Detection method for variable detection of alarm objects in scanning results of a scanning device for pieces of luggage

Abstract

The present invention relates to a detection method for variable detection of alarm objects (AO) in scanning results (SE) of a scanning device (100) for pieces of luggage (G), involving the following steps: - acquiring a scanning result (SE) of the scanning device (100); - analyzing the scanning result (SE) for the presence of alarm objects (AO); - determining a detection probability (EW) that the at least one detected alarm object (AO) is actually an alarm object (AO); - specifying a variable probability limit value (WG); - comparing the determined detection probability (EW) for the at least one detected alarm object (AO) with the specified variable probability limit value (WG); - outputting an alarm signal (AS) should the detection probability (EW) for the at least one alarm object (AO) exceed the probability limit value (WG).

Description

[0001] Detection method for variable detection of alarm objects in scan results from a luggage scanning device

[0002] The present invention relates to a detection method for variable detection of alarm objects in scan results of a luggage scanning device, a detection device for carrying out such a detection method, and a computer program product for carrying out such a detection method.

[0003] It is well known that scanning devices are used to protect sensitive areas by checking people's luggage. Besides their use in secure areas at airports, this also applies to secure areas in high-rise buildings, event venues, stadiums, and similar locations. Especially in so-called non-aviation areas, i.e., situations outside of airports, it is crucial to ensure the highest possible throughput of people and, consequently, luggage per unit of time. While maximum security must be guaranteed for high-security applications, such as at an airport, other applications can operate with lower security levels to allow for a correspondingly higher throughput of luggage.

[0004] In known scanning devices, scan results are typically generated using X-ray technology and then displayed to the operator on a screen. It is also known that such scanning devices employ analysis software for the detection of alarm objects, capable of distinguishing alarm objects from harmless objects within the luggage based on the scan results. The alarm objects are identified with a definable accuracy or sensitivity and then displayed to the operator on the screen. Thus, when checking luggage with a scanning device, a distinction can be made between luggage containing only harmless objects and luggage containing at least one alarm object.Such luggage, also known as alarm luggage, is then subjected to a manual and time-consuming inspection, for example by being opened and searched in the presence of the owner of the luggage.

[0005] As the preceding explanation makes clear, the time required for manually examining an alarmed piece of luggage is significantly greater than the time required for luggage analyzed as harmless. Therefore, one goal of known detection methods is to keep the false alarm rate as low as possible. However, contrary to this first goal, the aim is to achieve the highest possible level of certainty in the detection of alarm objects, depending on the application.

[0006] A disadvantage of the known solutions is that the alarm rate level of the detection methods is fixed. This means that, for example, in less security-sensitive situations, the same very high level of accuracy must still be maintained, so that despite the lower security relevance, such as at an event, correspondingly high alarm rates and thus a high manual monitoring effort can be expected.

[0007] The object of the present invention is to at least partially overcome the disadvantages described above. In particular, the object of the present invention is to allow a more flexible use of a scanning device in a cost-effective and simple manner.

[0008] The foregoing problem is solved by a recognition method with the features of claim 1, a recognition device with the features of claim 12, and a computer program product with the features of claim 13. Further features and details of the invention will become apparent from the dependent claims, the description, and the drawings. Features and details described in connection with the recognition method according to the invention naturally also apply in connection with the recognition device and the computer program product according to the invention, and vice versa, so that the disclosure of the individual aspects of the invention always makes, or can make, reciprocal references. According to the invention, a recognition method serves for the variable detection of alarm objects in the scan results of a luggage scanning device. Such a recognition method comprises the following steps:

[0009] - Capturing a scan result from the scanning device,

[0010] - Analyzing the scan results for the presence of alarm objects,

[0011] - Determining a probability of detection for at least one detected alarm object, that it is indeed an alarm object,

[0012] - Specifying a variable probability threshold,

[0013] - Comparison of the specific detection probability for the at least one detected alarm object with the specified variable probability threshold,

[0014] - Output of an alarm signal if the detection probability for at least one alarm object exceeds the probability limit.

[0015] A recognition method according to the invention can, for example, be implemented in a software-based scanning device as a supplement to the analysis method. It is based on the fact that the scanning device, for example using an X-ray device, can scan a large number of pieces of luggage one after the other and thus generate scan results for each piece of luggage. The scan results can, in particular, be image-based scan results.

[0016] The detection method according to the invention captures the scan results generated by the scanning device and analyzes them for the presence of alarm objects. Known analysis methods can be used for this analysis, which are based, for example, on the use of trained neural networks, i.e., so-called artificial intelligence, or on classical algorithmic analysis methods. The result of this step is the detection of alarm objects within luggage. It should also be noted that, within the scope of the present invention, a piece of luggage can have a variety of different configurations. It can be hand luggage, such as backpacks or handbags. However, other forms of packaged objects, such as parcels, envelopes, or the like, are also to be understood as luggage and thus generally as object containers within the scope of the present invention.

[0017] Once at least one alarm object has been detected in the analysis step using a detection method according to the invention, the detection probability for this at least one detected alarm object is then determined according to the invention. This can also be referred to as the detection sensitivity for the respective alarm object. Depending on the analysis method used, this allows not only for a qualitative distinction between an alarm object and a harmless object, but also for a quantitative statement regarding the probability that this alarm object is indeed an alarm object. This quantitative analysis can, for example, include a multitude of different parameters, such as a material analysis, a shape analysis, a size analysis, a position analysis within the piece of luggage, or similar parameters. It makes it possible to determine the degree of detection probability for the detected alarm object.A very high detection probability therefore indicates a more likely actual alarm device than a lower detection probability. Determining the detection probability can be integrated into the analysis process.

[0018] One of the key concepts of the invention is based on the integration of an additional verification step into the recognition process for this specific recognition probability. This additional verification step is formed by comparing the specific recognition probability with a predetermined probability threshold. However, the predetermined probability threshold is not a fixed value, but rather a variable probability threshold. This means that the probability threshold can be varied within the recognition process.Thus, it is possible that with the same piece of luggage containing the same detected alarm object, the detection probability, which results from the alarm object, the analysis method used, and the design of the scanning device, may yield different comparison results depending on the variation of the probability threshold.

[0019] In the final step of the detection process, an alarm signal is now issued not only under the condition that the analysis process has detected an alarm object, but also that the detection probability is above the variable probability threshold.

[0020] As can be seen from this now twofold condition, a detection method according to the invention will no longer trigger an alarm signal for every analyzed and detected alarm object. Rather, the alarm signal is only issued if the second condition, namely exceeding the variable probability threshold, is also met. In other words, alarm objects with a low detection probability, which lies below the variable probability threshold, are now ignored and no longer trigger an alarm signal within the detection method. Thus, with low detection probabilities, the piece of luggage is classified as harmless despite a positive analysis result for the presence of an alarm object and is no longer subjected to manual inspection when passing through a security checkpoint.

[0021] However, for one and the same piece of luggage, lowering the variable probability threshold can lead to a different comparison result for the same piece of luggage with the same alarm object and the same probability of detection, compared to a varied and lowered probability threshold. Therefore, with a different, downwardly varied probability threshold, the alarm signal is now issued because both conditions are met.

[0022] As explained above, with an identical scanning device, identical piece of luggage, identical alarm object, and, most importantly, identical analysis method, the sensitivity for issuing the alarm signal can be varied by adjusting the probability threshold. In other words, it becomes possible to deliberately and consciously influence the alarm rate, as explained later, by adjusting the probability threshold. If the probability threshold is raised, fewer detected alarm objects will have a detection probability exceeding this higher threshold, and the alarm signal rate will decrease accordingly.With this setting, the probability of alarm signals being triggered is reduced, but the throughput of such a scanning device is increased, since manual verification requires less effort with fewer alarm signals being triggered. Conversely, if, for example, a higher capacity is made available for manual verification, safety can be increased by lowering the variable probability threshold accordingly. Later, it will be explained how this basic variation can be further developed into a quantitative adjustment.

[0023] An inventive method now makes it possible to selectively vary and influence the alarm rate, and thus the efficiency and safety of the scanning device's operation, without interfering with the analysis method and / or the mechanical and structural design of the scanning device. This allows for adaptation to different safety and operational situations without structural modifications and, above all, without software changes to the analysis method or the analysis algorithm.

[0024] Advantages can arise if, in a detection method according to the invention, the scan result is captured in the form of a scan image. In such an embodiment, the scanning device is equipped as an imaging device, for example, based on X-ray technology. In such an embodiment, the alarm signal can be designed not only as a qualitative alarm signal but also as part of the scan image itself. For example, the alarm signal can take the form of a colored marking, a border, or a similar graphic highlight in the scan result.

[0025] Further advantages arise if, in a recognition method according to the invention, a distinction is made between at least two different types of alarm objects when analyzing the scan result, wherein, in particular, a separate variable alarm object type probability limit is specified for each type of alarm object.

[0026] While scanning devices generally improve security by searching for and detecting alarm objects, even more precise security distinctions can be achieved with different analysis methods. In particular, it is possible to differentiate between various types of alarm objects. These could include weapons, money, batteries, explosives, or similar items. It is therefore possible to define different alarm object categories for different types of alarm objects. Furthermore, different alarm object categories may be more or less important depending on the operational situation. For example, if a scanning device is used at an airport, it is crucial to detect batteries, which are either not permitted in checked baggage or are only allowed in very small sizes.However, if a scanning device is used, for example, at the entrance of an event venue, the detection of batteries is essentially irrelevant. In such a situation, the detection of weapons or explosives is significantly more relevant to security. By differentiating between different types of alarm objects, it now becomes possible to identify the individual alarm objects using one or even different analysis methods for each type, and to determine correspondingly different detection probabilities for each type of alarm object. With this further development of the present invention, variable and specific alarm object probability thresholds can now also be used for each type of alarm object.In the previously mentioned example of an event venue, the alarm object type probability threshold can now be significantly reduced for alarm object types irrelevant to this situation, such as money and batteries. In other words, alarm objects of the types "money" and "batteries" identified by analysis methods are indeed recognized and assigned a detection probability, but the comparison according to the invention with the significantly reduced alarm object type probability threshold for these alarm object types no longer results in an alarm signal being issued. It is therefore possible to include and / or exclude the alarm object types relevant to the respective application for each operational situation, particularly different event situations, and thus to very precisely reflect the security requirements.

[0027] Advantages also arise if, in a detection method according to the invention, the number of alarm signals relative to the analyzed scan results is determined and output as an alarm rate over the course of the detection process for a large number of scan results. The alarm rate is thus a non-specific correlation between all alarm signals and all scan results. For example, if 10 out of 100 scan results are marked with an alarm signal, this corresponds to an alarm rate of 10%. It is irrelevant whether these 10% of alarm signals actually reveal an alarm object during manual inspection or whether they are false alarms. However, the alarm rate correlates with the effort associated with issuing an alarm signal and the associated manual inspection of a relevant piece of luggage.The higher the alarm rate, the greater the effort required to process a manual baggage check, and the lower the corresponding throughput during scanning. In other words, since the alarm rate is based on past analysis and alarm signal output using this detection method on this scanner, it can predict the alarm rate to be expected at this precise setting between the probability threshold and the number of alarmed items.

[0028] It can be advantageous if, in a detection method according to the preceding paragraph, a varied alarm rate is determined when the probability threshold is varied, wherein stored, specific detection probabilities are compared with the varied probability threshold, and a varied number of alarm signals relative to all analyzed scan results is determined and output as the varied alarm rate. In other words, it is useful if, in a method according to the invention, not only the number of output alarm signals is stored over the course of a large number of scan results, but also the detection probabilities for the alarm objects are stored for all runs. This allows these stored detection probabilities from the past to be accessed at a later time point in time.It thus becomes possible to apply a varied probability threshold to past comparison results during the variation process. Consequently, reducing the probability threshold leads to an increasing varied alarm rate for previously determined detection probabilities, and vice versa. This allows an operator of a scanning device to obtain a forecast based on past scan results when varying the probability threshold. This forecast, in the form of a varied alarm rate, provides an indication of the alarm rate that the applied variation of the probability threshold will result in. This can be implemented, in particular, with a graphical display and / or a slider on a display device of the scanning device.The operator can reduce the probability threshold, for example using the slider, to ensure greater security. This increased security comes with a higher number of alarm signals, as the detection probability will more frequently exceed such a reduced probability threshold, even with the same probability of detection. Therefore, with the same number of analyzed scan results, a greater number of alarm signals can be expected, resulting in an increased variable alarm rate displayed next to the slider on the screen. Conversely, if the probability threshold is raised to reduce security, a correspondingly reduced variable alarm rate can be expected.This makes it possible not only to adjust the safety relevance by varying the probability threshold, but also to represent the effect of such a variation in safety in the form of a varied alarm rate. In other words, operators can now obtain information about the correlation between the variation in safety and the associated change in the alarm rate.

[0029] Advantages also arise if, in a detection method according to the invention, a distinction is made between at least two different alarm object type probability thresholds for different alarm object types when varying the probability threshold, wherein, in particular, a varied alarm object type alarm rate is determined and output for each varied alarm object probability threshold. This means that the prediction of a varied alarm rate can now be output not only absolutely for all alarm object types, but the varied alarm rate can be determined and output specifically for each alarm object type. Particularly for different types of events, this allows for a very specific adaptation of the security requirements.For example, at border crossings or embassy entrances, security can be set to be high for detecting money, weapons, and explosives, while security regarding batteries can be represented by correspondingly high and low specific probability thresholds for alarm object types. At large events, it may be more important to maintain high sensitivity solely for weapons and explosives, while batteries and money are less relevant. This makes it possible to ensure the highest possible level of security tailored to the specific application and to lower the probability thresholds for alarm object types of lower security relevance to such an extent that the alarm rate for unnecessary or uninteresting objects is correspondingly reduced.Preferably, a summation function can be provided to determine and output not only the varied alarm rates for the individual alarm object types, but also the overall alarm rate.

[0030] Furthermore, it can be advantageous if, in a detection method according to the invention, an overall probability limit is varied when the probability limit is varied, while maintaining the ratios of the alarm object type probability limits. This can be understood to mean that, depending on the application situation, the individual probability limits for the different alarm object types are now set, whereby their relative settings to each other are maintained when the overall probability limit is raised or lowered. In other words, the overall sensitivity can be raised and / or lowered while maintaining the safety characteristics of the intended application. Naturally, the relative values ​​for the individual alarm object type probability limits can also be stored and retrievable as presets.Further advantages can be achieved if, in addition to the alarm rate, at least one of the following parameters is determined and output in a detection method according to the invention:

[0031] - Expected number of scans per unit of time,

[0032] - Expected alarm effort.

[0033] The preceding list is not exhaustive. While the alarm rate is fundamentally a parameter that must be interpreted by the scanner operator, additional information such as scans per unit of time or alarm frequency can facilitate interpretation. For example, the average time required for manually checking an alarmed piece of luggage can be stored in the detection process. Based on the alarm rate and the number of possible scans with the scanner, the expected alarm frequency can then be predicted and compared, for example, with the number of operators at the respective scanner. This allows for prediction of whether, given the specific probability threshold setting, the number of operators is sufficient for manually checking alarmed luggage.

[0034] Further advantages arise when, in a recognition method according to the invention, the multitude of scan results originate from a single, specific scanning device. This allows the individual scanning device to permit internal variation of the recognition process or internal prediction. This leads to a very fast and, above all, automatically specific adaptation and variation of the probability thresholds for this scanning device and this application.

[0035] Additionally or alternatively, it is possible that in a detection method according to the invention, the multitude of scan results originates from at least two different scanning devices. These different scanning devices can communicate with each other, for example via wired or wireless connections, to exchange the necessary information. For example, at the entrance of an event venue or at a facility, such as a stadium, a multitude of scanning devices for similar purposes can be arranged to protect such an event, and thus together form a similar database for predicting an alarm rate.

[0036] It is also advantageous if, in a detection method according to the invention, an automatic variation of the probability limit value is carried out, based on a continuous determination of the alarm rate and a change of the probability limit value to comply with a target value for the alarm rate.

[0037] While the core concept of the invention was primarily based on a prediction for manual variation, this further development allows the probability threshold to be varied not only manually, but also automatically or semi-automatically. This ensures, for example, that the probability threshold adapts to a changing environment. For instance, an alarm rate can be set at a security gate as the maximum alarm rate for the gate's capacity. If the composition of the luggage changes such that more potential alarm objects are analyzed and detected with the same number of scans, this would lead to an increased manual verification effort.In this embodiment, this increase in the alarm rate would deviate from the predefined alarm rate value, so that an automatic variation in the form of raising the probability threshold could lead to a constant or even reduced alarm rate. The alarm rate can also be configured as a predefined range and, in particular, include maximum limits that must not be exceeded or fallen below during an automatic variation of the probability threshold.

[0038] A further aspect of the present invention is a detection device for carrying out a detection method for the variable detection of alarm objects in the scan results of a luggage scanning device. Such a detection device comprises a capture module for capturing a scan result from the scanning device. Furthermore, an analysis module is provided for analyzing the scan result for the presence of alarm objects. A determination module is used to determine the probability of detection for the at least one detected alarm object, indicating that it is indeed an alarm object. A preset module is used to specify a variable probability threshold. Finally, a comparison module is provided for comparing the determined probability of detection for the at least one detected alarm object with the probability threshold.An alarm signal is output via an output module if the detection probability for the at least one alarm object exceeds the probability threshold. The detection module, the analysis module, the determination module, the input module, the comparison module, and / or the output module are preferably configured for carrying out a detection method according to the invention. Thus, a detection device according to the invention offers the same advantages as those explained in detail with reference to a detection method according to the invention.

[0039] Another object of the present invention is a computer program product comprising instructions which, when the program is executed on a computer, cause the computer to perform a recognition method according to the invention. Thus, a computer program product according to the invention also offers the same advantages as those explained in detail with reference to a recognition method according to the invention.

[0040] Further advantages, features, and details of the invention will become apparent from the following description, in which exemplary embodiments of the invention are described in detail with reference to the drawings. The features mentioned in the claims and in the description can each be essential to the invention individually or in any combination. The drawings schematically show:

[0041] Fig. 1 shows an embodiment of a recognition device according to the invention during the execution of a recognition method according to the invention.

[0042] Fig. 2 shows an alternative run of a recognition method according to the invention,

[0043] Fig. 3 shows an alternative run of a recognition method according to the invention, Fig. 4 shows an alternative run of a recognition method according to the invention,

[0044] Fig. 5 shows an alternative run of a recognition method according to the invention,

[0045] Fig. 6 shows a possible representation of the variation of probability limits.

[0046] Figure 1 schematically depicts the situation at a security checkpoint, for example, at the entrance to an event venue. Luggage items G, such as backpacks, packages, handbags, or similar items, are X-rayed using a scanning device 100, and a scan result SE is generated via an imaging process. The respective scan result SE is schematically represented here as an X-ray image with a multitude of individual, different objects. The recognition method according to the invention now takes place in the recognition device 10. Within this recognition device 10, the scan result SE, which is designed as a scan image, is captured using the acquisition module 20. Using the analysis module 30, one or more alarm objects AO can then be detected in the scan result SE with the aid of an analysis algorithm or an analysis Kl (artificial intelligence).The embodiment shown in Figure 1 is a detected alarm object AO, which is depicted here as a circular cross-hatched area.

[0047] While previously known solutions would have directly triggered the output of an alarm signal AS based on the detected alarm object AO, the detection device 10 according to the invention checks a second condition. For this second condition, the sensitivity, and thus the detection accuracy, in the form of the detection probability EW for the detected alarm object AO is determined using the determination module 40. Additionally, a variable probability threshold WG is specified using a preset module 50, allowing the two values, the detection probability EW and the probability threshold WG, to be compared in the comparison module 60. In the example shown in Figure 1, the detection probability EW is above the specified variable probability threshold WG, so the output module 70 outputs an alarm signal AS on a display device 110.The alarm signal AS is provided here in the form of a coloring and marking within the scan image of the scan result SE.

[0048] During the processing of the recognition procedure according to Figure 1 across a large number of scan results SE, some of the luggage items G will trigger an alarm signal AS. By comparing the number of all alarm signals AS with all scan results SE, an alarm rate AR can be determined as a ratio and is displayed here. In the example of Figure 1, the alarm rate across the past scan results SE is 14%. This means that, in the past 1000 checks of scan results SE, 140 scan results SE resulted in an alarm object AO in analysis module 30 with a detection probability EW above the predefined variable probability threshold WG.

[0049] Figure 2 shows the embodiment of Figure 1, but with a variation of the probability threshold WG. The scanning result SE process, and in particular the hardware and the detection performed with the scanning device 100, is identical to Figure 1. The analysis procedure with a neural network, a computer, or an analysis algorithm in the analysis module 30 also remains unchanged, so that the circular alarm object AO is again detected with the same piece of luggage G as in Figure 1. The determined value for the detection probability EW for this detected alarm object AO is also identical, except that in this embodiment of the detection method, the variable probability threshold WG has been increased compared to Figure 1 to such an extent that the detection probability EW is now below this modified probability threshold WG.This results in the output module 70 no longer issuing an alarm signal AS in this specific situation, unlike in Figure 1, and the operator of the scanning device 100 recognizes an alarm-free and therefore harmless piece of luggage G on the display device 110. Since raising the probability threshold WG now results in fewer alarm objects AO triggering an alarm signal AS, a correspondingly lower alarm rate AR can be expected, which is determined and specified here, for example, as 10%. This clearly demonstrates that a change in the alarm rate AR can be achieved simply by varying the probability threshold WG without interfering with the analysis procedure in the analysis module 30 and / or the hardware of the scanning device 100.

[0050] Figure 3 defines different types of alarm objects. By analyzing the scan result SE, it is now possible to identify different types of alarm objects AO using a higher-level or different specific analysis methods. These two different alarm objects AO, representing different alarm object types, are indicated in Figure 3 by a circle and a square with differently sized crosshatching. Here, too, the detection probability EW is determined in the determination module 40 for all detected alarm objects AO, and these detection probabilities EW are then compared with the probability limit WG from the specification module 50.

[0051] In this embodiment of Figure 3, only the detection probability EW of one of the two alarm objects AO exceeds the predefined variable probability limit WG, so the alarm signal AS is only issued for this alarm object AO. Thus, an alarm rate AR of, for example, 14% can again be determined and output.

[0052] Figure 4 further specifies different alarm object types. Here again, two different alarm objects (AO) are detected and assigned detection probabilities (EW). However, instead of a common, overarching probability threshold (WG), there are now different variable alarm object type probability thresholds (AWG) for each alarm object type, which are each compared with the specific detection probability (EW). When setting and varying the alarm object type probability thresholds (AWG) in the predefined module 50, a varied alarm rate (VAR) is displayed, thus providing a forecast of how the alarm rate (AR) will behave with this setting of the alarm object type probability threshold (AWG).In Figure 4, the output remains unchanged compared to Figure 3, as both detection probabilities (AWG) maintain the same correlation with their respective alarm object type probability thresholds (AWG). In Figure 5, the operator aims to increase the security for an alarm object type by lowering its corresponding alarm object type probability threshold (AWG). The resulting varied alarm rate (VAR) therefore increases to 19%, which is reflected in the final output on display 110 if this setting is maintained. However, not only the alarm rate (AR) changes, but also the display itself, as the second alarm object (AO) is now also shown with an alarm signal (AS).

[0053] Figure 6 schematically illustrates an implementation in which operators can use digital sliders to adjust specific alarm object type probability thresholds (AWG) for different alarm object types. The varied alarm rates (VAR) are also shown, increasing or decreasing depending on the adjustment. High security levels, with sliders on the right, result in a higher trigger sensitivity and thus a higher alarm rate (AR), while lower trigger sensitivities, with sliders further to the left, result in a higher alarm rate (AR). An overall probability threshold (GWG) is also shown, along with the sum of all varied alarm rates (VAR), allowing for the output of a varied overall alarm rate.

[0054] The preceding explanation of the embodiments describes the present invention solely by way of examples. Naturally, individual features of the embodiments can be freely combined with one another, provided this is technically feasible, without departing from the scope of the present invention.

[0055] Reference symbol list

[0056] 10 Recognition device

[0057] 20 Data acquisition module

[0058] 30 Analysis Module

[0059] 40 Determination module

[0060] 50 Default module

[0061] 60 Comparison module

[0062] 70 Output module

[0063] 100 scanning device

[0064] 110 Display device

[0065] G piece of luggage

[0066] SE Scan result

[0067] AO Alarm Object

[0068] AS alarm signal

[0069] AR Alarm rate

[0070] VAR varied alarm rate

[0071] EW detection probability

[0072] WG probability limit

[0073] GWG Total Probability Threshold

[0074] AWG Alarm Object Type Probability Threshold

Claims

Patent claims 1. Detection method for variable detection of alarm objects (AO) in scan results (SE) of a scanning device (100) for luggage (G), comprising the following steps: - Capturing a scan result (SE) of the scanning device (100), - Analyzing the scan result (SE) for the presence of alarm objects (AO), - Determining a probability of recognition (PR) for the at least one detected alarm object (AO), that it is indeed an alarm object (AO), - Specifying a variable probability threshold (PV), - Comparison of the specific detection probability (DP) for the at least one detected alarm object (AO) with the specified variable probability limit (VF), - Output of an alarm signal (AS) if the detection probability (EW) for at least one alarm object (AO) exceeds the probability limit (WG).

2. Recognition method according to claim 1, characterized in that the scan result (SE) is captured in the form of a scan image.

3. Recognition method according to one of the preceding claims, characterized in that, during the analysis of the scan result (SE), a distinction is made between at least two different types of alarm objects (AO), wherein, in particular, a separate variable alarm object type probability threshold (AGW) is specified for each type of alarm object.

4. Recognition method according to one of the preceding claims, characterized in that, over the course of the recognition method for a plurality of scan results (SE), the number of alarm signals (AS) in relation to all analyzed scan results (SE) is determined and output as alarm rate (AR).

5. Recognition method according to claim 4, characterized in that a varied alarm rate (VAR) is determined when the probability limit (WG) is varied, wherein stored specific recognition probabilities (EW) are compared with the varied probability limit (GW) and a varied number of alarm signals (AS) in relation to all analyzed scan results (SE) is determined and output as a varied alarm rate (VAR).

6. Detection method according to claim 5, characterized in that when varying the probability threshold (WG) a distinction is made between at least two different alarm object type probability thresholds (AWG) for different alarm object types, wherein in particular a varied alarm rate (VAR) is determined and output for each varied alarm object type probability threshold (AWG).

7. Detection method according to claim 6, characterized in that when varying the probability limit (WG) an overall probability limit (GWG) is varied, whereby the ratios of the alarm object type probability limits (AWG) are maintained.

8. Detection method according to one of claims 4 to 7, characterized in that, in addition to the alarm rate (AR), at least one of the following parameters is determined and output: - Expected number of scans per unit of time Expected alarm effort 9. Recognition method according to one of claims 4 to 8, characterized in that the plurality of scan results (SE) originates from a single specific scanning device (100).

10. Recognition method according to one of claims 4 to 8, characterized in that the plurality of scan results (SE) originates from at least two different scanning devices (100).

11. Detection method according to one of the preceding claims, characterized in that an automatic variation of the probability limit (WG) is carried out, based on a continuous determination of the alarm rate (AR) and a change of the probability limit (WG) to comply with a target value for the alarm rate (AR).

12. Recognition device (10) for carrying out a recognition procedure for variable detection of alarm objects (AO) in scan results (SE) of a scanning device (100) for luggage (G), comprising a capture module (20) for capturing a scan result (SE) of the scanning device (100), an analysis module (30) for analyzing the scan result (SE) for the presence of alarm objects (AO), a determination module (40) for determining a detection probability (DP) for the at least one detected alarm object (AO) that it is indeed an alarm object (AO), a specification module (50) for specifying a variable probability limit (VL), a comparison module (60) for comparing the determined detection probability (DP) for the at least one detected alarm object (AO) with the probability limit (VL), and an output module (70) for outputting an alarm signal (AS).if the detection probability (DP) for the at least one alarm object (AO) exceeds the probability limit (VL), wherein the detection module (20), the analysis module (30), the determination module (40), the input module (50), the comparison module (60) and / or the output module (70) are configured for carrying out a detection method with the features of one of claims 1 to 11.

13. Computer program product comprising instructions which, when the program is executed on a computer, cause the computer to perform a recognition method having the features of any one of claims 1 to 11.

Images