Method and facility for online dimensional control of manufactured objects

Abstract

The present invention relates to a method and apparatus for automatically measuring the linear dimensions of a series of manufactured objects (2), comprising: - setting at least one X-ray focus (Fj) and one or more image sensors (Ci), the at least one X-ray focus (Fj) being on the same reference straight line parallel to the linear trajectory of the object's displacement; - acquiring a set of one-dimensional images for each object during its displacement, the set comprising a plurality of (NP) images obtained along at least three different projection directions (Dijk) in the cross-sectional planes containing the reference straight line; - determining, for each object and for each different cross-sectional plane (Pk), the contour of the object in the cross-sectional plane (Pk) under consideration from the acquired images.

Description

Technical Field

[0001] This invention relates to the field of dimensional X-ray inspection of manufactured objects that form a series of objects.

[0002] More specifically, the object of the present invention is to obtain X-ray measurements of linear dimensions, that is, X-ray measurements of lengths selected on manufactured objects in a general sense (e.g., containers, molded or machined parts, mechanical parts, packaging, body components, etc.). Background Technology

[0003] Existing technologies are known to allow for the dimensional inspection of objects using X-rays. Baggage inspection systems are also known, the purpose of which is not to measure the size of known objects, but to detect prohibited objects or numerous prohibited materials that are essentially randomly arranged, shaped, and numbered.

[0004] Therefore, axial rotation systems, including computer-aided tomography (CT) or computed tomography (CT), are known. JPKruth et al. (1) described this classic method in their article “Computed tomography for dimensional metrology” in CIRP Annals.

[0005] Volume 60, Issue 2, 2011, pp. 821-842, and implemented, for example, by tomographic scanning equipment sold by Werth Messtechnik or General Electric. The method involves positioning an object on a turntable about a vertical axis between an X-ray generating tube and a matrix or linear X-ray image sensor. A large number (at least 100, and typically more than 600) of 2D radiographic images of the object are acquired during rotation. If the image sensor is a matrix image sensor, the beam is conical. If the image sensor is linear, the beam is advantageously confined to a fan shape (fan beam) in a plane orthogonal to the axis of rotation, and for a full helical scan, rotation is accompanied by translation along the vertical axis of rotation. This technique can provide highly accurate three-dimensional measurements. However, even the fastest systems require at least one minute for acquisition, plus the time for loading and unloading the object, so at most 10 to 30 objects can be examined per hour.

[0006] For example, an alternative approach known as the rotating gantry approach has been proposed using a device known by the trade name "speed|scan CT 64" from General Electric. Similar to some 3D scanners used for luggage, the concept of this approach is analogous to medical imaging computed tomography in terms of the relative motion between the source, object, and image sensor. In practice, the manufactured object or luggage, positioned on a conveyor, translates within the device. They are transported via a projection plane orthogonal to the direction of displacement. Within a circular gantry containing said plane, an X-ray source and a typically curved image sensor, opposite to the source, rotate about a central axis of displacement to acquire projections slice by slice or via helical scanning, which are necessary for 3D reconstruction using algorithmic methods (e.g., "filtered projection" or ART methods). The aim of these devices is to allow the acquisition of a very large number of projections per revolution of the gantry, e.g., 100 to even 700 to 1,000 images per slice. For example, 3D reconstruction of the object is performed slice by slice. Having determined the attenuation of the slice at each point, the attenuation value at each volume element of the object is obtained by connecting the slices acquired during the object's displacement.

[0007] If these vertical axis rotation or rotating rack equipment are very accurate due to the large number of images provided, they are expensive and slow, and in practice they are reserved for offline inspection because they are not suitable for online dimensional inspection at a rate of 1 m / s travel per minute and more than 600 items.

[0008] Patent application DE102014103137 describes a method for determining the geometric characteristics of a workpiece using a tomodemometric detector system, which consists of an X-ray source, a plane detector, and a mechanical axis for rotating the workpiece or rotating the X-ray source and detector.

[0009] This method acquires radiographic images during rotation and ensures surface representation by using a surface model. This avoids the volumetric data reconstruction step, reducing computation time. However, this technique does not allow for the measurement of high-speed manufactured workpieces, as it requires loading the workpiece onto a turntable, rotating it at least 180°, and then unloading it to inspect another workpiece.

[0010] To overcome the drawbacks of tubes and image sensors embedded in rotating frames, US Patent 8971484 describes a baggage inspection system in which the rotating system is replaced by an array of stationary multi-beam X-ray sources, which are sequentially actuated to generate virtual displacements of the X-ray sources, thereby providing a large number of radiographic images with different projection angles. The number of "virtual rotations" is increased to 40 rotations per second compared to a physical rotating system limited to 4 rotations per second. This technology is implemented using a device known by the trade name Rapidan RTT from Rapidan Systems, and by generating tens of thousands of 2D images of baggage items, considering that the virtual rotations provide approximately 40 different projection angles, the technology is capable of inspecting 1,200 baggage items per hour.

[0011] The technology is very expensive due to the high cost of multi-source X-rays and the computing power required to process large amounts of data. Furthermore, the inspection rate remains limited and it is not suitable for online inspections.

[0012] Patents US7319737 and US7221732 disclose techniques for inspecting baggage items using a method known as digital tomography or tomographic synthesis. The baggage items are projected onto a series of conical projection planes called fan beams, each containing pairs of linear image sensors arranged in an L-shape. These techniques aim to detect weapons or explosives in baggage items containing objects of various shapes and materials by displaying the 3D positions of objects of various shapes and materials within the baggage and by evaluating, for example, the volume of a suspicious product. Multispectral techniques are typically used to also determine the atomic number of the material. Therefore, these systems attempt to determine the attenuation value at each point in the baggage. On the other hand, these systems cannot determine the dimensions of manufactured objects for quality inspection purposes with high speed and accuracy.

[0013] Patent application JPS60260807 proposes using measurements from X-rays originating from one or more focal points to measure the thickness of a tube wall translated along its axis, each of which is associated with a sensor. The focal points and sensors are positioned to generate radiographic projections along a plane orthogonal to the tube's displacement direction. Therefore, the radiographic projections are coplanar in a projection plane orthogonal to the tube's axis of symmetry. The directions of these radiographic projections form right angles (90°) with respect to the displacement direction. This technique does not allow for a complete understanding of the tube's inner and outer surfaces. The method described in this patent application only allows for the measurement of the cumulative thickness of the two walls of the tube in the projection direction, and cannot reconstruct a three-dimensional model of the tube that would allow for accurate measurements in other directions.

[0014] Similarly, patent US5864600 describes a method for determining the filling level of a container using X-ray sources and sensors laterally positioned on both sides of a container transport conveyor. This system does not allow for measurements on non-laterally oriented surfaces because the document does not provide a three-dimensional model of the container.

[0015] Patent application US2009 / 0262891 describes a system for detecting objects placed in luggage by X-rays, the luggage being moved by a conveyor. The system includes a pulse generator tube or a sensor with a large size parallel to the direction of travel. This document provides a method for reconstructing the object that is unsatisfactory because the lack of projection in the displacement direction does not allow for dimensional measurement in directions orthogonal to the displacement direction. The lack of radiographic projection in the angular sector does not allow for the generation of a suitable digital model to ensure accurate measurements.

[0016] Patent application DE19756697 describes an apparatus that has the same disadvantages as patent application US2009 / 0262891.

[0017] Patent application WO2010 / 092368 describes an apparatus for displaying a translated object using X-rays by means of a radiation source and three linear sensors.

[0018] Patent application US2010 / 220910 describes a method for detecting anomalies in an object by generating a 3D reference model representing an ideal object. The method then aims to compare a captured 2D image of the actual object with a 2D image corresponding to the reference model in order to infer anomalies. This method does not allow for accurate measurement of the object and only allows inspection of the object in the generated 2D image, thus limiting inspection to directions orthogonal to the projection direction.

[0019] Document WO2018014138 describes a method for performing inspection of manufactured articles. The method includes: acquiring a series of radiographic images of the article; determining the three-dimensional position of the article's outer shell for each acquired radiographic image; and performing a detailed three-dimensional model correction loop in the form of a mesh network, iteratively including: generating simulated radiographic images for each determined position of the article; and comparing the simulated radiographic images with the acquired radiographic images and generating matching results. If the matching results indicate a mismatch, the method includes: identifying differences between the simulated and acquired radiographic images and determining the cause of these differences in terms of material density or three-dimensional geometry; correcting one of the geometry and material density of the region of interest in the detailed three-dimensional model of the article based on each of the identified and characterized differences; and performing a new iteration. Therefore, this document works based on a three-dimensional model correction loop, i.e., achieving 3D geometry, which requires significant computation. Furthermore, the document teaches that in order to allow the three-dimensional model to be corrected in a loop, at least about 25 radiographic images, and preferably about 100 radiographic images, are required, which define a continuous image sequence, each image providing a unique viewing angle of the object.

[0020] The publication "Computed tomography for dimensional metrology" by KRUTH JP et al., CIRP ANNALS, Vol. 60, No. 2, December 31, 2011, pp. 821-842, ISSN: 0007-8506, DOI: 10.1016 / J.CIRP.2011.05.0, provides an overview of X-ray computed tomography techniques applied to dimensional metrology. Specifically, the document states that the basic principle of this technique involves the mathematical reconstruction of projected images to produce a 3D voxel model and involves post-processing of voxel data to detect workpiece edges (segmentation) and measure dimensions. The document also notes that the basic principle of this technique involves the rotation of the object around its axis.

[0021] Document US2009 / 262891 describes a system that performs standard imaging instead of geometric reconstruction, requiring a large number of angle samples and therefore multiple unit sensor assemblies along the object's trajectory. This necessitates a planar detector containing a high density of "light spots" along a line oriented along the object's trajectory. In other words, the 3D reconstruction system using a filtered back-projection algorithm has very high requirements in terms of the total number of unit sensor assemblies, as many are needed along the object's trajectory.

[0022] The object of this invention is to overcome the shortcomings of the prior art by providing a method that is inexpensive to implement and allows for accurate dimensional X-ray inspection of manufactured objects traveling at high speeds. In particular, the method aims to allow such inspections to be performed with reduced computational load, utilizing cost-effective equipment to achieve these high rates.

[0023] It is known in tomography that the absence of radiographic projections around a given direction prevents the reconstruction of surfaces parallel to that direction, resulting in a phenomenon known as "missing boundaries." This prevents the measurement and dimensional inspection of dimensions orthogonal to the missing radiographic projections.

[0024] Therefore, another object of the present invention is to provide a method that allows for accurate measurement of translated objects by constructing an accurate and complete three-dimensional digital model, when the number of radiographic projections is limited and it is not possible to collect data around the direction of object transport. Summary of the Invention

[0025] This invention relates to a method for automatically measuring the linear dimensions of a series of manufactured objects, comprising:

[0026] - Select a series of manufactured objects, each of which consists of one or more distinct parts, the number of parts being known, and each part being made of a material with a known attenuation coefficient that is uniform at every point of that part of the object;

[0027] - By means of a transport device, objects are transported in the direction of displacement along a straight trajectory in the transport plane, and these objects generate transport volume during their displacement.

[0028] - Set the following items in addition to the transfer volume:

[0029] • At least one focal point of the X-ray generating tube, each focal point being arranged on the same reference straight line parallel to the direction of displacement along a linear trajectory, and

[0030] • One or more image sensors, each of which is exposed to and sensitive to X-rays originating from a relevant focal point, which have at least passed through the area to be inspected, and produce a radiographic projection of the area to be inspected on each image sensor along the projection direction;

[0031] - A set of one-dimensional processed ray images is acquired for each object during its displacement using image sensors (Ci, Cik). Each one-dimensional processed ray image includes the projection of the object's cross-section along a cross-sectional plane (Pk) containing the reference line. This set includes:

[0032] • The one-dimensional processed ray radiographs for multiple (NK) different cross-sectional planes (Pk) containing the reference line;

[0033] • For each different cross-sectional plane (Pk), multiple (NP) one-dimensional processed radiographic images (Spk) of the area to be inspected are obtained along at least three different projection directions (Dijk) in the cross-sectional plane;

[0034] - For each object to be measured and each different cross-sectional plane (Pk), the contour of the object in the considered cross-sectional plane (Pk) is determined using a computer system from the one-dimensional processed radiographic images (Spk) of the area to be inspected obtained along at least three different projection directions (Dijk) in the cross-sectional plane, and

[0035] - For the object to be measured, at least one measurement result is obtained from the contour of the object in each different cross-sectional plane to determine the linear dimension of the area to be inspected of the object.

[0036] Other optional but combinable features of the method according to the invention are elaborated in the following paragraphs.

[0037] The outline of an object may include curves or sets of curves or formed by them, which represent the intersection of the object's boundary surface and the cross-sectional plane.

[0038] The curves of an object's outline, or each curve, can be planar curves modeled by a parametric system.

[0039] Determining the contour of an object in a cross-sectional plane can include a curve fitting algorithm that starts with the prior contour of the object in the cross-sectional plane.

[0040] Determining the contour of an object in a cross-sectional plane can include nonlinear recursive curve fitting algorithms.

[0041] Determining the contour of an object in a cross-sectional plane can include iterative curve fitting algorithms, which include:

[0042] - Treat the prior profile of the object in the cross-sectional plane as the computational profile of the first iteration level;

[0043] Then, the following operations are performed iteratively until the comparison result between the simulated one-dimensional radiographic image and the one-dimensional processed radiographic image reaches the predetermined optimization criterion:

[0044] • From the computed profile of the object at a given iteration level in the cross-sectional plane, compute at least three simulated one-dimensional radiographic images of the area to be inspected. These simulated one-dimensional radiographic images are computed in the cross-sectional plane along at least three different projection directions used to acquire one-dimensional processed radiographic images in the cross-sectional plane.

[0045] • Compare the simulated one-dimensional radiographic image with the processed one-dimensional radiographic image.

[0046] Based on the comparison results, the calculated profile is modified to a higher iteration level.

[0047] The method may include:

[0048] - Using an image sensor, at least three two-dimensional radiographic images of the area to be examined are acquired for each object during its displacement, each image obtained along a different projection direction.

[0049] - Extract one-dimensional processed radiographic images from two-dimensional radiographic images to form a collection of one-dimensional radiographic images.

[0050] A one-dimensional processed radiographic image of an object can be formed by sampling a point image acquired using a point image sensor during the object's displacement, during a scan duration corresponding to the duration of the object's displacement between the focal point and the point image sensor.

[0051] The method may include: using a computer system and from the contours of the object in each different cross-sectional plane, constructing a three-dimensional digital geometric model of the area to be inspected for the object to be measured, the model including:

[0052] - Three-dimensional points in space, each point belonging to the boundary surface of the area to be inspected of the object; and / or

[0053] - At least one three-dimensional surface of the area to be inspected.

[0054] For an object to be measured, at least one measurement result that determines the linear dimensions of the area to be inspected from the object’s profile in each different cross-sectional plane may include: determining the distance between at least two three-dimensional points of the three-dimensional digital geometric model of the area to be inspected.

[0055] The method may include providing the computer system with a priori contour of the object in each cross-sectional plane.

[0056] The prior contour can be obtained in the following ways:

[0057] - Computer-aided design digital models of the objects in this series; and / or

[0058] -Based on the measurement results of the measuring device on one or more objects in the same series; and / or

[0059] -Based on the values ​​and / or diagrams and / or shapes selected by the operator on the human-machine interface of the computer system.

[0060] The method may include: providing a computer system with a priori three-dimensional geometric model of the series of regions to be inspected, which can be obtained through the following means:

[0061] - Computer-aided design digital models of the objects in this series; and / or

[0062] - A digital geometric model obtained from the measurement results of a measuring device on one or more objects in the same series; and / or

[0063] - A digital geometric model generated by a computer system from values ​​and / or diagrams and / or shapes selected by the operator on the computer system's human-computer interface.

[0064] The method may include setting the focus in the transport plane.

[0065] The method may include: for the series of objects during their displacement and for each considered cross-sectional plane of the object, using an image sensor to acquire at least two one-dimensional processed ray images of the inspection area corresponding to a projection direction, the projection direction defining an effective angle greater than or equal to 45° and less than or equal to 90°, advantageously greater than or equal to 60° and less than or equal to 90° in the considered cross-sectional plane.

[0066] The method may include: using an image sensor to acquire at least one radiographic image of the inspection area corresponding to a projection direction for each of the series of objects during their displacement and for each considered cross-sectional plane of the object, the projection direction producing a projection in the transport plane and having an opening angle between 10° and 60° with respect to the displacement direction.

[0067] The method may not include: for each object in the series, during its displacement, using an image sensor to acquire any radiographic image of the inspection area corresponding to the projection direction, which has an angle of less than 10° with the displacement direction.

[0068] The method may include: performing and acquiring a radiographic projection of the area of ​​the object to be examined, such that X-rays originating from one or more focal points and reaching the image sensor do not pass through another object.

[0069] The method may include: acquiring one-dimensional processed radiographic images for each object in the series during its displacement and for each cross-sectional plane, consisting of 30 to 40 radiographic projections (preferably 4 to 15 radiographic projections) originating from different projection directions of the area to be inspected.

[0070] In some embodiments, the image sensor may be part of at least three physical sensor assemblies, each of which is linear and includes a linear array of X-ray sensitive elements distributed along a supporting line that, together with the focal point, defines a projection plane containing the projection direction. These image sensors are configured such that:

[0071] - Each of these physical sensor assemblies has at least m sensitive elements that receive radiographic projections of the area to be inspected via an X-ray beam originating from the focal point;

[0072] - The projection planes of different physical sensor components are different from each other and are not parallel to the transmission plane;

[0073] - Using each of at least three linear physical sensor components, at each incremental displacement of each object along the trajectory, acquire one-dimensional radiographic images of the area to be inspected, according to the selected number, such that for each object, the entire area to be inspected is fully represented in the set of one-dimensional radiographic images;

[0074] - Analyze at least three sets of one-dimensional radiographic images of the region to be examined for each object.

[0075] The present invention also relates to a facility for automatically measuring the linear dimension of at least one area to be inspected in a series of manufactured objects, the facility comprising:

[0076] - A means for transporting an object along a straight trajectory in a transport plane in a direction specified by a displacement vector, the object traveling through a transport volume extending in the displacement direction;

[0077] - At least one focal point of the X-ray generating tube is located outside the transverse volume and generates a diverging X-ray beam that is guided through at least one area to be inspected of the object, each focal point being arranged on the same reference straight line parallel to the direction of displacement along a straight trajectory.

[0078] - An image sensor, located outside the transport volume, is used to receive X-rays originating from a relevant focal point. The focal point and the image sensor are configured such that as the object passes through these rays, each image sensor receives a radiographic projection of the area to be inspected via rays originating from the focal point. The projection directions of these radiographic projections are different from each other.

[0079] - An acquisition system connected to an image sensor to acquire a set of one-dimensional processed radiographic images of each object during its displacement, the set comprising:

[0080] • One-dimensional processed ray radiographs of multiple different cross-sectional planes containing the reference line;

[0081] • For each different cross-sectional plane, multiple one-dimensional processed radiographic images of the area to be inspected are obtained along at least three different projection directions in the cross-sectional plane;

[0082] -A computer system configured as follows:

[0083] • For each different cross-sectional plane, determine the outline of the object in the considered cross-sectional plane from at least three one-dimensional processed ray radiographs.

[0084] Other optional but combinable features of the facility according to the invention are elaborated in the following paragraphs.

[0085] The facility may include: at least two X-ray generating focal points, positioned at two different locations on the same reference straight line parallel to the direction of displacement along the straight trajectory; and at least three image sensors, sensitive to X-rays and positioned such that:

[0086] - Each focal point emits a beam that passes through at least the area to be inspected to reach at least one associated sensor;

[0087] - Each sensor is associated with a focal point and receives X-rays originating from that focal point after they have passed through the area to be inspected.

[0088] The facility may include: at least one focal point from which a diverging X-ray beam has an opening of 90° or greater; or at least two focal points from which the sum of the openings of the diverging X-ray beams has an opening of 90° or greater.

[0089] The facility may include at least one focal point located in the transport plane.

[0090] At least one focal point and two image sensors can be configured such that the projection directions of the inspection area they receive have an effective angle greater than or equal to 45° and less than or equal to 90°, and advantageously greater than or equal to 60° and less than or equal to 90°.

[0091] At least one focal point and an image sensor can be configured such that when an object passes through the field of the sensor, the projection direction of the inspection area onto the image sensor forms an angle between the displacement direction and the angle between 10° and 60°.

[0092] Since the focal point of the X-ray generating tube is not in the transverse volume and the image sensor is not located in the transport volume, the projection direction of the inspection area onto the image sensor never forms an angle of less than 10° with the displacement direction.

[0093] The image sensor and focus are configured such that X-rays originating from (one or more) the focus and arriving at the image sensor and passing through the area of ​​the object do not pass through another object at a time.

[0094] The facility may include one to four focal points derived from one or more X-ray generating tubes.

[0095] The number and setting of the image sensors and associated focal points are such that, for each object in the series, the radiographic projection of the area to be inspected onto the image sensor during its displacement has thirty to forty different projection directions, preferably four to fifteen different projection directions.

[0096] Image sensors can be part of a linear type of physical sensor assembly, each physical sensor assembly comprising a linear array of X-ray sensitive elements distributed along a supporting straight line, which, together with the associated focal point, defines a projection plane containing the projection direction. These image sensors are configured such that:

[0097] - Each of these physical sensor assemblies has at least m sensitive elements that receive radiographic projections of the area to be inspected via an X-ray beam originating from the relevant focal point;

[0098] - The projection planes of different sensors are different from each other and are not parallel to the transmission plane.

[0099] In some embodiments, the supporting lines of at least three linear image sensors are parallel to each other.

[0100] In some embodiments, the supporting lines of at least three linear physical sensor assemblies are orthogonal to the transport plane. Attached Figure Description

[0101] [ Figure 1 ] Figure 1 This is a schematic top view showing a facility that allows for X-ray measurement of the dimensions of an object traveling along a straight line.

[0102] [ Figure 2 ] Figure 2 This is a schematic perspective side view showing part of a facility that allows for X-ray measurement of the dimensions of an object.

[0103] [ Figure 3 ] Figure 3 It shows more generally Figure 1 and Figure 2 A schematic perspective view of those types of facilities.

[0104] [ Figure 4 ] Figure 4 It is a schematic perspective view showing the volume that an object traverses or generates during its linear displacement.

[0105] [ Figure 5 ] Figure 5 This is a schematic top view illustrating an exemplary embodiment of a facility comprising two X-ray generating focal points according to the present invention.

[0106] [ Figure 6 ] Figure 6 yes Figure 5 A schematic cross-sectional front view of the facility shown.

[0107] [ Figure 7 ] Figure 7 This is a schematic diagram illustrating the definition of the effective angle between two projection directions.

[0108] [ Figure 8 ] Figure 8 This is a schematic diagram illustrating the definition of the effective angle between two projection directions.

[0109] [ Figure 9 ] Figure 9 This is a schematic perspective view illustrating another example of the positioning of an image sensor relative to the displacement of the object being inspected.

[0110] [ Figure 10 ] Figure 10 This is a schematic diagram of another exemplary embodiment of a facility for implementing a matrix image sensor according to the present invention.

[0111] [ Figure 11 ] Figure 11 It is a view of the X-ray sensing element matrix, corresponding to two different regions of the two matrix image sensors appearing on the X-ray sensing element matrix.

[0112] [ Figure 12 ] Figure 12 This is a flowchart of a method for determining the profile in a cross-sectional plane.

[0113] [ Figure 13 ] Figure 13 This is a flowchart of the method according to the present invention. Detailed Implementation

[0114] As a preliminary matter, some definitions of terms used within the scope of this invention are given below.

[0115] The focal spot Fj of the X-ray generating tube is a point X-ray source, preferably a "microfocal spot," having a diameter, for example, between 0.01 mm and 1 mm, producing a diverging X-ray beam. Any type of point or quasi-point X-ray source can be used.

[0116] The sensitive element of an image sensor for X-rays is an X-ray sensing element, in other words, a basic surface with dimensions of, for example, 0.2 x 0.2 mm or 0.02 x 0.02 mm, that converts the received X-rays into electrical signals. Typically, the sensitive element of an X-ray image sensor includes a scintillator that converts X-rays into visible light, and a photoelectric sensor that converts visible light into electrical signals. Techniques also exist for directly converting X-rays into electrical signals. A pixel specifies a point in a sampled image, characterized by, for example, its grayscale level between 0 and a maximum value. For example, for a 12-bit digital image, pixels use digital values ​​between 0 and 4,095.

[0117] A system for reading or acquiring radiographic images includes one or more X-ray sensitive surfaces, i.e., surfaces comprising one or more sensitive elements that convert X-rays into electrical signals to be transmitted to an analysis system, typically implemented by a computer and specified by the computer system SI in the remainder of the specification. Signals originating from a set of sensitive elements belonging to the same sensitive surface region are acquired by an acquisition device and transmitted together to the computer system, constituting a radiographic image. Preferably, for analysis by the computer system, the radiographic image is converted into a digital radiographic image as close as possible to the sensitive surface (e.g., in electronic circuitry integrated into a physical sensor assembly including the sensitive region), or remotely (e.g., as close as possible to the computer system SI or via the computer system SI).

[0118] Examples of computer system SIs in Figure 3 Symbolically shown in the diagram, and can be made in the form of at least one standard computer, thus including at least one microprocessor, one or more electronic memory units, and one or more displays (screen, projector, holographic display, etc.), inputs (keyboard, mouse, touchpad, touch screen, etc.) and / or communication interfaces (USB, etc.). Computer systems may include, for example, systems connected via the Internet or... The protocol and network are connected to one or more other computers or computer networks that share data with other networks. In addition to its obvious connection to image sensors, the computer system may also be connected to sensors that provide information about the status of the facility, and / or actuators (conveyors, exhausters, etc.) of the facility. The computer system may advantageously be connected to one or more X-ray tubes to acquire operational data from them and / or ensure their inspection. The computer system implements one or more software programs that are stored locally or remotely and / or executed, including implementation on one or more remote computer servers. Preferably, this or these software programs include one or more software programs programmed to implement the method according to the invention.

[0119] An X-ray beam originating from a focal point Fj passes through at least one inspection area and forms a radiographic projection of the inspection area on a sensitive surface. This radiographic projection is sometimes referred to as a radiation image and contains information about the attenuation of the X-rays by the material traversed.

[0120] Image sensors Ci and Cik refer to the X-ray-sensitive surface areas that receive the radiographic projection of the inspection area. Image sensors Ci and Cik are exposed to X-rays originating from the relevant focal point Fj. The image sensors convert this radiographic projection into a radiographic image of the inspection area.

[0121] When the sensitive surface region corresponding to an image sensor Ci contains a single row of photosensitive elements distributed along a supporting straight line segment, the radiographic image acquired during the sensor's integration time (also known as the acquisition time) is linear, and therefore one-dimensional, consisting of a row of pixels forming a one-dimensional value table. Thus, the image sensor Cik is called a linear sensor. Therefore, the sensitive surface region containing the single row of sensitive elements constituting a linear image sensor comprises a linear array of sensitive elements distributed along a supporting straight line segment. According to this definition, any set of sensitive elements aligned in a column or row on the sensitive matrix surface, or aligned sensitive elements (including along a diagonal or another diagonal line), is considered a linear image sensor. Each of several sensitive surface regions on the same surface contains a single row of sensitive elements, and the rows in different regions are different, thus constituting several linear image sensors.

[0122] When the sensitive surface area corresponding to the image sensor Ci contains a two-dimensional matrix of photosensitive elements, the radiographic image acquired during the sensor's integration time is a matrix, and therefore a two-dimensional image composed of a pixel matrix forming a two-dimensional value table. Thus, the image sensor Ci is called a two-dimensional sensor or a matrix sensor.

[0123] When the sensitive surface area corresponding to the image sensor Cik contains a single photosensitive element, the radiographic image acquired during the sensor's integration time is called a point image, which consists of pixels with a single value. Therefore, the image sensor Cik is called a point sensor.

[0124] Within the scope of this invention, image sensors Ci, Cik may correspond to physical sensor assemblies CC1, CC2, ..., CCnmax, a portion of physical sensor assemblies CC1, CC2, ..., CCnmax, or a combination of several physical sensor assemblies or portions of physical sensor assemblies. A physical sensor assembly is an assembly comprising one or more sensitive elements, which, if present, are fixed to each other, and includes a common connection for all its sensitive elements to interface with a computer system. The common connection interface may be analog or digital. The common connection interface typically forms part of the integrated electronic circuitry of the physical sensor assembly. In physical sensor assemblies CC1, CC2, ..., CCnmax comprising several sensitive elements Cik, the sensitive elements are arranged along a line or along a fixed surface, are not modifiable, are typically planar, but sometimes curved. When physical sensor assemblies CC1, CC2, ..., CCnmax contain rows of photosensitive elements, they are called linear sensor assemblies. When a physical sensor assembly contains a two-dimensional matrix of photosensitive elements, it is called a two-dimensional or matrix sensor assembly. When a physical sensor assembly contains a single photosensitive element, it is called a point sensor assembly.

[0125] Therefore, within the scope of this invention, it is understood that image sensors Ci and Cik are collections of one or more sensitive elements Cik that convert X-rays into electrical signals, and these sensitive elements are physical elements. This collection of physical sensitive elements may or may not correspond to physical sensor assemblies CC1, CC2, ..., CCnmax.

[0126] In this invention, a distinction can be made between radiographic images acquired by physical sensor components CC1, CC2, ..., CCnmax and processed radiographic images used by a computer system to calculate contours. These processed radiographic images may correspond to the radiographic images acquired by the physical sensor components, or a portion thereof, or a combination of several images or portions of radiographic images acquired by one or more physical sensor components. Specifically, a one-dimensional processed radiographic image of an object can be formed by sampling a point image acquired using a unique, given sensing element Cik during the object's displacement, during a scan duration corresponding to the duration of the object's displacement between the focal point and the sensing element. This sensing element Cik may correspond to a point physical sensor component, or may belong to a linear physical sensor component, or to a two-dimensional or matrix physical sensor component. Similarly, a two-dimensional processed radiographic image of an object can be formed by sampling a linear image acquired using a linear image sensor Cik during the object's displacement, during a scan duration corresponding to the duration of the object's displacement between the focal point and the linear sensor. The linear image sensor Cik can correspond to a linear physical sensor assembly, or belong to a two-dimensional or matrix physical sensor assembly. The image formed by sampling corresponds to a juxtaposition of several images acquired continuously over time. For radiographic images, the projection directions Dji, Djik are directional directions or vectors originating from the focal point Fj and passing through the center of the image sensors Ci, Cik implemented as the image acquisition sensors, i.e., the center of the X-ray sensitive region that receives the radiographic projection of the examined area during the acquisition time during the object displacement between the focal point and the image sensor. For associated image sensor-focal point pairs, the projection direction is a vector originating from the focal point and reaching the middle of the image sensor. The image sensor is positioned such that the sensitive surface is not parallel to the projection direction. In some cases, it may be advantageous for the sensitive surface of the image sensor to be orthogonal to the projection direction defined by the associated focal point. However, this is not mandatory; for example, if the sensitive surface of the same physical sensor assembly contains several sensitive regions, each forming the image sensor and cooperating for each image capture, each sensitive region has a different focal point and therefore along a different projection direction.

[0127] For a one-dimensional processed radiographic image of an object, it is formed by sampling a point image acquired using the sensing element Cik during the displacement of the object, during a scan duration corresponding to the duration of the object displacement between the focal point and the sensing element. Therefore, the projection direction Djik corresponds to the orientation direction starting from the focal point Fj and passing through the center of the realized sensing element.

[0128] For a one-dimensional processed radiographic image of an object, which is acquired using a linear image sensor Cik within a single integration time of the sensor, the projection direction corresponds to the orientation direction starting from the focal point Fj and passing through the center of the implemented linear image sensor Cik. Therefore, the projection direction Djik associated with the acquired linear processed radiographic image is the direction starting from the focal point and passing through the middle of the supporting straight line segment of the linear image sensor Cik during the image acquisition time corresponding to the integration time of the linear image sensor Cik.

[0129] For a two-dimensional processed radiographic image of an object formed by sampling a linear image, the linear image is acquired using a linear image sensor CCI during the object's displacement, during the scan duration corresponding to the duration of the object's displacement between the focal point and the linear image sensor. Therefore, the projection direction Dji corresponds to the orientation direction starting from the focal point Fj and passing through the center of the realized linear image sensor.

[0130] For a two-dimensional radiographic image of an object, which is acquired using a two-dimensional image sensor Ci, CCi during a single integration time of the sensor, the projection direction Dji corresponds to the orientation direction starting from the focal point Fj and passing through the center of the implemented two-dimensional image sensor.

[0131] If the projection directions Dji and Djik used in pairs together form a minimum angle of at least 3 degrees (preferably at least 5 degrees), then the projection directions Dji and Djik of the radiographic projection are considered to be different.

[0132] A sensitive surface region containing a matrix of sensitive elements constitutes a matrix or two-dimensional image sensor, which includes a matrix array of X-ray sensitive elements distributed in a matrix. For example... Figure 10 As shown, according to this definition, the sensitive matrix surface regions C1 and C1' belonging to the larger sensitive surface Ss are also matrix image sensors. Figure 10 In this example, the sensitive surface Ss corresponds to the sensitive surface of the matrix physical sensor component CC1. In some cases, several sensitive matrix surface regions C1, C1' of the same surface can be processed individually by the acquisition device. In any case, they thus constitute several matrix image sensors, each providing different matrix radiographic images M1, M1' (…). Figure 12 The projection directions D11 and D11' associated with the matrix ray radiograph images M1 and M1', respectively, are directions that start from the focal point F1 and pass through the middle of the sensitive matrix surface regions C1 and C1' during image acquisition. Therefore, the image sensors C1 and C1' may be non-intersecting regions and may be activated sequentially in time.

[0133] Of course, those skilled in the art can use techniques based on a physical matrix sensor assembly of a brightness amplifier or a “screen capture camera”, in which a scintillator plate receives a radiated image, converts it into visible light, and an image visible on the back of the scintillator is captured by a sensitive camera within the emission range of the scintillator (typically the visible light range), and an objective lens is provided where necessary.

[0134] This invention applies to a series of manufactured objects composed of one or more materials, such as objects obtained by methods of these types, including machining, molding, blow molding, sintering, injection, extrusion, or assembly, each of which consists of one or more distinct parts, the number of which is known, and each part is made of a material with a known and uniform attenuation coefficient μ (that is, having the same value at every point in the considered part of the area to be inspected of the object, and preferably, constant over time and the same for the series of objects).

[0135] In some embodiments, objects such as steel or aluminum cast mechanical parts, glass bottles, and plastic packaging can be referred to as single-material objects. In this case, the attenuation coefficient μ is known and uniform, meaning it has the same value at every point in the area to be inspected of the object. However, the present invention can be implemented for multi-material objects. In some cases, different materials have equal attenuation coefficients, so it is not necessary to subdivide the object into different parts, and the object can be regarded as a single-material object, as long as the attenuation coefficient is uniform in the sense that it is uniform across the entire inspection area.

[0136] However, the present invention can also be implemented for the measurement of linear dimensions of multi-material objects. Therefore, such an object is considered to be composed of a combination of compositionally homogeneous volumes, each of which is considered a part of the object. These volumes, or parts of the object, are defined by closed surfaces. For the purposes of this invention, the number of these parts will be considered known, at least for the area of ​​the object to be inspected. This number is countable for the area to be inspected, preferably a small number, for example, less than 50, preferably less than 20, more preferably less than 10, to limit the importance of the calculations and maintain convergence when using iterative adjustment methods.

[0137] Preferably, the topology of these parts of the object is known, i.e., particularly the relevant relative arrangements (the presence of common surfaces, the inclusion or exclusion of different parts in between, juxtaposition and relative positions, the presence of common surfaces, etc.). This is equivalent to saying that the X-rays through the object have passed through a finite number of volumes with different but known attenuation coefficients, and thus the path can be decomposed into segments, each segment connecting two points on the boundary surfaces of the parts belonging to the object, extending through regions of constant attenuation, even if the lengths of these segments are not known a priori. The attenuation of each X-ray depends only on the length of the consecutive segments traversed and the attenuation of each segment. Therefore, information about each radiographic image point is directly related to the actual size of the examined area with a known structure. Furthermore, the geometry of the object, and especially the geometry of its different parts, can be described a priori (before measurement) by a computer through a representation consisting of a set of closed surfaces, which can be called the boundary surfaces of the object. The boundary surfaces of the object are interface surfaces. Thus, the object can have one or more outer boundary surfaces, each of which is an interface between the surrounding air and the material of the part of the object. An object may also have one or more inner boundary surfaces, each of which is an interface between the two materials of two juxtaposed parts of the object. In the case of a hollow object including an inner cavity defined by the object's inner surface, the inner surface is the object's outer boundary surface because it is the interface between the object's material and the surrounding air, even in the case of a closed cavity in which the surrounding air would be trapped.

[0138] It should be noted that the attenuation coefficient μ of the material is strictly dependent on the wavelength λ of the X-rays or the spectral properties μ(λ) of their energy. This property need not be considered, as the X-ray source has its own emitted spectral components, and the attenuation μ can be considered a characteristic of the material with respect to the spectrum of the chosen source. Those skilled in the art will also appreciate how the present invention can be carried out using any method that takes into account the beam's spectrum or hardening attenuation.

[0139] Of course, local and / or temporal variations in the low-amplitude attenuation coefficient μ will not prevent the implementation of this method, but may cause slight or significant loss of accuracy in measurements performed by the facility, depending on the magnitude of the attenuation coefficient μ. Therefore, considering the uniqueness and constancy of material attenuation in each part of a verified object, such small variations are considered possible, for example, due to changes in object composition, manufacturing process parameters, environmental conditions, or X-ray source operation. On the other hand, the invention is not applicable to heterogeneous parts of an object, such as coarse-grained gravel or mortar containing pebbles, as long as the heterogeneous particles and attenuation are greater than the image resolution. The invention is not applicable to the inspection of objects whose shape and contents are unknown beforehand (e.g., luggage). For the same reason, the invention is generally not applicable to medical or biological imaging, except for objects that meet the criteria.

[0140] Compared to the attenuation of the material, the attenuation of air can be considered negligible. In this case, the attenuation of the X-ray beam passing through the object depends on the uniform attenuation of the emitted X-ray spectrum on the one hand, and on the other hand, on the cumulative material thickness traversed. Alternatively, the thickness of the traversed air can be considered large and uniform for all rays, and therefore can be considered known. The attenuation due to air can be subtracted from the measured total attenuation. Thus, for example, it can be considered that the gray level (possibly corrected) in each radiographic image depends only and directly on the total cumulative material thickness traversed. The boundary surface that serves as the transition between air and material can then be accurately determined.

[0141] Therefore, digital analysis of a radiographic image of an object allows us to know the relative positions of a certain number of points on the object's boundary surface in space.

[0142] Digital analysis of the radiographic images of each object may allow the construction of a three-dimensional digital geometric model of each object, referred to as a digital geometric model in the remainder of this specification. Alternatively, this digital geometric model may simply be a stack of two-dimensional digital geometric models. The generation of the digital geometric model involves how three-dimensional objects are represented and processed digitally in terms of mathematics, graphics, and data structures within the memory of a computer system. It should be understood that, in some of these embodiments, the invention allows for the determination of as many three-dimensional digital geometric models as radiographic objects, and the number of radiographic objects may be as many as the number of objects traveling on a transport system. In fact, a key feature of the invention is that it allows for the measurement of each object circulating within a facility, including the measurement of each object circulating within a facility at a high rate, if desired.

[0143] The surface model can be obtained directly from the radiographic image, meaning that there is no need to calculate the volume model.

[0144] In surface modeling, an object is defined by at least one three-dimensional surface, specifically a closed three-dimensional oriented surface corresponding to the outer boundary surface between the object's material and the external environment (typically air). The external environment allows for understanding the concepts of the object's interior and exterior. Different modeling methods can be used for such surfaces, including implicit surfaces, parametric surfaces (planar portions, B-splines, NURBS, etc.), and may be limited by arrays of curves. A simple model is a triangular mesh, which can be viewed as the boundary surface of a volume formed by tetrahedrons.

[0145] The cross-sections of a three-dimensional object, i.e., the intersection of the object with the cross-sectional plane, allow the definition of the object's outline within the cross-sectional plane. The cross-sections of the object's three-dimensional surface, and therefore its boundary surfaces, define one or more two-dimensional curves within the cross-sectional plane; these curves together form the object's outline within the cross-sectional plane. Knowledge of these two-dimensional curves in a series of cutting planes allows for the reconstruction of the three-dimensional surface; of course, the accuracy depends on the number of cross-sectional planes.

[0146] Several methods exist for performing measurements of linear dimensions such as length.

[0147] In a method called the surface method, line segments whose endpoints are the intersections of a straight line with the material / air boundary surface of the surface model can be calculated. Finally, the hybrid method involves converting the volumetric model into a surface model and then applying the second method.

[0148] The third method involves determining the distance between two points of one or two two-dimensional curves in the cutting plane, where any curve is the boundary between the material and the air.

[0149] A three-dimensional point is a point whose coordinates in three-dimensional space are known in any reference frame.

[0150] These three previous methods are examples of determining the distance between two three-dimensional points to determine the results of linear dimension measurements.

[0151] One object of the present invention is to perform more complete measurements than those possible with simple two-dimensional radiographic images. In fact, it is easy to obtain two-dimensional radiographic images corresponding to the projection of the area under inspection using a matrix image sensor, and it is easy to measure dimensions in a plane orthogonal to the projection direction (referred to as the "projection plane"). Similarly, it is easy to obtain two-dimensional radiographic images corresponding to the projection of the area under inspection obtained by juxtaposing continuous image lines acquired during displacement in the displacement direction using a linear image sensor, and it is easy to measure dimensions parallel to the displacement direction in the projection plane. On the other hand, according to the invention, linear dimensions can be measured along directions neither contained in nor parallel to the projection plane. The method according to the invention can practically include reconstructing and measuring dimensions along virtually all directions during processing a combination of radiographic images along at least three different projection directions. This can be achieved by any method that allows the determination of three-dimensional points in space belonging to the boundary surfaces included in the area under inspection of the object. Reconstruction of a three-dimensional model of the area under inspection, of surface or volume type or based on a cutting plane, is one possible method. In practice, it is possible to indirectly determine at least two three-dimensional points, or even more preferably three-dimensional point clouds, along directions that cannot be measured solely from two-dimensional radiographic images, from the three-dimensional volume model of the region to be inspected, and preferably from the three-dimensional surface model of the region to be inspected, by cutting the three-dimensional model of the region to be inspected.

[0152] Therefore, a digital geometric model, composed of geometric elements such as points, line segments, curves, and surfaces, is calculated based on radiographic projection by taking into account the attenuation of at least some X-rays passing through that point on the actual object. The aim is to ensure that the digital geometric model faithfully represents the geometry of the actual object, including deformations relative to the ideal object. In other words, even when these geometric elements cannot be distinguished in any 2D radiographic projection, the coordinates of the geometric elements can be determined by considering that the coordinates have been modified by the radiographic projection. Thus, dimensional measurements on the digital geometric model provide information about the dimensions of each modeled object based on geometric elements that are indistinguishable in any radiographic projection.

[0153] Therefore, one advantage of the method according to the invention is that it allows for the determination of a digital geometric model for each object consisting of at least two three-dimensional points, each of which still belongs to the boundary surface of the area to be inspected, even if the two points are not located in a plane orthogonal to the projection directions Dji, Djik, or in a plane parallel to the displacement direction.

[0154] Of course, the focus of this method is not only on providing measurement results in directions outside the plane orthogonal to the projection directions Dji and Djik and outside the plane parallel to the displacement direction, but also on providing a large number of measurement results distributed throughout the inspection area, thereby providing dimensions between multiple pairs of points in multiple directions. Preferably, the digital geometric model consists of the following:

[0155] - At least two three-dimensional points in space, each belonging to the boundary surface of the region to be inspected, and not located in a plane orthogonal to the projection directions Dji and Djik, nor in a plane parallel to the displacement direction T; and / or

[0156] - At least one three-dimensional surface of the area to be inspected contains points that do not belong to a plane orthogonal to the projection directions Dji and Djik and do not belong to a plane parallel to the displacement direction T; and / or

[0157] - At least one part of the area to be inspected is along a plane that is different from the plane orthogonal to the projection directions Dji and Djik and different from the plane parallel to the displacement direction.

[0158] A geometric model, often referred to as a "prior" geometric model, is a digital geometric model of a set of objects that can be used as initialization for reconstruction software to build digital geometric models of the objects. Its main function is to provide the computer system with information about the shape, geometry, and dimensions of the objects to be modeled computationally and / or different parts of the objects. However, this information is not accurate enough to measure the objects with the required accuracy.

[0159] Because of this information, the following become particularly likely:

[0160] - Attenuation is not modeled based on radiographic images in regions of the image space that are a priori lacking material, because attenuation is considered zero there; and / or

[0161] - Model only the surfaces to be dimensionally measured based on the radiographic images; and / or

[0162] - Only the deviation between the modeled surface from the radiographic image and the theoretical ideal surface is determined.

[0163] In the case of a single material object, the knowledge of the prior geometric model also allows for the determination of attenuation values ​​in the spatial region of the image containing the material from the radiographic image without relying on the prior model, since it is referred to as the material of the object.

[0164] However, it should be understood that, according to the present invention, the measurement results of the object are not derived from the measurement results on the prior geometric model, since the model is known to be independent of the object and to represent a theoretically ideal object that is not actual.

[0165] From the accompanying drawings and more specifically from... Figure 1 and Figure 2 It can be seen that the object of the present invention relates to a facility 1 that allows for the implementation of a method for automatically performing linear dimension measurements on a manufactured object 2 moving at high speed. The present invention relates to an inspection of a series of manufactured objects, known as an "online" inspection, which is considered identical after a conversion or manufacturing step, in order to check the quality of the objects or the conversion or manufacturing method. These objects are considered identical as long as no intentional action has been taken to make them different. However, it is well known that in a series, all objects are not identical due to hazards in the conversion or manufacturing process.

[0166] This method operates on the travel rate of object flow 2. Ideally, facility 1 is capable of processing production at a production rate, such as more than 100 objects per minute, preferably more than 300 objects per minute, and for example at a rate of at least 600 objects per minute.

[0167] However, the computation time may exceed the interval between the two objects. Similarly, the exposure time (also known as the integration time) of the image sensor and the readout sensor may be too long. If the fastest flow cannot be processed by a single facility according to the invention, multiple facilities can be implemented in parallel, each checking a portion of the production. Thus, the production flow can be divided into two or three parallel flows checked by two or three facilities according to the invention. Obviously, the economic benefits of the invention increase if the number of flows, and therefore the number of facilities according to the invention, remains small.

[0168] By measuring the moving object, the present invention brings considerable improvement by avoiding helical scanning and scanning on a turntable (which are unsuitable for production rates), as these two modes imply relative rotation of the object with respect to the focus and / or sensor, resulting in “journey interruption” or very slow displacement of the object within the facility.

[0169] The method according to the invention ensures the measurement of at least one and typically several linear dimensions (i.e., lengths) preferably on each object 2. A linear dimension is essentially the length measured along a line. This line along which the linear dimension is measured can be a straight line or a non-straight line, such as any curve, circular line, dashed line, etc. This line can be a flat line contained in a plane or a three-dimensional line not contained in a plane. Length is a measurement expressed in units of length (e.g., inches or meters). Linear dimensions of the manufactured object include, for example, the diameter, thickness, height, length, width, depth, distance (coordinated as the distance from the point to the origin), and perimeter of the manufactured object. At least one linear measurement of the inspected area is the distance between at least two three-dimensional points, each belonging to the boundary surface (particularly the outer boundary surface) of the area to be inspected and located in a plane, including a plane not orthogonal to the projection directions Dji, Djik.

[0170] According to the present invention, apart from dimensional variations, object 2 should be identical objects, forming a series of objects. In other words, when the objects are conformally identical, the series consists of theoretically identical objects. Dimensional checking involves measuring the actual dimensions and comparing them to the desired dimensions. A priori, any object in the series approximates an ideal reference object with the desired dimensions, but deviates from it due to dimensional variations.

[0171] According to an advantageous feature of the embodiment, at least one region of object 2 is selected as the region to be inspected, so that measurements can be performed on the dimensions of that region of the object corresponding to the dimensional characteristics of the region to be inspected. The object region in which at least linear dimensions are measured is inspected by X-ray. Therefore, the inspection region may correspond to the entire object or one or more regions of the object.

[0172] As indicated, all objects 2 in this series consist of a single part or several different parts, each part being made of a material having a uniform attenuation coefficient at every point of the considered part of the object.

[0173] According to an advantageous variation of the invention, the coefficient is known to the computer system for each portion of the area to be inspected. The method can provide an apparatus for making the attenuation coefficient value of the material available to the computer system. In the sense of a material property, the value can be a spectral value, defining the interaction of the material with radiation and depending on the wavelength of the radiation. In the sense of being independent of the radiation wavelength, the value can be non-spectral. The value can depend on the setup of the X-ray source. It can be provided through various input, communication, and memory devices. For example, the apparatus for making the attenuation coefficient value of the material available to the computer system is a mass storage device, a wired or wireless computer network, or a human / machine interface.

[0174] Facility 1 also includes means 5 for transporting object 2 in the transport plane PC (that is, along a flat trajectory whose direction is embodied by the displacement vector T). Preferably, the trajectory is straight, within limits generally allowing for the straightness of the transport line. Conventionally, the transport means 5 is a mesh or chain conveyor that ensures the linear translation of object 2 placed therein. Thus, objects 2 of the same series are essentially translated within the transport plane PC. Figure 1 and Figure 2 More specifically, for ease of description, it is generally assumed that the displacement direction of object 2 is established along the horizontal axis X of the reference system X, Y, Z, which includes a vertical axis Z perpendicular to the horizontal axis X and a transverse axis Y perpendicular to both the vertical axis Z and the horizontal axis X, and X and Y lie in a plane parallel to the transport plane PC (which is preferably, but not necessarily, horizontal).

[0175] The position of an object considered in a moving orthogonal reference frame translated along direction T is fixed during the object's displacement and the acquisition of radiographic images. Specifically, this fixed position means that there is no rotation of the object in the moving orthogonal reference frame translated along direction T, particularly, for example, no rotation of the object about any possible axis of symmetry. For example, placing the object on a conveyor belt, stably on a bearing, or, if possible, on a clean placement plane, such as the bottom of a container or the leg of a seat.

[0176] In one variant of the invention, a support may be provided for object 2. In this case, the support is fixed in a moving orthogonal reference frame translated along direction T, and it also keeps the object fixed in the moving orthogonal reference frame translated along direction T. According to the first variant, in order to prevent the support from affecting the measurement, the support is excluded from the inspection area so that the support does not appear in the superposition of the projections of the inspection area. According to the second variant, its attenuation coefficient is negligible relative to the attenuation coefficient of the object and can be assimilated to air or zero attenuation. According to a less advantageous third variant, the geometry of the support and its position in the moving reference frame are precisely known and repeatable for a series of objects, and its attenuation coefficient is precisely known and stable, and preferably the same as the attenuation coefficient of the objects in the series, so that the support is taken into account in the reconstruction and isolated from the geometric model of the object.

[0177] The position of the object is stable (during rolling and ray photography acquisition), and it is still preferred that the position is the same for each of the series of objects in a moving orthogonal reference frame translated along direction T.

[0178] If this is not the case, then according to a variation of the invention, means for determining the position of each object relative to a common reference mark of the facility in a moving orthogonal reference frame translated along direction T can be implemented, for example by means of means for calculating the contour of the object along the cross-sectional plane (which will be described later), taking this position into account. This preliminary step includes determining the position of each object. It can include matching an image acquired using a priori contour of the object in a virtual reference frame, which can be derived, for example, from a priori geometric model. In all cases, this is equivalent to determining the contour of the object in a moving orthogonal reference frame translated along direction T, and possibly a 3D model of the object from which it can be drawn.

[0179] On the other hand, it is understandable that if the position of an object is stable in a moving orthogonal reference frame that translates along direction T during rolling and ray photography, it is not necessary to determine the position of each object relative to a common reference frame of the facility.

[0180] like Figure 4 More specifically, during its translation, object 2 generates or passes through a volume called the transport volume Vt. The plane PS is the cutting plane of the transport volume Vt, orthogonal to the transport plane PC and parallel to the displacement direction T. For example, the plane PS can be an intermediate plane that divides the transport volume Vt into two equal sub-volumes. In the case where the transport plane is horizontal, the cutting plane PS is a vertical plane.

[0181] like Figure 1 and Figure 2 As shown, facility 1 also includes at least one focal point Fj of X-ray generating tube 7 (where j varies from 1 to NF), which generates a diverging X-ray beam guided through the transport volume Vt and, more specifically, through at least the area to be inspected of object 2. In variations of the facility that include several focal points Fj, such as... Figure 5 and Figure 6 As shown, all focal points used in the method according to the invention will be arranged on the same reference line B, which is parallel to the displacement direction T along the straight line trajectory. This does not prevent the possibility of having one or more auxiliary focal points (not shown) that can be used to acquire other images.

[0182] Facility 1 also includes image sensors Ci and Cik (where i varies from 1 to N, and N can be greater than or equal to 3 in some cases), which are sensitive to X-rays and are positioned to be exposed to X-rays originating from the focal point Fj and having passed through the transport volume Vt, and more specifically, at least through the area to be inspected of object 2. Of course, tube 7 and image sensors Ci and Cik are located outside the transport volume Vt to allow the object to move freely within that volume. Typically, the X-ray generating tube 7 and image sensors Ci and Cik are housed within an X-ray shielding enclosure.

[0183] In some embodiments, in an implementation of the method, the image sensor Ci is configured such that the image it delivers, and the image considered in the method, is formed solely by rays originating from the associated focal point Fj, and that the image sensor Ci, Cik is associated with a single focal point Fj. For example, the facility can be configured such that only rays originating from a given focal point can reach the associated image sensor, for example by setting an absorption mask in a suitable manner. According to another example that can be combined with the preceding example, it can be specified that image sensor acquisition of an image is triggered only when only a single associated focal point is activated.

[0184] However, in some embodiments, multiple image sensors may be associated with the same focus Fj, and / or multiple focus Fj may be associated with the same image sensor. In a preferred embodiment, multiple image sensors are associated with the same focus Fj.

[0185] As described above, image sensors Ci and Cik correspond to physical sensor components CC1, CC2, ..., CCnmax, ..., and correspond to a part of a physical sensor component, or to a combination of several parts of one or more physical sensor components.

[0186] The X-ray beam originating from the focal point Fj passes through at least the examination area and forms a radiographic projection of the examination area on the image sensor along the projection directions Dji and Djik. Figure 1 and 2 The projection directions Dji and Djik are the orientation directions of the vectors, starting from the focal point Fj and passing through the centers of the image sensors Ci and Cik implemented for acquisition. The focal point Fj and the image sensors Ci and Cik are configured such that each image sensor receives a ray-photographic projection of the area to be inspected along the projection direction.

[0187] Facility 1 also includes an acquisition system connected to image sensors Ci and Cik to acquire a set of one-dimensional processed ray radiographs of each object 2 during its displacement, wherein each one-dimensional processed ray radiograph includes a projection of a cross-section of the object along a cross-sectional plane Pk containing a reference line B. More specifically, the image set includes:

[0188] • This one-dimensional processed ray radiograph image is generated for different cross-sectional planes Pk containing a number of reference straight lines, NK.

[0189] • For each different cross-sectional plane Pk, the number of regions to be inspected obtained along at least three different projection directions Djik in the cross-sectional plane Pk is NP. This one-dimensional processing of radiographic images Spk, Sp'k, Sp"k...

[0190] Therefore, for each of the multiple cross-sectional planes Pk containing the reference line, the image set includes at least three one-dimensional processed ray radiographs Spk of the area to be inspected of the object, each obtained along a different projection direction Djik in the cross-sectional plane Pk.

[0191] Figure 2 The diagram shows the labels of several different cross-sectional planes Pk, Pk' that contain the reference line B and therefore the focus Fj. It should be noted that in embodiments containing several different focuses Fj, since these focuses lie on the same reference line B, a given cross-sectional plane Pk, Pk', Pk'', ..., contains all focuses Fj. It is worth noting that, by definition, the different cross-sectional planes Pk, Pk', ..., are not parallel to each other, but are arranged in a fan shape around the reference line B. Therefore, the reference line B is the intersection of all cross-sectional planes Pk that together form a family of planes.

[0192] Therefore, the cross-sectional plane Pk consists of a reference straight line B and at least one sensing element Cik, Ci'k, Ci"k, ..., of an image sensor contained in the cross-sectional plane Pk, and / or linear image sensors Cik, Ci'k, Ci”k, ..., (e.g., ...) contained in the cross-sectional plane Pk. Figure 9 (as in the example). In the case of a linear sensor, it can be oriented parallel to the reference line B, or along a direction intersecting the reference line B. The reference line B is fixed for the facility, so the cross-sectional plane Pk is associated with the sensing element Cik contained in the cross-sectional plane Pk, or with the linear image sensor Cik contained in the cross-sectional plane Pk. Therefore, for at least some cross-sectional planes Pk, the cross-sectional plane can cut through object 2.

[0193] The intersection of the cross-sectional plane Pk with object 2 defines the cross-section of object 2. As seen above, each cross-sectional plane Pk defines the outline of object 2 in the cross-sectional plane at its intersection with the object, which is formed by one or more two-dimensional curves considered together in the cross-sectional plane.

[0194] It is observed that only one cross-sectional plane Pk is parallel to the transport plane PC. As seen above, at least in some embodiments, the transport plane is considered horizontal. In some applications of the invention, a series of objects can be a series of containers, specifically bottles, particularly glass bottles. For these containers, it is generally observed that they have a central axis along which their overall shape is elongated; this central axis may be an axis of symmetry, or even a rotational symmetry, for some bottles. Typically, these objects are transported with their central axes in a vertical position. As a result, different cross-sectional planes Pk intersect the object by forming different angles with the object's central axis. In the case of objects primarily having a cylindrical shell or outer surface, the cross-sectional plane Pk cuts the object such that the object's outer contour appears as an elliptic curve.

[0195] The cross-sectional plane Pk also defines the projection of the object's cross-section onto the relevant sensitive element Cik contained in the plane Pk, or onto the linear image sensor Cik oriented parallel to the reference line B and contained in the plane Pk.

[0196] In operation, the sensitive element Cik, which intersects with plane Pk, or the linear image sensor Cik, which intersects with plane Pk, allows the acquisition of a one-dimensional radiographic image Spk, which is a projection of the cross section along the plane of the cross section.

[0197] In both cases, whether acquired directly by a linear image sensor or indirectly by sampling a single sensitive element, the one-dimensional radiographic image will be used for processing according to the invention, and will therefore be suitable as a one-dimensional processed radiographic image.

[0198] A single sensing element, Cik, is implemented to acquire this one-dimensional processed radiographic image. Figure 2 , Figure 3 and Figure 5In the example shown, it is necessary to form a one-dimensional processed ray radiograph image by sampling point images acquired using the sensitive element Cik during the object displacement, within a scan duration corresponding to the duration of the object displacement between the focal point and the sensitive element Cik. This juxtaposes several images continuously acquired over time by the implemented sensitive element Cik (which forms a point image sensor) during the scan duration. It is then noted that each image continuously acquired over time by the implemented point image sensor Cik is a pixel of the one-dimensional processed ray radiograph image. The time between two consecutively acquired point images by the implemented point image sensor Cik corresponds to the incremental displacement of the object along the displacement trajectory. Therefore, the scan duration corresponds to the time required for the entire cross-section obtained by the cross-sectional plane Pk to pass through the line supporting the projection direction. It is also noted that in this case, each pixel is obtained with a strictly identical projection direction corresponding to the vector direction connecting the focal point Fj to the sensitive element Cik, since all pixels of the processed ray radiograph image are obtained by the same sensitive element Cik with only time displacement.

[0199] exist Figure 9 In the case where the linear image sensor Cik shown is implemented to acquire such a one-dimensional processed ray radiograph, acquisition can be performed within a single acquisition time or an integration time of the sensor. The one-dimensional linear processed ray radiograph then corresponds to several point images acquired simultaneously, each point image delivered by a set of sensitive elements belonging to a set of sensitive elements continuously aligned on a straight line contained in the cross-sectional plane Pk. It is then noted that in this case, each pixel of the processed ray radiograph corresponds to a different sensitive element, and as a result, each pixel is obtained by projection using the direction corresponding to the direction of the vector connecting the focus Fj to the sensitive element specific to that pixel. However, by convention, the ray radiograph projection direction of this one-dimensional processed ray radiograph is considered to be the direction connecting the focus Fj to the center of the linear image sensor, which is implemented to acquire this one-dimensional processed ray radiograph. Therefore, the projection direction Djik is considered here to be the average projection direction of the one-dimensional processed ray radiograph.

[0200] Each pixel of a one-dimensional processed radiographic image Spk consists of a value representing the signal collected by the corresponding sensing element during the integration time of that pixel. Therefore, the value of this signal depends on the intensity of the received X-rays, and thus on the cumulative attenuation experienced by the X-rays between the focal point Fj and the corresponding sensing element. This depends on the thickness and coefficient of the material, or each layer of material, that the X-rays traverse between the focal point Fj and the corresponding sensing element. Therefore, each one-dimensional processed radiographic image Spk of the area to be examined can be represented by a set of these digital or analog values ​​of the entire cross-section of the object obtained by means of the corresponding cross-sectional plane, acquired over a single acquisition time of a linear sensor or a continuous acquisition time of a point image sensor.

[0201] This invention specifies that, for each object to be measured and each different cross-sectional plane Pk, the contour of the object in the considered cross-sectional plane Pk is determined. In each cross-sectional plane, this determination is performed based on multiple one-dimensional processed radiographic images of the area to be inspected obtained along at least three different projection directions Djik in the cross-sectional plane. Therefore, this determination is performed based on at least three (preferably between 3 and 40, more preferably between 8 and 15) one-dimensional processed radiographic images Spk of the cross-section of the area to be inspected of the object, each obtained along a different projection direction Djik in the cross-sectional plane.

[0202] It should be noted that this must be repeated for each cross-sectional plane that intersects the area to be inspected of the object. Therefore, this must be repeated for NK cross-sectional planes Pk.

[0203] In addition, Figures 1 to 3 The example provides the use of linear physical sensor components ( Figure 2 Only two of these linear physical sensor assemblies are shown in the image. In these examples, each physical sensor assembly CC1, CC2, ..., CCnmax comprises a linear array of X-ray sensitive elements distributed along a supporting line Ln, which defines the associated focal point Fj and projection plane PPji. Figure 2 In this example, the supporting lines Ln of the linear physical sensor assembly are parallel to each other and orthogonal to the transport plane PC. Figure 1In the example, there are eight linear physical sensor assemblies. These physical sensor assemblies CCI are configured such that at least m sensing elements Cik of each of these physical sensor assemblies receive radiographic projections of the area to be inspected of the object via an X-ray beam originating from the focal point Fj during object displacement between the focal point Fj and the sensor. In each cross-sectional plane Pk, each linear physical sensor assembly CCI includes a single sensing element Cik. Therefore, each sensing element Cik of each linear physical sensor assembly CCI is capable of acquiring a one-dimensional processed radiographic image Spk by sampling during a scan duration corresponding to the duration of object displacement between the focal point and the sensing element, thereby juxtaposing several images acquired sequentially over time. It is understood that in a given cross-sectional plane Pk, each sensing element present in that plane belongs to a different linear physical sensor assembly and defines different projection directions by the focal point Fj, and it is also understood that a one-dimensional processed radiographic image Spk is obtained along different projection directions Djik (in this case, along as many different projection directions as the sensing elements in the cross-sectional plane Pk). As described above, the linear physical sensor assembly is preferably configured such that it defines different projection directions Dji, Djik in each realized cross-sectional plane Pk, and these projection directions, taken in pairs, form a minimum angle of at least 3 degrees, preferably at least 5 degrees, in the interface plane Pk between them.

[0204] exist Figure 5 The example illustrates the same principle, but with two foci Fl and F2 aligned on a reference line B, each associated with several linear physical sensor assemblies (five in this case) having supporting lines Li parallel to each other and orthogonal to the transport plane PC. Therefore, for each cross-sectional plane Pk, ten one-dimensional processed ray radiographs Spk are obtained by sampling point images acquired by the sensing elements Cik of the linear physical sensor assemblies included in the considered cross-sectional plane for each linear physical sensor assembly, along ten different projection directions Dji.

[0205] exist Figures 9 to 10The example provided illustrates a physical sensor assembly used as a matrix or two-dimensional sensor assembly. In this example, three such physical sensor assemblies are present. Each physical sensor assembly CC1, CC2, CC3 comprises a matrix array of X-ray sensitive elements distributed along a support plane. In this example, the support planes of the matrix physical sensor assemblies CC1, CC2, CC3 are not parallel to each other. In fact, in this example, each matrix physical sensor assembly CC1, CC2, CC3 is configured such that its support plane is orthogonal to the projection direction defined by the focal point Fj and the center of the matrix physical sensor assembly CC1, CC2, CC3. However, the matrix physical sensor assemblies CC1, CC2, CC3, or at least some of them, may be parallel to each other. In this example, their respective support planes are perpendicular to the transport plane PC. For each matrix physical sensor assembly CC1, CC2, CC3, when the object is located between the focal point Fj and the sensor, at least m rows of sensitive elements Cik receive the radiographic projection of the area to be inspected from the X-ray beam originating from the focal point Fj. In each cross-sectional plane Pk, each matrix physical sensor assembly CCI includes rows of sensitive elements forming a linear image sensor Cik. Therefore, the sensitive elements of the linear image sensor Cik are capable of acquiring a one-dimensional processed radiographic image Spk together within a single acquisition time or integration time of the sensor. Typically, the sensitive elements on the matrix sensor assembly are arranged in a matrix configuration, with the sensitive elements vertically (columns) and horizontally (rows) aligned. When the matrix sensor assembly CCI forms a plane that is not parallel to the reference line B, the intersection of plane Pk and the plane of the matrix sensor element CCI is a straight line that is not aligned with the physical arrangement of pixels. In this case, the linear image sensor Cik is a subset of sensitive elements that do not follow the horizontal or vertical alignment of the matrix sensor assembly arrangement. The linear image corresponding to the linear image sensor Cik (and thus part of the matrix sensor assembly CCI) can be obtained by combining the values ​​delivered by the sensitive elements cut by the virtual straight line representing the linear image sensor Cik, or possibly by considering pixels provided by adjacent sensitive elements. For example, those skilled in the art know that interpolation and resampling are used in the combination of pixels.

[0206] It is understandable that, in a given cross-sectional plane Pk, each row of sensing elements forms a linear image sensor Cik in the matrix physical sensor assembly, and different projection directions are defined by the focal point Fj. It is also understandable that a one-dimensional processed ray radiograph image Spk is obtained along different projection directions Djik (in this case, along as many different projection directions as the matrix physical sensor assembly CCI).

[0207] From the contours of the object in each different cross-sectional plane, at least one measurement result of the linear dimensions of the area to be inspected of the object can be determined for the object to be measured, each of these contours being obtained by processing radiographic images along at least three different projection directions in the cross-sectional plane. For example, such a measurement result of the linear dimensions of the area to be inspected of the object can be determined as the distance between at least two three-dimensional points, each belonging to the boundary surface of the area to be inspected. For this purpose, the acquisition system is connected to a computer system, a non-limiting example of which is shown in... Figure 3 The image is symbolically shown, but it can have all types known in itself. According to an advantageous feature of the embodiment, the computer system uses image sensors Ci and Cik to record radiographic images for each object in the series during its displacement, the radiographic images being derived from a defined number of radiographic projections of the area to be inspected along different projection directions.

[0208] As we have seen, the contour of an object comprises curves or sets of curves representing the intersection of the object's boundary surface with the cross-sectional plane Pk. In such a contour, the curves of the contour, or each curve, are planar curves that can be modeled by a parametric system (particularly a system of one or more parametric equations). Preferably, the curves of the contour, or each curve, are modeled by a fixed number of parameters. Such contour curves can, for example, be polygonal curves. In this case, the coordinates of the polygon vertices can be used as parameters.

[0209] Among the possible methods for determining the contour of an object in a cross-sectional plane, a curve fitting algorithm can be implemented starting from the prior contour of the object in the cross-sectional plane.

[0210] Such algorithms can be iterative algorithms, especially nonlinear recursive algorithms.

[0211] like Figure 12 As shown, the iterative algorithm 100 that may be used can therefore iteratively implement the simulation step 110, the comparison step 120, and the step 130 of reducing the error by adjusting one or more parameters in the parameter system.

[0212] As an initial step 101, such an algorithm can advantageously consider the prior profile DLkl of the object in the cross-sectional plane Pk as the computational profile of the first iteration level. This prior profile DLkl is equivalent to defining an initial curve or set of initial curves in the given cross-sectional plane, preferably sufficiently close to the expected profile, even though this is unknown. Preferably, the initial profile will allow determination of the quantity and order of material cut by multiple X-rays emitted from the focus Fj, which are contained within the considered cross-sectional plane and collected by an image sensor after passing through the object. The prior profile DLk1 can be drawn based on: a prior geometric model of the object, and / or measurement results of the measuring device on one or more objects in the same series, and / or values ​​entered by the operator on the human-machine interface of the computer system and / or drawn diagrams and / or selected shapes. In the case of a series of hollow cylindrical objects comprising only one material and having a theoretical central axis perpendicular to the transport plane, the profile in the cross-sectional plane can consist of initial inner curves and initial outer curves (preferably closed curves), for example, of circular or elliptical type, or polygonal curve type. Such curves are particularly effective for objects rotating about the theoretical central axis. However, they can also be used for objects with prismatic cross-sections obtained by means of a plane perpendicular to the theoretical central axis, and satisfactory results can be obtained.

[0213] The iterative algorithm can then iteratively implement the following steps.

[0214] The iterative step can be a simulation step 110, which involves calculating simulated one-dimensional radiographic images SSpkr, SSp'kr, SSp"kr of the area to be inspected from the calculated profile DLkr in the cross-sectional plane of the object with a given iteration level r, each image being calculated in the cross-sectional plane Pk along one of different projection directions Djik, which are used to acquire one-dimensional processed radiographic images Spk, Sp'k, Sp”k in the cross-sectional plane.

[0215] Therefore, it is assumed here that one of the one-dimensional processed radiographic images Spk, Sp'k, and Sp'k is an image whose actual value of the image signal is known, such as that collected by an image sensor. This image necessarily corresponds to a given projection direction Dijk, and thus to a given focal point Fj and a given linear image sensor Cik. The principle is to calculate estimates of the signal SSpkr, SSp'kr, and SSp”kr at each given iteration level. These signals are collected by the same sensor for an X-ray beam emitted from the same focal point Fj but having passed through the object. The contour of the object will be the calculated contour DLkr at a given iteration level r. During the first iteration, the calculated contour DLk1 of the first contour level, i.e., the initial contour, can be used. For subsequent iterations, the contour calculated during the previous iteration is used.

[0216] At each iteration, the calculation is performed for a plurality of one-dimensional processed radiographs considered by the method in the cross-sectional plane, i.e., at least three, for example, between three and forty (in the sense of including three and forty), more preferably between four and fifteen (in the sense of including four and fifteen) one-dimensional processed radiographs.

[0217] In this way, a comparison 120 can then be performed at each iteration between the simulated one-dimensional radiographic images SSpkr, SSp'kr, SSp"kr and the one-dimensional processed radiographic images Spk, Sp'k, Sp"k. This comparison can be performed, for example, for each simulated one-dimensional radiographic image as a comparison function COMP, such as a function of the difference between the signal value of the simulated one-dimensional radiographic image SSpk and the value of the linear processed radiographic image Spk corresponding to the same radiographic projection. This difference function can be a pixel-to-pixel difference function. The comparison can be performed, for example, by considering several or all of the simulated one-dimensional radiographic images SSpkr, SSp'kr, SSp"kr and the associated one-dimensional processed radiographic images Spk, Sp'k, Sp"k together. The comparison can, for example, include calculating a comparison value COMPVAL, which can be an error value, such as a squared error value.

[0218] It is important to note that this iterative algorithm is applicable to linearly processed ray-photographed images within a given cross-sectional plane Pk, and therefore suitable for images belonging to the same plane. Thus, the calculations implemented in the iterative algorithm involve data belonging to the determined plane Pk, corresponding to the two-dimensional planar entities that constitute the contour. This greatly simplifies the calculations associated with the iterative algorithm, which would involve data corresponding to three-dimensional entities.

[0219] Based on the comparison, the iterative algorithm can then, for example, predict, at each iteration prior to the last iteration, a modification 130 to a new computational profile DLk(r+1) from the computational profile with the considered iteration level r to a higher iteration level (r+1), which will be used for the next iteration and thus have a higher iteration level. The implemented modification can be a modification function MOD, which can take into account the current profile DLkr, the comparison value COMPVAL calculated by the comparison function COMP at comparison step 120, and / or other possible calculations performed in previous iterations, in order to decrease the value of the function COMPVAL in subsequent iterations according to the principles of optimization descent. The modification function that can be used is, for example, least squares, such as linear least squares.

[0220] The above steps can indeed be repeated iteratively until, for the last iteration, the comparison meets the predetermined optimization criterion CRIT. The achievement of this predetermined optimization criterion can be verified during verification step 125. For example, it can be verified whether the verification function returns a limit value. One or different criteria can be verified, such as the value of the function COMP relative to a threshold, the number of iterations r relative to a maximum threshold, etc. This verification can be based on the comparison value COMPVAL, for example, by verifying whether it has reached or exceeded a predetermined squared error value.

[0221] Typically, for each cross-sectional plane, the number of different projection directions Dji, Djik is between thirty and forty (in the sense of including three and forty), more preferably between four and fifteen (in the sense of including four and fifteen). Furthermore, according to an advantageous variation of the embodiment, facility 1 includes between thirty and forty image sensors Ci in a given cross-sectional plane. According to a preferred variation of the embodiment, facility 1 includes between four and fifteen image sensors Ci in a given cross-sectional plane.

[0222] As will be explained in detail in the remainder of the specification, the computer system is programmed to analyze for each object at least three one-dimensional ray-photographed images derived from at least three ray-photographed projections in different directions in each cross-sectional plane, thereby determining the three-dimensional geometry of the object.

[0223] In some cases, it is therefore possible to construct a digital geometric model of each object being measured. This digital geometric model can be generated in any suitable manner, and its accuracy depends on the required accuracy of the distance measurement. Therefore, the digital geometric model can consist of at least two three-dimensional points, each belonging to the boundary surface of the area to be inspected on the object, and not located in a plane orthogonal to the projection directions Dji and Djik, nor in a plane parallel to the displacement direction T. The at least two points can belong to two different boundary surfaces, for example, to measure thickness or air gap.

[0224] The digital geometric model can also consist of one or more (preferably, several) cross-sections of the region to be inspected, each cross-section lying along a plane different from the plane orthogonal to the projection directions Dji, Djik. This cross-sectional plane can be one of the cross-sectional planes Pk implemented for image acquisition, or it can even be different planes. Further, the digital geometric model can consist of at least one three-dimensional surface of the region to be inspected, which is different from the plane orthogonal to the projection directions Dji, Djik and different from the plane parallel to the displacement direction T.

[0225] According to the method described above, for each measured object, the contour of the object is obtained in a series of cross-sectional planes Pk containing the reference line B (and thus in a sector plane surrounding the reference line). Each contour can be obtained in the form of a parametric system, for example, in the form of a set of points and / or a set of line segments, particularly in the form of a set of points and / or line segments belonging to the outer boundary surface of the object. The set of contours thus obtained can be considered as a geometric model of the object obtained by measurement. Alternatively, the geometric model of the object can be constructed from this set of contours, for example, by interpolation methods. Thus, the contours in plane Pk can be merged into 3D curves (boundaries), for example, to obtain an STL-type model. The 3D surface model thus obtained can then be cut again by cutting the plane corresponding to the plane on which the measurement was performed.

[0226] In summary, the above method allows the use of a computer system to construct a three-dimensional digital geometric model of the area to be inspected for the object to be measured, based on the object's contour in each different cross-sectional plane Pk. This three-dimensional digital geometric model includes:

[0227] - Three-dimensional points in space, each point belonging to the boundary surface of the area to be inspected of the object; and / or

[0228] - At least one three-dimensional surface of the area to be inspected.

[0229] Given a defined three-dimensional geometric model, the linear dimensions of the area to be inspected can be determined by measuring the distance between at least two three-dimensional points of the three-dimensional digital geometric model of the area to be inspected.

[0230] Of course, this invention allows for the construction of digital geometric models using a large number of three-dimensional points or three-dimensional point clouds.

[0231] Digital geometric models are constructed by using the attenuation coefficients of the materials of the objects in this series.

[0232] In some embodiments of the invention, it has been seen that the profile in each cross-sectional plane, and therefore possibly a digital geometric model, can be constructed using a prior geometric model of the region to be inspected from the series of objects. In other words, in this case, the computer system constructs a digital geometric model of each object using, on the one hand, a prior geometric model of the region to be inspected from the series of objects, and on the other hand, using the attenuation coefficient of the material or different attenuation coefficients of different parts of each object in the series.

[0233] Therefore, for this calculation operation, the computer system considers the attenuation coefficient of the material of the object under inspection. Advantageously, facility 1 includes means for making the attenuation coefficients of the materials of a series of objects available to the computer system.

[0234] The supply device can be made from a large-capacity memory, a human-machine interface, or a wired or wireless computer network.

[0235] Similarly, in some embodiments, the computer system has a geometric model, referred to as a prior geometric model of the region to be inspected, in order to perform the computational operation. Therefore, facility 1 may include means for making a prior geometric model of the region to be inspected of a series of objects available to the computer system.

[0236] Devices used to make the prior geometric model of the area to be inspected available to a computer system include, for example, a mass storage device, a wired or wireless computer network, or a human-machine interface.

[0237] As defined in the definition section, the prior geometric model is a digital model of a series of objects that can be positioned in a reference frame linked to the device and can be used as an initialization for determining the contours of the objects in each cross-sectional plane Pk.

[0238] Without prior geometric models, reconstruction can be computationally very expensive because attenuation must be calculated for every point in 3D space. Therefore, implementing prior geometric models allows for the measurement of linear dimensions of objects with good accuracy in a very short time and at a low cost.

[0239] According to the first variant, the prior geometric model is obtained through a series of computer-aided design digital models of objects generated during the object design (3D CAD). In this case, it can be available to the computer system, for example, through various possible means, such as by means of a computer network connected to a database containing several CAD models corresponding to various series of objects that can be measured in production, which can be selected by the operator from a database within the facility, and so on.

[0240] According to the second variation, the prior geometric model is obtained from a digital geometric model, which is constructed based on measurements of one or more objects in the same series by a measuring device (e.g., a sensor measuring machine or an axial tomographic scanning device with a slower reproducibility compared to the present invention). The prior geometric model can be constructed by combining the measurement results of several manufactured objects in the same series.

[0241] According to the third variant, the a priori geometric model is a digital geometric model generated by a computer system from values ​​and / or diagrams and / or shapes selected by the operator on the system's human-computer interface.

[0242] For example, to provide a priori geometry for a standardized type of nut with threaded holes and six outer M13s, the following approach is sufficient: The operator inputs the number and height of the sides, diameter, and pitch on the keyboard, and the system is configured to inspect metric nuts. No additional specific sides are specified. In another example, for the inspection of a single-material container made of glass or a synthetic polymer material (e.g., polyethylene or polyester), the operator provides only the following information: the object is a cylinder with a closed bottom and a conical top; two diameters, two heights, and one thickness are sufficient for the computer system to know the priori geometry of the object to be inspected. According to another example, the computer system can receive a technical description of the priori model through its interface, such as the number, diameter, depth, and location of various holes present in the surface that will be part of the area to be inspected of a larger object. This description can be geometric; for example, the computer system receives information describing the number and overall appearance of its boundary surfaces, the number of cavities, or the number of faces or edges of a polyhedron. In summary, it should be understood that the prior geometric model must contain at least sufficient technical, geometric, topological, and / or digital information to inform the computer system of the 3D structure of the object. The exhaustiveness and accuracy of this information can be very low without compromising the accuracy sought by linear measurements.

[0243] One of the advantages of determining the geometric model is that, for each object in the series, at least one linear measurement of the area to be inspected along any direction can be determined from the digital geometric model of the area to be inspected corresponding to the object in the series by the same computer system or another system that will provide the model, so that the direction is not necessarily contained in a plane orthogonal to the projection direction, nor necessarily contained in a plane parallel to the displacement direction.

[0244] Inspect at least one dimension (and usually several dimensions) on object 2. The purpose is typically to compare the measurements obtained on the object with desired values, for example, as defined by the quality department. These dimensional measurements, or their deviations from the desired values, can be displayed, recorded, etc. They can also be used to make decisions regarding the consistency of objects that can be automatically categorized.

[0245] Measurement results can be derived from measurements taken from a digital geometric model of the inspection area established for each object. For example, the inspection area may include holes. In the digital geometric model, the diameter or depth of a hole can be determined by calculating the distances between surface elements with opposite diameters. When the object is made of a single material, the positions of surface elements can be determined more accurately with minimal computation.

[0246] Another device for determining the measurement results of a hole's diameter or depth is to compare a digital geometric model of the area being inspected with a reference or theoretical geometric model.

[0247] A reference geometric model is an ideal model of a series of objects to be inspected. To perform dimensional checks, the digital geometric model of the area to be inspected is compared to the reference geometric model using an algorithm that includes model matching, and then the measured deviations between the models are compared. The reference geometric model can be exported from CAD.

[0248] Therefore, an operation can be performed to match the digital geometry model of the inspection area with a reference geometry model, and then the dimensional deviation can be determined by measuring the distance between surface elements belonging to the reference model and surface elements belonging to the digital geometry model. In the example of hole measurement, the cylinder with the largest diameter inscribed in the modeled inner surface of the hole, and the cylinder with the smallest diameter encompassing the modeled inner surface, can be virtually located, and the diameters of the inscribed and circumscribed cylinders can be considered as the measured diameter of the hole in the inspection area. This type of analysis can also verify the inner diameter of the neck of glass or plastic bottles obtained by pressure blowing or blow molding processes.

[0249] According to a variation of the invention, the reference geometric model and the prior geometric model are the same geometric model.

[0250] According to another variation of the invention, the prior geometric model is less accurate, less complete, and / or different from the reference geometric model.

[0251] In order to perform such measurements, the facility advantageously includes means for making the linear dimensions and / or the tolerances of these dimensions and / or the values ​​of the reference geometric model available to a computer system.

[0252] According to an advantageous feature of the embodiment, the computer system is connected to means for displaying linear measurements of the area to be inspected and / or dimensional deviations relative to reference values ​​and / or deviations between a digital geometric model of the area to be inspected and a reference geometric model. For example, for nuts, measurements such as thread depth, average pitch, average thread root radius, height, minimum or maximum inner diameter, and the flatness of one or more of its outer surfaces are displayed. For glass or plastic containers, the system displays the total height, and, for example, the minimum and maximum diameters of cylindrical portions at a predetermined height determined by the setting of the side to be verified. The side can be displayed in different colors depending on whether it conforms or not.

[0253] According to an advantageous feature of the embodiment, the computer system is connected to a device for classifying objects based on linear measurements of the area to be inspected. Thus, the classification device can, for example, use a discharge device to remove objects deemed defective from the transport device, taking into account the measured linear dimensions.

[0254] According to an advantageous feature of the embodiment, the computer system can be connected to a device for marking objects based on linear measurement results of the area to be inspected. This marking device can record, for example, the measured linear dimensions of the object, or conformal or defect conditions.

[0255] The computer system knows the relative positions of the focal point Fj and sensors Ci and Cik in the facility's fixed reference frame X, Y, Z. These positions can be obtained through assumption or calibration. Calibration includes, for example, placing or transferring precisely machined gauges within the facility.

[0256] Of course, considering that the focal point Fj and image sensors Ci and Cik are located outside the transmission volume Vt, the relative positions of the focal point Fj and image sensors Ci and Cik are different.

[0257] According to a variation of the embodiment, facility 1 includes a single focal point Fj = Fl disposed along one side of the transport volume Vt and a series of linear physical sensor assemblies CC1, CC2, ..., CCnmax, whose supporting lines are configured perpendicular to the transport plane. For multiple cross-sectional planes, each linear physical sensor assembly includes an image sensor disposed along the opposite side of the transport volume Vt to receive rays from focal point F1 that have passed through the area to be inspected. In this example, the focal point has at least in any plane (e.g., such as...) Figure 1 An opening of 120° or greater, measured in the planes X and Y, is considered to be at the output of the focal point. In cases where the facility includes a screen (which limits the beam to the only useful beam) located between the focal point and volume Vt or between volume Vt and the image sensor to reduce scattered beams, this opening of 120° is considered to be at the output of the focal point.

[0258] According to another variation of the embodiment, at least two X-ray generating focal points Fj (F1 and F2) are respectively positioned at two different locations along a reference straight line B parallel to the object's linear trajectory, and at least three physical sensor assemblies (each physical sensor assembly including an X-ray sensitive image sensor for multiple cross-sectional planes) are positioned such that each focal point is associated with at least one image sensor in each cross-sectional plane, and each image sensor is associated with a focal point and receives X-rays originating from said focal point and passing through the area to be inspected. In this example, each focal point has an opening of greater than or equal to 60°, such that the sum of the openings of the two focal points is greater than or equal to 120°.

[0259] exist Figure 5 and Figure 6In the exemplary embodiment shown, facility 1 includes two focal points F1 and F2 aligned on a reference straight line B parallel to the trajectory of object 2. Each of the two focal points F1 and F2 is associated with a different generator tube 7. Facility 1 also includes five linear physical sensor assemblies CC11, CC12, CC13, CC14, and CC15, whose supporting lines are configured perpendicular to the transport plane. For multiple cross-sectional planes, each linear physical sensor assembly includes an image sensor, each image sensor being sensitive to ray X originating from a first associated focal point F1. Facility 1 also includes five linear physical sensor assemblies CC21, CC22, CC23, CC24, and CC25, whose supporting lines are configured perpendicular to the transport plane. For multiple cross-sectional planes, each linear physical sensor assembly includes an image sensor, each image sensor being sensitive to ray X originating from a second associated focal point F2.

[0260] According to this exemplary embodiment, it should be noted that the focal points from which the diverging X-ray beam originates (F1 and F2 in this example) are positioned on one side of the cutting plane PS such that the beam passes through the cutting plane PS and the area to be inspected, while at least one image sensor Ci associated with said focal point Fj for receiving X-rays originating from said focal point Fj is positioned on the opposite side relative to the cutting plane PS. (In this example, there are five image sensors C11, C12, C13, C14, and C15, each sensitive to X-rays originating from the associated focal point F1, and five image sensors C21, C22, C23, C24, and C25, each sensitive to X-rays originating from the associated focal point F2).

[0261] According to an advantageous variation of the embodiment, the focal points Fj are disposed in the transport plane PC. Preferably, these focal points cooperate with an associated image sensor located opposite the cutting plane PS. Therefore, in the case of transporting an object disposed on a planar conveyor, this arrangement allows the projection of the object in the radiographic image to not be superimposed on the projection of the conveyor. Thus, in the digital geometric model of the object, the portion of the object in contact with the conveyor can be accurately determined.

[0262] An advantageous feature of the embodiment is that the image sensor Ci and the focal point are configured such that X-rays originating from the focal point Fj and arriving at the image sensor Ci pass through the area to be inspected only at a time. In other words, the X-rays can only pass through one object at a time. It should be noted that the facility may include a system for inspecting the spacing between consecutively moving objects.

[0263] One object of the present invention is to obtain a method that is not only fast but also inexpensive, capable of calculating the three-dimensional geometry of each transported object on a line with the accuracy required for dimensional inspection. The invention aims to reduce the number of images required for reconstruction to a minimum that allows the desired dimensional accuracy to be achieved. For example, the invention allows the measurement of the inner diameter of a cylinder between 10 and 120 mm with an accuracy of + / -0.05 mm using nine projections of the inspection area and a limited number of images. Advantageously, the facility according to the invention comprises one to four focal points Fj, and preferably one or two focal points Fj, and preferably multiple image sensors for acquiring radiographic projections in four to fifteen different projection directions in each cross-sectional plane.

[0264] According to the invention, taking into account that the transverse volume Vt should be unrestricted for the sake of object circulation, the image sensor and focus should be configured such that a combination of at least three projection directions optimizes the determination of the contour of the inspection area of ​​the object being transported online on each of a plurality of cross-sectional planes including a reference straight line B parallel to the object's straight trajectory T. Within the scope of the invention, the following rules are advantageously implemented, which are effective for linear or matrix image sensors.

[0265] In the following text, angles are absolute values. Figure 7 and Figure 8 Two projection directions, Dji and D'ji, are shown; they are also vectors. Here, we consider two projection directions, Dji and D'ji, contained within the same cross-sectional plane Pk. These figures illustrate the angle α between these two projection directions, i.e., And the angle s complementary to angle a, i.e., s = 180° - a. By definition, the effective angle α between two different projection directions Dji and D'ji in the cross-sectional plane Pk is the smallest of angles a and s, i.e., α = Min(a,s). Therefore, the effective angle α is the smallest angle formed by two straight lines with projection directions Dji and D'ji.

[0266] According to an advantageous variation of the invention, in at least three radiographic images derived from radiographic projections along at least three different projection directions, for each object and for each cross-sectional plane Pk, at least two images are acquired from two radiographic projections along two different directions Dji and D'ji, forming an effective angle α greater than or equal to 45° and less than or equal to 90° between the two different directions Dji and D'ji. According to another advantageous variation of the invention, in at least three radiographic images derived from radiographic projections along different directions, for each object and for each cross-sectional plane, at least two images are acquired from two radiographic projections along two different directions, forming an effective angle α greater than or equal to 60° and less than or equal to 90° between the two different directions.

[0267] Therefore, the facility 1 according to the invention includes at least one focal point and two image sensors, which are configured such that the projection directions of the inspection area they receive have an effective angle α greater than or equal to 45° and less than or equal to 90°, and advantageously greater than or equal to 60° and less than or equal to 90°.

[0268] For example, such as Figure 5 As shown, the effective angle α between directions D15 and D11, and between directions D13 and D25, is greater than 45°. It is clear that at least one effective angle is greater than or equal to 45° and less than or equal to 90°, and advantageously, at least one effective angle is greater than or equal to 60° and less than or equal to 90°, while the other effective angles between the two directions Dji and D'ji are arbitrary. Those skilled in the art using this rule will know how to find a setting that provides the most complete possible distribution of the projected directions of the inspection area.

[0269] In certain cases, the focus can be specified to be in the transport plane PC, and the physical sensor assembly is positioned such that the transport plane PC is one of the cross-sectional planes Pk. Then, in the cross-sectional plane Pk corresponding to the transport plane PC, the condition applies that, for each object and for the horizontal cross-sectional plane Pk coinciding with the transport plane PC, at least two images are acquired from two radiographic projections along two different directions, forming an effective angle α greater than or equal to 60° and less than or equal to 90° between the two different directions, in at least three radiographic images derived from radiographic projections along different directions.

[0270] Preferably, the effective angle α between two consecutive projection directions Dijk in a given cross-sectional plane Pk is the same for all consecutive projection directions Dijk implemented to acquire a one-dimensional processed image in the given cross-sectional plane Pk. In other words, the projection directions Dijk implemented to acquire at least three one-dimensional processed images in the given cross-sectional plane Pk are angularly and uniformly spaced in the given cross-sectional plane Pk. Preferably, the projection directions implemented to acquire at least three one-dimensional processed images in the given cross-sectional plane Pk are angularly distributed in the cross-sectional plane Pk so as to cover the angular amplitude measured in the cross-sectional plane Pk and covered by the opening Of of the focal point Fj, preferably, so as to cover at least 50% of the angular amplitude, preferably at least 75% of the amplitude.

[0271] According to another advantageous feature, for each object, the computer system acquires at least one radiographic image of the inspection area corresponding to the projection direction, which forms an angle β determined by the displacement direction T on an orthogonal projection on the transport plane PC.

[0272] like Figure 9 As shown, consider the angle p between the projection direction (vector Dji) that produces an orthogonal projection on the transport plane PC and the trajectory of the object (vector T), that is, angle p = (Dji, T), which means that in Figure 9 In the example shown, p = (D11, T) and p = (D12, T). An angle q complementary to angle p makes q = 180° - p. By definition, the angle β between the projection direction Dji that produces an orthogonal projection on the transport plane PC and the trajectory T is the smallest of angles p and q, i.e., β = Min(p, q). Therefore, the angle β is the smallest of the angles formed by two straight lines, one carrying the projection direction Dji that produces an orthogonal projection on the transport plane PC, and the other carrying the trajectory T.

[0273] According to another advantageous feature, for each object, the computer system acquires at least one radiographic image of the examination area corresponding to the projection directions Dji, Djik, which have an angle β between 10° and 60° with the displacement direction T. In other words, the facility according to the invention includes at least one focal point and an image sensor Ci, which is configured such that when the object passes through the field of the image sensor, the projection directions Dji, Djik of the examination area on the image sensor Ci form an angle β between 10° and 60° with the displacement direction T.

[0274] In other words, the configuration of facility 1 is optimized to reduce its volume in the displacement direction, while maintaining the transverse volume Vt suitable for the object and maintaining good reconstruction quality.

[0275] Because it traverses the volume Vt, the facility does not produce a projection around the displacement direction T. The traverse volume Vt produces a minimum subtraction angle βmin. For example, the minimum subtraction angle βmin = 10°. There is no sensor configured to provide a projection with a subtraction angle β less than 10°.

[0276] As can be seen from the above, the distribution of the projection angle of each object is not necessarily uniform.

[0277] like Figure 9 As shown, the distribution of projection angles can present gaps of twice 2x10° (i.e., 20°), known as blind spot areas, rather than complete coverage with more than 180°.

[0278] For example, such as Figure 9 As shown, the facility according to the invention includes at least one focal point Fl and two physical sensor assemblies CC1, CC2, here three and, for example, two-dimensional sensor assemblies. For multiple cross-sectional planes Pk, each physical sensor assembly includes image sensors C1k, C2k, C3k. For the image sensors, the projection directions D11, D12 that produce orthogonal projections on the transport plane PC and the displacement direction T define an angle β ranging from 10° to 60°, corresponding to angles p and q, respectively. Similarly, Figure 5 The facility shown includes an image sensor CC11 associated with a focal point F1, whose projection direction D11 forms an angle β between 10° and 60° relative to the displacement direction T.

[0279] The physical sensor components that form an image sensor Ci are typically matrix or linear.

[0280] According to a preferred variant of the embodiment, facility 1 includes linear physical sensor assemblies. According to this preferred variant, each physical sensor assembly CCI, CCI' includes a linear array of X-ray sensitive elements distributed along supporting lines Li, Li', which, together with the associated focal point Fj, define projection planes PPji, PPji' containing the projection direction Dji. Figure 2These physical sensor assemblies Ci are configured such that at least m sensitive elements of each of these image sensors receive radiographic projections of the area to be inspected via an X-ray beam originating from a relevant focal point Fj, wherein the projection planes PPji for different physical sensor assemblies are different from each other and not parallel to the transport plane PC. The number of sensitive elements m in each physical sensor assembly is greater than 128, preferably greater than 512. The distance between adjacent sensitive elements (referred to as the spacing) and / or the size of the sensitive elements is preferably less than 800 μm. The frequency of the image reading lines is preferably greater than 100 Hz, advantageously greater than 1 kHz. Of course, these parameters can be adjusted according to the size of the object, the required accuracy, and the travel speed. Thus, by properly arranging the physical sensor assemblies, planes Pk perpendicularly spaced by the spacing value can be obtained, the spacing distance being measured at the sensor. The number of cross-sectional planes Pk can reach 128, or even 512, or even more.

[0281] According to an advantageous feature of the embodiment, the supporting lines Li of at least three linear physical sensor components Ci are parallel to each other.

[0282] According to another advantageous feature of the embodiment, the supporting straight line Li of at least three linear physical sensor assemblies Ci is orthogonal to the transmission plane PC.

[0283] According to one variation, the focal point Fj is positioned such that its beam passes through the inspection area and then through the transmission plane PC. Furthermore, at least one associated linear physical sensor assembly Ci is positioned opposite the focal point Fj relative to the transmission plane PC, such that its supporting line Li is parallel to the transmission plane PC.

[0284] According to these variations of embodiments with linear physical sensor assemblies, the acquisition system uses each of at least three physical sensor assemblies Ci to acquire linear radiographic images of the area to be inspected at each incremental displacement of the object along the trajectory, based on a selected number, such that for each object, the entire area to be inspected is completely represented in a set of linear radiographic images. Thus, during object displacement, each image sensor is able to acquire linear radiographic images, thereby completely representing the entire area to be inspected of the object in a set of linear radiographic images obtained from the image sensors. Therefore, for each object, at least three sets of linear radiographic images of the area to be inspected are obtained and then analyzed. A matrix radiographic image of the area to be inspected can be constructed by juxtaposing the sets of linear radiographic images. However, the reconstruction and measurement of the geometric model are not necessarily imposed upon it.

[0285] It should be noted that, given a transverse volume Vt, no radiographic projection is acquired in the blind spot region (β < ±10°) on either side of the displacement direction T. Although no radiographic projection is acquired within this angular range, the method according to the invention allows for the reconstruction of an accurate and complete digital geometric model of the object, for example, due to the prior geometric model. Therefore, linear dimensions can be measured on the entire digital geometric model, and particularly along directions not orthogonal to possible projection directions, including measurements of linear dimensions along measurement directions orthogonal to the missing projection directions corresponding to the blind spot region on either side of the displacement direction T. In fact, without the method according to the invention, for example using methods intended for conventional "complete" axial tomography, if no radiographic projection is acquired along the blind spot direction, the reconstructed model will also exhibit reconstruction errors in angular sectors orthogonal to the blind spot, making it impossible to accurately determine the surface, and thus impossible to measure the linear dimensions of the object, even for a single material object.

[0286] Therefore, as Figure 10 and 11 As shown, according to the present invention, projection is impossible in a blind spot of, for example, 20° (βmin = 10°). According to the prior art, accurate measurements cannot be made on direction A, which is not orthogonal to any projection direction. In the sense that direction A has an angular deviation of at least 10 degrees from any projection direction in orthogonal projection on the transport plane PC, direction A is not close to orthogonality to any projection direction within at least 10°. Through reconstruction from a priori models of a series of objects and constant and uniform attenuation, the measurements of the inner diameter along direction A (distance a1) and the distance between the two outer sides perpendicular to direction A (distance a2) are correct and accurate. In other words, the three-dimensional geometry determined for the area to be inspected has no missing or blurred boundaries in direction A.

[0287] Of course, the number of focal points, the number of image sensors associated with each focal point, and their relative settings can be chosen in any suitable manner based on the desired measurement accuracy, the shape of the objects, and their spacing on the conveyor.

[0288] In short, such as Figure 13 As illustrated schematically, the method of the present invention provides:

[0289] -Transportation of objects (TRANS OBJ)

[0290] - During this transmission process, image sensors Ci and Cik acquire a set of one-dimensional processed radiographic images of each object during its displacement (ACQUIMAGE), including:

[0291] • One-dimensional processed ray-photographs of different cross-sectional planes Pk containing NK reference straight lines;

[0292] • For each different cross-sectional plane Pk, the number of one-dimensional processed radiographic images Spk of the region to be inspected obtained along at least three different projection directions in the cross-sectional plane is NP;

[0293] - For each object to be measured and each different cross-sectional plane Pk, the contour DLk of the object in the considered cross-sectional plane Pk is determined using a computer system from at least three one-dimensional processed radiographic images Spk of the area to be inspected obtained along at least three different projection directions Dijk in the cross-sectional plane.

[0294] - For the object to be measured, determine at least one linear dimension measurement result of the area to be inspected of the object from the contour of the object in each different cross-sectional plane (MES).

[0295] This method allows for real-time computer processing of images to create 3D models for measuring rapidly moving objects in production. Setting up translational movement allows manipulation within a cross-sectional plane Pk. A reconstruction algorithm can be used in Pk, which manipulates 2D geometry by computer, much faster than manipulating 3D data. Therefore, parametric curves can be manipulated by computer, much faster than manipulating the fundamental volume elements of a volumetric model. The method can consider prior information about the object to further limit the computational power required for implementation. Although the image lacks a projection direction close to the displacement direction, the proposed method also allows reconstruction in directions orthogonal to the displacement without regions known as "missing edge" error areas.

[0296] Because of the acquisition geometry proposed by this method, the problem of 3D reconstruction of an object from acquired radiographic data is decoupled into an NK-order reconstruction problem of the object's tilted 2D cross-section, which limits the computational power required for implementation. In fact, the problem of 3D reconstruction from A acquisitions on C linear sensors generates AxCxNK data points, and by solving a linear system of order AxCxNK multiplied by AxCxNK, it becomes possible to estimate voxels of order AxCxNK in an algebraic method. The cost of this operation is calculated using a Gaussian method according to order (AxCxNK)^3, and the iterative method utilizing the hollow matrix property is iter3Dx(A+C+NK)xAxCxNK, where iter3D iterations are performed.

[0297] In one method for solving the NK - fold reconstruction problem of an object's decoupled tilted 2D cross - sections, NK linear systems of order AxC by AxC should be solved. That is, the cost is (obtaining the NK^2 factor) in accordance with the order NKx(AxC)^3 by Gaussian - type methods and iter2D*(A + C)xAxCxNK, where usually iter2D < iter3D, and of course (A + C) < (A + C+NK). Therefore, this method requires less computation and thus less computing power. This increase in complexity is because curves can be reconstructed in each 2D cross - section, which is much simpler than reconstructing a 3D surface: the amount of data in local memory is lower, and the complexity of curve management is much lower than that of a surface.

[0298] It should be noted that since the object's trajectory is a translation, and since each cross - section plane contains a reference line and is thus parallel to the object's trajectory, each cross - section plane Pk is invariant with respect to the object during translation. Therefore, all one - dimensional processing radiograph images Spk of the region to be inspected obtained in the same cross - section plane Pk are one - dimensional radiograph images of the same cross - section (and thus, the same contour) of the object. Thus, it can be seen that the acquisition geometry proposed by this method allows direct processing of data belonging to the same plane, which greatly simplifies the calculation.

[0299] It should be noted that in industrial mass production, there may be several series present simultaneously on the same manufacturing or inspection line. In this case, the facility includes an indication system for indicating to the computer system the series to which each object belongs, in order to implement the method of the present invention for all objects of the same series. In fact, the facility according to the present invention can be used to inspect a flow of manufactured objects consisting of objects of many different series (e.g., a first series and a second series). These series may differ in the shape of the objects or their own attenuation coefficients, or both. In this case, the facility must be equipped with means for making available to the computer system the prior geometric model of each series of objects, the attenuation coefficients of each series of objects, and a means for associating the radiograph image of each object in the computer system with the series to which it belongs.

Claims

1. A method for automatically measuring linear dimensions of a series of manufactured objects (2), comprising: - selecting a series of manufactured objects (2), wherein each of said objects is made up of one or several distinct portions, the number of said portions being known, and each portion being made of a material having a known attenuation coefficient, said attenuation coefficient being uniform at each point of said portion of the object; - transporting the objects by means of transport means in a displacement direction (T) along a rectilinear trajectory in a conveying plane (PC), these objects generating a conveying volume (Vt) during their displacement; - arranging outside said conveying volume (Vt): ■ at least one focal point (Fj) of an X-ray generating tube, each focal point being arranged on a same reference straight line parallel to the displacement direction (T) along said rectilinear trajectory, and ■ one or several image sensors (Ci), each image sensor being exposed to and sensitive to X-rays originating from the associated focal point (Fj), these X-rays having at least passed through a region to be examined, generating on each image sensor a radiographic projection of the region to be examined along a projection direction (Dji, Djik); - using said image sensors (Ci, Cik) to acquire, for each object, a set of one-dimensional processed radiographic images during the displacement of the object, each one-dimensional processed radiographic image comprising a projection of a cross section of said object along a cross section plane (Pk) containing said reference straight line, said set comprising: ■ said one-dimensional processed radiographic images for a plurality (NK) of distinct cross section planes (Pk) containing said reference straight line; ■ for each distinct cross section plane (Pk), a plurality (NP) of said one-dimensional processed radiographic images (Spk) of the region to be examined obtained along at least three distinct projection directions (Dijk) in said cross section plane; - using a computer system to determine, for each object to be measured and for each distinct cross section plane (Pk), a contour of said object in the considered cross section plane (Pk) formed by one or more two-dimensional curves from said one-dimensional processed radiographic images (Spk) of the region to be examined obtained along said at least three distinct projection directions (Dijk) in the cross section plane, and - determining, for an object to be measured, at least one measurement of a linear dimension of the region to be examined of said object to be measured from the contours of said object to be measured in each distinct cross section plane determined from the one-dimensional processed radiographic images of the region to be examined.

2. The method of claim 1, wherein, The contour of said object comprises a curve or a set of curves representing the intersection of the boundary surface of said object with said cross section plane.

3. The method of claim 2, wherein, The curve or each curve of the contour of said object is a planar curve modeled by a parametric system.

4. The method of claim 1, wherein, Determining the contour of said object in said cross section plane comprises a curve fitting algorithm starting from a prior contour of said object in said cross section plane.

5. The method of claim 1, wherein, Determining the contour of said object in said cross section plane comprises a curve fitting algorithm of the nonlinear recursive type.

6. The method of claim 1, wherein, Determining the profile of the object in the section plane comprises an iterative curve fitting algorithm comprising: - considering a prior profile of the object in the section plane as a computed profile of a first iteration rank; - then iteratively performing the following operations until the comparison of the simulated one-dimensional radiograph images (SSpk) with the one-dimensional processed radiograph images (Spk) reaches a predetermined optimization criterion: ■ from the computed profile of the object in the section plane at a given iteration rank, computing a number (NP) of simulated one-dimensional radiograph images (SSpk) of the region under examination equal to at least three, calculated in the section plane along at least three different projection directions (Dijk) used to acquire one-dimensional processed radiograph images (Spk) in the section plane, ■ comparing the simulated one-dimensional radiograph images (SSpk) with the one-dimensional processed radiograph images (Spk), • modifying the computed profile into a computed profile of a higher iteration rank as a function of the comparison result.

7. The method according to claim 1, further comprising: - using the image sensors (Ci, Cik) to acquire a number (NP) of two-dimensional radiograph images (Ri) of the region under examination for each object equal to at least three during the displacement of the object, each two-dimensional radiograph image being obtained along a different projection direction (Dji), - extracting the one-dimensional processed radiograph images (Spk) in the two-dimensional radiograph images (Ri) to form a set of one-dimensional radiograph images.

8. The method of claim 1, wherein, The one-dimensional processed radiograph images (Spk) of the object are formed by sampling point images acquired using the point image sensors (Cik) during a scanning duration corresponding to the duration of the displacement between the object and the point image sensors (Cik).

9. The method of claim 1, further comprising: Using the computer system and according to the profile of the object in each different section plane (Pk), a three-dimensional digital geometric model of the region under examination for the object to be measured is constructed, the three-dimensional digital geometric model comprising: - three-dimensional points in space, each three-dimensional point belonging to the boundary surface of the region under examination of the object; and / or - at least one three-dimensional surface of the region under examination.

10. The method of claim 9, wherein, Determining at least one measurement of the linear dimensions of the region under examination of the object to be measured from the profile of the object in each different section plane comprises determining the distance between at least two three-dimensional points of the three-dimensional digital geometric model of the region under examination.

11. The method of claim 1, further comprising: The prior profile of the object in each section plane is provided to the computer system.

12. The method of claim 11, wherein, The prior profile is obtained by: - the series of computer-aided design digital models of the objects; and / or - measurements of one or several objects of the same series according to a measuring device; and / or - values and / or graphs and / or shapes selected by an operator entered on a human-machine interface of the computer system. - values and / or graphs and / or shapes selected by an operator entered on a human-machine interface of the computer system.

13. The method of claim 1, further comprising: - a computer-aided design digital model of the series of objects; and / or - a digital geometric model obtained from measurements of one or several objects of the same series by a measuring device; and / or - a digital geometric model generated by the computer system from values and / or drawings and / or selected shapes entered and / or made by an operator on a human-machine interface of the computer system. The focal point is set in the transport plane (PC).

14. The method of claim 1, further comprising: At least two one-dimensional processed radiographic images of the examination region corresponding to projection directions (Djik) defining an effective angle (a) greater than or equal to 45° and less than or equal to 90°, and advantageously an effective angle (a) greater than or equal to 60° and less than or equal to 90° in the considered cross-sectional plane (Pk), are acquired using the image sensor (Ci, Cik) for each object of the series during the displacement of this object and for each considered cross-sectional plane (Pk).

15. The method of claim 1, further comprising: At least one radiographic image of the examination region corresponding to a projection direction (Djik) producing a projection in the transport plane (PC) and having a spread angle (b) comprised between 10° and 60° with the displacement direction (T) is acquired using the image sensor (Ci, Cik) for each object of the series during the displacement of this object and for each considered cross-sectional plane (Pk).

16. The method of claim 1, further comprising: Any radiographic projection of the examination region corresponding to a projection direction (Dji, Djik) having a spread angle (b) less than 10° with the displacement direction (T) is acquired using the image sensor (Ci, Cik) for each object of the series during the displacement of this object.

17. The method of claim 1, excluding: The radiographic projections of the examination region of the objects are performed and acquired so that the X-rays originating from the focal point and reaching the image sensor (Ci) do not pass through another object.

18. The method of claim 1, further comprising: Between three and forty one-dimensional processed radiographic images of the radiographic projections of the examination region originating from different projection directions are acquired for each object of the series during the displacement of this object and for each cross-sectional plane.

19. The method of claim 1, further comprising:

20. The method according to claim 1, characterized in that: - the image sensor (Ci) forms part of at least three physical sensor assemblies (CCi), each of linear type, each comprising a linear array of X-ray sensitive elements distributed along a support straight line (Li) defining, with the focal point (Fj), a projection plane (PPji) containing the projection direction (Dji, Djik), these image sensors being set so that: ■ at least m sensitive elements of each of these physical sensor assemblies receive a radiographic projection of the examination region by the beam of X-rays originating from the focal point (Fj); ​ ■the projection planes (PPji) of the different physical sensor assemblies are different from each other and are not parallel to the transport plane (PC); - using each of the at least three linear physical sensor assemblies (CCI), at each incremental displacement of each object along the trajectory, a one-dimensional radiographic image of the region to be examined is acquired according to a selected number, so that for each object the entire region to be examined is represented in its entirety in a set of one-dimensional radiographic images; - for each object, at least three sets of one-dimensional radiographic images of the region to be examined are analyzed.

21. A facility for automatically measuring linear dimensions of at least one region to be examined of a series of manufactured objects, the facility comprising: - means for transporting objects along a rectilinear trajectory in a transport plane (PC) in a direction specified by a displacement vector, the objects travelling through a transport volume (Vt) extending in the displacement direction (T); - at least one focal point (Fj) of an X-ray generating tube, located outside the volume (Vt) and generating a diverging X-ray beam directed through at least one region to be examined of the objects, each focal point being arranged on a same reference straight line parallel to the displacement direction (T) along the rectilinear trajectory; - image sensors (Ci, Cik) located outside the transport volume (Vt) so as to receive X-rays originating from the associated focal point (Fj), the focal points (Fj) and the image sensors (Ci) being arranged so that, as the objects pass through these rays, each image sensor receives radiographic projections of the region to be examined through rays originating from the focal points (Fj), the projection directions of these radiographic projections being different from each other; - an acquisition system connected to the image sensors (Ci, Cik) so as to acquire, for each object, during the displacement of the object, a set of one-dimensional processed radiographic images, the set comprising: • the one-dimensional processed radiographic images for a plurality (NK) of different cross-sectional planes (Pk) containing the reference straight line; • for each different cross-sectional plane (Pk), a plurality (NP) of the one-dimensional processed radiographic images (Spk) of the region to be examined obtained along at least three different projection directions (Dijk) in the cross-sectional plane; - a computer system configured to: • for each different cross-sectional plane (Pk), determine, from the one-dimensional processed radiographic images (Spk) obtained along the at least three different projection directions (Dijk) in the cross-sectional plane, a contour of the object in the cross-sectional plane under consideration formed by one or more two-dimensional curves in the cross-sectional plane under consideration, and • for the object to be measured, determine at least one measurement of a linear dimension of the region to be examined of the object to be measured from the contours of the object to be measured in each different cross-sectional plane determined from the one-dimensional processed radiographic images of the region to be examined.

22. The facility of claim 21, further comprising: at least two X-ray generating foci (F1, F2) positioned at two different positions on the same reference straight line parallel to the direction of displacement (T) along a rectilinear trajectory; and at least three image sensors (Ci) sensitive to X-rays and positioned so that: - each focus emits an X-ray beam that passes through at least the region to be examined to reach at least one associated image sensor (Ci, Cik); - each image sensor (Ci) is associated with a focus and receives the X-rays originating from said focus after they have passed through the region to be examined.

23. The facility of claim 21, further comprising: at least one focus from which the diverging X-ray beam has an opening greater than or equal to 90°; or at least two foci from which the sum of the openings of the diverging X-ray beams is greater than or equal to 90°.

24. The installation according to claim 21, further comprising at least one focus positioned in the plane of conveyance (PC).

25. The facility of claim 22, wherein, The at least one focus and the two image sensors are positioned so that there is an effective angle (a) between the projection directions of the region to be examined received by them that is greater than or equal to 45° and less than or equal to 90°, and advantageously greater than or equal to 60° and less than or equal to 90°.

26. The facility of claim 21, wherein, The at least one focus and one image sensor (Ci) are positioned so that the projection direction of the region to be examined on said image sensor (Cik, Ci) forms an opening angle (b) with the direction of displacement (T) that is included between 10° and 60° when the object passes through the field of the sensor.

27. The facility of claim 21, wherein, Since the foci (Fj) of the X-ray generating tubes are not located in the transiting volume (Vt) and the image sensors (Ci) are not located in the transiting volume (Vt), the projection direction (Dji, Djik) of the region to be examined on the image sensors (Ci) never forms an opening angle (b) with the direction of displacement (T) that is less than 10°.

28. The facility of claim 21, wherein, The image sensors (Ci) and the foci (Fj) are positioned so that the X-rays originating from a focus and reaching an image sensor (Cik, Ci) and passing through the region of the object do not pass through another object at the same time.

29. The installation according to claim 21, further comprising between one and four foci (Fj) originating from one or several X-ray generating tubes.

30. The facility of claim 21, wherein, The number and positioning of the image sensors (Cik) and associated foci are such that, for each object of the series, during the displacement of the object, the radiographic projection of the region to be examined on the image sensors has between three and forty different projection directions.

31. The facility of claim 21, wherein, The image sensors (Ci) form part of physical sensor assemblies (CCi) of linear type, each comprising a linear array of X-ray sensitive elements distributed along a support straight line (Li) that, with the associated focus (Fj), delimits a projection plane (PPji) containing the projection direction (Dji, Djik), these image sensors being positioned so that: - at least m sensitive elements of each of these physical sensor assemblies receive the radiographic projection of the region to be examined by the X-ray beam originating from the associated focus (Fj); - the projection planes (PPji) of the different sensors are different from each other and not parallel to the conveying plane (PC).

32. The facility of claim 31, wherein, The support straight lines (Li) of the at least three linear image sensors (Ci) are parallel to each other.

33. The facility of claim 31, wherein, The support straight lines (Li) of the at least three linear physical sensor assemblies (CCi) are orthogonal to the conveying plane (PC).

Images