Method for inspecting a side wall of an object

Abstract

A method for inspecting a side wall (21) of an object (2) comprises the steps of: conveying the object (2) on a conveyor (3); receiving the object (2) to be inspected in an inspection station (4) positioned along the conveyor (3), wherein an axis (A) of the object is oriented along a longitudinal direction (D); capturing a first image (410) by a first camera (41) observing the object (2) from above along an optical path having at least one component parallel to the longitudinal direction (D); deriving a three-dimensional representation of the side wall (21) of the object (2) based on the first image (410) and reference data (60); capturing a second image (420) representing a portion of the side wall (21) of the object (2), the second image (420) having a perspective distortion; deriving a third image (50) representing the portion of the side wall of the object, the third image (50) being free of perspective distortion.

Description

Technical Field

[0001] This invention relates to a method and apparatus for inspecting the sidewalls of an object. The invention also relates to a method for processing images of the sidewalls of an object. Background Technology

[0002] This invention relates to the field of inspecting objects (e.g., containers or vials for substances in powder, gel, or liquid form). More specifically, this invention relates to inspecting the sidewalls of these objects using views to detect any defects in the object or labels affixed thereto. The sidewalls are inspected by capturing images of them; typically, these images have perspective distortion, which must be taken into account when processing the images in order to correctly identify defects.

[0003] A container inspection method is known in the art from the applicant's patent document WO2011013153A1. This method involves capturing two or more images of a container to define a reference axis of the container, and using a previously stored three-dimensional model of the container as a reference to form a planar unfolded image of the container's side surface from the two or more images. This method has the following disadvantages: it requires knowledge of and possession of a three-dimensional model of the container available in computer format; furthermore, the need to identify the container's axis makes the method computationally intensive; and additionally, the method is not suitable for inspecting containers that are not bodies of revolution (i.e., do not have central symmetry about the reference axis).

[0004] Another inspection method is described in patent document WO2007110372; this method also involves constructing an unfolded image from multiple images of the sidewall, and is therefore computationally intensive. Furthermore, the system requires a large number of cameras, which inevitably increases the cost of the equipment.

[0005] Other methods and systems for inspection are disclosed in patent documents JP5375488B2, FR2939201A1, WO2018 / 217077A1, and EP2297674A1. Summary of the Invention

[0006] The purpose of this invention is to provide a method and apparatus for inspecting the sidewalls of an object, and a method for processing images of the sidewalls of an object, to overcome the aforementioned disadvantages of the prior art.

[0007] These objectives are fully achieved by the methods and apparatus for inspecting objects as described in this disclosure and characterized in the appended claims, as well as the methods for processing images.

[0008] More specifically, this disclosure relates to a method for inspecting the sidewalls or side surfaces of an object. The object may be made of plastic, glass, or metal. The object may be a container for a substance in powder, liquid, or gel form; for example, the object may be a bottle, jar, box, or can. The object has sidewalls (or side surfaces) extending about a respective axis. The object also has a bottom wall connected to the sidewall along the lower outer edge of the sidewall. More specifically, the bottom wall may be oriented perpendicular to the axis of the object. The object preferably has a top wall; the top wall is connected to the sidewall along the upper outer edge of the sidewall. The lower outer edge surrounds the axis of the object; the upper outer edge surrounds the axis of the object.

[0009] The method includes the step of conveying objects on a conveyor. The conveyor can be, for example, a belt conveyor configured to convey objects along a conveying direction; the conveyor can also be a turntable configured to convey objects along a circular trajectory about an axis of rotation. Preferably, the conveyor conveys objects continuously. In one embodiment, the conveyor conveys a single object; in another embodiment, the objects may come into contact with each other. In either case, it is preferable to examine the objects one at a time (in fact, the first and second images further described in this disclosure were captured one after another in a rhythmic sequence for each sample).

[0010] Preferably, the conveyor transports objects with the bottom wall resting on the conveyor; the conveyor can also transport objects while holding them through a portion of the object (e.g., the neck), or can hold objects by suction.

[0011] The method includes the step of receiving an object to be inspected into an inspection station. The inspection station is positioned along a conveyor. At the inspection station, the object is received with its axis oriented longitudinally.

[0012] The method includes the step of capturing (or photographing) a first image. The first image is captured by a first camera that observes the object from above along an optical path (or observation axis) having at least one component parallel to the longitudinal direction. Thus, the first image represents at least a portion of the top wall and / or upper outer edge of the object.

[0013] The method includes the step of deriving a three-dimensional representation of the sidewalls of an object based on a first image and reference data. Therefore, the first image is processed according to the reference data to derive the three-dimensional representation. In an embodiment, the three-dimensional representation is (or includes) a three-dimensional model of the object.

[0014] In this embodiment, reference data is stored in memory and selectively retrieved based on input data. In this embodiment, reference data is received in the form of input data. The reference data represents one or more geometric features of an object; therefore, the reference data depends on the size and format of the object.

[0015] The method includes the step of capturing a second image; the second image represents a portion of the sidewall. The second image has perspective distortion. The perspective distortion of the second image is random and depends on the orientation of the object in a plane perpendicular to the longitudinal direction. In practice, the transmitter ensures that the object is received in the inspection station with its axis aligned longitudinally, but each object can rotate about its corresponding axis; therefore, its sidewall can have a random angle relative to the inspection station; this angle produces perspective distortion in the second image that changes from one object to another.

[0016] The method includes the step of processing a second image; the second image is processed based on a three-dimensional representation. More specifically, the processing steps include identifying a first plurality of points on the three-dimensional representation and identifying a second plurality of points (or pixels) on the second image. Thus, the processing includes applying the correlation between the first plurality of points and the second plurality of points.

[0017] More specifically, in an embodiment, each point on the side surface of an object in the second image is associated with a point in a three-dimensional representation.

[0018] In another embodiment, only a subset of points in the second image is associated with a corresponding subset of points in the 3D representation; in this way, the processing steps are simpler, but the results may be less accurate. In this embodiment, the method provides identifying a second plurality of points in the second image (which includes only a subset of points in the second image), identifying corresponding first plurality of points in the 3D representation, and associating points in the second plurality of points with points in the first plurality of points.

[0019] Therefore, the process includes a perspective transformation, which involves associating each point in the three-dimensional representation with a (unique) point in the second image.

[0020] The process involves deriving a third image representing a portion of the object's sidewalls. This third image is derived from the 3D representation using a perspective transformation of the second image. The third image has no perspective distortion (or less distortion than the second image).

[0021] It should be noted that the third image corresponds to the (second) image that would be obtained if the object has a predetermined nominal orientation in the checkpoint and / or if the object's axis coincides with a predetermined longitudinal axis; therefore, the third image is independent of the actual random orientation of the object in a plane perpendicular to its axis (and longitudinal direction), and / or of any deviation (if any) between the object's axis and longitudinal axis (or direction). Thus, the third image is spatially homogeneous.

[0022] This enables the third image to be used to identify defects in the sidewalls of an object or in labels affixed to it.

[0023] In one embodiment, the method includes the step of processing a third image to identify defects; for example, the third image may be compared with a reference image representing the sidewall of a defect-free object (oriented according to a nominal orientation and whose axis coincides with a predetermined longitudinal axis); in another example, the third image may be compared with an image of the sidewall of a defective object (also oriented according to a nominal orientation and whose corresponding axis coincides with a predetermined longitudinal axis).

[0024] In an embodiment, the third image is processed by a self-learning system; preferably, the self-learning system comprises a neural network (e.g., a convolutional neural network). The neural network is trained to identify defects (if any) in an object; the neural network can be trained from objects with defects and / or objects without defects.

[0025] Preferably, the first camera for capturing the first image is positioned above the object in the inspection station. More specifically, the optical path (or observation axis) of the first camera is preferably parallel to the longitudinal direction. In an embodiment, the optical path of the first camera is perpendicular to the surface of the transmitter.

[0026] Preferably, the second image is captured by a second camera. The second camera is different from the first camera. More specifically, the second camera has a different optical path (or observation axis) than the first camera. The optical path of the second camera is incident on the sidewall of the object; for example, the optical path of the second camera may be orthogonal to the longitudinal direction.

[0027] It should be noted that the first camera and the second camera form part of the same optical device. Preferably, the first camera and the second camera are calibrated according to the same spatial reference frame. Therefore, the reference frame is common to both the first and second cameras. In an embodiment, the first camera and the second camera are connected to each other in a stereo configuration. In an embodiment, the method may include the step of calibrating the first camera and the second camera according to the same spatial reference frame.

[0028] In this embodiment, both the first and second images are captured by a first camera; in this case, the optical path of the first camera is tilted relative to the longitudinal direction so that the object can be viewed from above and from the side. For example, in this case, the first camera may have an optical path oriented at an angle between 40° and 50° (specifically, 45°) to the longitudinal direction. In this case, the first camera also takes a single snapshot, which is processed to derive both the first and second images.

[0029] The reference data used to derive a three-dimensional representation of the object preferably includes the distance from the upper outer edge of the object's side surface to the first camera. This distance is defined longitudinally. The upper outer edge refers to the edge that defines the boundary of the object when viewed from above. The step of deriving the three-dimensional representation therefore includes identifying a line on the first image representing at least a portion of the upper outer edge of the object's side surface; depending on the geometry of the object, the line can be a curve, a straight line, or a polygon. This line is a geometric representation of the upper outer edge (or a portion thereof).

[0030] In this embodiment, the line also represents the boundary of the object's cover (viewed from above); in this embodiment, therefore, the line includes a first portion representing the upper outer edge of the sidewall and a second portion representing the boundary of the cover; in this example, the second portion may be surrounded by the first portion.

[0031] Preferably, the reference data includes the height of the object's sidewalls (i.e., the distance from the upper outer edge to the base of the object). It should be noted that the reference data can also include, in the same manner, the distance from the first camera to the bottom wall and the height of the object's sidewalls, rather than the distance from the first camera to the upper outer edge and the height of the object's sidewalls (in fact, the distance from the first camera to the upper outer edge is obtained from these by subtraction). When objects are placed on a conveyor that transports them, the distance from the first camera to the bottom wall of the object is defined by the distance from the first camera to the surface of the conveyor on which the object is placed. It should be noted that, generally, the reference data includes at least two of the following quantities: the distance from the upper outer edge of the object's side surface to the first camera, the height of the object's sidewalls, and the distance from the first camera to the bottom wall (or to the conveyor) of the object; in practice, a third can be obtained from two of these quantities.

[0032] A three-dimensional representation is derived from the aforementioned line by projecting it along the longitudinal direction at a height equal to the height of the object's sidewall. In cases where the line represents only a portion of the upper edge, a local three-dimensional representation (representing only a portion of the object's sidewall—the part to be examined) can be obtained, and / or the remaining portion of the upper edge can be reconstructed (e.g., by assuming it has a certain symmetry), and then the line representing the edge (including the portion shown in the second image and the reconstructed portion) is projected downwards to obtain the three-dimensional representation.

[0033] More specifically, projecting a line downwards means projecting it away from the first camera toward the bottom wall of the object. Thus, the line obtained from the first image defines or approximates the upper boundary (or boundary portion) of the object, and a three-dimensional representation is obtained by projecting (or highlighting) the boundary at a known height equal to the height of the object. Therefore, it should be noted that the inspection method according to this disclosure does not require providing a three-dimensional model of the object from the outset; in practice, a three-dimensional representation (in some cases, an approximation) is constructed using reference data such as the height of the object and the distance of the object from the camera; the storage and processing of reference data is less burdensome than that of a three-dimensional model.

[0034] In an embodiment, the method includes a learning step, including a step of storing the contours of a line and / or updating the stored contours; thus, in successive objects, lines are identified not only based on a first image but also based on the stored contours.

[0035] It should be noted that the orientation of the side surfaces of the object is preferably substantially parallel to the longitudinal direction (and to the axis of the object). "Substantially parallel" means that variations in the orientation of the side surfaces relative to the longitudinal direction are acceptable; however, the extent of such variations must be limited. For example, variations in orientation relative to the longitudinal direction between +30° and -30° (preferably between +20° and -20°, or between +15° and -15°) are acceptable.

[0036] It should be noted that the method of this disclosure is advantageously applicable to objects that are not bodies of revolution; for example, it is applicable to containers that are oval, elliptical, or rectangular in shape and have sidewalls that are parallel to or substantially parallel to the axis of the object.

[0037] This disclosure also provides an apparatus for inspecting the sidewalls of an object.

[0038] The device includes a conveyor configured to transport objects. The conveyor may include a belt or a turntable. The conveyor is preferably configured to transport objects resting thereon. The conveyor is preferably configured to transport multiple objects, holding each object with its respective axis parallel to the longitudinal direction. The conveyor is preferably configured to transport multiple objects continuously. In an embodiment, the conveyor is configured to transport a single object.

[0039] The device includes an inspection station positioned along an inspection path. The inspection station is configured to receive an object to be inspected, the object being inspected being oriented with its axis along the longitudinal direction.

[0040] The inspection station includes an optical device for capturing images. The optical device includes a first camera configured to capture a first image of an object from above. The first camera has an optical path having at least one component parallel to the longitudinal direction.

[0041] The device includes a control unit. The control unit accesses a memory containing reference data. In one embodiment, the device includes an interface connected to the memory and configured to receive the reference data.

[0042] The control unit is configured to derive a three-dimensional representation of the object's sidewalls based on the first image and reference data.

[0043] The control unit is configured to perform the step of processing the second image according to the three-dimensional representation. The processing includes perspective transformation; in other words, the processing includes: identifying a first plurality of points on the three-dimensional representation, identifying a second plurality of points on the second image, applying the correlation between the first plurality of points and the second plurality of points; and deriving a third image representing the portion of the sidewall of the object; the third image having no perspective distortion and / or less distortion than the second image.

[0044] Furthermore, the size of objects in the third image can be adjusted based on the object's actual position within the checkpoint. For example, if an object is positioned closer to or further away from the second camera (or the camera capturing images of the sidewalls) compared to its nominal position, the effect is to magnify or reduce the object to compensate for the difference from the nominal position. In other words, at the nominal position, the object's axis coincides with a predetermined longitudinal axis; if the object's axis, although parallel to the longitudinal axis, is spaced apart from the predetermined longitudinal axis, the third image shows the object as if it were positioned with its axis coinciding with the predetermined longitudinal axis.

[0045] The control unit is configured to enable the third image to be used to identify defects in portions of the sidewall of an object (or in labels affixed thereto). In an embodiment, the control unit is configured to process the third image and identify defects.

[0046] It should be noted that the reference data preferably includes the distance from the upper outer edge of the side surface of the object to the first camera. The control unit is configured to identify a line representing at least a portion of the upper outer edge of the side surface of the object on the first image and derive a three-dimensional representation from that line.

[0047] Preferably, the reference data includes the height of the sidewalls of the object. The control unit is configured to derive the three-dimensional representation by projecting the line downwards at a distance equal to the height of the sidewalls of the object.

[0048] The optical device preferably includes a second camera, which has an optical path different from that of the first camera and is configured to capture a second image.

[0049] In another embodiment, the first camera is configured to capture both the first image and the second image.

[0050] In one embodiment, the device includes an illuminator configured to illuminate an object (specifically, its side surface) located in a checkpoint. More specifically, the illuminator may be a structured illuminator configured to illuminate the object with structured light (i.e., the illuminator projects a light pattern onto the object). Structured illuminators are particularly useful when only one camera with a light path tilted relative to the longitudinal direction is present, enabling accurate identification of the upper edge.

[0051] This disclosure also provides a method for processing an image of the sidewall of an object. The processing method includes the step of deriving a three-dimensional representation of the object's sidewall starting from a first image of the object viewed from above and reference data. The processing method includes the step of identifying a first plurality of points on the three-dimensional representation. The processing method includes the step of identifying a second plurality of points (pixels) on a second image. The processing method includes the step of deriving another (or third) image representing the portion of the sidewall of the object. The other image is obtained based on the three-dimensional representation through a perspective transformation of that image. Therefore, the other image has no perspective distortion or less distortion than the first image. More generally, the processing method corresponds to the step of processing the second image in the inspection method described in this disclosure.

[0052] This disclosure also provides a computer program including operating instructions configured to perform steps of a processing method according to one or more aspects of this disclosure when run on a computer. Attached Figure Description

[0053] These and other features will become more apparent from the following detailed description of the preferred embodiments illustrated by way of non-limiting example in the accompanying drawings, wherein:

[0054] - Figure 1 The illustration shows an apparatus for inspecting the sidewall of an object according to the present disclosure;

[0055] - Figure 2 The diagram shows... Figure 1 Equipment inspection station;

[0056] - Figure 3 An example of the first image is illustrated;

[0057] - Figure 4 An example of the second image is illustrated;

[0058] - Figure 5 An example of the third image is shown;

[0059] - Figures 6-9The diagram shows what can be used Figure 1 An example of a container being inspected by the equipment;

[0060] - Figure 10 It shows that it is not suitable for use Figure 1 The equipment inspects the containers;

[0061] - Figure 11 It schematically represents Figure 1 The equipment. Detailed Implementation

[0062] Referring to the accompanying drawings, reference numeral 1 indicates a device for inspecting the sidewall 21 of an object (or container) 2. The object (or container) 2 includes a bottom wall, sidewalls connected to the bottom wall, and a top wall connected to the sidewalls. The top wall may define an opening in the container. The sidewall 21 extends about axis A of the object 2.

[0063] Device 1 includes a conveyor 3; in one embodiment, the conveyor 3 includes a conveyor belt, and objects 2 are conveyed with their bottom walls resting on the conveyor belt and with their axis A aligned parallel to the longitudinal direction D. The longitudinal direction D is orthogonal to the supply direction of the conveyor 3 and / or the surface of the conveyor 3 on which the objects 2 rest. Preferably, the conveyor 3 continuously conveys multiple objects 2 and holds them with their axis A aligned parallel to the longitudinal direction D. In an embodiment, the conveyor 3 is configured to continuously move and / or supply objects 2 along the supply direction of the conveyor 3. In another embodiment, the conveyor 3 is configured to move in a series of movement periods alternating with stop periods.

[0064] In one embodiment, the surface of the transmitter 3 has a color that provides high contrast with the color of the object 2. In another embodiment, the surface of the transmitter 3 includes a portion having a first color and a portion having a second color different from the first color; these two portions can be placed side-by-side with each other along the supply direction of the transmitter; thus, a portion of the object rests on the first portion, and a portion of the object rests on the second portion; in this way, depending on the color of the object, one portion or the other is more clearly visible, and the portion with greater contrast to the transmitter is used to construct a three-dimensional representation from one moment to the next.

[0065] Device 1 includes inspection station 4. Inspection station 4 is positioned along conveyor 3. Inspection station 4 is configured to receive one object 2 at a time. The axis A of the object 2 received in inspection station 4 is oriented in (or parallel to) the longitudinal direction D.

[0066] Inspection station 4 (or more generally, device 1) includes a first camera 41. The first camera 41 is located above the object 2 positioned in inspection station 4; therefore, camera 41 observes the object 2 from above. Preferably, the first camera 41 has an optical path oriented parallel to the longitudinal direction D. The first camera 41 is configured to capture a first image 410. The first image 410 is preferably a view of the object 2 from above. Thus, the first image 410 represents the top wall of the object 2; the upper outer edge B of the side surface 21 is also represented in the first image 410. In effect, the upper outer edge B of the side surface forms the boundary of the object 2 in the first image 410 taken from above.

[0067] Inspection station 4 (or more generally, device 1) includes a second camera 42. The second camera 42 is located next to the object 2 positioned in inspection station 4 and has an incident light path in the longitudinal direction D. The second camera 42 is configured to capture a second image 420. The second image 420 represents the sidewall 21 (or a portion or surface thereof) of the object 2. Because each object 2 has a random orientation about axis A (and different from the nominal orientation), the second image 420 of each object 2 has a corresponding perspective distortion that changes from one object 2 to another.

[0068] Preferably, the object 2 is in the same position at the moment the first camera 41 captures the first image 410 and the moment the second camera 42 captures the second image. More specifically, if the conveyor 3 transports the object in continuous motion, the first image 410 and the second image 420 can be captured simultaneously or at consecutive moments; in fact, in the latter case, the control unit can receive the supply speed of the conveyor 3 (which is typically tracked by an encoder) as input, and can then process the first image 410 and the second image 420 according to that speed, taking into account the spatial translation experienced by the object 2 from one moment to the next.

[0069] If the conveyor 3 transports the object in moving periods that alternate with stop periods, the conveyor 3 stops when the object 2 is in the checkpoint, and during the stop period, the first image 410 and the second image 420 are captured (not necessarily simultaneously).

[0070] Device 1 includes a control unit (or processing unit) 5. Control unit 5 is connected to a first camera 41 and a second camera 42 to receive a first image 410 and a second image 420. Control unit 5 is configured to identify a line L on the first image 410 representing at least a portion of the upper outer edge B of the side surface 21 of object 2. Control unit 5 is connected to a memory or database 61 (which may itself be part of device 1). Memory 61 contains reference data 60. Reference data 60 includes at least two of the following data items: the distance H1 of the upper edge B along the longitudinal direction D from the first camera 41; the height H2 of the sidewall 21 along the longitudinal direction D (i.e., the distance of the upper edge B from the base or from the transmitter 3); and the distance of the camera 41 from the base or from the transmitter 3 (equal to the sum of H1 and H2). Control unit 5 is configured to receive these reference data items 60 from memory 61. In an embodiment, control unit 5 is configured to process the reference data 60 to obtain distance H1 and / or height H2 (e.g., if the reference data 60 includes H2 and the sum of H1 and H2, then control unit 5 derives H1 by subtraction).

[0071] The control unit 5 is configured to derive a three-dimensional representation (i.e., a three-dimensional model) of the object 2 from the first image 410, from the distance H1, and from the height H2. More specifically, the control unit 5 zooms in and out of line L according to the first height H1, and then projects line L downward at a height equal to the height H2 (enlarging or reducing in size as needed by scaling).

[0072] The control unit 5 then performs a perspective transformation on the second image 420 based on the three-dimensional representation thus constructed. The perspective transformation includes identifying a second plurality of points in the second image 420, identifying corresponding first plurality of points in the three-dimensional representation, applying a correlation between the first plurality of points and the second plurality of points, and deriving a third image 50 based on the correlation. More specifically, in an embodiment, the control unit 5 associates each point or pixel on the second image 420 with a point in the three-dimensional representation (or model), thereby deriving the third image 50. The third image 50 represents the sidewall 21 of the object 2 without perspective distortion or with less perspective distortion than the second image 420.

[0073] Additionally, the axis A of object 2 may not be completely aligned with the longitudinal axis (or direction) D, and may be offset from it. In this case, the control unit 5 transforms the second image 420 (by zooming in or out) so that the generated third image 50 represents the sidewall 21 of object 2 as if its axis A were aligned with the longitudinal axis (or direction) D.

[0074] Then, the control unit 5 can make the third image 50 available for subsequent analysis to detect defects in the sidewall 21 or in the labels affixed thereto, or it can perform the analysis itself.

Claims

1. A method for inspecting the sidewalls of an object, wherein the sidewalls of the object extend about a corresponding axis, the method comprising the steps of: - The object is transported on a conveyor; - Receive the object to be inspected at an inspection station located along the conveyor, wherein the axis of the object is oriented along the longitudinal direction. - A first image is captured by a first camera, which observes the object from above along a light path having at least one component parallel to the longitudinal direction; - Derive a three-dimensional representation of the sidewall of the object based on the first image and reference data; - A second image of the object, representing a portion of the sidewall of the object, is captured by the first camera or the second camera, the second image having perspective distortion; - Process the second image according to the three-dimensional representation, wherein the processing steps include: Identify the first plurality of points on the three-dimensional representation. Identify a second plurality of points on the second image. Applying the correlation between the first plurality of points and the second plurality of points, A third image representing the portion of the sidewall of the object is derived, the third image being obtained based on the three-dimensional representation through a perspective transformation of the second image, the third image having no perspective distortion or less distortion than the second image; The third image enables the identification of defects in the portion of the object's sidewall.

2. The method of claim 1, wherein the first camera is located above the object positioned at the checkpoint.

3. The method of claim 2, wherein the second image is captured by a second camera having an optical path different from that of the first camera.

4. The method of claim 3, wherein the first camera and the second camera are calibrated according to the same spatial reference frame.

5. The method according to any one of the preceding claims, wherein the reference data includes the distance from the upper outer edge of the side surface of the object to the first camera.

6. The method of claim 5, wherein the step of deriving the three-dimensional representation comprises identifying on the first image a line representing at least a portion of the upper outer edge of the side surface of the object.

7. The method of claim 6, wherein the reference data further includes the height of the sidewall of the object, wherein the three-dimensional representation is derived from the line by projecting the line along the longitudinal direction at a height equal to the height of the sidewall of the object.

8. The method according to any one of claims 1 to 4, wherein the conveyor continuously supplies the objects and holds each object with its axis oriented parallel to the longitudinal direction.

9. The method according to any one of claims 1 to 4, wherein the side surface of the object is oriented substantially parallel to the longitudinal direction.

10. The method according to any one of claims 1 to 4, wherein in the processing step, each point of the second image is associated with a point of the three-dimensional representation.

11. An apparatus for inspecting the sidewall of an object, wherein the sidewall of the object extends about a respective axis, the apparatus comprising: - A conveyor configured to transport the object; - An inspection station, which is disposed along the transmitter and configured to receive an object to be inspected, the object being inspected being oriented with its axis along the longitudinal direction; wherein the inspection station includes optical means for capturing images; The optical device includes a first camera configured to capture a first image of the object from above, wherein the first camera has an optical path having at least one component parallel to the longitudinal direction. The optical device is configured to capture a second image of the object, representing a portion of the sidewall of the object to be inspected, via the first camera or the second camera; - Control unit, the control unit being configured to: A three-dimensional representation of the sidewall of the object is derived based on the first image and reference data; Performing the step of processing the second image according to the three-dimensional representation, wherein the processing step includes: Identify the first plurality of points on the three-dimensional representation. Identify a second plurality of points on the second image. Apply the correlation between the first plurality of points and the second plurality of points; A third image representing the portion of the sidewall of the object is derived, the third image being obtained based on the three-dimensional representation through a perspective transformation of the second image, the third image having no perspective distortion or less distortion than the second image; The third image is used to identify defects in the portion of the sidewall of the object.

12. The device of claim 11, wherein the reference data includes the distance of the upper outer edge of the side surface of the object from the first camera, and wherein the control unit is configured to identify, on the first image, a line representing at least a portion of the upper outer edge of the side surface of the object, and derive the three-dimensional representation from the line.

13. The device of claim 12, wherein the reference data further includes the height of the sidewall of the object, wherein the control unit is configured to derive the three-dimensional representation by projecting the line downwards at a height equal to the height of the sidewall of the object.

14. The device according to any one of claims 11 to 13, wherein the optical device includes a second camera configured to capture the second image, wherein the second camera has an optical path different from that of the first camera.

15. The device according to any one of claims 11 to 13, comprising a structured illuminator configured to illuminate the object positioned in the checkpoint with structured light.

16. A method for processing an image of the sidewall of an object, wherein the image has perspective distortion, the method comprising the steps of: - Derive a three-dimensional representation of the sidewalls of the object based on a first image of the object viewed from above and reference data; - Identify the first plurality of points on the three-dimensional representation; - Identify a second plurality of points on a second image of the object, the second image representing a portion of the sidewall of the object, and the second image having perspective distortion; - Apply the correlation between the first plurality of points and the second plurality of points; - Derive another image representing the portion of the sidewall of the object, the other image being obtained based on the three-dimensional representation through a perspective transformation of the second image, the other image having no perspective distortion or less distortion than the second image.

17. The method of claim 16, wherein the step of deriving the three-dimensional representation comprises identifying lines on the first image representing at least a portion of the upper outer edge of the side surface of the object.

Images