KR102063859B1 - Systems and methods for security search at airport based on AI and deep learning - Google Patents

Abstract

According to an embodiment of the present invention, an airport security screening system based on artificial intelligence (AI) and deep learning comprises: an image acquisition unit to acquire a transmission image obtained by photographing baggage passing through a baggage screening point of an airport at one or more angles; an artificial intelligence server to detect a baggage area in the transmission image photographed at the one or more angles, then divide the detected baggage area into a plurality of image block units (patches), then generate a plurality of object groups by normalizing and clustering features of the image block units, use a deep learning algorithm to classify and subdivide the plurality of object groups into normal items and prohibited items, and re-verify and learn the subdivided prohibited items; and an image display unit to display a prohibited item in a boundary color corresponding to a type of the prohibited item if one or more among the detected articles are classified as the prohibited item.

Description

Systems and methods for security search at airport based on AI and deep learning}

The present invention relates to an AI and deep learning based airport security screening system and method.

In general, an airport, a port, or various conference halls install a conventional metal detector to detect metallic materials such as firearms and prohibit the introduction of dangerous goods. In the metal detector, the magnetic field generated is changed depending on the presence or absence of the metallic material. It is a device that detects firearms or other metallic dangerous objects by using.

Metal detectors are widely used for portable metal detectors, which are carried by police officers or security personnel to search for possession of metals, and sentence detectors for teaming metals as they pass through the gates. to be.

However, the metal detector alone is difficult to detect plastic bombs, but not metallic materials, and it is difficult to prepare for the rapidly changing terrorism. there was.

In order to solve this problem, the baggage retrieval time can be reduced with the introduction of an X-Ray retrieval system that can read the baggage as it passes through. X-Ray search equipment installed in airports and harbors to search for illegal belongings in baggage can be read in real time, and search equipment with a recording function has been introduced and used.

However, the X-Ray scanner scans baggage continuously moving at a speed of more than 12 meters per minute, and the security screening agent visually discriminates the displayed image.

Therefore, as the work fatigue of security screening personnel increased, the search reading rate could decrease. In addition, there could be a difference in the readability of the goods according to the skill of the security screening personnel. In addition, X-Ray scanners require a long time for security checkers to precisely read the items when they are manually checked for scanned images, which can take a long time to search, increase congestion in airport buildings, and cause delays in flight operations. there was.

Patent Publication Number: 10-2017-0121088

An object of the present invention is to provide an AI and deep learning based airport security search system and method that can solve the conventional problems.

AI and deep learning-based airport security screening system according to an embodiment of the present invention for solving the above problem is an image acquisition unit for obtaining a transmission image of the baggage passing through the baggage screening of the airport from at least one angle ; After detecting a baggage area in the transmitted image photographed from the at least one angle, the detected baggage area is divided into a plurality of image block units, and then the features of the image block units are normalized and clustered. An artificial intelligence server for generating an object group, classifying and subdividing the plurality of object groups into regular products and prohibited items, and re-verifying and learning the detailed prohibited items by using a deep learning algorithm; And an image display unit for displaying one or more of the detected items as a prohibition item, wherein the image display unit displays a border color corresponding to a type of the prohibition item, wherein the artificial intelligence server binarizes and sessionizes the transmitted image of the target area. After performing one or more processes of noise reduction, the image pre-processing unit for generating a window coordinate value of each image block unit (patch) by grid division by a predetermined image block unit (patch); A baggage area extraction unit for extracting a baggage area in a transmission image of a target area grid-divided into image block units using a deep learning algorithm; Normalizing and clustering the image block unit of the luggage area, and extracting by specifying objects in the luggage area using the deep learning learning algorithm, and extracting the extracted objects in the database. The extracted objects are grouped into regular goods and prohibited items, and the objects grouped into prohibited items are classified into prohibited items and prohibited items. It includes a classification learning unit.

AI and deep learning-based airport security search method according to an embodiment of the present invention for solving the above problems comprises the steps of obtaining a transmission image taken at least one angle of the baggage passing through the baggage claim of the airport; After detecting a baggage area in the transmitted image photographed from the at least one angle, the detected baggage area is divided into a plurality of image block units, and then the features of the image block units are normalized and clustered. Creating an object group, classifying and subdividing the plurality of object groups into regular products and prohibited items, and re-verifying and learning the detailed prohibited items by using a deep learning algorithm; And if at least one of the detected items is classified as a prohibited entry item, displaying a border color corresponding to a type of the prohibited entry item, wherein the re-validation and learning are pre-processed on the transmission image of the target area. Generating a unique window coordinate value of each block by grid-dividing each of the transmitted images of the plurality of target areas in units of predetermined blocks; Extracting a baggage area in a transmission image of the target area grid-divided into the image block units using a deep learning learning algorithm; Normalizing and clustering image block units of the luggage area in the detected luggage area, and then specifying and extracting objects in the luggage area using the deep learning learning algorithm; The extracted objects are learned by deep learning learning algorithms in a database and compared with the stored objects. The extracted objects are grouped into regular items and prohibited items, and the objects grouped into prohibited items are prohibited from carry-on items and entrusted items. Subdividing into baggage prohibited items; And extracting the image block unit of each of the fine-grained prohibited items, batch normalizing, and verifying and learning the characteristics of each object classified as the prohibited items.

Using the AI and deep learning based airport security search system and method according to an embodiment of the present invention, it is possible to easily determine the baggage and the types of the goods in the baggage using the feature points of the goods in the X-ray perspective image, In addition, the AI learning algorithm, which is a deep learning algorithm, provides the advantage of minimizing the false detection rate and improving the detection accuracy of the prohibited items.

Through the above-described advantages, it is possible to automatically block the security search by increasing the read rate by performing the security search task automatically, and reduce the time required for the security search through the rapid security search.

In addition, it provides an advantage that the human error according to the skill of the security search personnel can be reduced by generating an alarm when searching prohibited items.

1 is a block diagram of an airport security search system based on AI and deep learning according to an embodiment of the present invention.
2 is an exemplary diagram of a CNN architecture among deep learning learning algorithms applicable to the present invention.
3 is an illustration of batch normalization.
FIG. 4 is an exemplary view illustrating a screen on which regular items and prohibited items are displayed on the image display unit of FIG. 1.
5 is a flowchart illustrating an AI and deep learning based airport security search method according to an embodiment of the present invention.
6 illustrates an example computing environment in which one or more embodiments disclosed herein may be implemented.

Hereinafter, exemplary embodiments will be described in detail with reference to the accompanying drawings. In adding reference numerals to the components of each drawing, it should be noted that the same reference numerals are assigned to the same components as much as possible even though they are shown in different drawings. In addition, in describing the embodiments, when it is determined that a detailed description of a related well-known configuration or function interferes with the understanding of the embodiment, the detailed description thereof will be omitted.

In addition, in describing the components of the embodiment, terms such as first, second, A, B, (a), and (b) may be used. These terms are only for distinguishing the components from other components, and the nature, order or order of the components are not limited by the terms. If a component is described as being "connected", "coupled" or "connected" to another component, that component may be directly connected or connected to that other component, but there is another component between each component. It will be understood that may be "connected", "coupled" or "connected".

Method according to the embodiment is implemented in the form of program instructions that can be executed by various computer means may be recorded on a computer readable medium. The computer readable medium may include program instructions, data files, data structures, and the like, alone or in combination. Program instructions recorded on the media may be those specially designed and constructed for the purposes of the present disclosure, or they may be of the kind well-known and available to those having skill in the computer software arts. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tape, optical media such as CD-ROMs, DVDs, and magnetic disks, such as floppy disks. Magneto-optical media, and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine code generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like. The hardware device described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

Prior to describing the present invention, baggage at the airport described in the present invention includes belongings and articles that are requested to be carried or checked by a passenger using an aircraft. In addition, prohibited items are items that are prohibited from being transported by aircraft, and may be determined according to national or operator policy. For example, non-carry-on articles include firearms, weapons, firearms, explosives, flammables, toxic substances, ornamental items, and drugs.

There are many kinds of prohibited items that can be enumerated by the law or the airport operator. However, the items discovered through X-ray inspection show a certain proportion of patterns. 70% of disposable gas lighters, 20% of cutter knives, and 10% of daily necessities such as stationery scissors and disposable butane gas. It is near level.

However, for the safety of the passengers at the airport, searching all the baggage and finding a disposable gas lighter can provide a lot of concentration and fatigue for the workers, but from the airport operation side, delays in baggage processing, flight delays and passenger delays And inconvenience.

1 is a view showing a conventional image of the cargo by X-ray retrieval, installed inside the X-ray scanner to find the prohibited goods to carry the image obtained by X-ray imaging in two directions of the upper and side This is the screen of the monitor.

It is recommended that the two screens pass through the inside of the X-ray scanner at the speed of conveying the conveyor, usually 12 to 18m per minute, and the distance between cargoes is recommended. The worker must read the images of the baggage being put in place. Normally, the reading time per cargo is less than 16 seconds. If the reading is not finished within the worker's mind, the Xray scanner may be stopped or sometimes reversed. In this case, all the conveyors in front of the X-ray scanner are stopped, and operation of the conveyors at the check-in counters is stopped.

Therefore, the main purpose of X-ray workers is to quickly read the presence of prohibited items without stopping the X-ray scanner, or conveyor line.

Hereinafter, an AI and deep learning based airport security search system and method according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram of an AI and deep learning based airport security search system according to an embodiment of the present invention, Figure 2 is an illustration of a CNN architecture according to an embodiment of the present invention, Figure 3 is batch normalization 4 is an exemplary view illustrating a screen on which a regular product item and a prohibited item are displayed on the image display unit of FIG. 1.

As shown in FIG. 1, the security search system 100 of a smart airport according to an embodiment of the present invention includes an image acquisition unit 200, an artificial intelligence server 300, and an image display unit 400.

Meanwhile, the image acquisition unit 200, the artificial intelligence server 300, and the image display unit 400 described herein may be configured in the form of a terminal and / or a server, and may communicate with each other through a network.

The network may include a local area network (LAN), a wide area network (WAN), a value added network (VAN), a personal local area network (PAN), or a mobile communication network (PAN). It can be implemented in all kinds of wired and wireless networks such as mobile radio communication network (Wibro), wireless broadband Internet (Wibro), mobile WiMAX, high speed downlink packet access (HSDPA), or satellite communication network.

More specifically, each component of the present invention will be described in more detail.

First, the image acquisition unit 200 radiates X-rays to a target area (for example, an area for detecting baggage) of the security checkpoint of the smart public security, and then acquires an X-ray scan image of the target area. The image acquisition unit 200 may be an X-ray imaging apparatus.

Here, the X-ray scan image of the target area may include a transmission image of the scan object (baggage) according to the scan direction. For example, the scan image may be a transmission image in the vertical and horizontal directions of the scan object.

In the embodiment, the image acquisition unit 200 radiates X-rays to scan a scan object (baggage), for example, but the present invention is not limited thereto. Any way you can be possible.

Next, the artificial intelligence server 300 detects a baggage area from the transmitted image photographed from the at least one or more angles, divides the detected baggage area into a plurality of image block units, and then performs the image block unit. To generate a plurality of object groups through normalization and clustering of features, classify and subdivide the plurality of object groups into regular items and prohibited items using deep learning algorithms, and re-verify the detailed prohibited items. Learn.

More specifically, the artificial intelligence server 300 includes an image preprocessing unit 310, a baggage area extracting unit 320, a classification learning unit 330, and an AI machine learning unit 340.

The image preprocessor 310 performs preprocessing on the transmitted image of the target area.

The image preprocessor 310 performs one or more preprocessing of binarization, sessionization, and noise removal on the transmitted image of the target region.

The image preprocessing unit 310 uses a Gaussian filter, a Laplacian filter, a Difference of Gaussian (DoG), and Canny edge detection to detect the transmitted image of the target area. It can be modified or improved.

For example, the image preprocessor 310 removes noise using a Gaussian filter, etc., performs a gradient operation for edge component detection, and performs a gradient operation to interpolate a broken edge line. Non-maximum suppression may be performed, and hysteresis thresholding for binarizing the edge map may be performed.

In addition, the image preprocessing unit 310 grid-divides each of the transmission images of the plurality of target areas photographed in the vertical and horizontal directions in a predetermined image block unit, and generates a unique window coordinate value of each block. .

Next, the baggage area extractor 320 extracts a baggage area in the transmission image of the target area grid-divided into the image block units by using a region proposal network (RPN).

Next, the classification learning unit 330 normalizes and clusters an image block unit of the luggage area, and then specifies objects in the luggage area by using the R-CNN algorithm. Extract, and extract the extracted objects in a database and compare them with other stored objects, group the extracted objects into regular items and prohibited items, and carry out the prohibited items and carry-on items After grouping them into prohibited items, the grouped objects are subdivided into subcategories.

Here, prohibited items on board are liquids (liquids such as pepper / kimchi) foods, dangerous goods (windows, swords, electronic impactors, firearms, martial arts signal equipment), tools (hammers, wrenches, etc.), dangerous goods (flammable gas liquids, Radioactive substances, etc.).

In addition, items prohibiting checked baggage are classified into inorganic materials and dangerous goods (lithium ion batteries, flammable gas liquids, radioactive materials), and the like.

Meanwhile, the classification learning unit 330 generates a descriptor including information defining each feature detected in each block unit through a speed up robust feature (SURF) and records the same in block units.

In addition, the classification learning unit 330 classifies the plurality of block units in which the feature information is recorded into a regular product object group and a prohibited item object group through a normalization and clustering process.

For reference, the clustering process is divided into partitioning methods and hierarchical methods. Clustering of partitioning methods requires that each group have at least one data, and each data is exactly one. It is a method of dividing a data set into small groups with the rule that it should belong to a group of. The clustering of such divided companies includes techniques such as K-means, K-medoids, and DBCAN.

Hierarchical methods are hierarchical decomposition of data sets, which are further classified into agglomerative and divisional clustering.

On the other hand, the classification learning unit 330 upgrades the existing learning results by storing the objects subdivided into inorganic, harmful chemicals (liquids), drugs, and flammables in the database.

Next, the AI machine learning unit 340 extracts the image block units of each of the objects of the prohibited entry items, and then batch normalizes each of the objects classified as the prohibited items. Validate and learn features.

For reference, referring to FIG. 3, batch normalization refers to an operation of normalizing an activation value or an output value of an activation function. The task of normalizing the distribution of data (placements) in each layer of a neural network. As a kind of noise addition, this can be normalized from batch to batch, so that the variance and value of the mean over the entire data can vary.

While normalizing in each hidden layer, the input distribution is constant. Accordingly, even if the learning rate is set large, the learning speed is increased as a result.

In other words, batch normalization means normalizing the mini-batch during learning as a unit and normalizing the distribution to have an average of 0 and a variance of 1.

As an example of batch normalization, first, an average and a variance of a minibatch used as an input are calculated, and normalization is performed such that an average is 0 and a variance is 1 with respect to an activation value / output value of a hidden layer. As a result, the data distribution is less skewed, and scale-up and shift transformations are performed for each batch normalization step.

Deep learning is a technology used to cluster or classify things or data. It's a technique that allows you to enter a lot of data into your computer and sort similar things together.

At this point, many machine learning algorithms have emerged regarding how to classify data. Deep learning is an artificial intelligence learning method proposed to overcome the limitations of artificial neural networks.

For reference, the artificial intelligence server 300 disclosed in the present invention employs a deep learning learning algorithm, and the deep learning learning algorithm may include a deep belief network, an autoencoder, a convolutional neural network (CNN), a recurrent neural network (RNN), and a deep Q. It may include a network, etc., the deep learning algorithms enumerated in the present invention is only an example, but is not limited thereto.

In the present invention, an example of applying the CNN (Convolutional Neural Network), which is one of the deep learning learning algorithms listed above, to an AI and deep learning based airport security search system has been described, but the present invention is not limited thereto. Various kinds of deep learning algorithms may be used.

Next, referring to FIG. 4, the image display unit 400 receives and displays an article in a bag from the artificial intelligence server 300 on the transmission image of the target area, and receives the item and the prohibited item from the item. The border color of prohibited items is indicated differently.

The image display unit 400 includes a liquid crystal display (LCD), a thin film transistor liquid crystal display (TFT LCD), an organic light-emitting diode (OLED), a flexible display ( The display device may include at least one of a flexible display, a 3D display, an e-ink display, and a light emitting diode (LED).

5 is a flowchart illustrating an AI and deep learning based airport security search method according to an embodiment of the present invention.

Referring to FIG. 5, in the airport security search method S700 based on AI and deep learning according to an embodiment of the present invention, the image acquisition unit 200 first detects a target area (eg, baggage) of an airport security checkpoint. After the X-rays are radiated to the target area, an X-ray scan image of the target area is acquired (S710), and a baggage area is detected from the transmission image photographed from at least one or more angles. After dividing the image block unit into a plurality of image block units, a plurality of object groups are generated through normalization and clustering of the characteristics of the image block unit, and the plurality of object groups are imported and imported using a deep learning algorithm. Classified and subdivided into prohibited items, re-validated and fine-grained prohibited items (S720), and in the baggage on the transmission image of the target area from the artificial intelligence server 300 Take the product provides normal items and Prohibited items include a series of processes to display (S730). In this case, the boundary color of the regular product and the prohibited items may be displayed differently, for example, the inorganic may be displayed in red, the liquid in orange, the flammable material in purple, and the like.

Here, the X-ray scan image of the target area includes a transmission image of the scan object (baggage) according to the scan direction, and may be, for example, a transmission image in the vertical and horizontal directions of the scan object.

More specifically, the process of S720, after performing one or more of binarization, sessionization, noise removal on the transmission image of the target area in the image pre-processing unit 310, the transmission of the plurality of target areas photographed in various directions Each image is grid-divided into predetermined block units to generate a unique window coordinate value of each block (S721).

Subsequently, the baggage area extractor 320 extracts a baggage area in the transmission image of the target area grid-divided into the image block units by using a region proposal network (RPN) (S722).

Subsequently, after normalization and clustering of image block units of the baggage area in the baggage area detected by the classification detection unit 330, an object in the baggage area using the R-CNN algorithm. Extract and classify the extracted objects into regular products and prohibited items, and compare the extracted objects with various objects stored in the database by using the R-CNN algorithm. Subdivided into inorganics, hazardous chemicals (liquids), drugs, and flammables (S723).

After that, the AI machine learning unit 340 extracts the image block unit of each of the objects of the prohibited items to be imported, batch normalizes them, and then features the characteristics of each object classified as the prohibited items. By verifying and learning (S724), the false detection rate is reduced.

Therefore, using the AI and deep learning based airport security screening system and method according to an embodiment of the present invention, using the features of the goods in the X-ray perspective image, the items in baggage and items in the baggage items and prohibited items It is easy and quick to determine and classify, and furthermore, the AI learning algorithm, which is a deep learning algorithm, can minimize the false detection rate of goods and improve the detection accuracy of the prohibited items.

Through the above-described advantages, it is possible to automatically block the security search by increasing the read rate by performing the security search task automatically, and reduce the time required for the security search through the rapid security search.

In addition, it provides an advantage that the human error according to the skill of the security search personnel can be reduced by generating an alarm when searching prohibited items.

FIG. 6 illustrates an example computing environment in which one or more embodiments disclosed herein may be implemented, and illustrates an example of a system 1000 that includes a computing device 1100 configured to implement one or more embodiments described above. Shows.

For example, computing device 1100 may be a personal computer, server computer, handheld or laptop device, mobile device (mobile phone, PDA, media player, etc.), multiprocessor system, consumer electronics, mini computer, mainframe computer, Distributed computing environments, including, but not limited to, any of the systems or devices described above. The computing device 1100 may include at least one processing unit 1110 and a memory 1120. The processing unit 1110 may include, for example, a central processing unit (CPU), a graphics processing unit (GPU), a microprocessor, an application specific integrated circuit (ASIC), field programmable gate arrays (FPGA), and the like. It may have a plurality of cores. The memory 1120 may be volatile memory (eg, RAM, etc.), nonvolatile memory (eg, ROM, flash memory, etc.), or a combination thereof.

In addition, computing device 1100 may include additional storage 1130. Storage 1130 includes, but is not limited to magnetic storage, optical storage, and the like. The storage 1130 may store computer readable instructions for implementing one or more embodiments disclosed herein, and other computer readable instructions for implementing an operating system, an application program, and the like. Computer readable instructions stored in storage 1130 may be loaded into memory 1120 for execution by processing unit 1110.

In addition, computing device 1100 may include input device (s) 1140 and output device (s) 1150. Here, input device (s) 1140 may include, for example, a keyboard, mouse, pen, voice input device, touch input device, infrared camera, video input device, or any other input device. Also, output device (s) 1150 may include, for example, one or more displays, speakers, printers, or any other output device. In addition, computing device 1100 may use an input device or output device included in another computing device as input device (s) 1140 or output device (s) 1150.

In addition, computing device 1100 may include communication connection (s) 1160 that enable computing device 1100 to communicate with another device (eg, computing device 1300). Here, the communication connection (s) 1160 may be a modem, a network interface card (NIC), an integrated network interface, a radio frequency transmitter / receiver, an infrared port, a USB connection, or other for connecting the computing device 1100 to another computing device. It may include an interface. In addition, communication connection (s) 1160 may include a wired connection or a wireless connection.

Each component of the computing device 1100 described above may be connected by various interconnections such as a bus (eg, peripheral component interconnect (PCI), USB, firmware (IEEE 1394), optical bus structure, etc.). And may be interconnected by the network 1200.

As used herein, terms such as "component", "unit" and the like generally refer to a computer-related entity that is hardware, a combination of hardware and software, software, or running software.

For example, a component may be, but is not limited to being, a process running on a processor, a processor, an object, an executable, a thread of execution, a program, and / or a computer. For example, both an application running on a controller and the controller can be a component. One or more components may reside within a thread of processes and / or execution, and the components may be localized on one computer and distributed between two or more computers.

The present invention is not limited by the above-described embodiment and the accompanying drawings. It is intended that the scope of the claims be defined by the appended claims, and that various forms of substitution, modification, and alteration can be made without departing from the spirit of the invention as set forth in the claims. Will be self explanatory.

100: smart airport security screening system
200: image acquisition unit
300: artificial intelligence server
310: image preprocessing unit
320: luggage area extraction unit
330: classification learning unit
340: AI machine learning unit

Claims (8)

An image acquisition unit for acquiring a transmission image of at least one angle of the baggage passing through the baggage claim area of the airport;
After detecting a baggage area in the transmitted image photographed from the at least one angle, the detected baggage area is divided into a plurality of image block units, and then the features of the image block units are normalized and clustered. An artificial intelligence server for generating an object group, classifying and subdividing the plurality of object groups into regular products and prohibited items, and re-verifying and learning the detailed prohibited items by using a deep learning algorithm; And
If one or more of the detected items are classified as prohibited items, and includes an image display unit for displaying a border color corresponding to the type of prohibited items,
The artificial intelligence server
After performing one or more processes of binarization, sessionization, and noise reduction on the transmitted image of the target area, grid division is performed on a predetermined image block unit to generate window coordinates of each image block unit. An image preprocessor;
A baggage area extraction unit for extracting a baggage area in a transmission image of a target area grid-divided into image block units using a deep learning algorithm; And
After normalizing and clustering an image block unit of the luggage area, the deep learning learning algorithm is used to specify and extract objects in the luggage area, and extract the extracted objects into the database. Compared with the objects stored by learning with a learning learning algorithm, the extracted objects are grouped into regular items and prohibited items, and the objects grouped into the prohibited items are classified into prohibited items and prohibited items. AI and deep learning based airport security screening system with classification learning.
delete delete delete The method of claim 1,
The classification learning unit
AI and deep learning based airport security search system for upgrading the existing learning results by storing the fragmented objects of the carry-on prohibited items and the fragmented objects of the checked baggage prohibited items in the database.
The method of claim 1,
The artificial intelligence server
After extracting the image block unit (patch) of the objects of the fine-grained prohibited items, AI and AI including a machine learning unit for verifying and learning the mis-detection of the details of the prohibited items using a batch normalization algorithm and Deep learning based airport security screening system.
Obtaining a transmission image of at least one angle of the baggage passing through the baggage claim area of the airport; And
After detecting a baggage area in the transmitted image photographed from the at least one angle, the detected baggage area is divided into a plurality of image block units, and then the features of the image block units are normalized and clustered. Creating an object group, classifying and subdividing the plurality of object groups into regular products and prohibited items, and revalidating and learning the detailed prohibited items by using a deep learning algorithm; And
If one or more of the detected items are classified as prohibited items, displaying a border color corresponding to the type of prohibited items;
The revalidation and learning step
Performing pre-processing on the transmitted image of the target region, and generating a unique window coordinate value of each block by grid-dividing each of the transmitted images of the plurality of target regions by a predetermined block unit;
Extracting a baggage area in a transmission image of the target area grid-divided into the image block units using a deep learning learning algorithm;
Normalizing and clustering the image block units of the luggage area in the detected luggage area, and then specifying and extracting objects in the luggage area using the deep learning learning algorithm;
The extracted objects are learned by deep learning learning algorithms in a database and compared with the stored objects. The extracted objects are grouped into regular items and prohibited items, and the objects grouped into prohibited items are prohibited from carry-on items and entrusted items. Subdividing into baggage prohibited items; And
After extracting the image block unit of each object of the fine-grained prohibited items, batch normalization, and verifying and learning the characteristics of each object classified as prohibited items, AI and deep learning Based airport security screening method.
delete

Images