Segmentation method for delineating object as segment in luggage item during x-ray diffraction analysis

Abstract

The invention relates to a segmentation method for delimiting an object (O) in a package (G) into segments (S) during an X-ray diffraction analysis, comprising the following steps:-detecting a pulse transmission function (20) for a plurality of voxels (30) of the package (G),-comparing the detected pulse transmission function (20) with at least one predetermined material function (40), the invention relates to a method for comparing voxels (33) in a spatial direction, and generating similarity values (50) as a result of the comparison,-selecting at least one voxel (30) as a starting voxel (32) on the basis of the similarity values (50),-generating at least one voxel cluster (34) starting from the at least one selected starting voxel (32) using a combination method having the following steps: e) combining the starting voxel (32) with adjacent voxels (33) in all spatial directions, f) for each of these combinations, determining a similarity value (50) with the at least one material function (40), g) adding to the voxel clusters (34) those adjacent voxels (33) resulting in the improvement of the similarity value (50), h) repeating steps a) to c) until a termination criterion (60),-combining overlapping voxel clusters (34) into at least one segment (S), an increase in the similarity value (50) is taken into account as a result of the combination.

Description

[0001] The invention relates to a segmentation method for delimiting an object into segments in a baggage item during X-ray diffraction analysis, and to a computer program product for performing such a segmentation method.

[0002] It is known that in security-related environments, such as at airports, items of luggage should be checked for the presence of unauthorized or dangerous objects. These may be, for example, illegal materials, such as drugs, cash or similar objects. It is also known that dangerous objects, such as objects made of explosive materials or containing such materials, should be reliably detected. In order to solve this problem, so-called transmission scanners are used, for example at airports, for checking luggage, which can perform a transmission scan of the items of luggage by means of X-ray transmission or as computed tomography (CT).

[0003] However, with these transmission methods, information about the material being detected can only be extracted to a very limited extent. Instead, the transmission scan essentially involves a check on the relevant form or forms defined as dangerous. Material information can only be obtained in a relatively imprecise manner with these transmission methods, for example by indicating the material density or the effective nuclear charge.

[0004] In order to ensure that, in the case of an object being classified as an alarm object by a transmission scan, it does not have to be unnecessarily separated from the baggage handling system for a time-consuming manual check, additional scanning equipment is usually provided in a two-stage process at airports. One possibility for this involves a diffraction scanner that can perform an X-ray diffraction analysis. In this case, the transmission is no longer determined, but the diffracted X-ray radiation is determined, so that the specific material of the individual objects in the baggage items can be inferred from this. The disadvantage of X-ray diffraction analysis is that, although it can provide material-specific information with very high accuracy, this is only possible if the scanning process is performed relatively slowly in order to have a large number of photons available for the diffraction analysis. However, this contradicts the need to check a high throughput of baggage in security-related areas (such as airports). Therefore, during the X-ray diffraction analysis, known solutions attempt to compromise between the fastest possible scanning and the most accurate possible material-specific evaluation. The faster the scanning is performed during the X-ray diffraction analysis, the fewer photons are contained in the individual voxels. As a result, the information value at the level of the individual voxels is usually reduced to such an extent that a reasonable segmentation is no longer possible. Therefore, in some cases, an attempt is made to extract a 3D data set from the 4D data, so that in principle an evaluation can be performed, but the transformation results in a loss of data in the 4D data set. In the classification after the segmentation, this can lead to poorer results and therefore to an increased false alarm rate. It is therefore an object of the present invention to at least partially remedy the above-mentioned disadvantages. In particular, it is an object of the present invention to obtain the most accurate possible delineation of objects in a baggage item in a cost-effective and simple manner.

[0005] The above objects are achieved by a segmentation method having the features of claim 1 and a computer program product having the features of claim 14. Further features and details of the invention are disclosed in the dependent claims, the description and the drawings. Of course, features and details described in conjunction with the segmentation method according to the invention also apply in conjunction with the computer program product according to the invention and vice versa, so that with respect to the disclosure, reference is made to each other or may always be made to various aspects of the invention.

[0006] According to the present invention, a segmentation method for delimiting an object into segments in a baggage item during X-ray diffraction analysis is proposed, comprising the following steps:

[0007] - recording the pulse transmission function for a large number of voxels of the luggage item,

[0008] - comparing the recorded pulse transmission function with at least one specified material function and generating a similarity value as a comparison result,

[0009] - selecting at least one voxel as a starting voxel based on the similarity value,

[0010] - creating at least one voxel cluster starting from said at least one selected starting voxel using a combined method comprising the following steps:

[0011] a) Combine the starting voxel with its neighboring voxels in all spatial directions,

[0012] b) determining for each of these combinations a similarity value to the at least one material function,

[0013] c) adding those neighboring voxels that produce an improved and / or preserved similarity value to the voxel cluster,

[0014] d) Repeat steps a) to c) until the termination criteria are met,

[0015] - combining overlapping voxel clusters into at least one segment, the increase in the similarity value being taken into account by the combination.

[0016] The core idea of ​​the invention is based on utilizing all data, in particular the data in the 4-dimensional data set, for segmentation and, despite this, performing the segmentation in a material-specific manner with high precision, even with a small number of photons. This is possible because, in contrast to known solutions, no reduction to a 3-dimensional data set occurs; instead, the segmentation is directly based on the data obtained from the diffraction analysis. In order to compensate for the lack of photons in individual voxels, a targeted combination of individual voxels is performed, so that the voxel clusters created in this way are based on a larger number of photons than individual voxels and thus also allow a more precise segmentation. At the same time, these steps preserve the information content of the 4-dimensional data. The arrangement according to the invention is based on the core idea that the pulse transmission function is specific to a specific material. The pulse transmission function is preferably recorded individually for all individual spatial volumes of the luggage item, i.e., preferably for all voxels of the luggage. The resolution of the voxels can be selected differently as required. The larger the selected size of the respective voxel, the higher the photon count of this larger voxel at the same scanning speed, but the less precise the resolution in the subsequent segmentation.

[0017] Preferably, when recording the pulse transmission function, this step is performed for a large number of identical or substantially identical voxels. This should be understood to mean that all voxels have the same or substantially the same form and / or size. At the end of this first step, there are now pulse transmission functions for an arbitrary number of voxels. For example, in the case of 1000 voxels, there are also 1000 pulse transfer functions, which can now be compared with at least one specific material function in the second step of the method according to the invention.

[0018] In the sense of the present invention, a material function is a pulse transmission function that is specific to a particular material. For example, for a large number of different materials or material components, a corresponding large number of material functions that are specific to these components can be stored in a database. Then, a suitable material function for the comparison step can be selected from this database. For example, a search for a specific explosive can be performed so that for such a specific explosive material, a corresponding material function of the specific explosive material is selected from such a material library and the corresponding material function is specified for comparison. In the comparison step, the pulse transmission function of all voxels is now checked for similarity with the specified material function. This is preferably a mathematical comparison, so that a similarity value between each pulse transmission function and a specific material function is generated as a comparison result in a qualitative manner, but also in a quantitative manner. In particular, vectors of pulse transmission functions and material functions are compared with each other. The result of such a vector comparison can be, for example, a scalar value and can be compared in particular with a limit value. The comparison result can indicate a relationship with such a limit value qualitatively or quantitatively. For example, a qualitative similarity value can assume a value of "1" for "similar" and "0" for "dissimilar". The quantitative formation of the similarity value can assume a degree of similarity, for example, between a maximum value of "same" and a minimum value of "completely different". In other words, the comparison step requires a comparison result which provides an indication for each individual voxel for which a pulse transmission function has been recorded whether and / or to what extent a similarity exists with at least one specified material function. In other words, it can indicate in which voxels the material being searched for is present with a certain probability and in which voxels this is not the case.

[0019] However, due to the preference for fast scanning, which is usually stipulated as a technical requirement, this comparison step is still subject to a high degree of inaccuracy, since due to the fast scanning process during the X-ray diffraction analysis, only a very small number of photons can be recorded for each individual voxel. In order to avoid false alarms and at the same time minimize the undesired non-identification of relevant materials, the segmentation method according to the invention further consists in selecting all voxels that have a certain similarity to the material function as starting voxels. This can be a purely qualitative selection as well as a quantitative selection. Preferably, as will be explained later, each similarity value for each individual voxel is a single scalar value, so that a very simple distinction can be made between similarity and dissimilarity to a specified material function.

[0020] The distinction between similar and dissimilar voxels and the selection of the starting voxel in this way now leads to a combined method incorporated into the segmentation method, which can also be referred to as a region growing method. The term region growing method is used here because starting from each voxel defined as the starting voxel, growth and combination with other neighboring voxels can occur, as will be explained below.

[0021] Therefore, each starting voxel is selected in this way because it meets a certain minimum similarity defined by the comparison result and therefore by the similarity value with respect to the material function. Now, a check is performed on each starting voxel to determine whether it can be combined with an adjacent voxel. In other words, the starting voxel is first combined with the first adjacent voxel, thereby doubling the volume of the combination to be checked. Of course, this also leads to an increase in the number of photons to be checked, which forms the basis of this step. Therefore, the combination of the starting voxel and the adjacent voxel produces a pulse transmission function specific to the combination, for which a similarity value can also be recorded according to the above-mentioned comparison step. This similarity value again allows to indicate whether the combination shows a similarity with the material function (qualitative statement), or even quantitatively, whether the similarity is increased, remains the same or decreased by the combination. Therefore, in a three-dimensional inspection of a spatial volume, starting from the starting voxel, there are up to 26 adjacent voxels, all of which are individually checked with respect to this combination, so that subsequently in step c) of the combination method, those adjacent voxels are combined with the starting voxel to form a voxel cluster that matches the specified material function due to the determined similarity value. Neighboring voxels that do not match the material function are excluded and are therefore defined as not belonging to the voxel cluster. These combination steps with neighboring voxels and the further growth of the resulting voxel cluster are performed until a termination criterion is reached, which will be explained in more detail later. In the simplest case, this termination criterion can be a maximum number of iterations, but it can also include a substantial evaluation of the quality of the growth and / or the similarity values. It should be pointed out at this point that, of course, in each case the resulting voxel cluster undergoes steps a), b) and c) with respect to all neighboring voxels, so that for each iterative run of the combination method, the number of neighboring voxels used for the combination and for determining the similarity values ​​also increases.

[0022] As a result, at least one voxel cluster is generated, in fact usually a large number of separate voxel clusters, which can be different from each other, but can also overlap. A voxel cluster can be interpreted as the connected voxels with the minimum similarity to the material function and thus the most likely part of the object that actually comprises the specific material associated with the specified material function. This has the great advantage that now not only a single voxel but a specifically selected combination of a plurality of voxels as a voxel cluster can be used as the basis for each voxel cluster, which leads to a significantly increased accuracy of the correlation with the specified material function at a significantly increased number of photons and, therefore, a high probability that hazardous materials can be detected and can be distinguished from non-hazardous materials.

[0023] In the final step of the segmentation method according to the invention, a final combination of overlapping voxel clusters into at least one segment is performed. In other words, after the step of growing starting voxels, which can also be referred to as seeds, all voxel clusters based on these seeds and grown thereon are combined with each other as long as they overlap and as long as they are associated with the same material function. In this way, like individual, overlapping jigsaw puzzle pieces, voxel clusters are combined to form a segment that can finally be distinguished and defined as an independent object from other segments or other objects within the luggage item.

[0024] As can be seen from the previous explanations, the method according to the invention is still based on the clear and specific comparability of the pulse transmission function with the corresponding material function specific to the material during the X-ray diffraction analysis. However, the disadvantage of the low number of photons per voxel is reduced in this evaluation method, because in the first part of the method, the assignment of each voxel to the material is performed, and through algorithmic iterations in the form of optimization problems and therefore the generation of voxel clusters, the enlargement of those identified relevant starting voxels occurs, which actually leads to voxel clusters during the combination and therefore during the growth of a single seed voxel. On the other hand, starting voxels whose neighboring voxels do not lead to an increase in the similarity value relative to the material function, or lead to very small voxel clusters, are excluded again, and therefore, this very small voxel cluster is excluded in the final step of combining into segments. This process allows the segmentation to be performed using all the information of the 4-dimensional data, while still obtaining a high-precision segmentation through the combination of the target and the optimization. Therefore, at a given scanning speed, a maximized evaluation is possible during the segmentation process. Since a further reduction in the number of photons can be at least partially compensated by the combination during the segmentation according to the invention, the scanning process can even be further accelerated. This results in more accurate subsequent classification and / or reduction of undesirable false alarms.

[0025] In the segmentation method according to the invention, it may be advantageous if at least one of the following is used as termination criterion for the combined method:

[0026] -Default number of iterations;

[0027] - the similarity value is not improved by steps a) and b);

[0028] - The similarity value is degraded by steps a) and b).

[0029] In the simplest case, the growth occurs for a maximum of, for example, three or four iterations. This is particularly simple and defines the specification of the termination criterion precisely above all in terms of the required computing time. However, in order to increase accuracy, more complex termination criteria can also be used, which take into account in particular the development of the similarity value through the combination. For example, the process can be terminated whenever the similarity value no longer improves, because the added or to-be-processed neighboring voxels will no longer lead to an improvement and therefore no longer lead to an increase in the similarity to the material function. A stronger termination criterion is if the combination will actually lead to a deterioration of the similarity value, i.e. the combination with the neighboring voxels will reduce the similarity value and therefore also reduce the similarity to the specified material function. The maximum number of voxels in a voxel cluster can also be used as a termination criterion and / or exclude all neighboring voxels from the combination. Thus, if no improvement or even deterioration of the similarity value is defined for all neighboring voxels examined for combination, the iteration is terminated, so that no single neighboring voxel is retained for the combination, and therefore, the iteration steps and iteration loops can be defined to terminate here.

[0030] In the segmentation method according to the invention, it is also advantageous if the generated voxel clusters are sorted according to their similarity values ​​before the step of combining them into at least one segment. In other words, the voxel clusters are sorted in such a way that the combination starts with those voxel clusters that appear most promising for the combination. This means that the combination for forming at least one segment can then be performed based on this sorting, i.e. in a first step the voxel clusters with the greatest similarity values ​​are considered, and then, as the similarity values ​​decrease, other voxel clusters are considered. This leads to a faster combination and the final step of the combination can additionally be improved.

[0031] Further advantages can be achieved if, in the segmentation method according to the invention, a group of several adjacent voxels is selected as starting voxel. In addition to individual starting voxels, such a group of starting voxels can also be defined, in particular if the scan of the X-ray diffraction analysis has been performed so quickly that the individual voxels have such a low photon density that a reasonable comparison in order to generate a similarity value is not possible. This step can also be understood to mean that the size of the individual voxels is increased to such an extent that the small initial voxel is enlarged to an enlarged final voxel, which then allows the individual now enlarged starting voxels to be used for performing the method according to the invention as a group of starting voxels according to the invention. This size or grouping is preferably adaptable, in particular when performing the X-ray diffraction analysis, to the size of the voxels and / or the scanning speed.

[0032] Furthermore, advantages are brought about if, in the segmentation method according to the invention, after combining overlapping voxel clusters into one segment, the step is repeated for the remaining voxel clusters. The remaining voxel clusters are voxel clusters that do not overlap with the voxel clusters of the first step of combination but are not connected to them. In other words, these can be other objects that include the same material, which causes them to be combined into voxel clusters. In principle, the size of the voxel clusters can also be taken into account, i.e., whether the combination step should also be performed again for these, or whether they should be discarded as too small. With this addition, two or more objects with the same material function similarity can also be detected. At this point, it should be pointed out again that the method according to the invention can of course also be performed separately for two or more different material functions. In particular, if there is no indication as to which specific substance may be present in the luggage item, a very extensive analysis can be performed here, so that not only two, but even a large number of different material functions are specified, and the method according to the invention can perform the steps for a large number of different specified material functions and thus identify the segmentation of the object.

[0033] This also brings advantages if, in the segmentation method according to the invention, the similarity value is formed as a scalar value. Such a scalar value can, for example, form a single and therefore very simple comparable parameter and can be the result of the vector comparison already described. Thus, in the course of the segmentation method according to the invention, the iterative optimization steps and the comparison steps become clearer and are optimized, in particular, with regard to the required number of calculations.

[0034] Furthermore, it is advantageous if, in the segmentation method according to the invention, similarity values ​​can be used for at least two different material functions. As already explained, these similarity values ​​can be performed sequentially one after the other, so that the evaluation of different materials can be performed simultaneously or sequentially. In particular, this integrates the exclusion of voxels and voxel clusters that have been assigned to materials, so that, in order to avoid double comparisons, only those voxels that are left over from the first segmentation method are passed to the segmentation method according to the invention for the run with the second specified material function. In the segmentation method according to the invention, advantages are also brought about if at least one material function is specified based on the scan results of a previous scanning method. Such a preceding scanning method can in particular be an X-ray transmission method. Basically, information such as the density of a single object can be obtained from such a transmission method, thereby containing basic information indicating which material may be involved. However, this information is not so precise that an alarm can be triggered or excluded. On the contrary, this means that in the case of a system combination (system of systems), all luggage items are scanned using this very fast transmission method. Baggage items that are identified as potentially dangerous because they may contain alarm objects are only sent to the X-ray diffraction analysis, so that the X-ray diffraction analysis only needs to process a portion of all baggage items. In order to make the X-ray diffraction analysis, which usually works much slower than the transmission technology, work more accurately and faster, when performing the segmentation method according to the invention, information about the materials that may be present in the baggage item can be selected for evaluation, so that a material type or even one or more defined specified material functions are assigned based on the first scan result.

[0035] Further advantages result if, in the segmentation method according to the invention, the steps are restricted to self-contained areas of the luggage item. In other words, for example, a partial area of ​​the luggage item can be defined as an area containing the alarm object, so that while the scanning process with X-ray diffraction analysis is carried out completely or at least on the disk-shaped part of the luggage item in which the object is arranged, the analysis is carried out only on a partial area and only on a portion of the voxels in the luggage item.

[0036] Furthermore, it can be advantageous if, in the segmentation method according to the invention, an alarm signal is output and / or an existing alarm signal is eliminated depending on the combination into at least one segment. If the segmentation method is used in a separate system with an X-ray diffraction analysis, an alarm signal can be generated or suppressed depending on whether the object has already been segmented, is sufficiently large and corresponds to a corresponding material function in terms of the similarity value. However, it is preferred if such an X-ray diffraction analysis is used in conjunction with a previously fast-working transmission system and is therefore only used to check luggage items that are identified as potentially dangerous. Thus, by means of the segmentation method according to the invention, an alarm signal output by the first scan of the transmission device is confirmed or eliminated by the X-ray diffraction analysis.

[0037] It is also advantageous if, in the segmentation method according to the invention, segment coordinates of the baggage item in a coordinate system are determined for the at least one segment. In this case, the coordinate system of the baggage is a relative coordinate system, so that the segment and the objects associated with the segment can be precisely defined within the baggage item. This is particularly advantageous if these segment coordinates are to be transferred together with the baggage item to a manual evaluation or are to be transferred for further processing. The coordinate system of the baggage item moves together with the baggage item, so that the objects therein are defined relative to the segment of the baggage item, regardless of the absolute orientation of the baggage item. In addition to further analysis, this also allows superposition with images generated for the same baggage item, for example using different scanning devices, such as, for example, transmission devices. In particular, this also allows correlation with position-specific warning signals for the respective baggage item.

[0038] Further advantages can be achieved if, in the segmentation method according to the invention, at least one transmission segment has already been generated by means of a preceding transmission method carried out on the same luggage item, in particular in the form of a CT method (Computed Tomography method), the transmission segmentation coordinates of the at least one transmission segment being determined in the coordinate system of the luggage item. This means that, so to speak, a transmission segmentation method is carried out, i.e. a segmentation method based exclusively on the data of the transmission method, in order to segment one or more objects within the luggage item. A superposition or even further use and comparison between the transmission segmentation and the segmentation according to the segmentation method according to the invention can then take place. The simplest case is if the transmission segmentation of the segmentation method according to the invention is transferred as a mark or marker to the evaluation of the X-ray diffraction analysis. However, in principle more complex embodiments are also conceivable.

[0039] Furthermore, if, in the segmentation method according to the invention, information (in particular the boundaries of the transmission segmentation) is taken into account when generating at least one voxel cluster and / or during the combination to form at least one segmentation, this brings advantages. This consideration allows additional considerations to be implemented during the combination method, i.e. during the growth of a single voxel cluster and / or during the combination of voxel clusters to form one or more segments, so that the boundaries of the segments identified in the transmission method can, for example, be taken into account as the boundaries of the segments in the X-ray diffraction analysis. In the segmentation, penalty terms and / or reward terms, for example in the cost function, can, for example, allow such considerations. For example, it is mathematically advantageous to follow the segmentation boundaries from the transmission during segmentation by X-ray diffraction analysis. Similarly, violations of the segmentation boundaries from the transmission during segmentation by X-ray diffraction analysis are penalized, whereby the accuracy of the X-ray diffraction analysis can be taken into account in particular.

[0040] The subject of the invention also includes a computer program product comprising instructions which, when the program is run on a computer, cause the computer to perform the steps of the segmentation method according to the invention. The computer program product according to the invention therefore has the same advantages as explained in detail with reference to the segmentation method according to the invention.

[0041] Further advantages, features and details of the invention are explained in the following description, in which exemplary embodiments of the invention are described in detail with reference to the accompanying drawings. The features mentioned in the claims and in the description may in each case be essential to the invention individually or in any combination. Schematically in each case:

[0042] Figure 1 An embodiment of the segmentation method according to the invention on an X-ray diffraction device is shown,

[0043] Figure 2 The steps of the segmentation method according to the present invention are shown,

[0044] Figure 3 Another step of the segmentation method according to the present invention is shown,

[0045] Figure 4 shows a representation of the combined method,

[0046] Figure 5 shows the first iteration step of the combined method,

[0047] Figure 6 Another iterative step of the combined method is shown,

[0048] Figure 7 Another iterative step of the combined method is shown,

[0049] Figure 8 shows the possible sequences in two-dimensional space for the combinatorial method,

[0050] Fig. 9 shows the steps of combining voxel clusters,

[0051] Fig.10 A combination of a transmission device and an X-ray diffraction device is shown.

[0052] Figure 1Schematically, it is shown how an X-ray diffraction device 100 may be constructed. Here, a baggage item G passes through an inclined scanning channel, through which diffracted X-rays may be detected in one or more detectors. Depending on the speed at which the baggage item G passes through the X-ray diffraction device 100, the detectors may record more or fewer photons in the form of one or more voxels 30 per spatial volume in the form of diffraction. In this representation, an object O is schematically represented as an alarm object, which, for example, will be detected by the X-ray diffraction device 100 with a high degree of certainty as an explosive object.

[0053] In order to be able to provide the most accurate evaluation possible by segmentation, in particular by taking advantage of the fast scanning of the X-ray diffraction device 100 and the associated low number of photons per voxel 30, in a first step, e.g. Figure 2 As shown, for all individual voxels 30 (i.e. all individual spatial volume units), a pulse transmission function 20 is recorded, here schematically represented as a spatial cube. This pulse transmission function 20 is specific to each individual voxel 30 and is still relatively inaccurate because it is based on a very small number of photons at a high scanning speed.

[0054] In the next step, if Figure 3 As shown, it is now checked for each of these individual voxels 30 to what extent the pulse transmission function 20 has a sufficient similarity to the material function 40, and if this is the case, they are defined as starting voxels 32. Voxels 30 with insufficient similarity are marked with an "X" and are excluded from further methods since they are sufficiently dissimilar to the specific material function 40. In particular, a similarity value of 50 (to be explained later) is used herein, whereby it is still possible to operate with high inaccuracies due to low photon densities in the first step, for example, a large number of starting voxels 30 can be selected as starting voxels to increase the reliability.

[0055] For each of these starting voxels 32, the combination method is now performed, as follows with reference to Figure 4 , 5 , 6 and 7 are explained in more detail. Figure 4 It is shown how, from a single starting voxel 32, neighboring voxels 33 in each of the six spatial directions are examined to determine to what extent a combination with this neighboring voxel 33 will lead to an improvement or deterioration of the similarity value 50 to the material function 40. This is a regular spatial grid, whereby the procedure according to the invention can of course also be used for irregular spatial divisions.

[0056] Figure 5 The first iteration step is shown. The seed cell (i.e., the starting voxel 32) has a single pulse transmission function 20 that is sufficiently similar to the material function 40 to be selected as the starting voxel 32. Figure 5In a first iteration step, a combination is now performed with each of the six nearest neighboring voxels 33, so that they also form six different possible combinations of new combined pulse transmission functions 20 for subsequent evaluation, all of which have their similarity values ​​50 compared with the material function 40. Figure 5 In the representation of , only one of the four combined pulse transmission functions 20 corresponds to a similarity value 50 that is large enough to be selected for combination with the voxel cluster 34 . Figure 6 and Figure 7 A further step is shown, whereby again only a single adjacent voxel 30 leads to a combined pulse transmission function 20, which in Figure 6 The lower right and Figure 7 This results in the selection and addition of voxel cluster 34 as outlined below.

[0057] Figure 8 Schematically shows the Figure 5 , 6 The result of the combination of steps 1 and 7. For simplicity, Figure 8 is shown two-dimensionally, whereby of course the function can and preferably should be three-dimensional. Starting from a single starting voxel 32 at the top left, four neighboring voxels 33 are examined in two dimensions, whereby only the two neighboring voxels 33 at the top right result in a combined pulse transmission function 20 with a sufficiently high similarity value 50. In the following steps, the enlargement is continued or individual neighboring voxels 33 are excluded from the voxel cluster 34 due to decreasing similarity values ​​50 until the termination criterion 60 is reached. Figure 8 In the embodiment of the present invention, the termination criterion is reached if a voxel cluster 34 is completely surrounded by neighboring voxels 33, which do not further increase the similarity value 50 or even lead to a combination whose combined pulse transmission function 20 would lead to a decrease in the similarity value 50 relative to the material function 40. This means that at least one voxel cluster 34 is already available.

[0058] Fig. 9 It is shown how two separate segments S are formed in this case. In one step, or in a combination step, overlapping voxel clusters 34 are combined with each other to form separate and common segments S. These segments S are shown on the right-hand side. In the case of the lower segments S on the right-hand side in combined form, the correlation with the transmission segment TS can also be seen, so that a restriction can be made on whether the segment S actually found from the X-ray diffraction device 100 is smaller than or equal to the corresponding transmission segment TS found in the transmission device 200 or is reduced in size.

[0059] Fig.10Once again, a schematic combination of several systems as a so-called system of systems arrangement is shown. In a first step, all luggage items G in an airport are scanned, for example, in a transmission device 200, and the detected objects are located in the coordinate system GKS of the luggage items, for example based on their shape or their specific density. Here, an inverse transformation is performed from the detection in the coordinate system TKS of the transmission device. The transmission segment coordinates TKO can thus be transferred to the evaluation device and can be taken into account in particular when carrying out the segmentation method according to the invention. Similarly, when carrying out the segmentation method and scanning in the X-ray diffraction device 100, starting from the detection in the coordinate system RKS of the X-ray diffraction device, the object O can be transformed back into the segment coordinates SKO in the coordinate system GKS of the luggage items by segmentation, so that a combined evaluation can be performed.

[0060] The above description describes the invention only in the context of examples. Of course, the individual features can be freely combined with one another where technically advantageous, without departing from the scope of the present invention.

[0061] Reference numerals list

[0062] 20 Pulse transmission function

[0063] 30 voxels

[0064] 32 starting voxels

[0065] 33 adjacent voxels

[0066] 34 voxel cluster

[0067] 40 Material Functions

[0068] 50 Similarity value

[0069] 60 Termination criteria

[0070] 100X-ray diffraction equipment

[0071] 200 Transmission Equipment

[0072] O Object

[0073] S segment

[0074] TS Transmission Segment

[0075] G. Luggage items

[0076] SKO segment coordinates

[0077] TKO launch segment coordinates

[0078] GKS Baggage item coordinate system

[0079] Coordinate system of TKS transmission equipment

[0080] Coordinate system of the RKS X-ray diffraction facility

Claims

1. A segmentation method for delimiting an object (O) into segments (S) in a baggage item (G) during X-ray diffraction analysis, comprising the following steps: - recording a pulse transmission function (20) for a plurality of voxels (30) of said luggage item (G), - comparing the recorded pulse transmission function (20) with at least one specified material function (40) and generating a similarity value (50) as a result of the comparison, - selecting at least one voxel (30) as the starting voxel (32) based on the similarity value (50), - creating at least one voxel cluster (34) starting from the at least one selected starting voxel (32) using a combined method comprising the following steps: a) combining the starting voxel (32) with neighboring voxels (33) in all spatial directions, b) determining, for each of these combinations, said similarity value (50) to said at least one material function (40), c) adding to said voxel cluster (34) those neighboring voxels (33) which result in an improvement and / or maintenance of said similarity value (50), d) repeating steps a) to c) until the termination criterion (60) is met, - combining overlapping voxel clusters (34) into at least one segment (S), the increase in the similarity value (50) being taken into account by the combination.

2. The segmentation method according to claim 1, characterized in that: At least one of the following is used as a termination criterion (60) for the combined method: -Default number of iterations; - the similarity value (50) is not improved by steps a) and b); - deteriorating the similarity value (50) by steps a) and b).

3. The segmentation method according to one of the preceding claims, characterized in that Prior to the step of combining the generated voxel clusters (34) into at least one segment (S), the generated voxel clusters (34) are sorted according to similarity values ​​(50) of the voxel clusters (34).

4. Segmentation method according to one of the preceding claims, characterized in that A set of several neighboring voxels (30) is selected as a starting voxel (32).

5. Segmentation method according to one of the preceding claims, characterized in that After combining overlapping voxel clusters (34) into segments (S), this step is repeated for the remaining voxel clusters (34).

6. Segmentation method according to one of the preceding claims, characterized in that The similarity value (50) is formed as a scalar value.

7. Segmentation method according to one of the preceding claims, characterized in that A similarity value (50) is applied to at least two different material functions (40).

8. Segmentation method according to one of the preceding claims, characterized in that The at least one material function (40) is specified based on scanning results of a previous scanning method.

9. Segmentation method according to one of the preceding claims, characterized in that The steps are confined to a self-contained area of ​​the luggage item (G).

10. Segmentation method according to one of the preceding claims, characterized in that Depending on the combination into at least one segment (S), an alarm signal is output and / or an existing alarm signal is eliminated.

11. Segmentation method according to one of the preceding claims, characterized in that Segment coordinates (SKO) in a coordinate system (GKS) of the luggage item (G) are determined for the at least one segment (S).

12. The segmentation method according to claim 11, characterized in that: At least one transmission segment (TS) has been generated by a previous transmission method performed on the same luggage item (G), the transmission segment coordinates (TKO) of which are determined in said coordinate system (GKS) of said luggage item (G).

13. Segmentation method according to one of claims 11 to 12, characterized in that When generating the at least one voxel cluster (34) and / or during the combination to form the at least one segment (S), information, in particular the boundaries of the transmission segment (TS), is taken into account.

14. A computer program product comprising instructions which, when said program is run on a computer, cause said computer to execute the steps of the segmentation method having the features of one of claims 1 to 13.

Images