System and method for detecting irregularities of object

Abstract

The invention relates to the observation and inspection of objects, for example for quality control purposes. A solution for camera-based inspection of an object to detect the presence of irregularities of the object assumes that the object and the camera move relative to each other. The camera is implemented as a DVS camera configured to generate an event signal pixel by pixel if a change in luminous flux over a pixel exceeds a given threshold. In the event that the constellation of the event signal generation pixel deviates from the predefined pattern, irregularities are deemed to be present.

Description

[0001] The invention relates to the viewing and inspection of objects, respectively, eg for quality control purposes.

[0002] Optical inspection in series production suffers from the following problem: When objects are inspected at their conveying speed in production with high lateral resolution (i.e. in the conveying direction and perpendicular to it), large amounts of data are created. Typical data rates caused by the corresponding camera systems may be in the range of 1 or even above 5 Gbit / s. This amount of data needs to be transmitted and pre-evaluated in smart cameras or edge devices in order to make the data flow to the end device (such as a superordinate production control system) manageable. Lower data rates lead to smaller resolutions and thus reduced inspection accuracy, which ultimately has a negative impact on the quality of the production process.

[0003] Prior art inspection systems capture large amounts of image data using suitable cameras and transmit the captured data to an edge device that performs a pre-evaluation of the data or even a full evaluation. In alternative solutions, the data is transmitted to a cloud system for analysis. In some cases, smart cameras are used that contain a CPU or FPGA in order to identify defects at an early stage of the process and crop relevant areas of interest in the image, or aggregate data by packaging, etc. In many applications, line scan cameras are used in order to obtain a more constant data rate than with full-field cameras, but for the same resolution, the general theme of high data volume remains exactly the same. Therefore, there is no suitable concept in place on the market that, on the one hand, reduces the data flow and, on the other hand, in applications where the object to be inspected and the inspection camera system move relative to each other, still covers a large dynamic range of, for example, 80 to 100 db or even higher. A high dynamic range is necessary to cover different lighting conditions in the camera's field of view and to maintain good imaging characteristics to ensure high inspection accuracy and quality. Typical full-field or line cameras have a maximum dynamic range of only 40 to 60 dB.

[0004] Therefore, a solution is needed that allows fast and reliable detection of irregularities of an object. This is solved by the method proposed in claim 1 and by the system according to claim 13. The dependent claims describe advantageous embodiments of the method and the system.

[0005] The method for camera-based inspection of an object to detect the presence of irregularities (e.g. defects or production errors) of the object assumes that the object and the camera move relative to each other in a conveying direction TD during the inspection. Therein, the expression "the object moves relative to the camera" covers the case of a moving object and a stationary camera, the case of a stationary object and a moving camera, and the case of a moving object and a moving camera. The camera is implemented as a DVS camera, which includes a sensor chip with a plurality of sensor elements SE(i), i.e. pixels, where i=1, ..., I and I represents the total number of sensor elements of the sensor chip. The DVS camera is configured to generate an event signal SIG(i) selected from a predefined group GRP of event types for each sensor element SE(i) individually, i.e. pixel by pixel. The group GRP may include at least positive events and negative events. An event signal SIG(i) is generated only if the light intensity or light flux irradiated on the corresponding sensor element SE(i) changes by more than a given threshold value THRES, e.g. 15%, and no event signal is generated if the flux change is less than the threshold value THRES. The method comprises an irregularity detection phase during which an object irregularity is considered to be present if one or more event signals SIG(i) generated by the DVS camera during relative movement of a portion of the object through the field of view FOV of the DVS camera satisfies a predefined condition.

[0006] The DVS camera is configured and arranged such that it captures a given field of view and such that the field of view covers a portion of the object surface.

[0007] The predefined group GRP of event types comprises positive events and negative events, wherein in case of a change in the light flux by an amount DELTAFX where DELTAFX>THRES to a higher light flux, the event signal SIG(i) in the case of generation of the corresponding sensor element SE(i) is a positive event. In case of a change in the light flux by an amount DELTAFX where DELTAFX>THRES to a lower light flux, the event signal SIG(i) in the case of generation is a negative event. In case of a change in the light flux DELTAFX below or equal to a threshold value THRES, i.e. DELTAFX≤THRES, no event signal SIG(i) is generated at all for the corresponding sensor element SE(i). Thus, an event signal is only generated in case of a significant flux change, i.e. DELTAFX>THRES.

[0008] An irregularity is only considered to be present if one or more event signals SIG(i) are generated for one or more sensor elements SE(i).

[0009] To complicate matters further, irregularities are only considered to be present if the temporal and / or spatial constellation of one or more sensor elements SE(i) (for which corresponding event signals SIG(i) are generated during the irregularity detection phase) forms a pattern in time (i.e. considering the series of event signals of the same sensor element over time) and / or in space (i.e. considering the event signals of different sensor elements) that corresponds to or deviates from a predefined reference pattern.

[0010] Therein, the reference pattern may be a pattern in space. The reference pattern may be such a spatial constellation of sensor elements that instantaneously generate event signals that would be caused by a certain part of the object that is instantaneously located in the field of view, in the absence of significant irregularities of the object in this certain part. In this case, an irregularity may be considered to exist if the constellation of one or more sensor elements SE(i) (for which the corresponding event signals SIG(i) are instantaneously generated) forms a pattern in space that deviates from a predefined reference pattern.

[0011] In other words, such a reference pattern corresponds to and represents the expected instantaneous appearance or spatial constellation of event signal generating sensor elements of the sensor chip, respectively, provided that there are no irregularities in the part of the object instantaneously located in the field of view of the DVS camera. This exploits the fact that the optical appearance of the object and its parts in the field of view is known, so that the event signals generated by the sensor elements of the DVS camera during the movement of the parts through the field of view, i.e. the "constellation of sensor elements", can be easily modeled. As soon as the instantaneous constellation of event signal generating sensor elements deviates from such a modeled constellation, an irregularity can be considered to be present.

[0012] The use of a well-defined reference pattern is particularly beneficial in the case of an object surface that is non-uniform, for example in the case of an object consisting of multiple components, such as in a batch production system, or in the case of a texture on the object surface showing significant roughness or contrasting elements. Inhomogeneities can lead to multiple event signals even in the absence of irregularities. Therefore, in contrast to the case of an essentially uniform object surface, the use of a threshold value THRES alone is not sufficient to ensure that irregularities can be detected with high reliability.

[0013] However, the constellation of sensor elements generating the event signal depends on known arrangement or formation inhomogeneities. Therefore, it can be assumed that the constellation is known and deviations from this known constellation can be interpreted as the presence of an irregularity.

[0014] The method may also include a selection phase performed before the irregularity detection phase, wherein in the selection phase at least one spatial reference pattern is automatically determined. For this purpose, the DVS camera may capture the object during its relative movement through the field of view FOV to generate at least one spatial constellation of sensor elements SE(j) from event signals SIG(j), and the at least one spatial constellation of event signals thus generated is used as at least one spatial reference pattern.

[0015] The reference pattern can also be at least a pattern in time, where for the same sensor element SE(i) a first event signal is at time point t1 and a different second event signal is at time point t2. An irregularity can be considered to be present if, during a given time span dT1>t2-t1 and between two time points t1, t2, for the same sensor element SE(i), the generated event signal SIG(i) changes from a first event type of group GRP to a second event type of group GRP, the second event type being different from the first event type, i.e. from a positive event to a negative event or vice versa. The maximum reasonable time span dT depends on the conveying speed, the maximum size of the irregularity to be detected, and the size of the sensor element.

[0016] The reference pattern may also be essentially a pattern in space, wherein an irregularity may be considered to exist if, during the time span dT2, event signals SIG(j1), SIG(j2), SIG(j3) of the same event type (e.g. positive events) are generated for a plurality of N>2 different sensor elements SE(j1), SE(j2), SE(j3) for different time points t1, t2, t3, wherein the different sensor elements SE(j1), SE(j2), SE(j3) are located on an essentially straight line. Thus, the reference pattern will include a plurality of points located on such a straight line.

[0017] In the case where the reference pattern is a pattern in space, an irregularity may be considered to exist if an event signal SIG(j1) of a first event type of the group GRP is generated for a first sensor element SE(j1) and an event signal SIG(j2) of a different second event type of the group GRP is generated for a second sensor element SE(j2) (which is positioned adjacent to the first sensor element SE(j1)) at substantially the same time t0. That is, the first event type and the second event type are different from each other.

[0018] The method can be applied to objects with a substantially uniform surface, so that in the absence of irregularities, no event signal is generated, i.e., neither positive events nor negative events are generated, because the fluctuations in intensity and flux do not exceed the noise level, which is generally kept below a given threshold. However, "substantially uniform" still applies to a certain extent to textures at the surface characterized by a certain roughness. A given threshold THRES can be set based on the roughness of the texture at the surface, for example, instead of 15% as described above, a lower threshold is set for finer textures and a higher threshold is set for coarser textures. The roughness of the texture may depend on the type or pattern of tissue included in the object.

[0019] The given threshold THRES can be set automatically in a selection phase of the method, which will be performed before the irregularity detection phase. In the selection phase, the threshold THRES can be changed until the relationship REL=f(NGEN, REFPARA) depending on the number of generated event signals NGEN related to the reference parameter REFPARA reaches the expected value X. For example, the reference parameter REFPARA can be a time span REFTIME, for example in seconds, or an area REFAREA, for example in square meters. Therefore, the relationship REL can represent the number of event signals NGEN generated per square meter or per second, i.e. REL=NGEN / REFTIME. Once this relationship REL reaches the expected value X (which in extreme cases may be X=0 or any other value representing the potential false alarm tolerance), the instantaneous threshold THRES that has a direct impact on N will be selected to set the threshold THRES.

[0020] Different methods of illuminating the object can be applied to different areas of the camera field of view, such as bright field transmitted light illumination in a first area, a combination of transmitted and reflected light in a second area, and / or only reflected light in a third area. For some inspection tasks, even dark field type illumination in reflection or transmission can also be used. The different illumination areas of the field of view can be arranged one behind the other in the transport direction TD.

[0021] Among them, if a first event signal of a first event type of a group GRP is generated for a first sensor element in a first of different illumination regions at a first time point t1, and a second event signal of an event type corresponding to the first event type is generated for a second sensor element at a second time point t2, and the second sensor element is located in a second of different illumination regions and corresponds to the first sensor element, an irregularity can be considered to exist. Among them, sensor elements in different regions correspond to each other if the mutual distance DIST between them in the conveying direction TD is substantially proportional to the distance traveled by a point on the object surface relative to the field of view between time points t1 and t2, i.e., DIST = PROP * VEL * (t2 - t1). Among them, the proportionality factor PROP depends on the imaging characteristics of the DVS camera and its optics, such as the scaling of the imaged object. Therefore, the distance DIST and the correspondence depend on the speed VEL of the relative movement of the DVS camera and the object.

[0022] The sensor chip can be a rectangular chip, and its sensor elements are arranged in rows and columns perpendicular to each other, where the DVS camera and its sensor chip are arranged such that 0° < ALPHA < 90° applies to the angle ALPHA between a row or a corresponding column and the conveying direction TD, preferably ALPHA < 10°. This non-parallel alignment improves the sampling accuracy of irregularities to the sub-pixel level. Multiple events from different rows of the sensor chip can be used to fuse data regarding the feeding motion to create a higher-resolution effective irregularity image.

[0023] A corresponding system for camera-based inspection of an object to detect the presence of irregularities in the object when the object and the camera move relative to each other in the conveying direction TD during inspection includes a DVS camera and an inspection system configured to perform the above method.

[0024] The DVS camera can be arranged such that the rows of the sensor elements of its sensor chip are aligned at an angle of 0° < ALPHA < 90° relative to the conveying direction TD.

[0025] The system can include an illumination system having a plurality of light sources for illuminating the field of view. The light sources can be arranged such that different light sources of the illumination system illuminate different regions and / or the same region of the field of view at different illumination incident angles.

[0026] The present invention advantageously implements a camera applying the principle of a "Dynamic Vision System" (DVS). Typically, such a DVS camera (also referred to as an "event camera" or a "neuromorphic camera") is a full-field camera with a chip having an aspect ratio of 4:3 or 16:9, etc., and is currently available with >10 6The basic characteristics of the DVS camera used in the solution described herein consist in that the DVS camera detects events, i.e. changes in the light flux exceeding a predefined threshold THRES (e.g. 15%), and detects the light flux pixel by pixel with an actual value FX, so that in practice the DVS camera has an effective dynamic range of 120 db or even higher, which is a result of the full well capacity of the pixel and the exposure time chosen according to the minimum and maximum allowed values.

[0027] In summary, the main differences between DVS camera-based inspection systems and conventional inspection systems are based on: the characteristics of DVS cameras described below that detect events instead of full-frame imaging; the significantly higher speed of DVS camera data acquisition compared to the typical frame rate of full-frame cameras; the tight and precise timing coupling due to the significant speed advantage of the DVS principle over conventional cameras, although such coupling is not required with the accuracy of conventional cameras; the significant reduction in the amount of data, since only relevant events are reported instead of a complete image capture of the entire surface of the inspected object; the color view mode of inspection utilizing event-based DVS cameras plus the possibility to distinguish between different imaging modes; and event-based data fusion plus the creation of "normal" area type images.

[0028] It should be understood that the elements and features listed in the attached claims can be combined in different ways to produce new claims that also fall within the scope of the invention. Therefore, although the dependent claims attached below are dependent on specific independent or dependent claims, it should be understood that these dependent claims can be alternatively dependent on any previous or subsequent claims, whether independent or dependent, and this new combination should be understood to form a part of this specification. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] In the following, possible embodiments of different aspects of the present invention will be described in more detail with reference to the accompanying drawings. The purpose and further advantages of this embodiment will become more obvious and easy to understand from the following description of the preferred embodiment in conjunction with the accompanying drawings, in which: Figure 1 shows a side view of a mass production facility for producing an object, Figure 2 shows a flow chart of a method METH for object inspection, Figure 3 shows a top view of the sensor chip, Figure 4 shows a top view of a production object suitable for a first embodiment, FIG5 shows a top view of a portion of a sensor chip illustrating a first situation, FIG. 6 shows a top view of a portion of a sensor chip illustrating a second situation, Figure 7 A top view of a portion of a sensor chip illustrating a third aspect is shown. Figure 8 shows a top view of a production object suitable for a second embodiment, Fig. 9 shows a top view of a portion of a sensor chip illustrating a constellation of event signal generating sensor elements of a production assembly, Fig.10 Shown is a top view of a portion of a sensor chip illustrating a constellation of event signal generating sensor elements for a production component having irregularities. DETAILED DESCRIPTION

[0030] Figure 1 A side view of a mass production facility 100 for producing an object 10 is shown. The object 10 may be Figure 8 A batch of a plurality of mass-produced components 10' is shown, for example any kind of components, such as solid parts made of metal or plastic, machined, printed, injection-molded, etc., or even soft materials, such as tissue, paper, fabric or food, such as any product from cheese, meat, sausage, biscuits, bread, tofu and agricultural products such as eggs, plants, fish, etc. The produced objects 10 are transported from a first station 110 of the mass production facility 100 to a second station 120 in a conveying direction TD, for example, by a conveyor belt 130. The stations 110, 120 can be machines or other components of the mass production facility 100, which are used to implement the mass production process applied in the mass production facility 100 to produce the objects 10.

[0031] Between the stations 110, 120, a DVS camera 210 of an inspection system 200 for camera-based inspection of an object 10 is installed. The inspection system 200 is configured to detect irregularities IRR, such as defects or any production errors, of the object 10 while the object 10 moves along and relative to the DVS camera 210, respectively. The DVS camera 210 is configured and arranged such that it captures a given field of view FOV and such that the field of view FOV covers a portion of the surface of the object 10. Preferably, the field of view FOV covers the entire width W of the object 10 and has an extension H in the conveying direction TD. Thus, at a certain point in time t(i), the DVS camera 210 captures the portion of the object 10 that is located in the field of view FOV of the DVS camera 210 at that point in time t(i). Since the object 10 moves relative to the DVS camera 210, different portions of the surface of the object 10 are captured at different points in time t(i).

[0032] The inspection system 200 comprises an evaluation system 220. The evaluation system 220 is connected to the DVS camera 210 via a wired or wireless connection 221 and is configured to control the DVS camera 210 and to receive imaging data SIG(i) from the DVS camera 210 via the connection 221. Furthermore, the evaluation system 220 is configured to determine whether an irregularity is present at the object 10 based on an evaluation of the imaging data SIG(i) provided by the DVS camera 210. For the purpose of the evaluation, the evaluation system 220 applies a method METH, which comprises an INSPECT phase for detecting irregularities (e.g. defects or production errors) of the object 10 and, optionally, an ADOPT phase for selecting evaluation parameters applied by the evaluation system 220. Figure 2 The method METH described in more detail in the context of can be implemented as corresponding software executed by the processor 222 of the evaluation system 220. The evaluation system 220 may also include a memory 223 for storing any data to be processed (e.g., imaging data SIG(i)), as well as software to be executed by the processor 222, and / or any other suitable or required data. The evaluation system 220 and the inspection system 200 are configured to create an alarm signal and / or interfere with the batch production process performed by the batch production facility 100 if an irregularity is considered to exist based on the evaluation of the imaging data SIG(i).

[0033] The DVS camera 210 has a sensor chip 211 with a plurality of light-sensitive sensor elements SE(j), where j=1, . . . , J and J represents the total number of sensor elements SE(j), for example J>10 6 As a simplified example, Figure 3 A top view of a sensor chip 211 having only 16*24=384 sensor elements SE(j) is shown in . The sensor chip 211 is a rectangular chip 211 whose sensor elements SE(j) are arranged in rows and columns perpendicular to one another. Only a few sensor elements SE(j) are labeled with reference numerals.

[0034] In contrast to conventional cameras, the signal SIG(j) generated by the DVS camera 210 for a given sensor element SE(j) does not directly or, for example, proportionally represent the light flux (e.g., in lumens) onto the respective sensor element SE(j), but it depends on the relationship between the light flux onto the sensor element SE(j) at the time point t(i) and the light flux onto the same sensor element SE(j) at the earlier time point t(il). More specifically, it indicates whether the light flux FX at the sensor element SE(j) at the time point t(i) differs from the light flux FX at the same sensor element SE(j) at the earlier time point t(i-1) by more than a given threshold value THRES. In case such a difference is indeed above the threshold value THRES, a positive or negative event is generated for the respective sensor element SE(j).

[0035] Therefore, the DVS camera 210 is configured to generate a signal SIG(j), hereinafter referred to as an "event signal" SIG(j), for each sensor element SE(j) individually in case the change in the light flux onto the corresponding sensor element SE(j) exceeds a given threshold value THRES (e.g. 15%), and not to generate an event signal in case the flux change is less than or equal to the threshold value THRES. Therein, the event signal SIG(j) is not assigned a free floating value, but each event signal SIG(j) is selected from a predefined group GRP of event types, which group GRP may include at least positive events and negative events. Therefore, the event signal SIG(j), in case it is generated, is a positive event only in case the flux changes to a higher flux by an amount DELTAFX where DELTAFX>THRES, or is a negative event in case the flux changes to a lower flux by an amount DELTAFX where DELTAFX>THRES. If the change DELTAFX in the flux onto the sensor element SE(j) is below or equal to the threshold THRES, i.e. DELTAFX≤THRES, no event signal SIG(j) is generated for the corresponding sensor element SE(j). Thus, an event signal SIG(j) is only generated in the case of a significant flux change, i.e. DELTAFX>THRES.

[0036] More specifically, consider a specific sensor element SE(j) of the sensor chip 211 of the DVS camera 210: In the case where the instantaneous flux FX(t(i)) onto the sensor element SE(j) is higher than an earlier flux FX(t(i - l)) onto the same sensor element SE(j) by more than THRES, i.e., in the case where FX(t(i)) - FX(t(i - 1)) > THRES, the event signal SIG(j) generated for the specific sensor element SE(j) is a positive event, which is shown by the symbol “+” in the corresponding figure. In the case where the instantaneous flux FX(t(i)) onto the sensor element SE(j) is lower than an earlier flux FX(t(i - 1)) onto the same sensor element SE(j) by more than THRES, i.e., in the case where FX(t(i - 1)) - FX(t(i)) > THRES, the event signal SIG(j) generated for the specific sensor element SE(j) is a negative event, which is shown by the symbol “-” in the corresponding figure. In all other cases where the amount of the difference |FX(t(i)) - FX(t(i - 1))| between the fluxes FX(t(i)) and FX(t(i - 1)) is less than THRES, i.e., in all other cases where |FX(t(i)) - FX(t(i - 1))| < THRES, no signal is generated for the corresponding sensor element SE(j). Correspondingly, the sensor element SE(j) for which no event signal is generated is not assigned any symbol in the corresponding figure.

[0037] The method METH for detecting the irregularity IRR of the object 10 advantageously utilizes this characteristic of the DVS camera 210. Very basically, in the case where one or more positive or negative event signals SIG(i) are generated for one or more sensor elements SE(i), an irregularity IRR can be considered to exist. This solution may be prone to false alarms. Therefore, in a more robust method, during the irregularity detection phase INSPECT, in the time and / or space constellation of those one or more sensor elements SE(i) for which the corresponding event signals SIG(i) are generated forms a pattern that corresponds to or deviates from a predefined reference pattern (i.e., considering the series of event signals of the same sensor element over time) and / or in space (i.e., considering the event signals SIG(j1), SIG(j2), SIG(j3),... of different sensor elements SE(j1), SE(j2), SE(j3),...), an irregularity IRR can be considered to exist. Therefore, the assumption of the existence of the irregularity IRR depends on the comparison of the constellation formed by the sensor elements SE(j) generating the event signals SIG(j) with the predefined reference pattern. For example, the term “constellation” can be interpreted as the overall spatial arrangement of the sensor elements SIG(j) generating the event signals.

[0038] In general, such a reference pattern may correspond to and represent the expected instantaneous appearance or spatial constellation of sensor elements SE(j) generated by the event signal SIG(j) of the sensor chip 211, respectively, under the premise that the object 10 has no irregularities in the portion thereof instantaneously located in the field of view FOV of the DVS camera 210. This takes advantage of the fact that the optical appearance of the object 10 and its portion in the field of view FOV is known, so that the event signal SIG(j), i.e., the “constellation of sensor elements”, generated by the sensor elements SE(j) of the DVS camera 210 during the movement of the portion through the field of view FOV may be easily predicted or modeled. As soon as the instantaneous constellation of sensor elements SE(j) generated by the event signal SIG(j) deviates from such a modeled constellation, it may be considered that an irregularity IRR exists.

[0039] Figure 4 , a top view of an object 10 suitable for a first embodiment of a method METH performed by an inspection system 200 to inspect an object 10 moving relative to a DVS camera 210 to detect irregularities IRR is shown. In the first embodiment, it is assumed that the surface of the object 10 captured in sections by the DVS camera 210 is substantially uniform, so that in the absence of irregularities no event signal SIG(j) will be generated, i.e. neither positive nor negative events, since regular fluctuations of the flux onto the sensor elements do not exceed a certain noise level.

[0040] Such noise level will usually remain below a given threshold value THRES, or correspondingly, the threshold value THRES is set such that it is not exceeded by such noise. However, "substantially uniform" still applies to a certain extent to the texture at the surface of the object 10, which texture is characterized by a certain roughness, which roughness may depend on the type or pattern of tissue included in the object 10. In this case, the threshold value THRES can be set in the optional optional stage ADOPT based on the roughness of the texture at the surface, for example, instead of the default value THRES=15% as described above, a lower threshold value can be set for finer textures and a higher threshold value can be set for coarser textures.

[0041] In an exemplary implementation of the object 10 applied in the first embodiment and described below, the object 10 is a large continuous fabric having a substantially continuous surface. As another example, the object 10 may be a large continuous sheet of metal or the like.

[0042] The method METH comprises an irregularity detection phase INSPECT for detecting irregularities IRR of the object 10, such as defects or production errors, and optionally an optional phase ADOPT executed before the phase INSPECT for detecting irregularities IRR, if applicable, for individually setting a threshold value THRES instead of applying a default value as threshold value.

[0043] In the case of execution, the stage ADOPT enables the automatic determination and setting of the threshold value THRES for the object to be inspected 10. For example, the threshold value THRES can be varied, starting from a starting point such as THRES(0)=0%, until the relationship REL=f(NGEN, REFPARA) depending on the number NGEN of generated event signals related to the reference parameter REFPARA reaches the expected value X.

[0044] For example, the reference parameter REFPARA may be a time span REFTIME, for example in seconds. Thus, the relationship REL may represent the number NGEN of event signals generated per second, i.e. REL=NGEN / REFTIME. This relationship REL of course depends on the actual threshold THRES, so a change in THRES results in a change in REL. Once this relationship REL reaches the expected value X (which in an extreme case may be X=0 or any other value representing a potential false alarm tolerance), the instantaneous threshold THRES will be selected as the set threshold THRES.

[0045] In another example, the reference parameter REFPARA may be the area REFAREA on the surface of the object 10, for example in square meters, which has passed through the field of view FOV during the execution phase ADOPT, depending on the transport speed VEL. In this case, the relationship REL represents the number N of event signals per square meter. Again, a change of THRES has an effect on N, and once this relationship REL reaches the expected value X (which in an extreme case may again be X=0 or any other value representing the tolerance for potential false alarms), the instantaneous threshold THRES will be selected as the set threshold THRES.

[0046] However, the irregularity detection phase INSPECT comprises using the DVS camera 210 to capture the field of view FOV during the movement of the object 10. Imaging data, ie event signals SIG(j), are provided to an evaluation system 220 for evaluation as soon as they occur to determine if an irregularity IRR is present.

[0047] Basically, the evaluation system 220 considers the presence of an irregularity IRR based on the provided imaging data SIG(j) only if there is a generated event signal SIG(j), i.e. a positive event or a negative event, for at least one sensor element SE(j). As described above, a more robust method considers the constellation of sensor elements SE(j) generated by the event signal SIG(j) of the sensor chip 211 and compares such a constellation with a reference pattern.

[0048] In a first approach, a reference pattern can be understood as a pattern at least in time, wherein for a specific sensor element SE(j1) at a time point t1 there is a first event signal SIG1(j1) and for the same sensor element SE(j1) at a later time point t2 there is a different second event signal SIG2(j1). For example, the first event signal SIG1(j1) may be a positive event and the second event signal SIG2(j1) may be a negative event, or vice versa. An irregularity IRR can be considered to exist when, during a given time span dT1≥t2-t1 during the irregularity detection phase INSPECT and between two time points t1, t2, the event signal SIG(j1) generated for the specific sensor element SE(j1) changes from a first event type of the group GRP to a different second event type of the group GRP, the second event type being different from the first event type, i.e. from a positive event to a negative event, or vice versa. This includes that before the time point t1, a sequence of a plurality of first event signals SIG1(j1) has been generated, and after the time point t2, a sequence of a plurality of second event signals SIG2(j1) has been generated. However, the event signal changes between t1 and t2. In other words, an irregularity IRR can be considered to exist in the case where, for the same sensor element SE(j1), the DVS camera 210 first provides at least one positive event and later provides at least one negative event, or vice versa. This actually means that a change has been detected at the sensor element SE(j1), which for example results in a positive event, and due to the movement of the object 10, this change disappears again for this same sensor element SE(j1) at a later point in time, which for example results in a negative event of SE(j1). This is illustrated in FIGS. 5a, 5b, which for the sake of clarity only show a limited number of sensor elements SE of the sensor chip 211: FIG. 5a represents a first point in time t1, at which a positive event is generated for the sensor element SE(j1). FIG5 b shows a second time point t2, where t2>t1, at which a negative event is generated for the same sensor element SE(j1). Therefore, it can be considered that an irregularity IRR exists.

[0049] The maximum reasonable time span dT1 within which a change in the event signal type should occur depends on the conveying speed VEL, the maximum size of the irregularity to be detected, and the size of the sensor element SE.

[0050] In a second method that can be applied in addition to or as an alternative to the first method, the reference pattern can be understood essentially as a pattern in space, where an irregularity IRR is considered to exist when a plurality of N>2 different sensor elements SE(j1), SE(j2), SE(j3) at different time points t1, t2>t1, t3>t2 generate event signals SIG(j1), SIG(j2), SIG(j3) of the same event type (e.g. a positive event) during a time span dT2. Wherein, it is required that those different sensor elements SE(j1), SE(j2), SE(j3) are located on a substantially straight line. Of course, the consideration can be extended to more than three time instances, and the irregularity IRR is assumed when pixels of different images with the same event type are located on a straight line. Therefore, the spatial reference pattern will include a plurality of points located on such a straight line.

[0051] Irregularities IRR lead to positive or negative events, and it must be assumed that due to the relative movement, the same event type will be generated at different positions and sensor elements at a later point in time, respectively. These sensor elements must be located on a straight line, practically parallel to the conveying direction TD. Figures 6a-6c show the corresponding situation, wherein for the sake of clarity, Figures 6a-6c again only show a limited number of sensor elements SE of the sensor chip 211. Figure 6a shows the first point in time t1 at which a positive event is generated for sensor element SE(j1), Figure 6b shows the first point in time t2 at which a positive event is generated for a different sensor element SE(j2), and Figure 6c shows the first point in time t3 at which a positive event is generated for a different sensor element SE(j3). It can be seen that the sensor elements SE(j1), SE(j2), SE(j3) are practically located on a substantially straight line, for example when the illustrations of Figures 6a-6c respectively at least virtually completely coincide or superimpose.

[0052] In a third method, which can be applied in addition to or as an alternative to the first and / or second method, a reference pattern can be understood as a pattern in space, wherein an irregularity IRR is considered to exist when an event signal SIG(j1) of a first event type (e.g. a positive event) of the group GRP is generated for a first sensor element SE(j1) and an event signal SIG(j2) of a different second event type (e.g. a negative event) of the group GRP is generated for a second sensor element SE(j2) (which is positioned adjacent to the first sensor element SE(j1), preferably adjacent to the first sensor element SE(j1)) at substantially the same time t0. That is, the first event type and the second event type are different from each other. This third method again takes into account the movement of the object 10 and the irregularity IRR relative to the DVS camera 210.

[0053] This situation is Figure 7 For the sake of clarity, Figure 7 Again only a limited number of sensor elements SE of the sensor chip 211 are shown. A negative event is generated for a first sensor element SE(j1) while a positive event is generated at the same time t0 for a second sensor element SE(j2) located adjacent to the first sensor element SE(j1).

[0054] The above method works best for objects 10 having a substantially uniform surface. However, this is not necessarily guaranteed for every object 10. Figure 8 1 shows a top view of an object 10 suitable for a second embodiment of a method METH, which is performed by an inspection system 200 to inspect an object 10 moving relative to a DVS camera 210 to detect irregularities IRR. In the second embodiment, it is assumed that the surface of the object 10 captured in sections by the DVS camera 210 is not of a substantially uniform type as in the case of a fabric or, for example, a large area of ​​sheet metal, etc. described in the first embodiment. Therefore, due to the uneven surface of the object 10, it must be assumed that the DVS camera 210 detects a plurality of events even in the case of no defects in the object 10: for example, Figure 8 As shown, the object 10 may be a batch of a plurality of mass-produced components 10', such as food or agricultural products, such as any product from cheese, meat, sausage, biscuits, bread, tofu, eggs, plants or fish, but may also be components from other industrial mass production, such as screws, circuit boards, fabrics with textures, etc. For the sake of clarity, only a few components 10' of the object 10 are marked with reference numerals. The entire surface of such an object 10 consisting of a plurality of individual components 10' includes the surfaces of a plurality of components 10' that are more or less adjacent to each other, so that such an entire surface is not substantially uniform as in the example of the first embodiment, but the entire surface will generally be characterized by a high-contrast texture, at least with a flux gradient in the area between adjacent components. The same applies to objects 10 that are not composed of individual components but whose surface has intrinsic inhomogeneities and correspondingly the texture on the surface shows significant roughness or contrast elements. In both cases, due to the high-contrast texture of the entire surface, the DVS camera 210 will detect a plurality of event signals SIG(j) of positive and negative event types even in the absence of irregularities. Therefore, contrary to the case where the surface of the object 10 is substantially uniform, the use of the threshold value THRES alone is not sufficient to ensure that the irregularity IRR can be detected with high reliability. However, according to the regular properties of the component 10', it can still be assumed that the imaging data SIG(j) captured by the DVS camera 210 over time t includes certain features.

[0055] For the sake of clarity, Fig. 9 Only a limited number of sensor elements SE of the sensor chip 211 are depicted, and Fig. 9 Shown for Figure 8 The event signal of only one of the components 10 ′ shown in FIG. 1 , which may be a substantially circular sheet, generates a constellation of sensor elements SIG(j).

[0056] According to this, the reference pattern of the object 10 and the component 10' may respectively consist of one or more constellations, such as Fig. 9 As shown. More generally, the reference pattern can be a pattern in space, where the reference pattern is such a spatial constellation of sensor elements that instantaneously generate any type of event signal, which would be caused by being instantaneously located in a certain portion of the object 10 in the field of view FOV in the absence of significant irregularities in the object 10 in that portion of the field of view FOV. In a first embodiment, due to the assumed uniform surface in the first embodiment, such a "spatial constellation of sensor elements that instantaneously generate event signals" will not include a single sensor element. In a second embodiment, at least event signals are generated when the edge or edge of the component 10' moves in and out of the area within the FOV field of view captured by a specific sensor element.

[0057] The constellation of sensor elements SE(j) generating the event signal SIG(j) depends on known inhomogeneities, for example on the arrangement and form of the component 10'. It can therefore be assumed that the constellation is known for an object 10 without irregularities, so that such a known constellation can be used as a reference pattern. Deviations from the known constellation and the reference pattern, respectively, can be interpreted as the presence of an irregularity IRR.

[0058] In a second embodiment, the method METH comprises a phase INSPECT of detecting irregularities (for example defects or production errors) of the object 10 and, if applicable, an optional phase ADOPT carried out before the phase INSPECT of detecting irregularities IRR.

[0059] In the case of execution, the stage ADOPT comprises determining at least one spatial reference pattern. For this purpose, the DVS camera 210 can capture the object 10 during its relative movement through the field of view FOV to generate at least one spatial constellation of sensor elements SE(j) of the event signal SIG(j). The at least one spatial constellation of sensor elements SE(j) generated by the event signal SIG(j) is used as at least one spatial reference pattern.

[0060] The irregularity detection phase INSPECT comprises using the DVS camera 210 to capture the field of view FOV during the movement of the object 10. Imaging data, i.e., event signals SIG(j), are provided to the evaluation system 220 as soon as they occur for evaluation to determine whether irregularities are present. Irregularities IRR may be considered to be present if the instantaneous constellation of one or more sensor elements SE(i) instantaneously generated by the corresponding event signals SIG(i) provided by the DVS camera 210 during the inspection phase forms a pattern in space that deviates from a predefined reference pattern. For this purpose, the reference pattern may be stored in the memory 223 of the evaluation system 220 so that the processor 222 can compare the instantaneous constellation provided by the DVS camera 210 with the reference pattern.

[0061] Although Fig. 9 shows a constellation that can be used as a reference pattern, but Fig.10 The instantaneous constellation of sensor elements SE(i) that instantaneously generate event signals SIG(i) in the case where the component 10' does have an irregularity IRR is shown. Fig. 9 It can be seen from the comparison that Fig.10 The instantaneous constellation includes in addition Fig. 9 The other sensor elements outside the reference pattern, which means that the component 10 ′ has irregularities IRR.

[0062] The evaluation system 220 may apply an artificial intelligence (AI) system to compare the instantaneous constellation with the corresponding reference pattern. Such an AI system may have been trained earlier based on a batch of absolutely damage-free components 10'. In this case, the imaging data SIG(j) captured by the DVS camera 210 over time t as described above is provided to the trained AI system, which is able to detect irregularities IRR despite the unevenness of the entire surface of the object 10.

[0063] In both embodiments, different settings may be applied by the illumination system to properly illuminate the object 10 for better imaging. For example, the illumination system may include one or more light sources 231, 232, 233, such as LED systems, which may apply different colors of illumination and / or be arranged so that the object is illuminated in both transmission and reflection light settings.

[0064] In a transmitted light setting, which applies to the case where the object 10 is at least semi-transparent and where at least one of the light sources 233 of the illumination system and the DVS camera 210 can be arranged on different sides of the object 10, i.e. the object 10 is located between the light source 233 and the DVS camera 210, the irregularities IRR implemented as holes in the object 10 will allow a higher light flux to reach the DVS camera 210, thus creating a positive event or even a series of positive events at those pixels PX and corresponding sensor elements SE onto which the holes are imaged. When the holes pass and leave the field of view FOV, a negative event or a series of negative events is created. Thus, each sensor element SE directly captures part of its irregularity IRR, and a group of adjacent sensor elements SE will image the irregularity IRR as a structure of positive and negative events. This also holds true when the irregularity IRR is implemented as an obstructive element in the object 10 (such as some inclusions of knots or particles, which are less transmissive than the regular texture of the object 10), just with the type of event reversed. The corresponding threshold THRES of the signal change for triggering an event can be set according to the texture of the object 10, for example, to a lower value for very fine textures and to a higher value for coarser textures instead of the typical 15%, depending on the material or tissue type of the object 10 or the pattern included.

[0065] Independently of the transparency of the object 10 , a reflected light setup may be applied, wherein the light sources 231 , 232 of the illumination system and the DVS camera 210 are on the same side of the object 10 .

[0066] Preferably, the object 10 can be illuminated by the light sources 231, 232 and / or 233 of the illumination system of the inspection system 200, in particular in the field of view FOV, such as Figure 1As shown. Preferably, but not necessarily, the lighting system and its one or more light sources 231, 232, 233 are configured and arranged so that different methods of illuminating the object 10 are applied to different areas Z1, Z2, Z3, ... of the camera field of view, for example, bright field transmitted light illumination in the first area Z1, a combination of transmitted and reflected light in the third area Z2, and only reflected light in the second area Z2. For example, different lighting methods may also apply light of different colors. For some inspections, even dark field type illumination in reflection or transmission may also be used. Among them, the areas Z1, Z2, Z3, ... of the field of view FOV are arranged one after another in the conveying direction TD. Correspondingly, the image IMA(t(i)) can be divided into the same number of areas. Potential event signals SIG(i1), SIG(i2), SIG(i3), ... detected from various zones Z1, Z2, Z3, ... in the field of view FOV of the DVS camera 210 using different lighting conditions can be combined or merged by using the transport speed VEL of the object 10 to be inspected, for example by the speed VEL of the object 10 or of a conveyor belt or a robotic system transporting the object 10 and / or the DVS camera 210. For example, the speed VEL, which can be obtained from the control system of the batch production facility 100, can be used to derive which sensor element SE(i2) in the second Z2 of the zones Z1, Z2, Z3 corresponds to the first sensor element SE(i1) in the first Z1 of the zones Z1, Z2, Z3. Such sensor elements SE(i1), SE(i2) in different zones Z1, Z2 correspond to each other if their mutual distance DIST from each other in the transport direction TD is substantially proportional to the distance traveled by any point on the surface of the object relative to the field of view FOV between the time points t1 and t2, i.e. DIST=PROP*VEL*(t2-t1). The scaling factor PROP depends on the imaging characteristics of the DVS camera and its optical devices, such as the scaling of the imaged object.

[0067] In this way, the availability of different zones Z1, Z2, ... as described above can be used to make the detection of irregularities IRR more reliable. If a first event signal SIG(i1) of a first event type (e.g. a positive event) of the group GRP is generated for a first sensor element SE(i1) in the first zone Z1 at a first point in time t1, and a second event signal SIG(i2) of the same event type as the first event type is generated for a second sensor element SE(i2) in the second zone Z2 at a second point in time t2, then an irregularity IRR can be considered to exist if, according to the above definition, the second sensor element SE(i2) "corresponds" to the first sensor element SE(i1).

[0068] Thus, data and information from different areas can be considered fused. For more robust detection and / or classification of potential irregularities, it is advantageous to use fused inspection data for the corresponding lateral positions of the object 10. Preferably, some coordinate axes (such as the conveying length of the aircraft or the universal length coordinate of the conveyor) can be used to appropriately fuse the corresponding data also based on the actual speed information registered from the conveying system. The same concept can also be used with a robotic system.

[0069] In principle, events are used instead of representing flux or similar wide range values ​​as a first data or event filter to reduce the amount of measurement data right at the source, i.e. at the imaging system itself. From a very broad and general point of view, the method based on the use of DVS camera 210 helps to reduce the data created directly to the relevant part without the need for a long image processing chain and therefore also improves the performance of the inspection system 200 in terms of sustainability, such as the energy consumption involved due to less powerful computing and data transmission systems, materials, and the entire ICT infrastructure and the related carbon footprint. Regardless of these optimizations, the core of the inspection in terms of finding potential defect locations is more efficient and more detailed than with traditional cameras.

[0070] For illustrative purposes, the following considerations sometimes use again the example of the object 10 being a fabric, as exemplarily applied in the first embodiment of the present invention. However, the corresponding considerations and benefits are generally applicable to other implementations of the object 10, and specific figures and numbers may be selected as appropriate, and are not limited to the object 10 being a fabric. However, the specific example of the object 10 being a fabric is beneficial because it can clearly demonstrate the positive aspects of the present invention, such as the fact that the fabric 10 is generally at least translucent. Therefore, it can be inspected with transmitted light and / or reflected light under bright field or dark field conditions, and there are high requirements for lateral resolution, because individual fibers or threads may be as small as 10 μm and below. A typical inspection task for the fabric 10 is to inspect the quality of a piece of cloth, for example, 1 m or 2 m wide, with a lateral sampling accuracy of, for example, 20 μm. This requires 50,000 points or pixels per meter of width. For conventional cameras, this means about 40 such cameras per meter. The fabric 10 is typically moved at a speed of 200 to 800 mm / s. When the same lateral sampling is applied, this means having 10,000 to 40,000 pixels or 14 to 55 images. At higher speeds, this would mean 40 cameras with 55Mpxl images, or only 2200 camera frames per second. In the case of an 8-bit image resolution, it would already be 2.2GByte / s or almost 20Gbit / s. For a 120db resolution, a 12 or even 16-bit image representation would be required, which would again double the amount of data to about 40Gbit / s as a rough figure for conventional cameras. Corresponding similar considerations apply to the other object 10 types in the first and second embodiments.

[0071] From a data point of view, and independently of the specific embodiment of the object 10, the amount of data created from the inspection system 200 with the DVS camera 210 is significantly reduced compared to a conventional camera, since all areas of the object 10 without anomalies or defects will not create events, since the change in the luminous flux of the corresponding illumination modality does not exceed the predefined threshold THRES. Typically, the size of the irregularity area is only a few mm 2 or cm 2 , with 100-500m of fabric 10 for example 2 This is very low compared to 1cm 2 100 defects (which would be very large defects) would constitute 100cm 2, so this area is only 20 to 100 ppm of the surface of object 10. On the other hand, when the irregularity area crosses the illumination field of view of the DVS camera 210, each irregularity IRR creates multiple events when crossing the entire height in the transport direction TD. At a given moment, when the irregularity IRR is completely within one field of view FOV and also in one illumination mode, at least a complete contour view of the irregularity IRR is created because most of the changes occur at the edge regions of the corresponding irregularity IRR. Another advantage regarding significantly saving the amount of data to be processed comes from the basic principle of the DVS camera that only responds to changes in flux. While traditional cameras provide images of the entire, complete area of an object or component, including data and information about areas with no contrast, the DVS camera only responds to changes, such that only high-contrast areas like moving edges may contribute to the amount of data to be processed.

[0072] Typically, and as Figure 3 shown in a simplified manner in, the sensor chip 211 is a rectangular chip 211, and its sensor elements are arranged in rows and columns perpendicular to each other. To improve the sampling accuracy of the irregularity IRR, the fact can be utilized that: generally, the transport direction TD of the object 10 and the rows of the sensor elements SE of the DVS camera 210 are not necessarily completely parallel, that is, 0° < ALPHA < 90° applies to the angle ALPHA between the rows or corresponding columns and the transport direction TD. Therefore, multiple events from different rows of the sensor chip 211 can be used to fuse data regarding the feed motion to create a more highly resolved effective irregularity detection, which achieves the sub-pixel level. Subsequently, this can be used to reduce the physical resolution of the design, which means that the DVS camera 210 with a sensor chip 211 of a given size can capture a larger area of the object 10 and still meet the required resolution criteria. This technical concept is basically known and established for traditional cameras, for example, from the pixel displacement imaging cameras used in the past when the number of sensor elements on the chip was limited. By applying the DVS camera 210 to the inspection system 200 introduced herein, it can significantly contribute to optimizing the system parameters because effective sampling obtains valuable gains in terms of optical resolution through 3 or 5 displacement intervals to cover the width of one effective sensor element SE. This factor can be used to improve the optical resolution, thereby reducing the number of DVS cameras 210 still required to perform high-resolution quality inspection tasks. Reducing the inherent optical resolution of the imaging system of the camera 210 also helps to smooth the information, that is, the maximum achievable optical resolution, and thus also reduces error events, such as for the surface of the object 10 with a texture having a lower fill factor or a grid or a staring-like object 10 to be inspected. For the inspection system 200, a zoom lens can be employed to adapt the inherent optical resolution to the characteristics of the object 10 or the surface characteristics of the object 10.

[0073] The benefits achieved by introducing the DVS camera 210 for such inspection tasks come from the much faster event-based imaging capability plus the fact that only changes to an earlier state constitute an event. For the exemplary fabric 10, it can be any irregularity IRR in the texture that locally changes the optical appearance of the surface texture. Such changes may be triggered by irregularity IRRs, such as different spacings between lines, holes, knots, inclusions from other fibers or particles, etc. Due to the fact that the vast majority of the fabric 10 does not have such or other irregularities, it is important not to miss irregularities and potential irregularities in order to conduct a proper quality inspection. During the quality inspection under the stage INSPECT, the irregularity IRR can be graded in terms of grade according to the corresponding severity from non-critical to critical according to the general quality standards in the corresponding inspection domain. For a given length of fabric 10, the specification of the total number of irregularities in the fabric 10 may be sufficient. For example, in the case where the total number of irregularity IRRs is below a certain number, the quality of the object 10 can be graded as "good" or "first class" or similar. If it is below the second higher number, it may be graded as "acceptable" or "second class". If it is exceeded, the entire object 10 may be rejected. The same of course applies to other objects 10 to be inspected in both the first and second embodiments described below. The quality standards and quantities may be based on manual inspection, respectively. Their vision system is different from that of cameras, which can and will detect more potential defects. Therefore, a good match between manual inspection and inspection based on DVS cameras 210 is crucial. DVS cameras 210 are very suitable for this task because their operating concept is to a certain extent adapted to the human visual system, because only deviations from the actual normal state are detected as events and can subsequently be classified.

[0074] To create such quality information of the inspected object 10, the inspection system 200 may include a plurality of DVS cameras 210, each configured as described above. The data IMA from the different DVS cameras may be aggregated and fused, for example to show a full digital view or digital twin of the object 10 as a kind of digital twin information, and all known computer and machine vision-based data evaluation strategies may be applied to the event data or to conventional camera views derived from the event data, or by employing a type of DVS camera 210 of the "DAVIS" variety.

[0075] Another advantage of the DVS camera 210 based inspection system 200 is the much higher speed of the event output from the DVS camera 210, which is 1,000 or even 10,000 times faster than conventional cameras or more. This facilitates marking or labeling the inspected object 10 directly at the location of the potential irregularity or very close to the location of the potential irregularity without reducing the conveying speed of the inspected product. This is not possible with today's conventional camera based inspection systems in a satisfactory operating range. Typically, the conveyor must be stopped and must be rewound to the potential irregularity location or coordinates of the conveying system. For example, for fabric 10, the prior art is to place a physical mark, such as a sticker, at one edge of the fabric 10. On sheet metal, some markings are also used by spraying colored spots at the corresponding location or also at the edge.

[0076] If the inspection should be carried out in color mode, the lateral separation areas of the illumination can be multiplied so that there are separate areas for each main color. The powerful LED illumination can even be used to switch the camera sensitivity mode adapted to the features and properties of the object, for example to analyze a color print on a panel, making it visible in one color but not in another.

[0077] With regard to the method METH itself, the phase ADOPT can be carried out immediately before the execution of the phase INSPECT, each time an object 10 has to be inspected. However, the phase ADOPT can also be carried out at an earlier time, for example in the context of a general installation or commissioning, for example when an object 10 to be inspected is produced for the first time, or in the context of training a potential application AI system. The data determined in the phase ADOPT and the threshold value THRES can then be stored, respectively, in a memory 223 of the evaluation system 220, for example. As soon as an object 10 of the same kind is produced again in a subsequent phase, the stored threshold value THRES can be used again, instead of executing the phase ADOPT again.

[0078] exist Figure 1 In the exemplary situation shown and assumed in the subsequent explanation in the context of the first and second embodiments, it is foreseen that the object 10 is moving, while the DVS camera 210 is stationary. However, the method presented herein requires only a relative movement between the object 10 to be inspected and the DVS camera 210 with a relative velocity VEL. Therefore, the above explanation can be applied without further difficulties to the case of a stationary object and a moving camera, as well as to the case where both the object 10 and the DVS camera 210 are moving.

[0079] Although the present invention has been described above with reference to various embodiments, it should be understood that many changes and modifications may be made to the described embodiments. Therefore, the foregoing description is intended to be considered illustrative rather than restrictive, and it should be understood that all equivalents and / or combinations of the embodiments are intended to be included in this description. Therefore, the present invention is not limited to the above-described embodiments, and those skilled in the art may derive changes without departing from the scope of the invention.

Claims

1. A method METH for performing a camera (210)-based inspection of an object (10) to detect the presence of irregularities IRR of the object (10), wherein the object (10) and the camera (210) are moved relative to each other in a conveying direction TD during the inspection, - wherein the camera (210) is a DVS camera, -- comprising a sensor chip (211) having a plurality of sensor elements SE(j), and - configured to generate, for each sensor element SE(j), individually an event signal SIG(j) from a group GRP of event types including at least positive events and negative events, if the light flux FX impinging on the corresponding sensor element SE(j) changes by more than a given threshold value THRES, and - wherein the method METH comprises an irregularity detection phase INSPECT during which an irregularity IRR of the object (10) is considered to be present if one or more event signals SIG(j) generated by the DVS camera (210) during a relative movement of a portion of the object (10) through the field of view FOV of the DVS camera (210) satisfies a predefined condition.

2. The method according to claim 1, wherein: In case one or more event signals SIG(j) are generated for one or more sensor elements SE(j), an irregularity is considered to be present.

3. The method according to any one of claims 1 to 2, wherein: An irregularity is considered to be present if the constellation of those sensor elements SE(j) generating respective event signals SIG(j) during the irregularity detection phase INSPECT forms a pattern in time and / or in space that corresponds to or deviates from a predefined reference pattern.

4. The method according to claim 3, wherein: The reference pattern is a pattern in space, wherein the reference pattern is a spatial constellation of sensor elements SE(j) that instantaneously generate an event signal SIG(j) caused by a certain part of the object (10) instantaneously located in the field of view FOV when no irregularity of the object (10) exists, wherein an irregularity IRR is considered to exist when the constellation of one or more sensor elements SE(j) that instantaneously generate the corresponding event signal SIG(j) forms a pattern in space that deviates from a predetermined reference pattern.

5. The method according to claim 4, wherein: The method METH comprises an selection phase ADOPT performed before the irregularity detection phase INSPECT, wherein in the selection phase ADOPT at least one spatial reference pattern is automatically determined, wherein preferably, - the DVS camera (210) captures the object (10) during its relative movement through the field of view FOV to generate an event signal SIG(j) generating at least one spatial constellation of sensor elements SE(j), - the event signal SIG(j) thus generated generating at least one spatial constellation of sensor elements SE(j) being used as at least one spatial reference pattern.

6. A method according to any one of claims 1 to 5, wherein an irregularity IRR is considered to exist when, during a given time span dTl, for the same sensor element SE(j), the generated event signal SIG(j) changes from a first event type of the group GRP to a different second event type of the group GRP.

7. A method according to any one of claims 1 to 6, wherein an irregularity IRR is considered to exist when, during a time span dT2, event signals SIG(j1), SIG(j2), SIG(j3) of the same event type are generated for a plurality of N>2 different sensor elements SE(j1), SE(j2), SE(j3) at different time points t1, t2, t3, wherein the different sensor elements SE(j1), SE(j2), SE(j3) are located on a substantially straight line.

8. The method according to any one of claims 1 to 7, wherein an irregular IRR is considered to exist if: - for the first sensor element SE(j1), an event signal SIG(j1) of the first event type of the group GRP is generated and substantially at the same time t0, For a second sensor element SE(j2) located adjacent to the first sensor element SE(j1), an event signal SIG(j2) of a different second event type of the group GRP is generated.

9. The method according to any one of claims 1 to 8, wherein the method METH is applied to an object (10) whose surface is substantially uniform, wherein a given threshold THRES is set based on the roughness of the texture at the surface.

10. The method according to any one of claims 1 to 9, wherein the method comprises an activation phase ADOPT performed before the irregularity detection phase INSPECT, wherein in the activation phase ADOPT a given threshold THRES is automatically set, wherein: Preferably, - varying the threshold value THRES until the relation REL=f(NGEN, REFPARA) depending on the number NGEN of generated event signals associated with the reference parameter REFPARA reaches the expected value X, - When this relationship REL reaches the expected value X, the instantaneous threshold value THRES is selected as the set threshold value THRES.

11. The method according to any one of claims 1 to 10, wherein different methods of illuminating the object (10) are applied to different areas (Z1, Z2, Z3) of the field of view FOV of the DVS camera (210), wherein: The different illumination zones ( Z1 , Z2 , Z3 ) of the field of view FOV are arranged one behind the other in the transport direction TD.

12. The method according to claim 11, wherein: An irregularity is considered to be present if, at a first point in time t1, a first event signal SIG1(j1) of a first event type of the group GRP is generated for a first sensor element SE(j1) in a first (Z1) of the different illumination zones (Z1, Z2, Z3), and if, at a second point in time t2, a second event signal SIG2(j2) of an event type corresponding to the first event type is generated for a second sensor element SE(j2), and if: a second sensor element SE(j2) is located in a second (Z2, Z3) of the different illumination zones (Z1, Z2, Z3), and - The second sensor element SE(j2) corresponds to the first sensor element SE(j1), where sensor elements in different regions (Z1, Z2, Z3) correspond to each other if the mutual distance DIST between them in the transport direction TD is substantially proportional to the distance by which a point on the surface of the object (10) moves relative to the field of view FOV between time points t1 and t2.

13. A system (200) for camera-based inspection of an object (10) to detect the presence of an irregularity IRR of the object (10) when the object (10) and the camera (210) move relative to each other in the transport direction TD during inspection, comprising a camera (210) implemented as a DVS camera (210) and an evaluation system (220) configured to perform the method according to any one of claims 1 to 12.

14. The system (200) according to claim 13, wherein the DVS camera (210) is arranged such that the rows of the sensor elements SE(j) of its sensor chip (211) are aligned at an angle of 0° < ALPHA < 90°, preferably ALPHA < 10°, relative to the transport direction TD.

15. The system (200) according to any one of claims 13 to 14, wherein the system (200) comprises an illumination system having a plurality of light sources (231, 232, 233) for illuminating the field of view FOV, wherein the different light sources (231, 232, 233) of the illumination system illuminate different regions (Z1, Z2, Z3) of the field of view FOV and / or illuminate the same region of the field of view FOV at different angles of incidence on the object (10).

Images