System and method for optically inspecting objects

Abstract

The present invention relates to a system (100) for the optical inspection of objects (102), comprising: one or more detection units (104) for, in particular, continuous optical detection of at least one object (102) to be inspected in a measuring range (106) and for providing image information of the detected object (102); and a computing unit (110) which is designed and programmed to inspect the detected object (102) using the image information provided based on one or more, in particular positive, reference objects and to provide a corresponding inspection result about the object (102) in question. The invention also relates to a method for the optical inspection of objects (102).

Description

[0001] System and method for optical inspection of objects

[0002] The present invention relates to a system and a method for optical inspection of objects.

[0003] Increasing demands on already high quality standards, such as those required by the automotive industry, are increasingly challenging the detection of defects in raw materials, intermediates, and / or finished products using conventional measurement and testing technology. In addition, there are requirements such as 100% control and traceability of raw materials, intermediates, and / or finished products.

[0004] There is therefore a particular need to simultaneously analyze various parameters and quality characteristics in individual tests (e.g., test bench) and / or during ongoing production (e.g., inline). This requires maximum speeds, maximum data security, and minimal latency. Furthermore, standard solutions for (optical) quality assurance cannot always be adapted or applied to the often individually configured industrial systems for processing and / or manufacturing raw materials, intermediate products, and / or finished products, particularly with regard to lighting conditions and / or viewing angles.

[0005] The present invention is therefore based on the object of solving the aforementioned problems, in particular of providing a system and a method for the optical inspection of objects, which in particular enables a reliable and efficient inspection of objects.

[0006] This object is achieved according to the invention by the features of the independent system claim and by the features of the independent method claim. Advantageous developments of the invention are described in the respective dependent claims.

[0007] The system according to the invention for the optical inspection of objects comprises: one or more detection units for, in particular, the continuous optical detection of at least one object to be inspected in a measuring area and for providing image information of the detected object; and a computing unit which is designed and programmed to inspect the detected object using the provided image information based on one or more, in particular positive, reference objects and to provide a corresponding inspection result for the object in question. The invention is based in particular on the idea that the computing unit inspects or can inspect objects, in particular purely on the basis of positive reference objects, for example (actual) good examples or (actual) good patterns. In particular, at least part of the computing unit can inspect purely on the basis of said positive reference objects orthe good examples can be or are trained to do so. This part of the computing unit can, in particular, be configured as an artificial neural network. In other words, this can mean that negative reference objects or bad examples are not required for an inspection. The system can therefore be or be operated efficiently. In particular, the efficiency of the system can be increased because an inspection can be carried out solely on the basis of the said positive reference objects and, for example, a comparison with numerous negative reference objects is thus avoided. It should be understood that the positive reference objects are not artificially generated, but that the reference objects are, in particular, only generated or can be generated by recording original good objects.However, it should be understood that, additionally or alternatively, the positive reference objects can be recorded and additionally generated artificially, and that, additionally or alternatively, negative reference objects can optionally be provided, which can be generated artificially. This makes it possible to provide a flexibly configurable system. Furthermore, the system can be commissioned quickly, since, in particular, an inspection of objects can already be carried out with a positive reference object. In other words, the system can be trained using a positive reference object, particularly with regard to the object features to be inspected. Commissioning therefore requires (almost) no specialist knowledge of the object to be inspected, since the object features can be trained or acquired using the positive reference object according to the described “Teach & Go” principle.can be trained particularly by means of the artificial neural network.

[0008] It can be provided that an object can be any starting material, intermediate product and / or product. In particular, an object can be a plate-shaped object, a cylindrical object, a tubular object (e.g. with a circular or oval cross-section) and / or an elongated object. It is further conceivable that the object can be angled and / or curved. It is further conceivable that an object can have a symmetrical or an asymmetrical cross-section. It should also be understood that the object can also have a combination of the aforementioned exemplary lists. For example, an elongated object can comprise at least one of the following: cables; fiber optics; wires; metal, wood and / or plastic profiles; hoses; ropes; yarns; chains; drills; threaded rods; screws; nails; and / or pins.Furthermore, it should be understood that an elongated object can also comprise several of the aforementioned elements, such as two or more interconnected, in particular twisted and / or tangled cables and / or wires. For example, a plate-shaped object can comprise a pastry, e.g., cookies. For example, a cylindrical object can comprise a wooden profile, e.g., tree trunks. It should be understood that the above lists are purely exemplary and that the object can be designed additionally or alternatively.

[0009] It can be provided that the computing unit has a locator module which is designed and programmed to determine an outer contour of the detected object in the image information and, optionally, to delete information content in the image information that can be assigned to an area outside the outer contour of the detected object. The locator module can be configured, in particular, as a pre-trained or already learned artificial neural network. Because the outer contour can be determined using the locator module, the amount of information that must be processed by the computing unit can be reduced. This can increase the efficiency of the computing unit and thus of the system. Furthermore, since the unnecessary information can be deleted, the computing unit can be operated more economically, particularly with regard to storage capacity.Furthermore, since the outer contour can be determined, the locator module can additionally or alternatively serve to optically eliminate movements, e.g. vibrations, of the detected object, which can also lead to an improvement in the performance of the system.

[0010] It can be provided that the computing unit has one or more recognition modules which are designed and programmed to analyze the detected object for correspondence with and / or deviations from one or more, in particular positive, reference objects based on the image information and to provide a corresponding measured value. The one or more recognition modules can each be configured as a combination of artificial neural networks and conventional algorithms. Since the detected object can be analyzed on the basis of positive reference objects by means of the one or more recognition modules and therefore only a correspondence and / or a deviation from these needs to be analyzed, the system can be operated or be operable efficiently.In particular, the efficiency of the system can be increased because an analysis can be carried out only on the basis of the said positive reference objects and, for example, a comparison with numerous negative reference objects is thus avoided.

[0011] It can be provided that at least one recognition module is designed and programmed for symbol recognition analysis, in particular text recognition analysis. For example, the recognition module for symbol recognition analysis can be designed and programmed as an OCR module, in particular to be able to analyze imprints (e.g., markings, codes, lettering, and / or serial numbers) on the objects in question. It can be provided that the recognition module for symbol recognition analysis is pre-trained or taught accordingly.

[0012] Additionally or alternatively, it can be provided that at least one recognition module is designed and programmed for geometric measurement analysis. Using the recognition module for geometric measurement analysis, for example, a specific dimension (e.g., a diameter) can be calculated at defined points within the image information of the respective objects, in particular, it can be calculated quantitatively.

[0013] Additionally or alternatively, it can be provided that at least one recognition module is designed and programmed for color analysis. In other words, the recognition module for color analysis can be configured to monitor the color of the objects in question. For example, a color deviation of the objects in question can be calculated using the recognition module for color analysis.

[0014] Additionally or alternatively, it can be provided that at least one detection module is designed and programmed for surface analysis and / or general defect detection. In other words, it can be provided that at least one detection module is designed and programmed for anomaly detection. For example, using the detection module for surface analysis and / or defect detection and / or anomaly detection, breaks, cracks, incisions, holes, dents, pimples, foreign particles, and / or air bubbles in and / or on the objects can be detected.

[0015] It can be provided that the computing unit has an evaluator module which is designed and programmed to determine whether the analyzed measured value lies within predetermined limit values, wherein the corresponding inspection result for the object can be generated on the basis of this determination. The evaluator module can therefore classify the measured values ​​and subsequently provide the corresponding inspection result. It can be provided that the limit values ​​in the computing unit can be flexibly changed by a user, in particular depending on an object to be inspected and / or a quality expectation for an object to be inspected. The limit values ​​can represent a tolerance range of an acceptable quality deficit of the inspected objects with regard to the respective measured value category of the respective detection modules.

[0016] It can be provided that the inspection result can be made available, in particular, to an output unit of the system for output to the user and / or to an interface for transmission to an additional system unit and / or a unit assigned or assignable to the system. An output unit can be, for example, a display unit, e.g. a screen, and / or an alarm unit, e.g. an acoustic and / or visual alarm unit. The additional system unit and / or the unit assigned or assignable to the system can be configured, for example, as a big data unit and / or as a cloud. Additionally or alternatively, the inspection result can be storable in a storage unit of the system. Additionally or alternatively, the interface can be configured as a programming interface (API) or an industrial interface (e.g. Profinet).

[0017] It can be provided that the computing unit has a trainer module which is designed and programmed to generate a first and / or at least one further, in particular positive, reference object, in particular for subsequent inspections, on the basis of the image information of the detected object, if the detected object substantially corresponds to a target state. Additionally or alternatively, it can be provided that the trainer module is designed and programmed to artificially generate potential error characteristics, preferably for subsequent inspections, in particular on the basis of the image information of the detected object. In particular, the trainer module can access the image information generated by the detection units in order to generate image information or image material for positive reference objects, if the detected object substantially corresponds to a target state.The target state can be defined and / or specified by a user in the computing unit. It can be provided that the target state has a tolerance range. In other words, the positive reference objects or their image information can only or predominantly comprise image information from positive reference objects, which consequently have no or only a few production errors / deviations within an acceptable tolerance range. If production errors / deviations are included, they must be significantly in the minority; otherwise, they are defined as permissible production characteristics. It can be provided that the computing unit has one or more training modules corresponding to the one or more recognition modules, which are designed and programmed to train the one or more recognition modules based on at least one, in particular positive, reference object.In other words, a training module can be assigned to each recognition module. The one or more training modules can each be configured as a combination of artificial neural networks and conventional algorithms. For example, it can be provided that at least one training module is designed and programmed for symbol recognition analysis, in particular text recognition analysis. Additionally or alternatively, it can be provided that at least one training module is designed and programmed for geometric measurement analysis. Additionally or alternatively, it can be provided that at least one training module is designed and programmed for color analysis. Additionally or alternatively, it can be provided that at least one training module is designed and programmed for surface analysis. By the training modules training the recognition modules orBy training on the basis of positive reference objects, the reliability and efficiency of the analyses of the recognition modules can be continuously increased depending on the number of positive reference objects by means of which the recognition modules can be trained.

[0018] For example, it may be provided that one or more or each recognition module, in particular designed as a (e.g., deep) neural network, can be trained, in particular by means of a corresponding training module, for example, so that one or more probability values ​​can be provided or output for each image area and / or image pixel. In this case, it may be provided that the provided values ​​or output values ​​can be set with regard to a defect probability, color deviations, shape deviations, and / or other desired or undesired optical features.

[0019] It can be provided that each detection unit has a trigger input by means of which an analog trigger signal can be received, wherein each detection unit is configured to start and / or synchronize optical detection, in particular with the computing unit, as soon as the analog trigger signal is received. This can ensure that optical detection begins reliably, in particular during an inline inspection, for example in a production plant. Furthermore, it can be provided that by means of the trigger input, for example by means of the trigger signal received via it, a lighting system assigned or assignable to the (e.g. each) detection unit can be started and / or synchronized. This can reduce and / or optimize the corresponding power and / or cooling requirements.It can be provided that the measuring area, in which the at least one object to be inspected can be optically detected by means of the one or more detection units, is arranged in an at least partially or completely closed space, in particular a measuring chamber, of the system. Additionally or alternatively, it can be provided that the measuring area, in which the at least one object to be inspected can be optically detected by means of the one or more detection units, is arranged in an open space. Additionally or alternatively, it can be provided that the measuring area is arranged at or on a conveyor device, e.g. a conveyor belt. Additionally or alternatively, it can be provided that a conveyor device, e.g. an unwinding device and / or a winding device, is arranged upstream and / or downstream of the measuring area.For example, this can be the case with elongated and / or continuous objects, such as a cable or extruded objects. In other words, an object to be inspected can be conveyed by unwinding and rewinding through the measuring area and optionally through a production system in which the conveyor device can be arranged. It should be understood that the one or more detection units can be arranged or can be arranged in a suitable manner with respect to the measuring area for optically detecting objects in the measuring area.

[0020] For example, the system may comprise a measuring chamber comprising an inlet opening for receiving the object to be inspected into the measuring chamber and an outlet opening for discharging the inspected object from the measuring chamber, wherein said measuring region is arranged between the inlet opening and the outlet opening in the measuring chamber.

[0021] It can be provided that the one or more detection units are configured for, in particular, continuous optical detection of the object in the measuring area of ​​the measuring chamber.

[0022] It can be provided that one or more or each detection unit is configured to coaxially illuminate the object, in particular during continuous optical detection of the object (e.g. in the measuring chamber). This can provide a bright field that can ensure uniform illumination and prevent reflections. This lighting technology can be particularly suitable for reflective surfaces of all kinds, such as metals, glass, smooth or polished surfaces, etc. Due to a homogeneous lighting atmosphere, coaxial lighting can be particularly suitable for defect inspection on uneven surfaces. For example, details can be made visible despite creasing and / or curvature. The reliability of the system can thus be improved. It should be understood that alternative lighting methods are also conceivable, such as by means of direct illumination, e.g. using a directional lighting device.

[0023] It can be provided that one or more or each detection unit comprises a camera for optically detecting the object and a light source for illuminating the object.

[0024] Each camera can be provided with a lens. Generally, any type of lens can be provided. For example, the lens can be a telecentric lens and / or a conventional lens, for example, an optically corrected lens, e.g., a conventionally distorted lens, in particular with at least one adjustable lens.

[0025] One challenge here can be the shallower depth of field of a conventional lens compared to a telecentric lens. Since an object to be captured and / or inspected may move at least slightly, e.g., vibrate slightly, the image may become blurred if the object to be captured and / or inspected leaves the focus area or approaches the edge of the focus area.

[0026] It is conceivable to motorize the classic, distorted lens with one or more adjustable lenses via a stepper motor so that the focus can be adjusted automatically (especially electronically). This method can be applied to the zoom and / or aperture. Software (e.g. an AI-supported algorithm) can ensure that the focus is always optimal and the image remains sharp. Using a limit switch, the software can determine which lens settings have been used. These can also be recalled later, for example if the object to be captured and / or inspected is changed and the size and / or dimension and / or diameter of the object to be captured and / or inspected or its position changes.

[0027] The aperture can be set so that it is closed as far as possible (small f-number). This increases the depth of field. At the same time, it ensures that the aperture setting does not block out too much light.

[0028] The advantages of such a classic, corrected lens compared to a telecentric lens include, among others, lower costs, less loss of light power (telecentric lenses have many lenses and mirrors that consume light power), as well as the possibility of further measurement (in particular size and / or dimension and / or diameter of the object to be captured and / or inspected), in particular via software calculation.

[0029] The distance of the object to be captured and / or inspected from the lens can be calculated from the zoom and / or focus positions, allowing the size and / or dimensions and / or diameter of the object to be determined. The built-in limit switch can be used to initially calibrate the lens settings (zoom and / or focus positions). This calculation can be performed independently with all available cameras to increase accuracy.

[0030] It can be provided that one or more or each detection unit comprises a camera for optically detecting the object, a light source for illuminating the object and a beam splitter for deflecting emitted light from the light source, wherein the camera, the light source and the beam splitter are arranged relative to one another such that a central optical axis of the camera and light rays that can be emitted by the light source run parallel and / or coaxially to one another, wherein in particular the camera and the light source are arranged substantially at right angles to one another and the beam splitter is arranged at an angle of substantially 45° to the central optical axis of the camera and a light emitting direction of the light source.By designing the system with a beam splitter, particularly a semi-transparent mirror, it is possible to ensure that the camera sees the object directly through the mirror glass of the beam splitter, which is transparent on one side, rather than "through" the light source, as is the case with direct illumination (e.g., with a ring light). This can prevent the camera from being "blinded," and image information about the captured object from being lost due to overexposure. Consequently, the system can be operated more reliably and efficiently.

[0031] It can be provided that one or more, or each, detection units have a diffuser, which can be arranged between the light source and the beam splitter. This can provide a highly diffuse bright field, which can ensure even more uniform illumination and even better prevent reflections.

[0032] Each light source may comprise at least one circuit board and a plurality of LEDs arranged in a regular two-dimensional pattern on the circuit board. This allows for uniform and reliable illumination. Each light source may have a power range of substantially 20 to 60 watts or 70 watts, preferably substantially 24 to 48 watts.

[0033] Alternatively, it's conceivable that each light source has a power range of essentially 100-150 watts. The exposure time of the camera sensors can then be reduced to avoid image distortion, especially for capturing higher speeds of moving objects.

[0034] It can further be provided that one or more, e.g., every, light source(s) can be operated in a pulsed mode. This can achieve powerful illumination for detecting even the smallest defects, which can also enable very fast image captures or very short exposure times by the camera. Furthermore, in the case of a measuring chamber, the enclosure cannot be completely closed due to the short, adjustable exposure times of the cameras. In other words, the entrance and exit openings of the measuring chamber do not have to be completely light-tight. The high luminosity of the installed light source can compensate for any residual ambient light that enters.

[0035] It can be provided that each light source has a color rendering index range of substantially 92 to 98, preferably substantially 94 to 96, in particular substantially 95. This can enable a reliable color analysis.

[0036] For example, it can be provided that the system comprises several, e.g. three, detection units, wherein the detection units are arranged at regular intervals from one another and in particular radially around the measuring area. Additionally or alternatively, it can be provided that several detection units are arranged at regular intervals from one another and in particular longitudinally along the measuring area. For example, the system can comprise six detection units, wherein three detection units forming a group are arranged radially around the measuring area and the respective groups are arranged longitudinally along the measuring area. In a sense, this can be seen as a 2x3 radial arrangement. This allows, for example, elongated objects, e.g. extrusion products or cables, to be optically detected completely and reliably.

[0037] It is also conceivable for the system to comprise one or two detection units. This may be the case, for example, if a complete analysis / inspection is not required and / or if a geometry and / or other property of an object to be inspected requires only one or two detection units for complete analysis / inspection.

[0038] In particular, it may be conceivable that the system comprises a detection unit advantageously for the inspection of one-dimensional features, in particular prints or a print result, of an object to be inspected.

[0039] Additionally or alternatively, it may be conceivable for the system to include a detection unit, preferably for the inspection of partially or fully transparent objects to be inspected. Especially for fully transparent objects to be inspected, one detection unit may be sufficient for complete analysis / inspection.

[0040] In particular, it is conceivable for the system to comprise two detection units, advantageously for objects to be inspected with a suitable geometry, such as flat profiles. The detection units can be oriented, for example, at 180° and / or opposite each other or at 90° and / or substantially perpendicular to each other. For example, one detection unit can capture a side view of the object to be inspected and / or one, for example the other, detection unit can capture a top or bottom view of the object to be inspected.

[0041] In particular, it is conceivable for the system to be configured to generate 3D image data, in particular based on the principle of triangulation using the light-section method. The system can then preferably further comprise at least one projection device that projects at least one light pattern, for example parallel black / white line pairs or points, onto the object to be inspected at a known angle. The projection device can comprise an illumination source, for example structured illumination or a laser.

[0042] The one or more detection units are arranged at a known angle to the projection device and / or illumination. In particular, the one or more detection units record the light pattern deformed by the surface shape of the object to be inspected, for example, a striped pattern or dot pattern. The system thus makes it possible to visualize the surface, e.g., curvatures or roundness of the object to be inspected. Possible dents or bumps can also be made visible. With a sufficiently high refresh rate, the projection can be performed by flashing in the same section of the one or more detection units in which the optical inspection is or is taking place, thus utilizing the full field of view.Alternatively, the projection can be limited to only a part of the image section, so that optical defect inspection and 3D measurement can be carried out simultaneously, i.e. during the same image capture.

[0043] It should be understood that to increase production speeds (for example, at a cable and / or object speed), one or more or each of the detection units can be used, which can be configured for a higher frame rate and / or which can be operated with an enlarged image section. Additionally or alternatively, it may be conceivable to increase the number of detection units, in particular to duplicate them, and to synchronize the respective detection units, for example, in pairs. For example, a higher measurement speed can be achieved by arranging the duplicated detection units one behind the other.

[0044] Furthermore, the present invention provides a method according to the invention for the optical inspection of objects.

[0045] It should be understood that the method according to the invention can be carried out by means of the system described herein.

[0046] It should further be understood that any structural and / or functional features and / or properties and / or advantages that are and / or will be described in connection with the system for optical inspection of objects according to the invention may also be part of and / or attributable to said method.

[0047] The method according to the invention for the optical inspection of objects comprises: providing an object to be inspected in a measuring area; optically detecting the object to be inspected by means of one or more detection units, in particular in a continuous manner; providing image information of the detected object by means of the one or more detection units; inspecting the detected object using the provided image information based on one or more, in particular positive, reference objects by means of a computing unit; and providing a corresponding inspection result for the object by means of the computing unit. The method can further comprise: projecting light, in particular light patterns, for example parallel black / white line pairs or points, at a known angle onto the object to be inspected by a projection device.The projection device may comprise illumination, for example, structured illumination or laser. Furthermore, the method may include detecting the surface shape of the object to be inspected using deformed light patterns, for example, deformed stripe patterns or dot patterns.

[0048] It can be provided that the method comprises: determining an outer contour of the detected object in the image information by means of a locator module of the computing unit; and, optionally, deleting an information content that can be assigned to an area outside the outer contour of the detected object in the image information by means of the locator module.

[0049] It can be provided that the inspection comprises: analyzing the detected object for correspondence with and / or deviations from one or more, in particular positive, reference objects based on the image information by means of one or more recognition modules of the computing unit; and providing a corresponding measured value by means of the one or more recognition modules.

[0050] It can be provided that at least one recognition module is designed and programmed for symbol recognition analysis, in particular text recognition analysis, and / or at least one recognition module is designed and programmed for geometric measurement analysis, and / or at least one recognition module is designed and programmed for color analysis, and / or at least one recognition module is designed and programmed for surface analysis.

[0051] It can be provided that the inspection comprises: determining by means of an evaluator module of the computing unit whether the analyzed measured value lies within predetermined limit values, wherein the corresponding inspection result about the object can be generated or is generated on the basis of this determination.

[0052] It can be provided that the method comprises: generating a first and / or at least one further, in particular positive, reference object, in particular for subsequent inspections, on the basis of the image information of the detected object by means of a trainer module of the computing unit, if the detected object substantially corresponds to a target state.

[0053] It can be provided that the method comprises: training the one or more recognition modules on the basis of at least one, in particular positive, reference object by means of one or more training modules of the computing unit.

[0054] Further preferred features and / or advantages of the present invention are the subject of the following description and the drawings of exemplary embodiments.

[0055] The figures show schematically:

[0056] Fig. 1 shows a system for optical inspection of objects according to a first embodiment;

[0057] Fig. 2 shows a system for the optical inspection of objects according to a second embodiment;

[0058] Fig. 3 is a detailed view of part of the system of Fig. 2;

[0059] Fig. 4 is a functional view of part of the system of Fig. 2;

[0060] Fig. 5 is a flowchart illustrating the operation of the system of Fig. 1 and the system of Fig. 2, or a method for optically inspecting objects according to a third embodiment; and

[0061] Fig. 6 is a side view of the detail view of Fig. 3.

[0062] Identical or functionally equivalent elements are provided with the same reference numerals in all figures.

[0063] Referring to Fig. 1 in conjunction with Fig. 5, a system 100 according to the invention for the optical inspection of objects 102 is schematically illustrated according to a first exemplary embodiment. The system 100 for the optical inspection of objects 102 has a plurality of detection units 104, in this case three detection units 104. It can be provided that the number of detection units 104 used depends on the size or dimensions of an object to be inspected. For example, four or more detection units 104 can also be included for larger diameters and / or comparable dimensional parameters.

[0064] By means of each detection unit 104, at least one object 102 to be inspected can be optically detected, in particular continuously optically detected, in a measuring area 106.

[0065] After an optical detection, corresponding image information of the detected object 102 can be provided by each detection unit 104, in particular for further utilization or processing.

[0066] In other words, each detection unit 104 is configured for, in particular, continuous optical detection of at least one object 102 to be inspected in a measuring area 106 and for providing image information of the detected object 102.

[0067] An optical single detection, in other words a discontinuous detection, may be conceivable additionally or alternatively.

[0068] In the present embodiment, the objects to be inspected are cylindrical objects, e.g., tree trunks, or plate-shaped objects, e.g., cookies. Any other object types are also conceivable.

[0069] The measuring area 106 is arranged in the present case on and / or on a conveyor device, here a conveyor belt 108.

[0070] The detection units 104 are suitably arranged and aligned with respect to the measuring area 106, e.g., as in the present case, above the conveyor belt 108, in order to be able to optically detect the objects 102 in the measuring area 106. Any suitable arrangement of the detection units 104 is conceivable.

[0071] Each detection unit 104 includes a camera 112 for optically capturing the object 102 and a light source 114 for illuminating the object 102 (not shown in Fig. 1).

[0072] Each camera 112 is provided with a lens 156. The lens 156 can be a telecentric and / or optically corrected, e.g., classically distorted, lens.

[0073] Each light source 114 comprises at least one circuit board and a plurality of LEDs arranged in a regular two-dimensional pattern on the circuit board.

[0074] In the present embodiment, the LEDs are arranged at least partially in a ring-like manner around the camera lens opening to provide a ring light (not shown in Fig. 1). Additionally or alternatively, it may be conceivable for the LEDs to be arranged at least partially or completely relative to the camera lens opening to provide a coaxial light (see, for example, Fig. 4).

[0075] Each light source 114 has a power range of substantially 20 to 70 watts, preferably substantially 24 to 48 watts.

[0076] Alternatively, it is conceivable that each light source 114 has a power range of essentially 100-150 watts.

[0077] Each light source 114 has a color rendering index range of substantially 92 to 98, preferably substantially 94 to 96, more preferably substantially 95 or 96.

[0078] Not shown in Fig. 1 is that the system 100 can further comprise at least one projection device which projects at least one light pattern, for example parallel black / white line pairs or points, at a known angle onto the object to be inspected. The projection device can comprise an illumination, for example structured illumination or laser. At least one detection unit 104 can be arranged at a known angle to the projection device and / or illumination. The system 100 thus enables the surface, e.g. curvatures or roundnesses, of the object to be inspected to be made visible and / or analyzed.

[0079] Furthermore, the system 100 has a computing unit 110, which is operatively connected to the detection units 104, in particular electrically and / or signal-wise connected.

[0080] The computing unit 110 is configured and programmed to inspect the captured object 102 based on the provided image information. This inspection is based on one or more positive reference objects.

[0081] A positive reference object is a good example and / or a good sample of the objects 102 to be inspected, in particular from which the inspection is based as the ideal state.

[0082] A corresponding inspection result for the object 102 in question can be provided by means of the computing unit 110, in particular for further use.

[0083] In other words, the computing unit 110 is designed and programmed to provide a corresponding inspection result about the object 102 in question, in particular for further use.

[0084] With reference to Fig. 5 in conjunction with Fig. 1, the computing unit 110 will now be described in particular:

[0085] The computing unit 110 is configured and programmed to receive image information from the detection units 104 as a data bundle provided with a time stamp and / or to summarize the image information from the detection units 104 as a data bundle and provide it with a time stamp (see field 1 of Fig. 5). This serves, in particular, to reliably assign the image information.

[0086] As shown in Fig. 5, the optical detection by the detection units 104 can optionally be synchronized or synchronized using a hardware trigger. In this case, an analog signal is simultaneously sent to a trigger input of the detection units 104, and the optical detection is synchronized within nanoseconds, although this is not absolutely necessary in the present embodiment.

[0087] In other words, it can be provided that each detection unit 104 has a trigger input by means of which an analog trigger signal can be received, wherein each detection unit 104 is configured to start an optical detection and / or to synchronize it, in particular with the computing unit 110, as soon as the analog trigger signal is received.

[0088] The computing unit 110 has a locator module 116 (see field 3 of Fig. 5). The locator module 116 is configured as a pre-trained or already learned artificial neural network.

[0089] The image information, in particular the bundled image information (cf. field 1 of Fig. 5), can be received by means of the locator module 116.

[0090] Additionally or alternatively, the image information can be provided on a separate stream in order to be retrievable and / or receivable by other processes and / or modules as needed (see field 2 of Fig. 5).

[0091] By means of the locator module 116, an outer contour of the detected object 102 can be determined in the image information.

[0092] An information content that can be assigned to an area outside the outer contour of the detected object 102 can be deleted from the image information by means of the locator module 116.

[0093] In other words, the locator module 116 is designed and programmed to determine an outer contour of the detected object 102 in the image information and to delete an information content in the image information that can be assigned to an area outside the outer contour of the detected object 102.

[0094] Furthermore, the computing unit 110 has several, here four, recognition modules 118, 120, 122, 124 (cf. fields 4 to 7 of Fig. 5).

[0095] The recognition modules 118, 120, 122, 124 are each configured as a combination of artificial neural networks and conventional algorithms.

[0096] The image information from the locator module 116 can be received by the recognition modules 118, 120, 122, 124.

[0097] By means of the recognition modules 118, 120, 122, 124, the detected object 102 can be analyzed based on the image information for agreement with and / or deviations from one or more positive reference objects.

[0098] A corresponding measured value can be provided by the respective recognition modules 118, 120, 122, 124. In other words, each recognition module 118, 120, 122, 124 is configured and programmed to analyze the detected object 102 based on the image information for correspondence with and / or deviations from one or more positive reference objects and to provide a corresponding measured value.

[0099] The detection modules 118, 120, 122, 124 include a first detection module 118, a second detection module 120, a third detection module 122 and a fourth detection module 124.

[0100] The first recognition module 118 is designed and programmed for symbol recognition analysis, in particular text recognition analysis.

[0101] In particular, the first recognition module 118 is designed and programmed as an OCR module to be able to analyze, in particular, imprints (e.g., markings, codes, lettering, and / or serial numbers) on the respective objects 102. The first recognition module 118 is pre-trained or taught accordingly.

[0102] The second recognition module 120 is designed and programmed for geometric measurement analysis.

[0103] By means of the second recognition module 120 for geometric measurement analysis, for example, a concrete dimension (e.g. a diameter) at defined points within the image information of the relevant objects 102 can be calculated, in particular concretely and quantitatively calculated.

[0104] The third recognition module 122 is designed and programmed for color analysis.

[0105] In other words, the third detection module 122 is configured to monitor the color of the objects 102 in question.

[0106] For example, a color deviation of the objects 102 in question can be calculated by means of the third recognition module 122, e.g. by specifying a percentage deviation as a measured value.

[0107] The fourth detection module 124 is designed and programmed for surface analysis. For example, using the fourth detection module 124 for surface analysis, it can detect breaks, cracks, cuts, holes, dents, pimples, foreign particles, and / or air bubbles in and / or on the objects.

[0108] The computing unit 110 further comprises an evaluator module 126 (see field 8 of Fig. 5).

[0109] The measured values ​​determined by means of the detection modules 118, 120, 122, 124 can be received by means of the evaluator module 126.

[0110] The evaluator module 126 can be used to determine whether the analyzed measured value lies within predetermined limit values.

[0111] The inspection result for object 102 can be generated on the basis of this determination.

[0112] In other words, the evaluator module 126 is designed and programmed to determine whether the analyzed measured value lies within predetermined limit values, wherein the corresponding inspection result about the object 102 can be generated on the basis of this determination.

[0113] In other words, the evaluator module 126 can be used to classify the measured values ​​and subsequently provide a corresponding inspection result.

[0114] The limit values ​​are stored in the computing unit 110 and / or are accessible from the computing unit 110.

[0115] The limit values ​​can be flexibly changed by a user, in particular depending on an object 102 to be inspected and / or a quality expectation for an object 102 to be inspected. This can be done, for example, via an input device of the system 100 assigned to the computing unit 110.

[0116] After the inspection result has been generated, it can be made available or made available for further use (see field 9 of Fig. 5).

[0117] A process run, in particular a process run between field 1 to field 9 of Fig. 5, lasts between 3 and 7 ms, in particular essentially 5 ms. An inspection result or the inspection result can be provided, in particular, to an output unit of system 100 for output to the user and / or to an interface for transmission to an additional system unit and / or a unit assigned or assignable to the system.

[0118] A non-exhaustive list of examples can be found in Fig. 5 (see fields 10 to 15 of Fig. 5). An output unit can be, for example, a display unit, e.g. a screen (see field

[0119] 11 of Fig. 5) and / or an alarm unit, e.g., an acoustic and / or visual alarm unit (cf. field 13 of Fig. 5). The additional system unit and / or the unit assigned or assignable to the system 100 can, for example, be a big data unit (cf. field

[0120] 12 of Fig. 5) and / or as a cloud (see field 15 of Fig. 5). Additionally or alternatively, the inspection result can be stored in a storage unit of the system 100 (see field 10 of Fig. 5). Additionally or alternatively, the interface can be configured as a programming interface (e.g., API (e.g., Profinet)) (see field 14 of Fig. 5).

[0121] Furthermore, the computing unit 110 has a trainer module 128 (see field 16 of Fig. 5).

[0122] The image information, in particular the bundled image information (cf. field 1 of Fig. 5), can be received by means of the trainer module 128.

[0123] A first, in particular very first, or further (in particular second, third, etc.), positive reference object for subsequent inspections can be generated on the basis of this image information of the detected object 102 by means of the trainer module 128.

[0124] In particular, these reference objects can be generated by means of the trainer module 128 if the detected object 102 essentially corresponds to a target state.

[0125] In other words, the trainer module 128 is designed and programmed to generate a first and / or at least one further positive reference object, in particular for subsequent inspections, based on the image information of the detected object 102, if the detected object 102 substantially corresponds to a desired state.

[0126] In particular, the trainer module 128 serves to access the image information generated by the detection units 104 in order to generate image information or image material for positive reference objects if the detected object 102 substantially corresponds to a target state. The target state can be defined and / or set by a user in the computing unit 110 (e.g., via an input device of the system 100).

[0127] It can be specified that the target state has a tolerance range. In other words, the positive reference objects or their image information can contain only or predominantly image information from positive reference objects, which consequently exhibit no or only a few production defects / deviations within an acceptable tolerance range. If production defects / deviations are included, they must be significantly in the minority; otherwise, they are defined as permissible production characteristics.

[0128] The positive reference objects can be stored and / or retained in the computing unit 110 and / or accessed for their further use (cf. field 17 of Fig. 5).

[0129] The image information of the reference objects can be transferred to the locator module 116 and / or can be received by the locator module 116 (cf. field 18 of Fig. 5) in order to determine an outer contour of the detected object 102, here the positive reference object, in the image information and to delete an information content that can be assigned to an area outside the outer contour of the detected object 102, here the positive reference object, in the image information.

[0130] Furthermore, the computing unit 110 has several, here four, learning modules 130, 132, 134, 136 (cf. field 19 of Fig. 5; divided separately: cf. fields 20 to 23 in Fig. 5).

[0131] The training modules 130, 132, 134, 136 are each configured as a combination of artificial neural networks and conventional algorithms.

[0132] Each training module 130, 132, 134, 136 is configured and programmed to access and / or utilize resources from: dedicated hardware, in particular GPU and / or FPGA, and / or a remote cloud and / or remote data centers. In particular, the dedicated hardware can be part of the system 100. This serves, in particular, to increase performance.

[0133] The learning modules 130, 132, 134, 136 correspond to the recognition modules 118, 120, 122, 124. That is, the learning modules 130, 132, 134, 136 comprise a first learning module 130, a second learning module 132, a third learning module 134, and a fourth learning module 136.

[0134] By means of the training modules 130, 132, 134, 136, the respective corresponding recognition modules 118, 120, 122, 124 can be trained on the basis of at least one positive reference object.

[0135] In other words, the training modules 130, 132, 134, 136 are designed and programmed to train the recognition modules 118, 120, 122, 124 respectively on the basis of at least one positive reference object.

[0136] Each detection module 118, 120, 122, 124 is assigned to a learning module 130, 132, 134, 136.

[0137] The first training module 130 is assigned to the first recognition module 118 and can train it. In other words, the first training module 130 is designed and programmed for a symbol recognition analysis, in particular text recognition analysis, or for corresponding training.

[0138] The second training module 132 is assigned to the second recognition module 120 and can train it. In other words, the second training module 132 is designed and programmed for geometric measurement analysis or for corresponding training.

[0139] The third training module 134 is assigned to the third recognition module 122 and can train it. In other words, the third training module 134 is designed and programmed for color analysis or for corresponding training.

[0140] The fourth training module 136 is assigned to the fourth recognition module 124 and can train it. In other words, the fourth training module 136 is designed and programmed for a surface analysis or for corresponding training.

[0141] By training the learning modules 130, 132, 134, 136 the recognition modules 118, 120, 122, 124 based on the positive reference objects, the recognition modules 118, 120, 122, 124 are continuously improved and become more reliable in their analysis of image information from detected objects 102. Referring to the system of Fig. 1 in conjunction with Fig. 5, this can be operated in particular as follows:

[0142] A reference object 102 or a good example or a good sample is fed to the measuring area 106 via the conveyor belt 108. It is also conceivable that a reference object 102 or a good example or a good sample can be fed to the measuring area 106 in a freely suspended manner, for example by means of a feed and discharge device spaced apart from the measuring area 106.

[0143] At least one detection unit 104 optically detects the object 102 and generates image information.

[0144] The image information is bundled and provided with a time stamp by means of the computing unit 110 (see field 1 of Fig. 5).

[0145] Based on this image information of the detected object 102, the trainer module 128 generates a first, positive reference object for subsequent inspections, since the detected object 102 is a good example or a good pattern and thus essentially corresponds to a target state (cf. field 16 of Fig. 5).

[0146] The first positive reference object is stored (see field 17 of Fig. 5) and further processed by the locator module 116 (see field 18 of Fig. 5).

[0147] By means of the locator module 116, an outer contour of the detected object 102 is determined in the image information.

[0148] An information content that can be assigned to an area outside the outer contour of the detected object 102 is deleted from the image information by means of the locator module 116.

[0149] The resulting image information is passed to the training modules 130, 132, 134, 136 (see fields 19 to 23 of Fig. 5), which convert the image information into a class model (see field 24 of Fig. 5) for training the recognition modules 118, 120, 122, 124.

[0150] Depending on the detected reference object, the corresponding recognition modules 118, 120, 122, 124 are now trained by the respective training modules 130, 132, 134, 136, so that subsequent inspections can be performed based on at least this first reference object. Any number of reference objects can be trained.

[0151] After training or teaching at least the first reference object, subsequent inspections can be carried out.

[0152] That is, at least one detection unit 104 optically detects the object 102 to be inspected, which is moved via the conveyor belt 108 into the measuring area 106, and generates image information.

[0153] The image information is bundled and provided with a time stamp by means of the computing unit 110 (see field 1 of Fig. 5).

[0154] This image information is further processed by the locator module 116 (see field 3 of Fig. 5).

[0155] By means of the locator module 116, an outer contour of the detected object 102 is determined in the image information.

[0156] An information content that can be assigned to an area outside the outer contour of the detected object 102 is deleted from the image information by means of the locator module 116.

[0157] The resulting image information is passed to the recognition modules 118, 120, 122, 124 (see fields 4 to 7 of Fig. 5), which are already trained with at least the first reference object.

[0158] The recognition modules 118, 120, 122, 124 analyze the detected object 102 based on the image information for agreement with and / or deviations from the positive reference object and provide a respective corresponding measured value.

[0159] The evaluator module 126 then determines whether this analyzed measured value lies within predetermined limit values ​​(see field 8 of Fig. 5).

[0160] In the present example of cookies as objects 102, for example, damaged and / or discolored, e.g., burnt, cookies can be reliably identified. In this case, geometric and / or color measurements would be outside the predetermined limits. In the present example of tree trunks as objects 102, for example, deformed and / or only partially stripped tree trunks can be reliably identified. In this case, geometric and / or color measurements would be outside the predetermined limits.

[0161] The inspection result for object 102 is generated based on this determination.

[0162] After the inspection result has been generated, it can be made available or made available for further use (see field 9 of Fig. 5).

[0163] The inspection result is then provided in particular to an output unit of the system 100 for output to the user and / or to an interface for transmission to an additional system unit and / or a unit assigned or assignable to the system (cf. fields 10 to 15 of Fig. 5) in order to be able to carry out corresponding quality assurance actions (e.g. sorting and / or further processing and / or warning and / or marking, etc.).

[0164] The system 100 is therefore based in particular on the idea that the computing unit 110 inspects or can inspect objects 102, in particular purely on the basis of positive reference objects, for example good examples or good patterns. In particular, at least some of the computing unit 110, i.e., the recognition modules 118, 120, 122, 124, can be trained or are trained to do so purely on the basis of said positive reference objects or good examples. In other words, this means that negative reference objects or bad examples are not required for an inspection. The system 100 can therefore be operated more efficiently. In particular, the efficiency of the system 100 is increased because an inspection can be carried out solely on the basis of said positive reference objects, thus avoiding, for example, a comparison with numerous negative reference objects.Furthermore, commissioning of the system 100 can be carried out quickly, since in particular an inspection of objects 102 can already be carried out with a positive reference object. In other words, the system 100 can be trained with an initial positive reference object, particularly with regard to object features to be inspected. Commissioning therefore requires (almost) no specialist knowledge of the object 102 to be inspected, since the object features can be trained or taught using the described "teach & go" principle based on the positive reference object. Furthermore, by the training modules 130, 132, 134, 136 training or teaching the recognition modules 118, 120, 122, 122training on the basis of positive reference objects, the reliability and efficiency of the analyses of the recognition modules 118, 120, 122, 122 can be continuously increased depending on a set of positive reference objects by means of which the recognition modules 118, 120, 122, 122 can be trained.

[0165] Referring to Figs. 2 to 4 in conjunction with Fig. 5, a system 100 according to the invention for the optical inspection of objects 102 according to a second embodiment is schematically shown.

[0166] The system according to the second embodiment essentially corresponds to the system according to the first embodiment, so that only the differences are described below.

[0167] In the present embodiment, the objects 102 to be inspected are longitudinally extending or endless objects, for example, a cable. Any other object types are also conceivable.

[0168] In other words, the system 100 can be regarded as a cable inspection system, in particular as a cable inspection device.

[0169] Referring to Fig. 2, the system 100 comprises a support device 138, which is designed as a profile frame, a housing 140, a combined display unit and input device in the form of a touch screen 142, and an enclosure 144 for a measuring chamber 146.

[0170] The housing 140, the touch screen 142 and the enclosure 144 are arranged on the support device 138.

[0171] The housing 144 is height-adjustable on the support device 138 by means of a corresponding connection, which is well known in the art.

[0172] The system 100, in particular the carrying device 138, is configured to be mobile, which in the present embodiment can be achieved by means of brakeable and / or lockable rollers.

[0173] The computing unit 110, e.g., in the form of a computer device, is arranged in the housing 140. Furthermore, the housing 140 contains, in particular, a GPU device that is operatively connected to the computing unit 110 and / or forms part of it and / or is assigned to it, a cooling device and / or a fan, a power supply (e.g., power supplies, fuses, cabling), one or more communication modules (e.g., Profibus, Profinet, 4G / 5G routers, etc.), and one or more control elements (e.g., main on / off switch, height adjustment UP / DOWN).

[0174] The housing 144 surrounds a measuring chamber 146 in which the measuring area 106 is arranged, as can be seen in Fig. 3 and / or Fig. 6.

[0175] A compressed air device may also be provided in the housing 144 for air measurement and / or dust protection (not shown in the figures).

[0176] A wall of the measuring chamber 146 is provided with a light-absorbing or light-absorbing coating.

[0177] The measuring chamber 146 has an inlet opening 148 for receiving the object 102 to be inspected into the measuring chamber 146 and an outlet opening 150 for discharging the inspected object 102 from the measuring chamber 146.

[0178] The said measuring area 106 is arranged here between the inlet opening 148 and the outlet opening 150 in the measuring chamber 146.

[0179] An object 102 to be inspected, here the cable, can be passed through the inlet opening 148 and the outlet opening 150 and can thus extend through the measuring chamber 146 in order to be optically detected there.

[0180] The plurality of, here three, detection units 104 are arranged in the housing 144 for the continuous optical detection of the object 102 in the measuring chamber 146, in particular in the measuring area 106.

[0181] The detection units 104 are arranged at regular intervals from one another and in particular radially around the measuring area 106.

[0182] As can be seen in Figs. 3 and 6, the detection units 104 are arranged at 120° intervals from one another around the measuring area 106, in particular around a longitudinal axis of the measuring area 106. In Figs. 3 and 6, the longitudinal axis can substantially coincide with the object 102, ie, the cable.

[0183] Additionally or alternatively, it can be provided that a plurality of detection units 104 are arranged at regular intervals from one another and in particular longitudinally along the measuring area 106. For example, the system 100 can comprise six detection units 104, wherein three detection units 104 forming a group are arranged radially around the measuring area, as already shown in Figs. 3 and 6, and the respective groups are arranged longitudinally along the measuring area 106. In a sense, this can be seen as a 2x3 radial arrangement, i.e., in the case of Figs. 3 and 6, three further detection units 104 would be arranged behind and / or in front of the already visible detection units 104.

[0184] Each detection unit 104 is configured to coaxially illuminate the object 102 during continuous optical detection of the object 102 in the measuring chamber 106.

[0185] Furthermore, as can be seen in Figs. 3 and 6, each detection unit 104 is assigned a surface element 158, in particular a surface element designed as a projection surface.

[0186] In particular, the respectively assigned detection units 104 and surface elements 158 are arranged on opposite sides of the object 102 with respect to the object 102.

[0187] The detection unit 104 and the surface element 158 ​​are in particular aligned with each other so that a viewing axis of the detection unit 104 is directed substantially perpendicularly to the surface element 158.

[0188] The surface element 158 ​​serves as a background for the object 102, for example, to capture the object 102, particularly its contours, more clearly. This can improve accuracy.

[0189] The surface element 158 ​​comprises a light-absorbing, or to a certain extent light-absorbing, material and / or is at least partially formed from such a material. For example, the surface element 158 ​​can be coated with a light-absorbing, or to a certain extent light-absorbing, material. As shown in Fig. 4, each detection unit 104 comprises a camera 112 for optically capturing the object 102, a light source 114 for illuminating the object 102, and a beam splitter 152 for redirecting the light emitted by the light source 114.

[0190] Each detection unit 104 further comprises a diffuser 154 arranged between the light source 114 and the beam splitter 152.

[0191] The diffuser 154 is used to provide a highly diffuse bright field, which can ensure even more uniform illumination and prevent reflections even better.

[0192] The camera 112, the light source 114 and the beam splitter 52 are arranged relative to one another such that a central optical axis of the camera 112 (cf. arrow from camera 112 to object 102 in Fig. 4) and light rays that can be emitted by the light source (cf. other arrows in Fig. 4) run parallel and / or coaxial to one another.

[0193] In particular, the camera 112 and the light source 114 are arranged substantially perpendicular to each other and the beam splitter 152 is arranged at an angle of substantially 45° to the central optical axis of the camera 112 and a light emitting direction of the light source 114.

[0194] The design with the beam splitter 152, in particular a semi-transparent mirror, allows the camera 112 to view the object 102 directly through the mirror glass of the beam splitter 152, which is transparent on one side, and not—as with direct illumination (e.g., with a ring light)—"through" the light source 114. This can prevent the camera 112 from being "blinded" and image information about the captured object 102 from being read out due to overexposure. Consequently, the system 100 can be operated more reliably and efficiently.

[0195] Each light source 114 comprises at least one circuit board and a plurality of LEDs arranged in a regular two-dimensional pattern on the circuit board (see Fig. 4).

[0196] Each light source 114 has a power range of substantially 20 to 60 watts, preferably substantially 24 to 48 watts.

[0197] Alternatively, it is conceivable for each light source 114 to have a power range of essentially 100-150 watts. This can achieve powerful illumination for detecting even the smallest defects, which can also enable very fast image capture or very short exposure times by the camera 112.

[0198] Furthermore, due to the short adjustable exposure times of the cameras 112, the housing 144 or the measuring chamber 146 cannot be completely closed. In other words, the inlet opening 148 and the outlet opening 150 of the measuring chamber 146 do not have to be completely light-tight. Due to the high luminosity of the installed light source 114, incoming residual ambient light can be compensated.

[0199] Furthermore, each light source 114 has a color rendering index range of substantially 92 to 98, preferably substantially 94 to 96, in particular substantially 95. This may enable a reliable color analysis.

[0200] With regard to the computing unit 110, which is arranged in the housing 140, reference is made to the explanations of the first embodiment.

[0201] Referring to the system 100 of Fig. 2 in conjunction with Fig. 5, in particular, it is operable substantially like the system 100 of Fig. 1 in conjunction with Fig. 5, which has already been described.

[0202] In the case of the system of Fig. 2, the object 102, here the cable, is continuously guided and conveyed through the measuring chamber 106.

[0203] That is, the detection units 104 continuously detect the object 102.

[0204] In other words, the object 102, here the cable, can be inspected or inspectable in-line.

[0205] For conveying through the measuring area 106, a conveying device is assigned to the system 100 (not shown in the figures).

[0206] The conveyor device is designed, for example, as an unwinding device and a winding device and is arranged upstream and downstream of the system 100. In other words, the cable can be conveyed by unwinding and winding through the measuring area 106 and optionally through a production facility in which the system 100 is arranged or can be arranged.

[0207] The detection modules 118, 120, 122, 124 are taught using a good example or good sample of a cable.

[0208] The inspection of cables in particular can include, for example:

[0209] - Analysis of the product surface and / or detection and classification of color deviations (monochrome / multi-color), inclusions, cracks, scratches, streaks, abrasions, stripes, smears, lack of material, excess material, geometric deviations (e.g., bumps, bubbles, kinks, constrictions, dents, holes), open core (e.g., visible strands of wire); and / or

[0210] Measurement of product geometry: diameter, linearity (e.g. product curvature); and / or

[0211] - Print control: Analysis of the quality and accuracy of prints (e.g. inkjet, laser, embossing, etc.), logos, serial numbers, barcodes and / or QR codes, matrix text, product names and parameters, length information (e.g. meter information for underground cables).

[0212] It should be understood that the above list is exemplary and not exhaustive.

[0213] Furthermore, it should be understood that all advantages of the system 100 according to the first embodiment are also advantages of the system 100 according to the second embodiment or can be assigned thereto.

[0214] Further advantages of the system 100 according to the second embodiment are in particular:

[0215] The training (deep neural networks for analysis) of new products is carried out purely on the basis of good examples. No bad examples (NoO) are required. Occasional errors may occur on the sample material, as long as they are only minor.

[0216] The cable (e.g., extruded product) may vibrate as long as it does not leave the camera's image field. The vibrations can be compensated using the locator module 116. The surface of the extruded product is completely (essentially 100%) captured and analyzed.

[0217] The detection area is shielded from ambient light by the housing 144 and therefore operates independently of the ambient light situation.

[0218] - Due to the short adjustable exposure time of camera 112, the enclosure 144 does not need to be completely closed. The entry and exit areas, in particular the entry and exit openings 148, 150, for the object 102, here a cable, do not need to be completely light-tight.

[0219] The high luminous intensity of the installed light source 114 compensates for any residual ambient light.

[0220] Water droplets from the cooling process are detected as such and not as product defects due to the detection modules used and in particular the illumination.

[0221] With reference to Fig. 5, a method according to the invention for the optical inspection of objects 102 will be described below, which can be carried out in particular by systems 100 already described according to the first and second embodiments.

[0222] It should be understood that any structural and / or functional features and / or properties and / or advantages described in connection with the system 100 according to the invention for optically inspecting objects 102 may also be part of and / or attributable to said method.

[0223] The method comprises providing an object 102 to be inspected in a measuring area 106.

[0224] The method further comprises optically detecting the object 102 to be inspected by means of one or more, here three, detection units 104, in particular in a continuous manner.

[0225] The optical detection comprises an optical detection of the object 102 by means of a camera 112 and a particularly coaxial illumination of the object 102 to be inspected by means of a light source 114, wherein in particular the optical detection by means of the camera 112 and the particularly coaxial illumination by means of the light source 114 take place simultaneously.

[0226] The method further comprises providing image information of the detected object 102 by means of the detection units 104 (see field 1 of Fig. 5). Not explicitly shown, the method may further comprise: projecting light, in particular light patterns, for example parallel black / white line pairs or dots, at a known angle onto the object to be inspected by a projection device. The projection device may comprise illumination, for example structured illumination or laser; detecting the surface shape of the object to be inspected by deformed light patterns, for example deformed stripe patterns or dot patterns.

[0227] The method further comprises inspecting the detected object 102 using the provided image information based on one or more, in particular positive, reference objects by means of a computing unit 110 (cf. fields 1 to 8 of Fig. 5).

[0228] The method further comprises providing a corresponding inspection result about the object 102 by means of the computing unit 110 (see field 9 of Fig. 5).

[0229] The method further comprises: determining an outer contour of the detected object 102 in the image information by means of a locator module 116 of the computing unit 110 and deleting an information content that can be assigned to an area outside the outer contour of the detected object 102 in the image information by means of the locator module 116 (cf. field 3 of Fig. 5).

[0230] The inspection comprises: analyzing the detected object 102 for correspondence with and / or deviations from one or more, in particular positive, reference objects based on the image information by means of one or more recognition modules 118, 120, 122, 124 of the computing unit 110; and providing a corresponding measured value by means of the one or more recognition modules 118, 120, 122, 124 (cf. fields 4 to 7 of Fig. 5).

[0231] The inspection further comprises: determining by means of an evaluator module 126 of the computing unit 110 whether the analyzed measured value lies within predetermined limit values ​​(cf. field 8 of Fig. 5), wherein the corresponding inspection result about the object 102 can be generated or is generated on the basis of this determination (cf. field 9 of Fig. 5).

[0232] The method also comprises: generating a first and / or at least one further, in particular positive, reference object, in particular for subsequent inspections, on the basis of the image information of the detected object 102 by means of a trainer module 128 of the computing unit 110, if the detected object 102 substantially corresponds to a desired state (cf. fields 1 and 16 with 17 of Fig. 5).

[0233] The method also comprises: determining an outer contour of the detected reference object 102 in the image information by means of the locator module 116 of the computing unit 110 and deleting an information content that can be assigned to an area outside the outer contour of the detected reference object 102 in the image information by means of the locator module 116 (cf. field 18 of Fig. 5).

[0234] The method further comprises: training the one or more recognition modules 118, 120, 122, 124 on the basis of at least one, in particular positive, reference object by means of training modules 130, 132, 134, 136 of the computing unit 110 (cf. fields 19 to 24 of Fig. 5).

[0235] The method may, for example, comprise the object 102 to be inspected being continuously movable or moving through the measuring area 106, e.g., of a measuring chamber 146.

[0236] List of reference symbols

[0237] System Object Detection unit Measuring range Conveyor belt Computing unit Camera Light source Locator module First detection module Second detection module Third detection module Fourth detection module Evaluator module Trainer module First training module Second training module Third training module Fourth training module Supporting device Housing Touch screen Enclosure Measuring chamber Entrance opening

[0238] Exit aperture Beam splitter Diffuser Lens Surface element

Claims

AMENDED CLAIMS received by the International Bureau on 08 March 2024 (08.03.2024) System (100) for optical inspection of objects (102), comprising: - one or more detection units (104) for the continuous optical detection of at least one object (102) to be inspected in a measuring area (106) and for providing image information of the detected object (102); and - a computing unit (110) designed and programmed to inspect the detected object (102) using the provided image information based on one or more, in particular positive, reference objects, and to provide a corresponding inspection result for the object (102) in question. The system (100) according to claim 1, characterized in that the computing unit (110) has a locator module (116) designed and programmed to determine an outer contour of the detected object (102) in the image information and, optionally, to delete information content in the image information that can be assigned to an area outside the outer contour of the detected object (102).System (100) according to claim 1 or 2, characterized in that the computing unit (100) has one or more recognition modules (118, 120, 122, 124) which are designed and programmed to analyze the detected object (102) based on the image information for correspondence with and / or deviations from one or more, in particular positive, reference objects and to provide a corresponding measured value. System (100) according to claim 3, characterized in that. - at least one recognition module (118) is designed and programmed for symbol recognition analysis, in particular text recognition analysis, and / or - at least one recognition module (120) is designed and programmed for geometric measurement analysis, and / or AMENDED SHEET (ARTICLE 19) - at least one recognition module (122) is designed and programmed for color analysis, and / or - at least one recognition module (124) is designed and programmed for surface analysis. System (100) according to claim 3 or 4, characterized in that the computing unit (100) has an evaluator module (126) which is designed and programmed to determine whether the analyzed measured value lies within predetermined limit values, wherein the corresponding inspection result for the object (102) can be generated on the basis of this determination. System (100) according to one of claims 1 to 5, characterized in that the computing unit (110) has a trainer module (128) which is designed and programmed to generate a first and / or at least one further, in particular positive, reference object, in particular for subsequent inspections, based on the image information of the detected object (102), if the detected object (102) substantially corresponds to a target state.System (100) according to claim 6 in conjunction with claim 3 or 4, characterized in that the computing unit (110) has one or more training modules (130, 132, 134, 136) corresponding to the one or more recognition modules (118, 120, 122, 124), which are designed and programmed to train the one or more recognition modules (118, 120, 122, 124) on the basis of at least one, in particular positive, reference object. System (100) according to one of claims 1 to 7, characterized in that each detection unit (104) has a trigger input by means of which an analog trigger signal can be received, wherein each detection unit (104) is configured to start and / or synchronize an optical detection, in particular with the computing unit (110), as soon as the analog trigger signal is received. AMENDED SHEET (ARTICLE 19) System (100) for optical inspection of linear objects (102) such as cables, wires, hoses and pipes, comprising: - a conveyor device for continuously passing a linear object (102) to be inspected through a measuring area (106); - one or more detection units (104) for optically detecting the linear object (102) in the measuring area (106) and for providing image information of the linear object (102); - a computing unit (110) which is designed and programmed to inspect the linear object (102) based on the provided image information and to provide a corresponding inspection result about the linear object (102), - wherein the computing unit (110) has at least one recognition module (118, 120, 122, 124) which is designed and programmed to analyze the image information for conformity with and / or deviations from a predetermined good pattern for the linear object (102) to be inspected. System (100) according to claim 9, characterized in that the recognition module (118, 120, 122, 124) comprises an artificial neural network for learning a good pattern. System (100) according to claim 9 or 10, characterized in that the computing unit (110) is designed and programmed to generate a good pattern from the acquired image information of a section of a linear object (102). System (100) according to one of claims 9 to 11, characterized in that a plurality of stationary detection units (104) are arranged distributed in the circumferential direction around the linear object (102).System (100) according to one of claims 9 to 12, characterized in that the linear object (102) is preferably guided freely through the measuring area from roller to roller. System (100) according to one of claims 9 to 13, characterized in that. AMENDED SHEET (ARTICLE 19) the measuring area (106) is at least partially shielded from ambient light by a housing (144). System (100) according to one of claims 9 to 14, characterized in that the image information can be provided as a high-speed image sequence with more than 500 images per second by means of a camera (112) of each of the detection units (104). System (100) according to one of claims 9 to 15, characterized in that the inspection result comprises at least one property of the linear object (102) from the group consisting of peripheral geometry, imprints, color, defects, and foreign particles. System (100) according to one of claims 9 to 16, characterized in that the inspection result can be provided on an output unit in-line during the passage of the linear object (102).System (100) according to one of claims 9 to 17, characterized in that the computing unit (110) is designed and programmed to determine the position of a detected defective property along the linear object (102). A method for the optical inspection of objects (102), the method comprising: - Providing an object to be inspected (102) in a measuring area (106); - optically detecting the object to be inspected (102) by means of one or more detection units (104), in particular in a continuous manner; - providing image information of the detected object (102) by means of the one or more detection units (104); - Inspecting the detected object (102) using the provided image information based on one or more, in particular positive, reference objects by means of a computing unit (110); and - Providing a corresponding inspection result about the object (102) by means of the computing unit (110). Method according to claim 19, characterized in that AMENDED SHEET (ARTICLE 19) the procedure has: - determining an outer contour of the detected object (102) in the image information by means of a locator module (116) of the computing unit (110); and - optionally, deleting an information content that can be assigned to an area outside the outer contour of the detected object (102) from the image information by means of the locator module (116). The method according to claim 19 or 20, characterized in that the inspection comprises: - analyzing the detected object (102) for correspondence with and / or deviations from one or more, in particular positive, reference objects based on the image information by means of one or more recognition modules (118, 120, 122, 124) of the computing unit (110); and - Providing a corresponding measured value by means of the one or more detection modules (118, 120, 122, 124). Method according to claim 21, characterized in that - at least one recognition module (118) is designed and programmed for symbol recognition analysis, in particular text recognition analysis, and / or - at least one recognition module (120) is designed and programmed for geometric measurement analysis, and / or - at least one recognition module (122) is designed and programmed for color analysis, and / or - at least one detection module (124) is designed and programmed for surface analysis. Method according to one of claims 21 to 23, characterized in that the inspection comprises: - Determining, by means of an evaluator module (126) of the computing unit (110), whether the analyzed measured value lies within predetermined limit values, wherein the corresponding inspection result for the object (102) can be generated or is generated based on this determination. Method according to one of claims 19 to 23, AMENDED SHEET (ARTICLE 19) characterized in that the method comprises: - Generating a first and / or at least one further, in particular positive, reference object, in particular for subsequent inspections, based on the image information of the detected object (102) by means of a trainer module (128) of the computing unit (110), if the detected object (102) substantially corresponds to a target state. Method according to claim 24 in conjunction with claim 21 or 22, characterized in that the method comprises: - Training the one or more recognition modules (118, 120, 122, 124) on the basis of at least one, in particular positive, reference object by means of one or more training modules (130, 132, 134, 136) of the computing unit (110). AMENDED SHEET (ARTICLE 19)

Images