batch of tests are entered into the test result confirmation device. The device confirms the test results by comparing the names of prohibited items in the preset, correct package images with the names of prohibited items in the identified package images. If the comparison results are the same, the device confirms the correct verification label configuration for the identification result with the manufacturer label by verifying the test results. When the comparison results are different, the test result confirmation device selects the misidentification review label and assigns it to the recognition result with the manufacturer label.

10. A testing method for an intelligent image recognition device according to claim 9, characterized in that, The specific calculation process for S8 is as follows: Count the number of correctly identified review tags and the number of incorrectly identified review tags respectively; The recognition rate is calculated according to Formula 2; Formula 2: Recognition rate = Number of correctly recognized labels / (Number of correctly recognized labels + Number of incorrectly recognized labels)%.

Images