Method for the security control of at least one person; control device for carrying out the method; computer program product; computer-readable storage medium

Abstract

The invention relates to a method for the security control of at least one person (50) with the aid of a control device (1), the control device (1) comprising a body scanner (5) for carrying out a body scan of the person (50), a sensor system (6) for detecting the person (50) by means of one or more cameras, and an AI system (8), wherein the method comprises the following steps, in particular in the listed order: • evaluating the data detected by the sensor system (6) with the aid of the AI system (8) in order to ascertain at least one risk feature of the person (50), wherein the risk feature cannot be identified by means of the body scanner (5), in particular cannot be identified with a predetermined probability, and / or wherein the risk feature is intentionally ignored and / or masked by means of the body scanner (5); • determining, on the basis of the evaluation, that at least one risk feature is present; • triggering a risk signal if the at least one risk feature has been ascertained, wherein the risk signal can be perceived optically and / or acoustically and / or haptically and / or electronically, in particular by the person (50), and / or wherein the risk signal prevents a triggering of the body scan.

Description

[0001] Method for security control of at least one person; control device for carrying out the method; computer program product; computer-readable storage medium

[0002] The invention relates to a method for security control of at least one person using a control device, the control device comprising a body scanner for carrying out a body scan of the person, a sensor system for detecting the person with one or more cameras and a Kl system.

[0003] Such devices are known, for example, from EP 4 266 268 A1.

[0004] It has been shown in public institutions such as airports, courts, railway stations and businesses that, given the current security situation, the procedure for security screening of persons should be optimized in terms of security and / or in terms of personnel requirements for security screening.

[0005] Against this background, the task is to provide the method for security control and a control device for carrying out the method, which are improved compared to the known methods and control devices.

[0006] The object is achieved by a security control method according to claim 1, a control device according to claim 15, a computer program product according to claim 16 and a computer-readable storage medium according to claim 17.

[0007] The method according to the invention for security control of at least one person using a control device, the control device comprising a body scanner for carrying out a body scan of the person, a sensor system for detecting the person with one or more cameras and a Kl system, wherein the method comprises the following steps, in particular in the listed sequence:

[0008] • Evaluating the data recorded by the sensor system using the AI ​​system to determine at least one risk characteristic of the person, wherein the risk characteristic is not recognizable by means of the body scanner, in particular not recognizable with a predetermined probability, and / or wherein the risk characteristic is intentionally ignored and / or hidden by means of the body scanner;

[0009] • Determining, based on the evaluation, that at least one risk characteristic is present; • Triggering a risk signal if the at least one risk characteristic has been determined, wherein the risk signal is visually and / or acoustically and / or haptically and / or electronically perceptible, in particular by the person, and / or wherein the risk signal prevents the body scan from being triggered.

[0010] The abbreviation "Ki" is used here for artificial intelligence. "Ki-System" is used here as an abbreviation for an artificial intelligence system. In particular, the "Ki-System" can use a machine learning model.

[0011] This makes the process more automated, reducing human error among security personnel. Human performance is subject to fluctuations and always depends on the day. Furthermore, AI systems do not have the same biases or subjectivity as human observers. Furthermore, the process can be carried out with fewer or no security personnel. For example, the process can be carried out at night with minimal personnel expenditure. Advantageously, the personnel's eyes can be replaced or supported, thus enabling an overall higher level of security, particularly with reduced personnel expenditure.

[0012] A staff check can be eliminated or supported. This reduces unnecessary, particularly energy-intensive, body scans, making the procedure more sustainable.

[0013] The body scanner is designed to detect dangerous objects on the person's body and / or clothing. Dangerous objects can include, for example, weapons and / or objects that can be used as weapons and / or other hazardous substances. The body scanner is preferably a full-body scanner. In particular, the body scanner can be configured to operate using millimeter wave and / or backscatter technology. These technologies use high-frequency waves to create an image of the body. Millimeter wave technology generates images using radiofrequency energy, whereas backscatter technology uses low-dose X-rays. Such a body scanner can also effectively detect hidden dangerous objects, for example, beneath clothing.With state-of-the-art body scanners, certain body areas, especially those clearly visible to security personnel, such as the hand or head area, are sometimes masked out to prevent a dangerous object from being detected, even if it is an everyday object, such as a wristwatch or glasses. However, this poses a certain security risk, as human personnel can make errors.

[0014] In particular, the hazardous objects can be marked on a generic body image, especially an avatar, of the person, particularly on a playback device. This allows the person to be easily and quickly informed of where a safety risk still exists.

[0015] The term "risk characteristic" refers to a characteristic that poses a potential risk because it cannot be reliably scanned and / or x-rayed by the body scanner. Alternatively, or cumulatively, the risk characteristic is automatically excluded from the body scanner because it most often leads to false detection. In principle, the risk characteristic does not necessarily pose a risk to the person, but merely jeopardizes the possibility of reliable screening by the body scanner. As an alternative to the term "risk characteristic," the term "risk potential" can also be used.

[0016] This does not mean that the risk feature cannot be detected due to the body scanner's detection range, but rather that it cannot be detected due to its detection technology, in particular due to its frequency range. The predetermined probability under which the risk feature cannot be detected can be higher than 50%, preferably higher than 75%, particularly preferably higher than 90%, in particular higher than 95%, preferably higher than 98%. This also enables application in high-security areas such as airports, since the risk feature can be reliably detected by the sensor system.

[0017] In particular, the predetermined probability can be configured to be adjustable, particularly by means of a control unit. This allows the security level of the method to be adapted. Furthermore, a predetermined probability can be assigned to the respective features of the risk feature, particularly by means of a control unit. This makes it even easier to adapt the method to requirements. In this respect, the method can be adapted, for example, to determine features of the risk feature that are intentionally ignored and / or masked out by the body scanner and / or are not detectable with a predetermined probability.

[0018] In particular, it can be provided that the sensor system comprises exclusively optical sensors, in particular cameras. Thus, the sensor system can be designed particularly simply. Furthermore, this can simplify the processing and / or evaluation of the data, especially when only one data type is to be processed.

[0019] In particular, the camera can be designed as a depth camera, grayscale, monochrome, black and white and / or RGB camera.

[0020] A grayscale camera and / or a monochrome camera and / or a black-and-white camera can help save data and / or reduce the computing power required for analysis. However, with the help of an image processing unit and / or the AI ​​system, the data may still be sufficient to reliably identify the features.

[0021] A depth camera and / or RGB camera provide more data, allowing more details to be detected during processing. This can increase security.

[0022] Alternatively or cumulatively, the camera can be configured as a 2D camera or a 3D camera. A 2D camera is advantageous because less data is captured. However, with the help of an image processing unit and / or the AI ​​system, the data may still be sufficient to reliably detect the features. 3D cameras have the advantage of providing more detail, allowing more details to be detected during processing.

[0023] The sensor system preferably comprises an image processing unit for processing the sensor data. The image processing unit can create a 3D view from the sensor data, particularly with the aid of the AI ​​system and / or another AI system. This can contribute to increased detection reliability.

[0024] The sensor system is preferably designed and configured to track the person and / or the risk characteristic within a detection range of the sensor system, in particular continuously. Thus, a successful passage and / or exit of the person and / or the risk characteristic can be detected. This tracking can preferably be supported and / or carried out using the AI ​​system and / or another AI system.

[0025] In particular, the data acquired by the sensor system can be processed prior to evaluation, particularly using an image processing unit and / or the AI ​​system. This step is considered pre-processing of the data. This can include filtering the data, in particular filtering out data errors and / or image errors such as noise, and / or generalizing and / or anonymizing the data. This can improve the evaluation because the data is already processed in a previous step. This allows computing power to be reduced and resources to be saved.

[0026] In particular, the sensor system can comprise a plurality of cameras. The sensor system preferably comprises at least two cameras, preferably three cameras, particularly preferably four cameras. The cameras can be configured, especially together, to achieve complete detection, in particular 360° detection, of the person.

[0027] In particular, a lighting device can also be arranged, especially in the form of a light strip. In particular, the lighting device can extend over substantially the entire height of the control device and / or be designed as LED lighting. This allows the optimal lighting ratio for the sensor system to be achieved. This serves to improve the visibility of features. In particular, the control device can comprise the lighting device.

[0028] Preferably, the risk characteristic comprises at least one of the following characteristics:

[0029] • at least partially closed hand position, especially fist position; and / or

[0030] • at least one glove on.

[0031] Since scanners are often unsuitable for scanning body parts, or they mask this area from the outset, determining this risk characteristic is advantageous. In particular, it may be sufficient to detect the risk characteristic in at least one of the person's hands. This can eliminate or assist personnel in performing a check. In particular, the glove may be a type of dummy hand.

[0032] Preferably, the risk feature comprises at least one of the following features: • at least one prohibited object in the area of ​​at least one hand, in particular comprising a wrist area, preferably excluding: ring and / or watch and / or bracelet; and / or

[0033] • at least one prohibited object in the area of ​​the head, in particular comprising a neck area, preferably excluding: glasses and / or hearing aid and / or earring and / or piercing and / or necklace and / or headscarf; and / or

[0034] • at least one prohibited item in the buttocks and / or waist area of ​​the person, preferably excluding: belts.

[0035] Since scanners are in most cases unsuitable for assessing these risk characteristics in terms of security, or they simply ignore these areas, determining these risk characteristics is advantageous.

[0036] The KL system can reliably identify body areas and the objects present there. This eliminates and / or supports personnel inspection. Sharp and / or pointed objects and / or objects made at least partially of metal can be identified as prohibited objects. The KL system is preferably designed and configured to classify objects. This allows, in particular, the aforementioned safe objects:

[0037] Ring and / or watch and / or bracelet and / or

[0038] Glasses and / or hearing aid and / or earring and / or piercing and / or necklace and / or headscarf and / or

[0039] Belts; can be reliably identified, so that these items are not determined to be at risk. This can eliminate and / or support further inspection by staff.

[0040] In particular, the prohibited object may be in the area of ​​both of the person's hands. In cases where the person only has one hand, e.g. because a hand or arm is amputated, checking the only hand may be sufficient. In particular, the AI ​​system can be set up to recognize such a case and / or other cases of a disability or handicap of the person, in particular a wheelchair and / or a visual impairment. In such cases, the AI ​​system or a control unit can cause exceptions to the procedure to be applied and / or for personnel to be consulted or called upon. Each individual feature can be understood as an embodiment of the invention in combination with one or more or all of the features previously described as part of the risk feature.The more characteristics are determined as part of the risk characteristic by the KL system, the more safety can be achieved through the procedure, regardless of whether further testing by the staff can be omitted and / or supported.

[0041] In particular, the risk feature may include the following features: at least one prohibited object in the area of ​​at least one hand and at least partially closed hand position, in particular fist position of the hand.

[0042] Preferably, the risk characteristic comprises at least one of the following characteristics:

[0043] • comprises at least one shoe sole, in particular a shoe heel, which exceeds a predetermined height and / or width; and / or

[0044] • a speed of the person, in particular at least one body part of the person, wherein the speed exceeds a predetermined speed.

[0045] Since scanners are often unsuitable for assessing these risk characteristics with regard to safety, determining these risk characteristics is advantageous. For example, the scanners cannot scan shoes of a certain thickness. Therefore, staff must currently instruct the person to remove such shoes before or after the scan. This can be improved. Furthermore, staff must currently instruct the person to maintain a certain position for a certain period of time so that the body scanner can perform the scan. By checking the speed, the AI ​​system can determine whether the person is exceeding a certain speed that would lead to an unreliable scan. This can eliminate the need for further checks by staff and / or support this.

[0046] Each individual feature, in combination with one or more or all of the features previously described as part of the risk feature, can be understood as an embodiment of the invention. The more features the AI ​​system determines as part of the risk feature, the greater the level of safety that can be achieved through the method, regardless of whether further personnel testing is eliminated and / or supported.

[0047] In particular, the risk feature can comprise the following features: at least one unauthorized object in the area of ​​at least one hand and / or at least partially closed hand position, in particular fist position of the hand and / or a speed of the person, in particular at least one body part of the person, wherein the speed exceeds a predetermined speed and / or comprises at least one shoe sole, in particular shoe heel, which exceeds a predetermined height and / or width.

[0048] Preferably, the risk characteristic comprises an abnormal behavior of the person, having at least one of the following characteristics:

[0049] • unusual facial expressions and / or unusual gestures; and / or

[0050] • an unusual movement pattern, in particular directly or indirectly before and / or directly or indirectly after the body scan is triggered; and / or

[0051] • Placing and / or leaving at least one object in the control device.

[0052] Unusual facial expressions and / or unusual gestures would be reflected, for example, in unusual eye and / or mouth and / or head and / or hand movements. In particular, the unusual movement pattern may include the person initially moving to the end or into a corner of the screening device before moving to the correct position. In particular, the person may also be stretching upwards or bending downwards. Detecting objects being put down and / or left lying around can not only indicate a risk but also serve as a reminder to avoid forgetting something. This can help identify conspicuous behavior patterns in people, which could also be recognized by staff. Abnormal behavior is characterized by the fact that it contradicts a person's usual behavior.This could indicate nervousness and imply that the person may want to tamper with the device. In such cases, staff could be notified.

[0053] Each individual feature can be understood as an embodiment of the invention in combination with one or more or all of the features previously described as part of the risk feature. This is because the more features are determined by the AI ​​system as part of the risk feature, the more security can be achieved by the method, regardless of whether further checking by the staff is unnecessary and / or can be supported. In particular, the risk feature can comprise the following features: at least one unauthorized object in the area of ​​at least one hand and / or at least partially closed hand position, in particular fist position of the hand and / or a speed of the person, in particular at least one body part of the person, wherein the speed exceeds a predetermined speed and / or placing and / or leaving at least one object in the control device.

[0054] Preferably, the method comprises the step:

[0055] - Evaluating the data acquired by the sensor system and / or the body scanner during the body scan using the Kl system and / or another Kl system in order to determine at least a deception potential of the person,

[0056] - where the deceptive potential includes a plurality of persons, at least two persons.

[0057] The term "deceptive potential" implies that this is a feature that could be suitable for manipulating the control device. Such deceptive potential could be determined from the data of the sensor system and / or the data of the body scanner, particularly after performing the body scan and / or after performing a pre-scan using the body scanner.

[0058] Since people are to be checked individually, this is particularly useful to recognize.

[0059] The evaluation of the data and the determination of the presence of at least one potential for deception can also be carried out using the AI ​​system and / or another AI system. The embodiments described with reference to the AI ​​system can also be applied to the other AI system.

[0060] In particular, a deception signal can be triggered when the at least one deception potential is present, wherein the signal is perceptible optically and / or acoustically and / or haptically and / or electronically, in particular by the person, and / or wherein the signal prevents the body scan from being triggered.

[0061] The potential for deception can be understood as an embodiment of the invention in combination with one or more or all of the features previously described as part of the risk feature. This is because the more features are part of the risk feature and / or the more potential for deception are determined by the AI ​​system and / or another AI system, the greater the security that can be achieved by the method, regardless of whether further personnel testing is eliminated and / or supported.

[0062] Preferably, the method comprises the step:

[0063] - Evaluating the data acquired by the sensor system and / or by the body scanner during the body scan using the Kl system and / or another Kl system to determine at least one deceptive potential of the person, wherein the deceptive potential comprises an upper part of the person's clothing which comprises a jacket and / or exceeds a predetermined proportion in relation to a size, in particular with regard to a width and / or a height, of the person.

[0064] This allows jackets and other items of clothing that could influence the result of the body scan to be reliably detected.

[0065] The evaluation of the data and the determination of the presence of at least one potential for deception can also be carried out using the AI ​​system and / or another AI system. The embodiments already described with regard to the AI ​​system can also be applied here.

[0066] In particular, a deception signal can be triggered upon a positive determination of the at least one deception potential, wherein the signal is optically and / or acoustically and / or haptically and / or electronically perceptible, in particular by the person, and / or wherein the signal prevents the body scan from being triggered.

[0067] The potential for deception can be understood as an embodiment of the invention in combination with one or more or all of the previously described potentials for deception and / or the features previously described as part of the risk feature. This is because the more features are determined as part of the risk feature and / or the more potentials for deception by the AI ​​system and / or another AI system, the greater the security that can be achieved by the method, regardless of whether further personnel testing can be omitted and / or supported.

[0068] In particular, the deceptive potential can include: a plurality of people, at least two people, and / or an upper part of the person's clothing, which comprises a jacket and / or exceeds a predetermined proportion in relation to a size, in particular with respect to a width and / or a height, of the person. Preferably, the method comprises the step:

[0069] - Evaluating the data acquired by the sensor system and / or the body scanner during the body scan using the Kl system and / or another Kl system in order to determine at least a deception potential of the person,

[0070] - wherein the deceptive potential comprises at least one item of clothing of the person made of a particularly dense, in particular difficult to penetrate, and / or of an at least partially particularly thick material.

[0071] Such clothing items may include, in particular, winter jackets, thick winter coats, and / or leather clothing. Preferably, such clothing items can be detected by the body scanner in combination with the AI ​​system.

[0072] The evaluation of the data and the determination of the presence of at least one potential for deception can also be carried out using the AI ​​system and / or another AI system. The embodiments already described with regard to the AI ​​system can also be applied here.

[0073] In particular, a deception signal can be triggered when the at least one deception potential is present, wherein the signal is perceptible optically and / or acoustically and / or haptically and / or electronically, in particular by the person, and / or wherein the signal prevents the body scan from being triggered.

[0074] The potential for deception can be understood as an embodiment of the invention in combination with one or more or all of the previously described potentials for deception and / or the features previously described as part of the risk feature. This is because the more features are determined as part of the risk feature and / or the more potentials for deception by the AI ​​system and / or another AI system, the greater the security that can be achieved by the method, regardless of whether further personnel testing can be omitted and / or supported.

[0075] In particular, the deceptive potential can include: a plurality of people, at least two people, and / or an upper part of the person's clothing which comprises a jacket and / or exceeds a predetermined proportion in relation to a size, in particular with regard to a width and / or a height, of the person and / or at least one item of clothing of the person made of a particularly dense, in particular difficult to penetrate, and / or of an at least partially particularly thick material.

[0076] Preferably, the method comprises the step:

[0077] Evaluating the data acquired by the sensor system and / or by the body scanner during the body scan using the AI ​​system and / or another AI system to determine at least one deception potential of the person, wherein the deception potential comprises a position of the person that is unsuitable for the body scan and / or for the most complete possible detection of the person by means of the sensor system and / or for a reliable evaluation by means of the AI ​​system, in particular with regard to the risk feature and / or with regard to further deception potentials.

[0078] The unsuitable position prevents a reliable body scan and / or the most complete detection of the person by the sensor system with a predetermined probability. Therefore, the unsuitable position must be corrected to ensure a high level of safety. A position unsuitable for the most complete detection of the person by the sensor system and / or for reliable evaluation by the AI ​​system may differ from or correspond to the position unsuitable for the body scan.

[0079] The evaluation of the data and the determination of the presence of at least one potential for deception can also be carried out using the AI ​​system and / or another AI system. The embodiments already described with regard to the AI ​​system can also be applied here.

[0080] In particular, a deception signal can be triggered when the at least one deception potential is present, wherein the signal is perceptible optically and / or acoustically and / or haptically and / or electronically, in particular by the person, and / or wherein the signal prevents the body scan from being triggered.

[0081] The potential for deception can be understood as an embodiment of the invention in combination with one or more or all of the previously described potentials for deception and / or the features previously described as part of the risk feature. This is because the more features are determined as part of the risk feature and / or the more potentials for deception by the AI ​​system and / or another AI system, the greater the security that can be achieved by the method, regardless of whether further personnel testing can be omitted and / or supported.In particular, the deceptive potential can include: a plurality of people, at least two people, and / or an upper part of the person's clothing which comprises a jacket and / or exceeds a predetermined proportion in relation to a size, in particular with regard to a width and / or a height, of the person and / or at least one item of clothing of the person made of a particularly dense, in particular difficult to penetrate, and / or of an at least partially particularly thick material and / or a position of the person which is unsuitable for the body scan and / or for the most complete detection of the person by means of the sensor system and / or for a reliable evaluation by means of the AI ​​system, in particular with regard to the risk feature and / or with regard to further deceptive potential.

[0082] Preferably, the body scanner is designed and configured to perform a body scan while the person remains in a substantially unchanged scanning position for a predetermined time.

[0083] The scan position is suitable for performing a reliable body scan. The presence of the scan position can also be determined using the KL system and / or another KL system. In particular, the body scan can be prevented as long as no scan position is present. This makes the process more efficient and saves energy, especially for energy-intensive body scans.

[0084] In particular, the risk feature can comprise at least one of the following features: a speed of the person, in particular of at least one body part of the person, wherein the speed exceeds a predetermined speed.

[0085] In particular, the potential for deception can include a person's position being unsuitable for the body scan. Such a combination enables reliable and / or energy-efficient screening of individuals.

[0086] Preferably, the body scanner is designed and configured to perform a body scan while the person moves through the control device, in particular, while the person's movement does not exceed a predetermined speed. In particular, such body scanners can use radiation that has different frequency ranges, in particular lower frequencies and thus radiation with a longer wavelength, than those body scanners that require a scanning position to perform the body scan.

[0087] In particular, the risk feature can comprise at least one of the following features: a speed of the person, in particular of at least one body part of the person, wherein the speed exceeds a predetermined speed.

[0088] Such a combination enables reliable testing even when people are moving.

[0089] Preferably, an output device is connected to the control device or the control device has the output device which can be controlled depending on a result of the determination of the at least one risk feature and / or on a result of the determination of the at least one deception potential and / or on a result of the body scan.

[0090] The exit gate can be opened if no risk features and / or no potential for deception and / or no security risk are identified in the body scan results. This allows only those persons who do not pose any security concerns or whose security concerns can be efficiently resolved during a follow-up check to pass through the exit gate.

[0091] This allows the process to be further automated, eliminating and / or supporting security personnel's monitoring.

[0092] In particular, an electromechanical drive can be provided for at least partially or completely automatically controlling the exit device. In particular, the control device can comprise the drive. The drive can move or release the exit device.

[0093] The exit device preferably comprises an exit from the control device into at least two separate exit paths depending on a result of the determination of the at least one risk feature and / or on a result of the body scan, preferably wherein: • the exit device comprises a barrier which is movable between at least two positions, in particular automatically, in particular wherein the barrier in a first position releases the exit into a first exit path and at the same time blocks the exit into a second exit path and / or wherein the barrier in a second position blocks the exit into the first exit path and at the same time releases the exit into the second exit path.

[0094] The barrier can be designed as a switch, which either opens the first exit path and simultaneously blocks the second exit path, or opens the second exit path and simultaneously blocks the first exit path. In particular, the barrier can be designed as a rotating arm barrier, turnstile, revolving door, sliding door, folding door, or pivoting door, in particular a double pivoting door. In particular, an electromechanical drive, in particular, can be provided for at least partially or completely automatically moving the barrier. In particular, the control device can comprise the drive. The drive can move the barrier into the two positions or release it.

[0095] A rotating arm barrier can comprise one, two, three, or more arms that are rotatable about a vertical axis of rotation or a non-vertical axis of rotation. A turnstile can comprise two or more locking elements that are rotatable about a vertical axis of rotation. A rotating leaf can comprise at least one leaf that is rotatable about a vertical axis of rotation. A pivoting leaf can comprise one leaf that is pivotable about a vertical axis of rotation or a non-vertical axis of rotation.

[0096] A sliding door can comprise one or more sliding panels. A folding door can comprise one or more sliding panels, with the individual panels pivoting relative to each other. This allows a folding door to move like an accordion.

[0097] Preferably, the at least two output paths are physically separated from each other. This prevents sabotage.

[0098] In particular, a barrier design as a rotating arm barrier, a rotating leaf, a pivoting leaf, a sliding door, or a folding leaf door can allow for advantageous movement of the barrier, particularly between the two positions. Alternatively or in addition to a barrier, the exit device and / or the control device can comprise a barrier, in particular a virtual barrier. The term "virtual barrier" refers to a barrier using an acoustic and / or optical signal. In particular, the signal can be perceptible from the immediate and / or distant surroundings, thus allowing for simple control.

[0099] In particular, the control device may comprise a control unit by means of which the method can be executed, in particular computer-implemented. In particular, the control unit may be embodied as a computer or may comprise the computer.

[0100] Preferably, the control device comprises a control unit for triggering the body scan, preferably wherein the control unit is configured to control the sensor system and / or the Kl system and / or the output device.

[0101] This allows the control unit to efficiently send the signal to trigger the body scan. This allows for efficient processing of the procedure. Preferably, the control unit is configured to control the body scanner.

[0102] In particular, data and / or signals can be received and / or transmitted wirelessly and / or wired by the control unit. Wired transmission can be more secure, faster, and / or more robust against errors. Wireless transmission can enable a flexible design of the elements, particularly without additional hardware.

[0103] In particular, the Kl system and / or the further Kl system can be designed as part of the control unit and / or as part of the sensor system or as part of a central processing unit or as part of an external processing unit, in particular in a cloud.

[0104] When implementing the AI ​​system as part of the control unit and / or as part of the sensor system, it is advantageous that the data and / or signals do not have to be transmitted over particularly long distances. Furthermore, this enables greater reliability and / or lower latency.

[0105] A central processing unit (CPU) means that the processing unit can act as the central hub for a larger building management system, such as an airport, to which multiple devices can be connected. Data can also be transmitted wired and / or wirelessly, particularly without the need for internet transmission. Transmission can take place via an internal network. This ensures greater data security.

[0106] When implementing the Kl system as part of an external unit, it is advantageous that less hardware is required on site.

[0107] In particular, it can be provided that the control device does not store any data from the sensor system, or that the data is stored only temporarily. This eliminates the need for a large storage unit. Alternatively, the data can be stored anonymously. This eliminates data protection considerations, allowing the method and / or the control device as a whole to be implemented in a technically simpler manner.

[0108] The described steps of evaluating to determine at least one potential for deception in the person and evaluating to determine at least one risk characteristic can be performed simultaneously, in particular at least partially simultaneously. This ensures that the test and / or the steps are performed on the same person.

[0109] Alternatively, these steps can be performed sequentially, whereby the sensor system and / or the AI ​​system and / or the control unit can be configured to continuously track the same person, in particular based on defined characteristics and / or patterns. This ensures that the test and / or the steps are performed on the same person. In particular, the evaluation step to determine at least one risk characteristic can be performed first.

[0110] The body scan can be initiated at least partially simultaneously with the described steps of the method. This ensures that the test and / or the steps are performed on the same person and that the same person is scanned using the body scanner. Alternatively, the body scan can be initiated after the described steps. In particular, in this case, the sensor system and / or the AI ​​system and / or the control unit can be configured to track the person, in particular based on defined characteristics and / or patterns. This ensures that the test and / or the steps are performed on the same person and that the same person is scanned using the body scanner. In particular, the body scan can be initiated

[0111] - if there is no risk characteristic; and / or

[0112] - if a risk feature is present which includes one or more of the following features: o at least one prohibited object in the area of ​​at least one hand and / or in the area of ​​the head and / or object in the buttocks area and / or waist area; and / or o at least one shoe sole, in particular shoe heel, which exceeds a predetermined height and / or width o unusual facial expressions and / or unusual gestures; and / or o an unusual movement pattern, in particular indirectly or immediately before the triggering and / or indirectly or immediately after the triggering of the body scan; and / or o placing and / or leaving at least one object in the control device.

[0113] Alternatively or cumulatively, the body scan can be triggered

[0114] - if there is no potential for deception; and / or

[0115] - if one or more of the deceptive potentials are present.

[0116] In particular, the remaining features of the risk feature and / or the remaining potential for deception can be checked during a follow-up check, in particular by security personnel. Preferably, the remaining features of the risk feature and / or the remaining potential for deception can be displayed, in particular to the security personnel, using the playback device and / or another playback device.

[0117] Preferably, the AI ​​system is based on a convolutional neural network (CNN) and / or on a recurrent neural network (RNN), preferably wherein the AI ​​system comprises and / or uses a machine learning model, particularly preferably wherein supervised learning and / or semi-supervised learning and / or unsupervised learning has been and / or is carried out in the learning model.

[0118] In particular, the AI ​​system can be implemented in the form of computer vision. Computer vision is a field of artificial intelligence that aims to give computers the ability to understand and interpret images and videos similarly to how a human would. Computer vision can be used to automatically extract, analyze, and understand useful information, particularly risk features and / or deceptive potentials, from a single image or a series of images.

[0119] The advantage is that computer vision systems can analyze images and videos on a large scale and in real time, something humans cannot do. They can operate around the clock, and their accuracy is not affected by fatigue or daily routine. Furthermore, computer vision systems do not have the same biases or subjectivity as human observers. They deliver consistent results. Therefore, computer vision and / or the AI ​​system in general can be used to automate and / or support security screening tasks, especially visual inspection of security screening. Such tasks have previously been performed primarily by humans.

[0120] According to one embodiment of the present invention, it is possible for the AI ​​system to have and / or implement a machine learning model, a deep learning model, a neural network, contour detection, and / or pattern recognition. The AI ​​system is configured and designed to reliably detect the at least one described risk feature, in particular all risk features. Furthermore, the AI ​​system can be configured and designed to detect the at least one deceptive potential.

[0121] In one embodiment, the AI ​​system comprises an underlying machine learning model. Suitable machine learning models are based on a convolutional neural network (CNN) or a recurrent neural network (RNN), or a combination thereof. CNNs and RNNs are both types of deep learning algorithms commonly used in machine learning applications. A CNN is a type of neural network particularly well-suited for image and video recognition tasks.

[0122] Generally, a CNN uses convolutional layers to extract features from input data, followed by pooling layers to reduce the dimensionality of the feature maps. The output of the convolutional and pooling layers is then flattened and fed into one or more fully connected layers, which are responsible for the final classification decision. The main advantage of CNNs is their ability to automatically learn hierarchical representations of input data, making them very effective at detecting complex patterns in images and videos. An RNN, on the other hand, is a type of neural network particularly well-suited to sequence data, such as time series or natural language. An RNN uses recurrent connections to pass information from one layer to the next.This allows the network to maintain an internal state or memory for previous inputs, which is useful for tasks such as speech recognition or language translation. The main advantage of RNNs is their ability to capture temporal dependencies in the input data, making them very effective at modeling sequential data.

[0123] A key difference between CNNs and RNNs is that CNNs are designed to process input data of fixed size (e.g., fixed-resolution images), while RNNs can process input data of variable length (e.g., sequences of different lengths). Another difference is that CNNs are better suited for tasks that involve spatial information (e.g., image classification), while RNNs are better suited for tasks that involve temporal information (e.g., motion detection).

[0124] In one embodiment of the present invention, an AI system based on CNNs is used. This enables efficient evaluation of image data and / or tends to require fewer computational resources than RNNs, as they do not have repeating loops. This simplifies on-site implementation of the AI ​​system, for example, as part of the sensor system and / or the control unit. A further advantage is that CNNs are robust against changes in the position of the features in the input space thanks to their convolutional design.

[0125] In one embodiment of the present invention, a combination of CNNs and RNNs can be provided, since both spatial and temporal information are important for video recognition tasks. A CNN can be used to extract spatial features from each frame of the video, which are then fed into an RNN to model the temporal dependencies between frames.

[0126] In this respect, a risk feature can be identified using CNN, while its tracking and continued presence can be checked using RNN.

[0127] The machine learning model can be trained on a large dataset of labeled examples, where each example consists of a time series of data points and a corresponding label indicating the type of risk feature and / or deceptiveness. The training data is used to adjust the model's weights and biases so that it can accurately classify new, unseen examples. The training process typically involves an iterative approach, where the model is presented with a set of training examples and the weights and biases are adjusted to minimize the difference between the predicted output and the actual label. This process is repeated over many epochs until the model has learned to accurately classify the risk feature and / or deceptiveness in the training data.

[0128] The type of machine learning model that can be used in this context is typically supervised learning, where the model is trained on a labeled dataset. However, unsupervised learning approaches such as clustering and anomaly detection can also be used within the scope of the present invention, for example, to populate the aforementioned database.

[0129] Embodiments of the invention may be based on the use of a machine learning model or a machine learning algorithm. Machine learning may refer to algorithms and statistical models that computer systems can use to perform a specific task without explicit instructions, relying instead on models and inferences. For example, instead of a rule-based transformation of data, machine learning may use a transformation of data derived from an analysis of historical and / or training data. For example, the content of images may be analyzed using a machine learning model or a machine learning algorithm. In order for the machine learning model to analyze the content of an image, the machine learning model may be trained with training images as input and training content information as output.By training the machine learning model with a large number of training images and / or training sequences (e.g., words or sentences) and associated training content information (e.g., labels or annotations), the machine learning model "learns" to recognize the content of the images, allowing the content of images not included in the training data to be recognized using the machine learning model. The same principle can also be used for other types of sensor data: By training a machine learning model with training sensor data and a desired output, the machine learning model "learns" a transformation between the sensor data and the output, which can be used to provide an output based on non-training sensor data provided to the machine learning model. The provided data (e.g.,Sensor data, metadata, and / or image data) can be preprocessed to obtain a feature vector that is used as input to the machine learning model.

[0130] Machine learning models can be trained using training data. In the examples above, a training method called "supervised learning" is used. In supervised learning, the machine learning model is trained using a large number of training samples, where each sample can include a large number of input data values ​​and a large number of desired output values, meaning that each training sample is associated with a desired output value. By specifying both training samples and desired output values, the machine learning model "learns" what output value it should produce based on an input sample that is similar to the samples provided during training. In addition to supervised learning, semi-supervised learning can also be used. In semi-supervised learning, some of the training samples lack a corresponding desired output value.Supervised learning can be based on a supervised learning algorithm (e.g., a classification algorithm, a regression algorithm, or a similarity learning algorithm). Classification algorithms can be used when the output values ​​are restricted to a limited set of values ​​(categorical variables), meaning the input is assigned to one of the limited set of values. Regression algorithms can be used when the outputs can have any numerical value (within a range). Similarity learning algorithms can be similar to both classification and regression algorithms, but are based on learning from examples using a similarity function that measures how similar or related two objects are. In addition to supervised or semi-supervised learning, unsupervised learning can also be used to train the machine learning model.In unsupervised learning, (only) input data can be provided, and an unsupervised learning algorithm can be used to find structure in the input data (e.g., by grouping or clustering the input data, finding commonalities in the data). Clustering is the assignment of input data, which includes a large number of input values, to subsets (clusters) such that input values ​​within the same cluster are similar according to one or more (predefined) similarity criteria, while differing from input values ​​contained in other clusters. Reinforcement learning is a third group of machine learning algorithms that can be used to train the machine learning model. In reinforcement learning, one or more software actors (so-called "software agents") are trained to perform actions in an environment.A reward is calculated based on the actions performed. In reinforcement learning, one or more software agents are trained to choose actions in such a way that the cumulative reward increases, resulting in the software agents becoming increasingly better at completing the task assigned to them (which translates into increasing rewards).

[0131] Furthermore, some techniques can be applied to some of the machine learning algorithms. For example, feature learning can be used. In other words, the machine learning model can be trained at least partially using feature learning and / or the machine learning algorithm can include a feature learning component. Feature learning algorithms, which can also be called representation learning algorithms, can preserve the information in their inputs but also transform it to make it useful, especially what is often done as a preprocessing step before performing classification or prediction. Feature learning can be based on, for example, principal component analysis or cluster analysis.

[0132] In particular, the Kl system can be designed and configured, in particular trained, to determine at least one, in particular several or all features of the risk feature, and / or at least one, in particular several or all deception potentials from the sensor data of the sensor system and / or the body scanner, in particular to separate them and / or to assign them accordingly.

[0133] Through the use of the KL, security checks can be largely automated and / or carried out in a particularly reliable manner, supporting staff.

[0134] In particular, learning can be achieved by providing the control device with new training data as an update. This allows the method to be subsequently improved.

[0135] In particular, a learning model can be applied in which initially data that can be assigned with insufficient probability, preferably from multiple control devices, is stored, particularly anonymized, on one or more storage devices. In the next step, this data can be assigned to a characteristic of the risk attribute and / or a deceptive potential. After assignment, this data can then be supplied, preferably to multiple control devices, as training data, particularly in the form of an update. This allows control devices to benefit from the data of other control devices and ensure greater security.

[0136] Preferably, an optical reproduction device is arranged and configured

[0137] • to display at least one identified risk characteristic; and / or

[0138] • to indicate at least one identified potential for deception; and / or

[0139] • to display at least one dangerous object detected by the body scanner; and / or

[0140] • to display a scanning position of the person (50), wherein the scanning position is suitable for successfully scanning the body of the person (50); and / or

[0141] • to display a test position, whereby the test position is suitable for carrying out the most complete possible detection of the person by means of the sensor system and / or for carrying out a reliable evaluation by means of the Kl system.

[0142] In particular, the control device may comprise the playback device.

[0143] The scanning position can include the position of the person within the interior area and / or the person's posture. In particular, the scanning position can include a frame within which the person must position themselves to enable successful scanning of the body. The scanning position can be adjusted depending on the person's size and / or contour and / or position. This option is created without the need for dedicated security personnel to indicate the problem to the person. This can prevent delays during personal checks. This can reduce waiting times while still achieving an advantageous level of security.

[0144] The test position is suitable for the most complete possible detection of the person using the sensor system and / or for a reliable evaluation using the AI ​​system, particularly with regard to the risk characteristic and / or the potential for deception. The test position can differ from the scanning position or correspond to the scanning position. Adopting the test position can improve security.

[0145] This allows the person to be easily and quickly informed of any remaining problems during the security check. These can be displayed using a generic body image, particularly an avatar. This allows the person to quickly identify and resolve the problem. This makes a successful check more efficient.

[0146] The scanning position is preferably displayed in animated form. In other words, it is displayed as an animation and / or a moving image. In particular, an image of the person can be used as the beginning of the animation. In particular, a motion animation leading to the appropriate scanning position can be displayed. In particular, the animation can end with the scanning position suitable for successful scanning, in particular where the appropriate scanning position is derived from the image of the person and / or includes a frame. This makes it easier for the person to understand which position they need to assume, thus saving time.

[0147] Preferably, the scan position of the person is displayed using the Kl system and / or another Kl system.

[0148] Preferably, an image and / or avatar of the person currently in the lock is used to display the scanning position, in particular wherein the image and / or avatar takes into account a spatial position of the person and / or a size of the person and / or a contour of the person and / or the actual appearance of the person, in particular in an attenuated or softened form. In particular, the image can be scaled to a standard size. Preferably, the image has at least partially one or more marked body areas, wherein the marked area indicates to the person whether the position is correct. The marking can be done, for example, by color markings. Alternatively or cumulatively, a status message, in particular as a color marking, can be provided preferably by means of the display device and / or the further display device, wherein the status message indicates whether the scanning position has been assumed.

[0149] This allows for a suitable display. The person can be interactively and effectively encouraged to move into the correct scanning position. This improves the efficiency of the airlock. The AI ​​system and / or the additional AI system can be used to create the image and / or avatar.

[0150] Another object of the invention is a control device designed and configured to carry out the method according to the invention, comprising:

[0151] - a sensor system;

[0152] - a control system. The control device can comprise an interior space that is or can be sealed off from the outside.

[0153] The features, embodiments, and advantages already described in connection with the method according to the invention can be applied to the control device according to the invention. In particular, the control device can comprise additional components, in particular all components, for carrying out the described steps.

[0154] The control device is preferably designed such that it can be used to control the flow of people and / or to conduct a person check. In particular, at least the exit can be at least partially closed, particularly by means of the exit device of the control device, so that people cannot pass through the control device unhindered. Furthermore, the entrance to the control device can preferably be at least partially closed.

[0155] Preferably, an access solution for an area and / or a building, in particular for an airport and / or a train station, is implemented using the control device. The control device can be used to manage and / or control the flow of people into a restricted area and / or a security area.

[0156] The control device can comprise an input, in particular controllable by means of an input device, and an output, in particular controllable by means of an output device.

[0157] Alternatively, the control device may include an input that also serves as an output from the control device. In this case, the input and output can be controlled by a combined input and output device.

[0158] The control device is preferably designed as a personnel checkpoint, with which a flow of people can be controlled and / or a person check can be carried out. In particular, at least the exit can be at least partially closed, in particular by means of the exit device of the checkpoint, so that people cannot pass through the checkpoint unhindered. Preferably, the entrance to the checkpoint is also at least partially closeable. Preferably, the checkpoint is used to create an access solution for an area and / or a building, in particular for an airport and / or a train station. With the help of the checkpoint, a flow of people into a cordoned-off area and / or a security area can be controlled and / or monitored.

[0159] Preferably, the airlock is designed and configured to allow entry into the airlock, in particular by means of the entrance door device, only for a single person, while an exit from the airlock is at least partially closed, in particular by means of the exit door device. After the single person has entered, the entrance can be closed so that the single person can be checked. If the check was successful, the exit, in particular exit path x or exit path y, can be opened. As soon as the checked person has left the airlock, the exit can be at least partially closed again and the entrance opened so that another single person can enter the airlock. If the check was not successful, the exit can remain closed as long as the check is successful. The person can then repeat the steps if necessary.Alternatively, if the check is unsuccessful, the person can only leave the lock via a specific exit route, especially for a follow-up check.

[0160] A further subject of the present invention is a computer-implemented method in which one, several, or all steps of the method according to the invention are computer-implemented and / or executed and / or triggered using a computer. In particular, the individual components of the control device are controlled by a computer. At least one computer can be arranged for this purpose. The computer can be embodied as part of the control unit.

[0161] A further subject matter of the present invention is a computer program product, wherein the computer program product comprises instructions which, when the computer program product is executed by a computer, in particular an electronic control unit, preferably of the control device, cause the computer to execute a method according to an embodiment of the present invention. The computer is preferably designed partially or completely as part of the control device and / or partially or completely external to the control device. The computer is preferably designed to control at least one, in particular several components of the control device. It is conceivable that the computer is a single computer device or that the computer comprises several distributed computer devices. The several computer devices can in particular

[0162] TI may be arranged at different locations, for example partly as part of the control device and / or partly as part of a monitoring device for security personnel.

[0163] Another subject of the present invention is a computer-readable storage medium comprising the computer program product.

[0164] For the computer-implemented method according to the invention, the computer program product according to the invention and the computer-readable storage medium according to the invention, the features, embodiments and advantages that have already been described in connection with the method according to the invention and / or in connection with an embodiment of the control device according to the invention can be applied.

[0165] In the following, advantages and details of the invention will be explained with reference to the exemplary embodiments shown in the figures.

[0166] Fig. 1 is a schematic representation of a control device according to a first embodiment of the present invention;

[0167] Fig. 2 is a schematic plan view of a control device according to a second embodiment of the present invention;

[0168] Fig. 3 is a schematic representation of a control device according to a third embodiment of the present invention;

[0169] Fig. 4 is a schematic representation of a method according to the invention.

[0170] Figure 1 shows a schematic representation of a screening device 1 used for security screening of persons according to a first embodiment of the present invention. The device 1 comprises an interior space accessible via an entrance device 2 designed as an automatic double sliding door. A person 50 to be screened is depicted in the interior space. The device 1 further comprises a sensor system 6 having four optical sensors. The sensors are designed as cameras and are configured to achieve complete detection, in particular 360° detection, of the person 50. Each camera is arranged in one of the four corners of the interior space and has a different perspective so that the person is recorded as completely as possible. Furthermore, an AI system 8 is shown, which is data-connected to the sensor system 6. The device 1 further comprises a body scanner 5 and a playback device 7.Furthermore, a control unit 9 is shown, which in turn is data-connected to the Kl system 8 and is set up to control the aforementioned elements.

[0171] The method according to the invention is carried out as follows. When the person 50 approaches the control device 1, the control unit 9 activates the sensor system 6. This saves energy when no person is present. Activation occurs when a sensor (not shown) located near the control device 1 detects the presence of objects. Alternatively, activation can occur when one of the cameras of the sensor system 6 is active and detects the presence of objects. If objects are present, an animated test position is displayed by the playback device 7, which enables the most complete detection of the person 50 by the sensor system 6. This increases the reliability of the evaluation.

[0172] Person 50 is detected by sensor system 6. The detected data is evaluated using AI system 8 to determine at least one risk characteristic of person 50. The risk characteristic cannot be detected by body scanner 5 with a predetermined probability because the scanner technology is not designed for this purpose. Based on the evaluation, it is determined that at least one risk characteristic is present or that no risk characteristic is present, and it is further determined whether or not there is any potential for deception. If a risk characteristic is present, a risk signal is generated, which triggers a risk signal, the risk signal being visually displayed on display device 7, and the risk signal preventing the body scan from being triggered. If there is any potential for deception, a deception signal is triggered, which in this example is designed like the risk signal.

[0173] In this example, the following features of the risk feature are examined: at least partially closed hand position; at least one prohibited object in the area of ​​at least one hand, comprising a wrist area, excluding: ring and / or watch and / or bracelet; at least one shoe sole, in particular shoe heel, which exceeds a predetermined height and / or width;

[0174] •Placing and / or leaving at least one object in the control device.

[0175] In combination with this, the following deception potentials are checked: a plurality of persons; a position of the person that is unsuitable for the body scan or for the most complete possible recording of the person by means of the sensor system or for a reliable evaluation by means of the KL system with regard to the risk characteristic and / or with regard to further deception potentials; at least one item of clothing of the person made of a particularly dense, in particular difficult to penetrate, and / or of an at least partially particularly thick material.

[0176] If no risk feature or deceptive potential is present, a scanning position suitable for the body scan is displayed using the display device 7. In this example, the scanning position corresponds to the test position. However, in other embodiments, the positions may differ. After assuming the scanning position, which is also detected by the AI ​​system 8 in conjunction with the sensor system 6, the body scan is initiated by the body scanner 5 via the control unit 9. If no hazardous objects are detected here either, the person 50 has passed the security check.

[0177] With such a screening device 1, the security screening of persons can be automated. This allows parts of the security personnel, for example at airports, to be replaced or relieved of their workload, since many security deficiencies can already be detected automatically and reliably.

[0178] Figure 2 shows a schematic plan view of a control device according to a second embodiment of the present invention, which essentially corresponds to the device according to Figure 1. In contrast to Figure 1, an entrance device 20 is designed as an upstream entrance device and comprises a ticket control device 21 and / or an identity control device 22 and / or a sensor device 23. The entrance device 20 further comprises a pivoting wing (not shown) as a barrier. The entrance device 20 is designed to allow only authorized persons to pass through, and only one person at a time. The body scanner 5 is designed and configured to perform a body scan while the person moves through the control device, wherein the person's movement does not exceed a predetermined speed.Therefore, in addition to Figure 1, the following feature is checked as a risk feature: a speed of the person, in particular of at least one part of the person's body, where the speed exceeds a predetermined speed. Barriers 32 are provided which are intended to prevent unauthorized leaving of the permitted areas. Also shown here is an exit device 3 in the form of a rotating arm barrier with an arm which can be moved into at least two positions, namely, as shown, into a first position with a solid line which releases path y and into a second dashed position which blocks path y and releases path x. Furthermore, a third dash-dotted position is shown which releases both paths x and y.The control unit will move to the position indicated by the solid line if there are still safety concerns during the security check, as security personnel will conduct a follow-up check along path y. If no safety concerns are identified, path x is released, allowing the person to proceed directly without any safety concerns.

[0179] This ensures that the flow of people through the control device 1 is always in one direction. This reduces congestion and allows the flow of people to be more efficient, particularly without reverse movements.

[0180] Figure 3 shows a schematic representation of a control device according to a third exemplary embodiment of the present invention, which essentially corresponds to the device according to Figure 2. In contrast, the exit device 30 is designed as an automatic sliding door. Depending on which of the x and y paths is to be opened or closed, the sliding door moves into one position or the other in order to block the respective path in conjunction with the barrier 32. In the illustration, however, both paths appear closed. Furthermore, in this case, only two cameras of the sensor system 6 are arranged in the upper region of the control device 1. This embodiment also enables a continuous flow of people through the control device 1.

[0181] Figure 4 shows a schematic sequence of the method according to the invention. In step 70, the person enters the control device. In step 71, the person is detected by the sensor system. In step 72, the data detected by the sensor system is evaluated using the AI ​​system in order to determine at least one risk characteristic of the person and, based on the evaluation, to determine that at least one or no risk characteristic is present. If a risk characteristic is present, a risk signal is triggered in step 73. The person can then eliminate the risk characteristic. In step 74, the data detected by the sensor system is evaluated using the AI ​​system in order to determine at least one potential for deception of the person and, based on the evaluation, to determine that at least one or no potential for deception is present. If potential for deception is present, a deception signal is triggered in step 75.The person can then eliminate the potential for deception. Steps 72 to 75 occur at least partially simultaneously.

[0182] As soon as no risk feature and no potential for deception are present, the body scan is triggered and carried out in step 76. If a dangerous object or a potential for deception or another security risk was identified during the body scan in step 77, step 76 must be repeated. If no security risks were identified in step 76, an exit device is controlled accordingly in step 78 so that the person can leave the control device according to the exit path permitted for them. The method can be carried out by reading a computer-readable storage medium comprising a computer program product by means of a computer. The commands of the computer program product are executed by the computer and cause the computer to carry out the method according to the invention. The control unit can comprise the computer.

[0183] List of reference symbols

[0184] 1 Control device

[0185] 2 entrance device

[0186] 3 Exit device

[0187] 5 body scanners

[0188] 6 Sensor system

[0189] 7 Playback device

[0190] 8 Kl (artificial intelligence) system

[0191] 9 Control unit

[0192] 20 entrance device

[0193] 21 Ticket control device

[0194] 22 Identity control device

[0195] 23 Sensor device

[0196] 30 Exit device

[0197] 32 Barrier

[0198] 50 Person x Path xy Path y

Claims

Claims 1. A method for security control of at least one person (50) using a Control device (1), the control device (1) comprising a body scanner (5) for performing a body scan of the person (50), a sensor system (6) for detecting the person (50) with one or more cameras and a Kl system (8), wherein the method comprises the following steps, in particular in the listed sequence: • Evaluating the data acquired by the sensor system (6) with the aid of the Kl system (8) in order to determine at least one risk characteristic of the person (50), wherein the risk characteristic is not recognizable by means of the body scanner (5), in particular not recognizable with a predetermined probability, and / or wherein the risk characteristic is intentionally ignored and / or hidden by means of the body scanner (5); • Determine based on the evaluation that at least one risk characteristic is present; • Triggering a risk signal when the at least one risk feature has been determined, wherein the risk signal is perceptible optically and / or acoustically and / or haptically and / or electronically, in particular by the person (50), and / or wherein the risk signal prevents the body scan from being triggered.

2. The method according to claim 1, wherein the risk feature comprises at least one of the following features: • at least partially closed hand position, especially fist position; and / or • at least one glove on.

3. The method according to claim 1 or 2, wherein the risk feature comprises at least one of the following features: • at least one prohibited object in the area of at least one hand, in particular comprising a wrist area, preferably excluding: ring and / or watch and / or bracelet; and / or • at least one prohibited object in the area of the head, in particular comprising a neck area, preferably excluding: glasses and / or hearing aid and / or earring and / or piercing and / or necklace and / or headscarf; and / or • at least one prohibited object in the buttocks area and / or waist area of the person (50), preferably excluding: belts.

4. Method according to one of the preceding claims, wherein the risk feature comprises at least one of the following features: • comprises at least one shoe sole, in particular a shoe heel, which exceeds a predetermined height and / or width; and / or • a speed of the person (50), in particular of at least one body part of the person (50), wherein the speed exceeds a predetermined speed.

5. Method according to one of the preceding claims, wherein the risk feature comprises an abnormal behavior of the person (50), having at least one of the following features: • unusual facial expressions and / or unusual gestures; and / or • an unusual movement pattern, in particular directly or indirectly before and / or directly or indirectly after the body scan is triggered; and / or • Placing and / or leaving at least one object in the control device (1).

6. A method according to any one of the preceding claims, wherein the method comprises the step: • Evaluating the data acquired by the sensor system (6) and / or by the body scanner (5) during the body scan using the Kl system (8) and / or another Kl system in order to determine at least one deception potential of the person (50), • where the deceptive potential includes a plurality of persons, at least two persons.

7. A method according to any one of the preceding claims, wherein the method comprises the step: • Evaluating the data acquired by the sensor system (6) and / or by the body scanner (5) during the body scan using the Kl system (8) and / or another Kl system in order to determine at least one deception potential of the person (50), • wherein the deceptive potential comprises at least one item of clothing of the person (50) made of a particularly dense, in particular difficult to penetrate, and / or of an at least partially particularly thick material.

8. Method according to one of the preceding claims, wherein the body scanner (5) is designed and arranged to carry out a body scan while the person (50) remains in a substantially unchanged scanning position for a predetermined time.

9. Method according to one of the preceding claims, wherein the body scanner (5) is designed and configured to carry out a body scan while the person (50) moves through the control device (1), in particular wherein the movement of the person (50) does not exceed a predetermined speed.

10. Method according to one of the preceding claims, wherein an output device (3, 30) is connected to the control device (1) or the control device (1) has the output device (3, 30) which can be controlled depending on a result of the determination of the at least one risk feature and / or on a result of the determination of the at least one deception potential and / or on a result of the body scan.

11. The method according to claim 10, wherein the output device (3, 30) enables an output from the control device (1) into at least two separate output paths depending on a result of the determination of the at least one risk feature and / or on a result of the body scan, preferably wherein: • the exit device (3, 30) comprises a barrier which is movable between at least two positions, in particular automatically, in particular wherein the barrier in a first position releases the exit into a first exit path and at the same time blocks the exit into a second exit path and / or wherein the barrier in a second position blocks the exit into the first exit path and at the same time releases the exit into the second exit path.

12. Method according to one of the preceding claims, wherein the control device (1) comprises a control unit (9) for triggering the body scan, preferably wherein the control unit (9) is configured to control the sensor system (6) and / or the KI system (8) and / or the output device (3, 30).

13. Method according to one of the preceding claims, wherein the Kl system (8) is based on a convolutional neural network (CNN) and / or on a recurrent neural network (RNN), preferably wherein the Kl system (8) comprises and / or uses a machine learning model, particularly preferably wherein supervised learning and / or semi-supervised learning and / or unsupervised learning has taken place and / or takes place in the learning model.

14. Method according to one of the preceding claims, wherein an optical reproduction device is arranged and set up, • to display at least one identified risk characteristic; and / or • to indicate at least one identified potential for deception; and / or • to display at least one dangerous object detected by the body scanner (5); and / or • to display a scanning position of the person (50), wherein the scanning position is suitable for successfully scanning the body of the person (50); and / or • for displaying a test position, wherein the test position is suitable for carrying out the most complete possible detection of the person (50) by means of the sensor system (6) and / or for carrying out a reliable evaluation by means of the Kl system (8).

15. Control device (1) designed and configured to carry out the method according to one of the preceding claims, comprising: • a sensor system (6); • a Kl system (8).

16. A computer program product, wherein the computer program product comprises instructions which, when the computer program product is executed by a computer, cause the computer to execute a method according to one of claims 1 to 14.

17. A computer-readable storage medium comprising the computer program product of claim 16.

Images