EP3959685A1 - Method and installation for the in-line dimensional control of manufactured objects - Google Patents

Abstract

The invention relates to a method and installation for automatically measuring linear dimensions of manufactured objects (2) of a run comprising: - arranging at least one X-ray focal point (Fj) on a given base line parallel to the rectilinear trajectory of movement of the objects and providing one or more image sensors (Ci); - acquiring, for each object over the course of its movement, a set of one-dimensional images comprising, for a number (NK) of distinct sectional planes (Pk) containing the base line, a number (NP) of said images obtained in at least three different directions of projection (Dijk) in the sectional plane; - for each object, and for each distinct sectional plane (Pk), determining, on the basis of the obtained images, a delineation of the object in the sectional plane (Pk) in question.

Description

Description

Title of the invention: CONTROL PROCESS AND INSTALLATION

ONLINE DIMENSIONAL OF MANUFACTURED OBJECTS

Technical area

The invention relates to the field of dimensional control by X-rays,

manufactured objects forming a series of objects.

[0002] The object of the invention is more particularly to obtain the measurement, by

X-rays, of linear dimensions, i.e. lengths taken on manufactured objects in the general sense, such as for example on containers, molded or machined parts, mechanical parts, packaging, bodywork components .

Prior art

[0003] The prior art knows various techniques for controlling

dimensional objects by X-rays. Also known are baggage inspection systems, which do not aim to measure the dimensions of known objects but to detect prohibited objects or quantities of prohibited materials occurring in arrangements, shapes. and virtually random amounts.

Axial rotation systems of which computer-assisted tomography or CT ("computed tomography") is a part are thus known. This classic method is described in the article "Computed tomography for dimensional metrology" by J.P. Kruth (1) & ail, in CIRP Annals.

[0005] Volume 60, Issue 2, 2011, Pages 821 -842 and implemented, for example, by tomography devices sold by the companies Werth Messtechnik or General Electric. This method consists in positioning between an X-ray generator tube and a matrix or linear X-ray image sensor, an object on a turntable around a vertical axis. A very large number (at least 100 and often more than 600) 2D x-ray images of the objects are acquired during the rotation. If the image sensor is matrix, the beam is conical. If the image sensor is linear, the beam is advantageously confined as a fan ("fan beam") in a plane orthogonal to the axis of rotation and the rotation is accompanied by a translation along the vertical axis of the rotation, for a complete sweep of the helical type. This technique can provide high precision three-dimensional measurements. However, the time

acquisition time requires at least a minute for the fastest systems, plus the time of loading and unloading objects so that 10 to 30 objects are inspected at most per hour.

Another so-called rotary gantry solution is proposed for example by

the device known by the trade name "speed | scan CT 64" from the General Electric Company. Like some 3D scanners for luggage, this concept of solution is similar to medical imaging tomographs in terms of relative movements between source, object and image sensor. In fact, manufactured objects or luggage, placed on a conveyor, are in translation in the device. They cross a projection plane orthogonal to the direction of displacement. In a circular gantry containing said plane, an X-ray source and a generally curved image sensor, opposite the source, are rotated around the central axis of movement, to obtain, slice by slice or by helical scan, the projections necessary for the 3D reconstruction by an algorithm implementing, for example, the “filtered rear projection” method or the ART method. The objective of these devices is to allow the acquisition of a very large number of projections at each turn of the gantry, for example 100, even 700 to 1000 images per slice. The 3D reconstruction of the objects is made for example slice by slice. Having in fact determined the attenuation at any point of the slice, by concatenating the slices obtained during the movement of the object, we obtain an attenuation value at any element of the object's volume.

If these devices with vertical axis rotation or rotating gantry are very precise thanks to the large number of images provided, these devices are expensive and slow and are in practice reserved for offline control because they are not suitable for online dimensional control for throughputs that can reach and exceed 600 articles per minute with scrolls of 1 m / s.

[0008] Patent application DE 10 2014 103137 describes a method for determining geometric characteristics on a workpiece using a system of tomodensitometric detectors, consisting of an X-ray source, a plane detector and a mechanical axis to rotate the part or to rotate the x-ray source and the detector.

[0009] The method acquires radiographic images during rotation and provides a representation of the surface using a model of the surface. Such a method eliminates the need for volume data reconstruction steps to reduce the computation time. Such a technique does not make it possible to measure parts manufactured at high speed because it requires loading the parts on a turntable then rotating at least 180 ° and then unloading it for the control of another part.

To overcome the drawbacks of the tube and image sensor on board a rotating gantry, US Pat. No. 8,971,484 describes a baggage inspection system in which the rotating system is replaced by a network of multibeam X-ray sources. stationary actuated successively to create a virtual displacement of the X-ray sources making it possible to provide a large number of radiographic images with different angles of projection. Compared to physical rotation systems limited to 4 revolutions per second, the number of "virtual rotations" is increased to 40 revolutions per second. This technique, which is implemented by the device known under the trade name Rapiscan RTT from the company Rapiscan Systems, is able to check 1,200 bags per hour, by producing tens of thousands of 2D images of the baggage, considering that virtual rotation provides around 40 different projection angles.

[0011] This technique is very expensive due to the high price for

multisource X-ray and the computing power to process a very large volume of data. In addition, the control rates are still limited and are not suitable for on-line control.

US Patents 7,319,737 and US 7,221,732 propose to control

baggage by a technique called digital laminography or tomosynthesis. The baggage passes through a succession of conical projection planes called "fan beam" and each containing a pair of linear image sensors arranged in an L. These techniques aim to search for weapons or explosives in the baggage which contains shaped objects and very diverse materials, by visualizing their 3D positions in the luggage and by evaluating for example the volume of suspect product. It is common to use multispectral technology to also determine the atomic number of matter. These systems therefore seek to determine an attenuation value at any point of a piece of luggage. On the other hand, these systems are not able to determine at high speed and with precision, the dimensions of manufactured objects for the purpose of quality control.

JP S60 260807 patent application proposes to measure the thickness of the walls of a tube moving in translation along the axis of the tube, using X-ray measurements from one or more foci to each of which are associated sensors. The foci and the sensors are positioned to produce radiographic projections in a plane orthogonal to the direction of movement of the tube. The radiographic projections are therefore coplanar in a projection plane which is orthogonal to the axis of symmetry of the tube. The direction of these radiographic projections forms a right angle (90 °) with respect to the direction of movement. This technique does not allow complete knowledge of the internal and external surfaces of the tube. The method described by this patent application makes it possible to measure only the cumulative thickness of the two walls of the tube in the direction of projection, without reconstructing a three-dimensional model of a tube which would allow precise measurements to be made in the other directions.

Likewise, US Pat. No. 5,864,600 describes a method for determining the

filling level of a container using an X-ray source and a sensor arranged transversely on either side of the container transport conveyor. This system does not allow measurements to be taken for a non-transversely oriented surface because this document does not provide for three-dimensional modeling of the containers.

[0015] Patent application US 2009/0262891 describes a system for detecting by X-rays, objects placed in baggage moved in translation by a conveyor. This system comprises pulsed generator tubes or a sensor having a large dimension parallel to the direction of travel. This document provides a method of reconstructing the object which is not satisfactory because the absence of projections in the direction of displacement does not allow the measurement of dimensions in the direction orthogonal to the direction of travel. The lack of radiographic projections in an angular sector does not make it possible to produce a digital model suitable for ensuring precise measurements.

Patent application DE 197 56 697 describes a device exhibiting the

same drawbacks as patent application US 2009/0262891.

[0017] Patent application WO 2010/092368 describes a device for viewing an object moving in translation by X-rays using a source of radiation and three linear sensors.

[0018] Patent application US 2010/220910 describes a method for detecting anomalies in an object by producing a reference 3D model representing an ideal object. The method then aims to compare a 2D image acquired of a real object with the 2D image corresponding to the reference model in order to deduce an anomaly. This method does not make it possible to take precise measurements of an object and only makes it possible to control an object in the 2D images produced, therefore only the directions orthogonal to the projection directions.

[0019] Document WO2018014138 describes a method for carrying out the inspection of a manufactured article. The method includes acquiring a sequence of radiographic images of the article; determining a three-dimensional position of the envelope of the article for each of the radiographic images acquired; and performing a detailed three-dimensional model correction loop in the form of a mesh network, which iteratively comprises: generating a simulated radiographic image for each determined position of the article; and comparing the simulated radiographic images and the acquired radiographic images and generating a match result. If the result of

the pairing is indicative of a mismatch, the method comprises identifying and distinguishing between the simulated radiographic images and the acquired radiographic images and determining their cause in terms of material density or three-dimensional geometry; correct one of geometry and material density of a region of interest of the model

detailed three-dimensional article on the basis of each of the differences identified and characterized; and perform a new iteration. This document therefore works on the basis of a three-dimensional model correction loop, that is to say implementing 3D geometric shapes, which requires significant calculations. In addition, this document teaches that it is necessary to implement, in order to allow this model correction loop

three-dimensional, at least about 25 x-ray images and preferably about 100 x-ray images defining a continuous sequence of images, each image providing a unique viewing angle of the article.

[0020] The publication of KRUTH JP et al, “Computed tomography for dimensional metrology”, CIRP ANNALS, vol. 60, no. 2, December 31, 2011, pages 821-842, ISSN: 0007-8506, DOI: 10.1016 / J. CIRP.2011.05.0 provides an overview of the technique of X-ray tomography applied to metrology

dimensional. This document indicates in particular that the basic principle of this technology implies that the mathematical reconstruction of the projected images leads to a 3D voxel model, and involves a post-processing of the voxel data, for the detection of the edges of the part (segmentation) and for dimensional measurement. This document also indicates that the basic principle of this technology involves the rotation of the object around its axis.

[0021] Document US 2009/262891 describes a system which does not

geometric reconstructions but standard imaging, which requires a large angular sampling and therefore many unitary sensor components in the direction of the object trajectory. This therefore requires plane detectors containing a high density of “photo sites” along lines oriented in the direction of the trajectory of the object. In other words, the 3D reconstruction system by filtered back projection algorithms is very demanding in terms of the total number of unit sensor components, since it takes many unit sensor components in the direction of the object trajectory.

The object of the invention aims to remedy the drawbacks of the prior art by proposing an inexpensive process to implement and allowing precise dimensional control by X-rays, of manufactured objects moving in translation at high speed . In particular, the method aims to allow such control with reduced calculation volumes, allowing these high rates to be achieved with equipment the cost of which remains reasonable. It is known in tomography that the absence of radiographic projections around a given direction prevents the reconstruction of surfaces parallel to this direction, creating the phenomenon of "missing border", which prevents for a dimensional control, the measurement of dimensions orthogonal to the missing radiographic projections.

[0024] Another object of the invention therefore aims to provide a method for making precise measurements on objects moved in translation,

possibly by building a precise and complete three-dimensional digital model, while the radiographic projections are limited in number and cannot be acquired around the direction of conveying the objects.

Disclosure of the invention

The invention relates to a method for automatically measuring linear dimensions of manufactured objects in a series comprising:

- the choice of a series of manufactured objects in which each of said objects consists of one or more distinct parts, the number of parts being known and each part being made of a material with a known and uniform attenuation coefficient at any point of the part of the object;

- the transport, by means of a transport device, of objects in a direction of travel along a rectilinear path in a conveying plane, these objects generating a conveying volume during their movement;

- the arrangement, outside the conveying volume,

of at least one focus of an X-ray generator tube, each focus being arranged on the same basic line parallel to the direction of

displacement along the rectilinear path, and

one or more image sensors each exposed and sensitive to X-rays from an associated focus, these X-rays having passed through at least the region to be inspected producing on each image sensor a radiographic projection of the region to be inspected in a direction projection;

- the acquisition, using the image sensor (s) (Ci, Cik), for each object during its movement, of a set of one-dimensional radiographic processing images, each one-dimensional radiographic image treatment comprising a projection of a section of the object according to a section plane (Pk) containing the base line, the assembly comprising said one-dimensional radiographic processing images for a number (NK) of section planes (Pk ) distinct containing the base line;

for each distinct section plane (Pk), a number (NP) of said one-dimensional radiographic processing images (Spk) of the region to be inspected, obtained according to at least three different projection directions (Dijk) in the section plane;

- for each object to be measured, and for each distinct section plane (Pk), the determination, using the computer system, of a delineation of the object in the section plane (Pk) considered, from the said one-dimensional treatment radiographic images (Spk) of the region to be inspected, obtained according to at least three different projection directions (Dijk) in the section plane,

- and determining, for an object to be measured, from the delineations of the object in each distinct section plane, at least one linear dimension measure of the region to be inspected of the object to be measured.

[0026] Other characteristics of a method according to the invention, which are optional but which can be combined with each other, are developed in the following paragraphs.

[0027] A delineation of the object may include or be formed by a curve or a set of curves which represent the intersection, with the section plane, of the boundary surfaces of the object.

[0028] The curve or each curve of the delineation of the object can be a

plane curve modeled by a parametric system.

[0029] The determination of a delineation of the object in the section plane may include a curve fitting algorithm starting from an a priori delineation of the object in the section plane.

The determination of a delineation of the object in the section plane can include a nonlinear regression type curve fitting algorithm. The determination of a delineation of the object in the section plane can comprise an iterative curve fitting algorithm comprising:

- taking into account an a priori delineation of the object in the section plane as a calculated delineation of the first iteration rank;

- then iteratively,

the calculation, from the calculated delineation of a given iteration rank of the object in the section plane, of a number at least equal to three simulated radiographic one-dimensional images of the region to be inspected, calculated in the section plane according to the at least three different projection directions which were used for the acquisition of the one-dimensional radiographic processing images in the section plane,

comparing the simulated one-dimensional radiographic images to the one-dimensional processing radiographic images,

depending on the comparison, the modification of the calculated delineation into a calculated delineation of higher iteration rank,

until the comparison of the simulated one-dimensional radiographic images to the one-dimensional treatment radiographic images reaches a predefined optimization criterion.

[0032] The method can include:

the acquisition, using the image sensors, for each object during its movement, of a number at least equal to three two-dimensional radiographic images of the region to be inspected, each obtained in a direction of different projection,

- extracting, from the two-dimensional radiographic images, the processing one-dimensional radiographic images to form the set of one-dimensional radiographic images.

A one-dimensional radiographic processing image of an object can be formed by sampling a point image acquired using a point image sensor, for a scanning period corresponding to the duration of the displacement of the object. object between the focus and the point image sensor, during its movement.

The method may include the construction, for an object to be measured, using the computer system and from the delineations of the object in each of the separate section plans, of a three-dimensional digital geometric model of the region to be inspected, comprising:

- three-dimensional points in space each belonging to a boundary surface of the region to be inspected of the object;

- and / or at least one three-dimensional surface of the region to be inspected.

The determination, for an object to be measured, from the delineations of the object in each distinct section plane, of at least one measure of linear dimension of the region to be inspected of the object to be measured may comprise determining the distance between at least two three-dimensional points of the three-dimensional digital geometric model of the region to be inspected.

[0036] The method may include providing the computer system, for each section plane, with an a priori delineation of the object in the section plane.

The a priori delineations can be obtained by:

- a digital computer design model of the objects in the series;

- and / or from the measurement of one or more objects of the same series by a measuring device;

- and / or from values entered and / or drawings made and / or shapes selected by an operator on a man-machine interface of a computer system.

[0038] The method may include providing the computer system with a

a priori three-dimensional geometric model of the region to be inspected of the series, which can be obtained by:

- a digital computer design model of the objects in the series;

- and / or a geometric digital model obtained from the measurement of one or more objects of the same series by a measuring device;

- and / or a geometric digital model generated by a computer system from values entered and / or drawings made and / or shapes selected by an operator on a man-machine interface of the computer system.

The method can include the arrangement of the hearth or foci in the conveying plane.

[0040] The method can comprise the acquisition, using the sensor (s)

of images, for an object of the series during its movement, and for each considered section plane of the object, of at least two one-dimensional radiographic processing images of the inspected region corresponding to projection directions defining, in the section plane considered, a useful 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 °.

[0041] The method can comprise the acquisition, using the sensor (s)

of images, for an object of the series during its displacement, and for each considered section plane of the object, of at least one radiographic image of the inspected region corresponding to a direction of projection having, brought back to projection in the conveying plane, an opening angle with the direction of movement of between 10 ° and 60 °.

[0042] The method can include the acquisition, using the sensor (s)

image, for each object in the series during its movement, no radiographic image of the inspected region corresponding to a direction of projection having an opening angle with the direction of displacement of less than 10 °.

[0043] The method may include making and acquiring X-ray projections of the inspected region of an object so that X-rays from the focal point (s) and reaching the image sensors do not pass through another object.

The method may include the acquisition, for each object of the series during its movement and for each section plane, one-dimensional radiographic treatment images from between three and forty radiographic projections of the region to be inspected from different projection directions, preferably from between four and fifteen projections

X-rays of the area to be inspected from different projection directions. In certain embodiments, the image sensors can be part of at least three physical sensor components which are each of linear type, each comprising a linear array of elements sensitive to X-rays distributed along a support line. , which defines with a focal point a projection plane containing the direction of projection, these image sensors being arranged so that:

- at least m sensitive elements of each of these physical sensor components receive the radiographic projection of the region to be inspected by the X-ray beam from a focal point;

- the projection planes for the various physical sensor components are distinct from one another and not parallel to the conveying plane;

- using each of the at least three linear physical sensor components is acquired, at each incremental displacement of each object along the trajectory, one-dimensional radiographic images of the region to be inspected according to a number chosen so that for each object, the the entire region to be inspected is completely represented in the set of one-dimensional radiographic images;

- the at least three sets of one-dimensional radiographic images of the region to be inspected are analyzed for each object.

[0046] The invention also relates to an installation for automatic measurement of linear dimensions of at least one region to be inspected of manufactured objects in a series, the installation comprising:

a device for transporting objects in a direction materialized by a displacement vector, along a rectilinear path in a conveying plane, the objects traversing an extended conveying volume in the direction of displacement;

- at least one focal point of an X-ray generator tube located outside the volume crossed, and creating a divergent beam of X-rays directed to pass through at least one region to be inspected of the object, each focal point being arranged on the same straight line base parallel to the direction of travel along a rectilinear path;

- image sensors, located outside the conveying volume, so as to receive X-rays from an associated focus, the focus (s) and the image sensors being arranged so that each image sensor receives the radiographic projection of the region to be inspected by the rays coming from the focus when the object passes through these rays, the directions of projection of these radiographic projections being different from one another;

- an acquisition system connected to the image sensors, so as to acquire for each object during its movement, a set of one-dimensional radiographic processing images, the set comprising:

one-dimensional processing radiographic images for a number of distinct section planes containing the base line;

for each distinct section plane, a number of one-dimensional treatment radiographic images of the region to be inspected, obtained according to at least three different projection directions in the section plane;

- a computer system configured for:

for each separate section plane, determine a

delineation of the object in the section plane considered, from at least three one-dimensional radiographic processing images.

[0047] Other characteristics of an installation according to the invention, which are

optional but which can be combined with each other, are developed in the following paragraphs.

The installation may include at least two X-ray production foci, positioned separately in two distinct positions on the same basic straight line parallel to the direction of movement along the rectilinear path, and at least three image sensors, X-ray sensitive and positioned so that:

- Each focus emits its beam through at least the region to be inspected to reach at least one associated sensor;

- Each sensor is associated with a focal point and receives the X-rays from said focal point after having passed through the region to be inspected.

The installation may include at least one focus from which comes a divergent X-ray beam with an aperture greater than or equal to 90 ° or at least two foci from which originate divergent X-ray beams, the sum of the openings of which is greater than or equal to 90 °. [0050] The installation may include at least one hearth arranged in the conveying plane.

At least one focus and two image sensors can be arranged in

so that the directions of projection of the inspected region that they receive have between them a useful 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 °.

At least one focus and one image sensor can be arranged so that, when an object crosses the field of the sensors, the direction of projection of the region inspected on the image sensor makes an opening angle with the direction of movement between 10 ° and 60 °.

No focus of an X-ray generator tube being located in the volume passed through, and no image sensor being located in the conveying volume, the direction of projection of the region inspected on the sensor d The images never make an opening angle with the direction of movement less than 10 °.

[0054] The image sensors and focal points are arranged so that x-rays from the focal point (s) and reaching the image sensors and passing through the region of one object do not pass through another object at a time.

[0055] The installation may have between one and four homes, from one or from

several x-ray generator tubes.

The number and arrangement of the image sensors and the associated foci are such that, for each object in the series during its movement, the radiographic projections of the region to be inspected on the image sensors present between three and forty different directions of projection, preferably between four and fifteen different directions of projection.

[0057] Image sensors can be part of sensor components

linear type physics each comprising a linear array of elements sensitive to X-rays distributed along a support line, which defines with an associated focus a projection plane containing the direction of projection, these image sensors being arranged so that:

- at least m sensitive elements of each of these physical sensor components receive the radiographic projection of the region to inspect by the X-ray beam from the associated focus;

- the projection planes for the various sensors are distinct from one another and not parallel to the conveying plane.

[0058] In some embodiments, at least three image sensors

linear have their support lines parallel to each other.

In some embodiments, at least three linear physical sensor components have their support lines orthogonal to the conveying plane.

Brief description of the drawings

[0060] [Fig. 1] Figure 1 is a schematic top view showing a

installation allowing the measurement, by X-rays, of dimensions on objects scrolling in line.

[0061] [Fig. 2] Figure 2 is a schematic perspective side view

showing part of the installation allowing the measurement, by X-rays, of dimensions on an object.

[0062] [Fig. 3] Figure 3 is a schematic perspective view showing more generally an installation of the type of those of Figs. 1 and 2

[0063] [Fig. 4] Figure 4 is a schematic perspective view showing the volume traversed or generated by the objects during their linear displacement.

[0064] [Fig. 5] Figure 5 is a schematic top view showing a

example of an installation according to the invention comprising two foci that generate X-rays.

[0065] [Fig. 6] Figure 6 is a schematic cross-elevation view of the installation shown in Fig. 5.

[0066] [Fig. 7] Figure 7 is a schematic view explaining the definition of

the useful angle between two directions of projection.

[0067] [Fig. 8] Figure 8 is a schematic view explaining the definition of

the useful angle between two directions of projection. [0068] [Fig. 9] Figure 9 is a schematic perspective view showing another example of positioning of image sensors with respect to the

moving objects to be inspected.

[0069] [Fig. 10] Figure 10 is a schematic view of another example of

realization of an installation according to the invention, using matrix image sensors.

[0070] [Fig. 11] Figure 11 is a view of an array of x-ray sensitive elements showing two distinct areas corresponding to two raster image sensors.

[0071] [Fig. 12] Figure 12 is a flow diagram of a method for determining a delineation in a section plane.

[0072] [Fig. 13] Figure 13 is a flow chart of a method according to the invention.

Description of embodiments

As a preliminary, some definitions of the terms used in the context of

the invention are given below.

A focus Fj of an X-ray generator tube is a point source of X-rays, preferably a “micro focus”, with a diameter for example between 0.01 mm and 1 mm, creating a divergent beam of X-rays. It is possible to use any type of point or near point X-ray source.

A sensitive element of an image sensor for X-rays is an element sensitive to X-rays, in other words an elementary surface, of dimension for example 0.2 x 0.2 mm or 0.02 x 0.02 mm, converting the x-rays it receives into an electrical signal. Typically, a sensitive element of an x-ray image sensor includes a scintillator that converts x-rays into visible light and then a photoelectric sensor converts visible light into an electrical signal. Techniques for the direct conversion of X-rays into an electrical signal also exist. A pixel designates an elementary value of a point of a sampled image, characterized for example by its gray level between 0 and a maximum value. For example, for a 12-bit digital image, a pixel takes digital values between 0 and 4095. A system for reading or acquiring radiographic images comprises one or more surfaces sensitive to X-rays, that is to say surfaces comprising one or more sensitive elements converting the X-rays into an electrical signal in order to be transmitted to an analysis system, conventionally implemented by a computer, and designated by the SI computer system in the remainder of the description. The signals originating from a set of sensitive elements belonging to the same sensitive surface zone, acquired by the acquisition device and transmitted together to the computer system, constitute a radiographic image. For analysis by the computer system, the x-ray images are preferably converted into images.

digital radiographic images, either as close as possible to the sensitive surface, for example in an electronic circuit integrated into a physical sensor component comprising the sensitive area, or at a distance, for example as close as possible to the computer system SI, or even by the computer system SI.

The SI computer system, an example of which is symbolically illustrated in FIG. 3, can be produced in the form of at least one standard computer, therefore comprising at least one microprocessor, one or more electronic memory units and one or more display interfaces (screen, projector, holographic display, etc.), input (keyboard, mouse, touchpad, touch screen, etc.), and / or communication (USB, Ethernet®, Wi-Fi®,

Bluetooth®, Zigbee®, ...). The computer system may include a computer network sharing data with one or more other computers in the network, or with other networks, for example by an internet or Ethernet® protocol. In addition to its obvious link to image sensors, the computer system can be linked to sensors giving information on the status of the installation, and / or to actuators of the installation (conveyors, ejectors, etc. ). The computer system can advantageously be connected to the X-ray tube (s), in order to acquire operating data therefrom, and / or to ensure control thereof. The computer system implements one or more software, stored and / or executed (s) locally or remotely, including on one or more remote computer servers. This or these software (s) preferably comprise one or more software (s) programmed to implement the method according to the invention. The X-ray beams from a focus Fj pass through at least one inspected region, and form on a sensitive surface, the radiographic projection of the inspected region, which is sometimes called the radiant image and which contains

X-ray attenuation information by the material passed through.

An X-ray sensitive surface area which receives the radiographic projection of the inspected region is called image sensor Ci, Cik. An image sensor Ci, Cik is exposed to X-rays from an associated focus Fj. The image sensor converts this X-ray projection into an X-ray image of the area being inspected.

When the sensitive surface area corresponding to an image sensor Ci contains a single line of photosensitive elements distributed along a support line segment, the radiographic image acquired in an integration time (also called time d acquisition) of the sensor is linear, therefore

one-dimensional, consisting of a row of pixels forming an array of one-dimensional values. The Cik image sensor is then called a linear sensor. A sensitive surface area containing a single line of sensitive elements, which constitutes a linear image sensor, therefore comprises a linear array of sensitive elements distributed along a support line segment. According to this definition, a column, or row, or any set of aligned sensitive elements (including along a diagonal or other inclined line) belonging to a matrix sensitive surface, is considered a linear image sensor. Several sensitive surface areas of the same surface and each containing a single line of sensitive elements, the lines of the different areas being different, therefore constitute several linear image sensors.

When the sensitive surface area corresponding to an image sensor Ci contains a two-dimensional matrix of photosensitive elements, the radiographic image acquired in an integration time of the sensor is matrix, therefore two-dimensional, composed of a matrix of pixels forming an array of two-dimensional values. The image sensor Ci is then called a two-dimensional sensor or a matrix sensor. When the sensitive surface area corresponding to a Cik image sensor contains a single photosensitive element, the radiographic image acquired in an integration time of the sensor is said to be point, composed of a pixel having a single value, The Cik image sensor is then called a point sensor.

In the context of the present invention, an image sensor Ci, Cik can

correspond to a physical sensor component CC1, CC2, ..., CCnmax, to a part of a physical sensor component CC1, CC2, ..., CCnmax, or to

the assembly of several physical sensor components or parts of physical sensor components. A physical sensor component is a component comprising one or more sensitive elements, integral if there are several, and comprising, for all of its sensitive elements, a common connection interface with a computer system. The common connection interface can be analog or digital. The common connection interface is usually part of an integrated electronic circuit of the physical sensor component. In a physical sensor component CC1, CC2, ..., CCnmax comprising several sensitive elements Cik, the sensitive elements are arranged along a line or along a fixed surface, not modifiable, generally flat, but sometimes curved. When the physical sensor component CC1, CC2, ..., CCnmax contains a line of photosensitive elements, it is said to be linear. When the physical sensor component contains a two-dimensional matrix of photosensitive elements, it is said to be two-dimensional or matrix. When the physical sensor component contains a single photosensitive element, it is said to be point.

In the context of the present invention, it is therefore understood that an image sensor Ci, Cik is a set of one or more sensitive elements Cik converting the X-rays into an electrical signal, the sensitive elements being elements physical. This set of physical sensitive elements can correspond to a physical sensor component CC1, CC2, ..., CCnmax, or not.

In the invention, it is possible to distinguish the radiographic images acquired by a physical sensor component CC1, CC2, ..., CCnmax, and the processing radiographic images used by the computer system for the calculation of the delineations, which may correspond to a radiographic image acquired by a physical sensor component, to a part of an image radiographic image acquired by a physical sensor component, or to the assembly of several images or parts of radiographic images acquired by one or more physical sensor components. In particular, an image

One-dimensional radiographic processing of an object can be formed by sampling a point image acquired using a single, given Cik sensitive element, for a scanning time corresponding to the duration of the object's movement between the focus and the sensitive element, during its movement. This sensitive element Cik can correspond to a point physical sensor component, or belong to a linear physical sensor component or to a two-dimensional or matrix physical sensor component. Likewise, a two-dimensional processing radiographic image of an object can be formed by sampling a linear image acquired using a linear image sensor Cik, for a scanning period corresponding to the duration of the displacement of the object. the object between the focal point and the linear sensor, during its movement. This linear image sensor Cik can correspond to a linear physical sensor component or belong to a two-dimensional or matrix physical sensor component. An image formed by sampling corresponds to the juxtaposition of several images acquired successively in time, For a radiographic image, the direction of projection Dji, Djik is the oriented direction or the vector, starting from the focus Fj to pass through the center of the sensor d 'Ci, Cik images implemented to acquire the image, i.e. through the center of an X-ray sensitive area which receives the radiographic projection of the region inspected at the time of acquisition during the movement of the object between the focus and the image sensor. For an associated image sensor-focal point pair, the projection direction is the vector from the focal point reaching the middle of the image sensor. The positioning of the image sensors is such that the sensitive surface is not parallel to the direction of projection. It may be advantageous in certain cases for the sensitive surface of the image sensor to be orthogonal to the direction of projection defined with the associated focus. But this is not compulsory, for example if a sensitive surface of the same physical sensor component contains several sensitive zones each forming an image sensor and which cooperate for each image capture, each with a different focus, therefore according to different directions of projection. For a one-dimensional radiographic processing image of an object which would be formed by sampling a point image acquired using a sensitive element Cik for a scanning period corresponding to the duration of the movement of the object between the focus and the sensitive element, during its movement, the direction of projection Djik therefore corresponds to the direction oriented starting from the focus Fj to pass through the center of the sensitive element used.

For a one-dimensional radiographic processing image of an object, acquired using a linear image sensor Cik during a single integration time of the sensor, the projection direction therefore corresponds to the oriented direction starting from of the focus Fj to pass through the center of the linear image sensor Cik implemented. The direction of projection Djik associated with the linear radiographic processing image obtained is therefore the direction starting from the focus and passing through the middle of the line segment supporting the linear image sensor Cik, at the time of acquisition of the image corresponding to the integration time of this linear image sensor Cik.

For a two-dimensional radiographic processing image of an object formed by sampling a linear image, the linear image being acquired using a linear image sensor CCi for a scanning period corresponding to the duration of the movement of the object between the focal point and the linear image sensor, during its movement, the projection direction Dji therefore corresponds to the direction oriented starting from the focal point Fj to pass through the center of the linear image sensor implemented.

For a two-dimensional radiographic image of an object, acquired using a two-dimensional image sensor Ci, CCi during a single integration time of the sensor, the projection direction Dji therefore corresponds to the oriented direction starting from the focus Fj to pass through the center of the two-dimensional image sensor implemented.

The directions of projection Dji, Djik of radiographic projections are considered to be different if the directions of projection Dji, Djik taken in pairs form between them a minimum angle at least equal to 3 degrees of angle, preferably at least equal to 5 degrees of angle. A sensitive surface area which contains a matrix of sensitive elements constitutes a matrix or two-dimensional image sensor, which comprises a matrix network of elements sensitive to X-rays, distributed in a matrix. As shown in Fig. 10, according to this definition, an area of matrix sensitive surface C1, C1 ′, which belongs to a larger sensitive surface Ss, is also a matrix image sensor. In this example of FIG. 10, the sensitive surface Ss corresponds to the sensitive surface of a physical sensor component CC1 matrix. Several matrix sensitive surface areas C1, C1 ′ of the same surface can, in certain cases, be treated separately by the acquisition device. In all cases, they therefore constitute several matrix image sensors providing different matrix radiographic images, respectively M1, M1 ′ (FIG. 12). The direction D11, D11 'of projection associated with the matrix radiographic image respectively M1, M1' is the direction starting from the focus F1 and passing through the middle of the zone C1, C1 'of the matrix sensitive surface, at the instant of l image acquisition. It is therefore possible that the image sensors C1, C1 ′ are non-disjoint regions, possibly activated successively over time.

Of course, those skilled in the art can use a matrix physical sensor component technology based on an image intensifier or else a "screen capture camera" in which a scintillator plate receives the radiant image, the converts into visible light, the visible image at the rear of the scintillator being photographed by a camera sensitive in the emission domain of the scintillator, generally the visible domain, and provided if necessary with a lens.

The invention applies to series of manufactured objects composed of one or more material (s), such as objects obtained by machining, molding, blow molding, sintering, injection, extrusion, or assembly of objects. obtained by these types of processes, each of said objects consisting of one or more distinct parts, the number of parts being known and each part being made of a material whose attenuation coefficient m is known and uniform, c ' that is to say having the same value at all points of the part considered of a region to be inspected of the object and preferably constant over time and identical for the objects of the series. In certain embodiments, for example mechanical parts of steel or aluminum foundry, glass bottles, plastic packaging, it may be so-called mono-material objects. In this case, the attenuation coefficient m is known and uniform, that is to say having the same value at any point of a region to be inspected of the object. However, the invention can be implemented for multi-material objects. In some cases, the different materials have an equal attenuation coefficient, so the subdivision of the object into different parts can be ignored and the object can be considered as if it were a mono-material object, provided that the attenuation coefficient is uniform, in the sense of uniform over the whole of the inspected region.

However, the invention can also be implemented for the measurement of

linear dimensions for multi-material objects. Such an object is therefore considered to consist of an assembly of volumes that are homogeneous in composition, each volume that is homogeneous in composition being considered as part of the object. These volumes, or parts of the object, are delimited by closed surfaces. For the invention, it will be considered that the number of such parts is known, at least for the region to be inspected of the object. This number can be counted for the inspected region, preferably low, for example less than 50, preferably less than 20, more preferably less than 10, in order to limit the importance of the calculations and to preserve convergence in the event of use of an iterative adjustment method.

Preferably, the topology of these parts of the object is known, namely in particular the relative arrangements for connectivity (existence of surface

common, situations of inclusion or exclusion of the different parts between them, juxtapositions and relative positions, existence of common surfaces, etc ...). This amounts to saying that we can consider that an X-ray passing through the object has crossed a finite number of volumes presenting different but known attenuation coefficients, therefore that the path can be broken down into segments, each segment joining two points belonging to boundary surfaces of part of the object, traversing a region of constant attenuation, even if the length of these segments is not known a priori.

The attenuation of each X-ray depends only on the length of the successive segments traversed and the attenuation for each segment. Consequently, the information of each x-ray image point is directly related to the actual dimensions of the inspected region whose structure is known. Moreover, the geometry of the object, and in particular the geometry of its different parts, can be described by computer a priori (before the measurement) by a representation made up of a set of closed surfaces, which can be called boundary surfaces. of the object. The boundary surfaces of an object are the interface surfaces. It is thus possible to have one or more external boundary surfaces of the object, which are each an interface between the surrounding air and the material of part of the object. It is also possible to have one or more internal boundary surfaces of the object, which are each an interface between the two materials of two juxtaposed parts of the object. In the case of a hollow object comprising an interior cavity delimited by an interior surface of the object, the interior surface is an exterior boundary surface of the object because it is the interface between the material of the object and the surrounding air, even in the case of a closed cavity in which the surrounding air would be trapped.

It should be noted that the attenuation coefficient m of a material is in all

rigor a spectral property m (l) according to the wavelength l or the energy of the X-rays. This characteristic is not necessarily taken into account insofar as the source of X-rays having a specific emitted spectral composition, it It is possible to consider that the attenuation m is a characteristic of the material for the spectrum of the chosen source. A person skilled in the art will also know how to carry out the invention using any method of taking into account the spectral attenuation or the hardening of the beams.

Of course, local and / or temporal variations of the coefficient

m low amplitude attenuation does not prevent the implementation of the method, but could possibly, depending on their amplitude, cause slight or significant losses of precision in the measurements carried out by the installation. It is therefore considered that such small variations, due for example to variations in the composition of objects, variations in parameters of the manufacturing process, modifications of environmental conditions, or even changes in the operation of X-ray sources, are possible while considering as verified the uniqueness and the constancy of the attenuation of the material in each part of the object. In On the other hand, the invention does not apply to heterogeneous parts of objects such as conglomerates of coarse grain size, mortars with rollers since the heterogeneous grains and attenuations are larger than the resolution of the images. The invention does not apply to checks of objects of shape and content unknown in advance, such as baggage. For the same reasons, the invention does not apply to medical or biological imaging in general, except for objects meeting the criteria.

The attenuation of the air can be considered negligible compared to that of the material (s). In this case, the attenuation of an X-ray beam passing through the object will depend only, on the one hand, on said uniform attenuation for the spectrum of X-rays emitted, and, on the other hand, on the thickness cumulative amount of material traversed. Alternatively, it is considered that the thickness of the air passed through is large and uniform for all radii, so it can be considered to be known. The attenuation due to air can be subtracted from the total measured attenuation.

Thus, it can for example be considered that the gray level in each radiographic image, possibly corrected, depends solely and directly on the total cumulative thickness of material traversed. It is then possible to determine with precision boundary surfaces which are the transitions between air and matter.

[0100] The digital analysis of the radiographic images of an object therefore makes it possible to know the relative position in space of a certain number of points of the boundary surfaces of the object.

[0101] The digital analysis of the radiographic images of each object possibly makes it possible to construct a three-dimensional digital geometric model of each object, referred to as a digital geometric model in the remainder of the description. Optionally, this digital geometric model can simply be a stack of digital geometric models.

two-dimensional. Making a digital geometric model is how - in mathematical, graphical and data structure terms - three-dimensional objects are represented and manipulated in digital form in a memory of a computer system. It should be considered that the invention, in some of these embodiments, makes it possible to determine as many three-dimensional digital geometric models as there are objects x-rayed, and that there can be as many objects x-rayed as the number of objects passing on the transport system. Indeed, a characteristic of the invention is that it makes it possible, if necessary, to carry out a measurement on each of the objects circulating in the installation, including at high speed.

[0102] It is possible to obtain a surface model directly from the radiographic images, that is to say without going through the calculation of a volume model.

In surface modelings, an object is defined by at least one

three-dimensional surface, in particular a closed oriented three-dimensional surface, corresponding to the external boundary surface between the material of the object and the external environment (generally air), which makes it possible to understand the concepts of interior and exterior of the object. Different modelizations are possible for such surfaces: implicit surfaces, parameterized surfaces (parts of plan, B-spline, NURBS, ...), possibly limited by a network of curves. A simple model is the triangular mesh, which can be seen as the boundary surface of a volume formed by tetrahedra.

[0104] A section of a three-dimensional object, that is to say the intersection of the object with a section plane, makes it possible to define a delineation of the object in the section plane. The section of the three-dimensional surfaces of the object, therefore of its boundary surfaces, determine one or more two-dimensional curves in the section plane which, taken together, form the delineation of the object in the section plane. Knowledge of these curves

two-dimensional in a succession of section planes allows the

reconstruction of three-dimensional surfaces, with a precision which obviously depends on the number of section planes.

[0105] In order to make measurements of linear dimensions such as lengths, there are several approaches.

In a so-called surface method, it is possible to calculate a segment whose ends are the intersections of a straight line with the material / air boundary surface of a surface model. Finally, a mixed method consists in transforming the solid model into a surface model, then in applying the second method. A third method consists in determining, in a section plane, the distance between two points of one or two two-dimensional curves, any curve being a border between the material and the air.

A three-dimensional point is a point whose coordinates are known.

in three-dimensional space, in any frame of reference.

[0109] These three previous methods are examples of determining a distance between two three-dimensional points, to determine a measure of linear dimension.

[0110] An objective of the invention is to achieve more comprehensive measurements than

those made possible by simple radiographic images

two-dimensional. In fact, it is easy, using a matrix image sensor, to obtain a two-dimensional radiographic image corresponding to a projection of the inspected region and to measure dimensions in a plane orthogonal to the direction of projection called "plane. projected ”. Likewise, it is easy with the aid of a linear image sensor to obtain a two-dimensional radiographic image corresponding to a projection of the inspected region obtained by juxtaposition of the successive image lines acquired during the displacement in the direction of. displacement, and measure dimensions in a projected plane, which is parallel to the direction of displacement. On the other hand, according to the invention, linear dimensions can be measured in directions that are neither contained in the projected planes, nor parallel to the projected planes. The method according to the invention may in fact include, during the processing of a combination of the radiographic images in at least three different projection directions, the reconstruction and measurement of the dimensions in virtually all directions. This is possible by any method

allowing the determination of three-dimensional points in space belonging to a boundary surface included in the region to be inspected of the object. The

One possible method is reconstruction of a three-dimensional model of the region to be inspected, of a surface or volume type or based on section planes. In fact, it is possible, indirectly, from a three-dimensional volume model of the region to be inspected, and preferably from a three-dimensional surface model of the region to be inspected, possibly by

determining sections of the three-dimensional models of the region, determining at at least two three-dimensional points, or even preferably clouds of three-dimensional points, distributed in directions that cannot be measured from the two-dimensional radiographic images alone.

[0111] A digital geometric model is therefore composed of elements

geometric shapes such as points, segments, curves, surfaces, calculated from radiographic projections, considering in order to calculate each element, the attenuation of at least some X-rays having passed through this point on the real object, with the aim that the digital geometric model is a faithful representation of the geometry of the real object, including deformations compared to an ideal object. In other words, the coordinates of the geometric elements are determined by considering that said coordinates have modified the radiographic projections, even when these geometric elements are not.

distinguishable in any of the 2D radiographic projections. The measurements of dimensions on the digital geometric model therefore give

information on the dimensions of each modeled object, from geometric elements not distinguishable in any of the radiographic projections.

[0112] Consequently, an advantage of the method according to the invention is that it makes it possible to determine, for each object, a digital geometric model consisting of at least two three-dimensional points, each of these two points belonging to a border surface of the region to be inspected, even if these two points are not located in a plane orthogonal to a direction of projection Dji, Djik, nor in a plane parallel to the direction of displacement.

[0113] Of course, the advantage of the method is not only to provide

measurements in directions outside a plane orthogonal to a projection direction Dji, Djik, and outside a plane parallel to the direction of displacement, but also to provide a large number of measurements distributed in the inspected region, therefore dimensions in many directions, between multiple pairs of points. Preferably, a digital geometric model is made up:

- at least two three-dimensional points in space each belonging to a border surface of the region to be inspected and not located in a plane orthogonal to a direction of projection Dji, Djik, and not located in a plane parallel to the direction T of displacement ; - and / or at least one three-dimensional surface of the region to be inspected, containing points not belonging to a plane orthogonal to a direction of projection Dji, Djik, and not belonging to a plane parallel to the direction T of displacement ;

- and / or at least one section of the region to be inspected, according to a plane different from a plane orthogonal to a direction of projection Dji, Djik and different from a plane parallel to the direction of displacement.

[0114] A so-called "a priori" geometric model is a geometric digital model of the series of objects, which can serve as an initialization for

reconstruction in order to build a digital geometric model of the object. Its role is then mainly to provide, to the computer system, information on the shape, geometry and dimensions of the object and / or of the different parts of the object to be modeled by the calculation, this information being however insufficiently precise for allow measurement of the object with the precision required for the measurement.

[0115] Thanks to this information, it becomes in particular possible:

- not to model, from the radiographic images, the attenuation in regions of the image space empty of material a priori because the attenuation is considered to be zero there;

and or

- to model from the radiographic images only the surfaces on which the measurements of dimensions are to be made;

and or

- to determine only the differences between the surfaces modeled from the radiographic images and the theoretical ideal surfaces.

[0116] In the case of mono-material objects, knowledge of the a priori geometric model also makes it possible not to determine, from the images

radiographic, attenuation values in regions of space of the image containing material according to the a priori model because it is known as that of the material of manufacture of the object.

However, it should be understood that according to the invention, no measurement of an object is deduced from a measurement on the a priori geometric model, since this model is known independently of said object and represents a non-real theoretical ideal.

[0118] As emerges from the drawings and more precisely from Figs. 1 and 2, the object of the invention relates to an installation 1 allowing the implementation of a method for automatically performing linear dimensional measurements on manufactured objects 2 moving in scrolling at high speed. The invention relates to a so-called "on-line" control of a series of manufactured objects, the objects of a series being assumed to be identical, after a processing or manufacturing step, in order to control the quality of the objects or of the manufacturing process. processing or manufacturing. The objects are assumed to be identical since no intentional action is taken to make them differ. However, it is well known that, in a series, all objects are not the same, due to the vagaries of transformation or manufacture.

The method operates for a scrolling rate of a stream of objects 2.

Ideally, the plant 1 is capable of processing production at the production rate, for example at 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. minute.

[0120] However, the calculation time can exceed the interval between two objects. Likewise, the exposure times, also called integration times, of the image and reading sensors may be too long. If the fastest flow cannot be handled by a single installation according to the invention, then several installations can be implemented in parallel, each controlling part of the production. Thus it is possible to divide the production flow into two or three parallel flows inspected by two or three installations according to the invention. Obviously, the economic interest of the invention is increased if the number of flows and therefore of installations according to the invention remains low.

[0121] The invention provides a considerable improvement thanks to the measurement of moving objects, avoiding the helical scanning and the scanning on a plate which are not adapted to the production rates because these two modalities, implying a relative rotation of the objects. compared to foci and / or sensors, create a “break in scrolling” or very slow movement of objects within the installation.

[0122] The method according to the invention ensures the measurement, preferably on each

object 2, of at least one and generally of several linear dimensions, i.e. lengths. A linear dimension is indeed a length measured along a line. This line according to which the linear dimension is measured can be a rectilinear line, or a non-rectilinear line, for example any curved line, a circular line, a broken line, etc ... This line can be a flat line, contained in a plane, or a three-dimensional line that is not included in a plane. A length is a measurement expressed in units of length, such as inches or meters. A linear dimension of a manufactured object is for example a diameter, a thickness, a height, a length, a width, a depth, a distance, co-ordinated as the distance of a point from an origin, a perimeter of the manufactured object. At least one linear measure of the inspected region is the distance between at least two three-dimensional points each belonging to a boundary surface, in particular an external boundary surface, of the region to be inspected and located in a plane, including a plane not orthogonal to a direction of projection Dji, Djik.

[0123] According to the invention, the objects 2 are supposedly identical objects, to

dimensional variations close, forming a series of objects. In other words, a series is composed of theoretically identical objects when they conform. Dimensional control consists of measuring actual dimensions and comparing them to the required dimensions. A priori, any object in a series is close to an ideal reference object with the required dimensions but deviates from them through dimensional variations.

[0124] According to an advantageous embodiment characteristic, at least one region of the object 2 is chosen to be inspected so as to be able to carry out measurements of dimensions in this region of the object, corresponding to a dimensional characteristic of the region to be measured. inspect. At least the region of the object in which the linear dimension (s) are to be measured is inspected by X-rays. Thus, the inspected region can correspond to the whole of the object or to one or more regions of this object. As indicated, all the objects 2 of a series consist of a single part, or of several distinct parts, each part being constituted by a material having a uniform attenuation coefficient at any point of the part considered. the object.

[0126] According to an advantageous variant of the invention, for each part of the

region to be inspected, this coefficient is known by the computer system. The method may provide a means of providing for the system

computing the value of the material's attenuation coefficient. This value can be spectral, in the sense of a value which is a property of the material, which defines the interaction of this material with a radiation and which depends on the wavelength of the radiation. This value can be non-spectral, in the sense of being independent of the wavelength of the radiation. This value can be made dependent on the settings of the X-ray sources. The provision is possible by various capture, communication and memory devices. For example, the device for making the value of the material's attenuation coefficient available to the computer system is a mass memory, a wired or wireless computer network or a man / machine interface.

[0127] The installation 1 also includes a device 5 for transporting objects 2 in a conveying plane PC, that is to say along a plane path, with a direction materialized by a displacement vector T. Preferably, the path is rectilinear, within the limits usually accepted for the straightness of a conveyor line. Conventionally, the transport device 5 is a conveyor belt or chain conveyor ensuring linear translation of the objects 2 which are deposited therein. Thus, the objects 2 of the same series are essentially translational movement in the conveying plane PC. As emerges more precisely from Figs. 1 and 2, we take, for the convenience of this description, the convention according to which the direction of movement of the objects 2 is established along a horizontal axis X of a frame X, Y, Z comprising a vertical axis Z perpendicular to the 'horizontal axis X and a transverse axis Y

perpendicular to the vertical axis Z and to the horizontal axis X, and X and Y being in a plane parallel to the conveying plane PC which is preferably, but not necessarily, horizontal. The position of the objects considered in a mobile frame orthonormal in translation in the direction T is fixed during their movement and the acquisition of the radiographic images. This fixed position implies in particular an absence of rotation of the object in a mobile frame orthonormal in translation in the direction T, in particular for example an absence of rotation of the object around a possible axis of symmetry of the object. For example, the objects are placed on the conveyor belt, resting stable, possibly on a clean laying surface such as the bottom of a container or the legs of a seat.

In a variant of the invention, it is possible to provide a support for the objects 2. In this case, this support is fixed in the orthonormal mobile frame, in translation in the direction T, and it holds the object. also fixed in the orthonormal mobile frame in translation in the direction T. So that the support does not influence the measurements, according to a first variant it is excluded from the inspected region so as not to appear in superposition of the inspected region in the projections. According to a second variant, its attenuation coefficient is negligible relative to that of the objects and can be likened to air or to zero attenuation. According to a third less advantageous variant, the geometry of the support, as well as its position in the moving frame, are precisely known and repeatable for the series of objects and its attenuation coefficient is known precisely and stable, and preferably identical to that of the objects of the series of objects, so that the support is taken into account in the reconstruction and isolated from the geometric model of the object.

[0130] The position of the objects being stable (during scrolling and acquisition of the radiographs), it remains preferable that this position in the orthonormal mobile frame in translation in the direction T, is also the same for each object in a series. objects.

If this is not the case, it is then possible, according to a variant of the invention, to implement a means for determining the position of each object in the mobile frame orthonormal in translation in the direction T with respect to a common reference mark of the installation, this position being taken into account, for example by the means for calculating the delineation of the object according to section planes which will be described later. This preliminary step consists in determining the position of each object. She will be able to understand the correspondence, in a virtual frame of reference, of the images acquired with a priori delineations of the object, which may for example be derived from an a priori geometric model. This amounts in all cases to determining the delineations of the objects, and possibly a 3D model of the objects which would be drawn from them, in the orthonormal mobile frame of reference in translation in the direction T.

On the other hand, it is understood that, if the position of the objects is stable in the mobile frame orthonormal in translation in the direction T during the scrolling and acquisition of the radiographs, it is not necessary to determine the position of each object. in relation to a common reference of the installation.

[0133] As emerges more precisely from FIG. 4, during their

displacement in translation, the objects 2 generate or pass through a so-called conveying volume Vt. The plane PS is the secant plane of the conveying volume Vt, orthogonal to the conveying plane PC and parallel to the direction of movement T. For example, the plane PS can be a median plane which separates the conveying volume Vt into two sub- equal volumes. The intersecting plane PS is a vertical plane insofar as the conveying plane is horizontal.

[0134] The installation 1 also comprises, as illustrated in Figs. 1 and 2, at least one focus Fj (with j varying from 1 to NF) of an X-ray generator tube 7 creating a divergent beam of X-rays directed to pass through the conveying volume Vt and more precisely to cross at least the region to be inspected of the object 2. In the variants in which the installation comprises several hearths Fj, as illustrated in Figs. 5 and 6, all the foci that will be used for the method according to the invention will be arranged on the same basic line B, parallel to the direction T of displacement along the rectilinear path. This does not preclude the possible presence of one or more auxiliary foci (not shown) which could be used for the acquisition of other images.

The installation 1 also comprises image sensors Ci, Cik (with i varying from 1 to N and N which may, in certain cases, be greater than or equal to 3) sensitive to X-rays and located so as to be exposed to X-rays coming from a focus Fj and having passed through the conveying volume Vt and more precisely, at least the region to be inspected of the object 2. Of course, the tube or tubes 7 and the image sensors Ci, Cik are located outside the conveying volume Vt to allow the free movement of objects in this volume. Conventionally, the X-ray generator tubes 7 and the image sensors Ci, Cik are placed in an X-ray sealed enclosure.

[0136] In certain embodiments, an image sensor Ci, Cik is associated with a single focus Fj in the sense that, in the implementation of the method, this image sensor Ci is provided so that the images that it delivers and which are taken into account in the method are formed only of rays from the associated focus Fj. For example, the installation can be designed so that only rays from a given focus can reach the associated image sensor, for example by arranging absorbent masks in a suitable manner. According to another example, which can be combined with the previous one, it can be provided that the acquisition of images by an image sensor is triggered only when only the single associated focus is activated.

[0137] However, in certain embodiments, several image sensors can be associated with the same focal point Fj, and / or several focal points Fj can be associated with the same image sensor. In a preferred embodiment, several image sensors are associated with the same focus Fj.

[0138] As indicated above, an image sensor Ci, Cik corresponds to a

physical sensor component CC1, CC2, ..., CCnmax, ...., to a part of a physical sensor component or the assembly of several parts of one or more physical sensor component.

The X-ray beams coming from a focus Fj pass through at least the inspected region, and form, on an image sensor, a radiographic projection of the inspected region, in a projection direction Dji, Djik (Fig. 1 and 2). The direction of projection Dji, Djik is the oriented direction of the vector starting from the focus Fj to pass through the center of the image sensor Ci, Cik used for the acquisition. The focal point (s) Fj and the image sensors Ci, Cik are arranged so that each image sensor receives the radiographic projection of the region to be inspected according to the direction of projection.

[0140] Installation 1 also includes an acquisition system connected to

image sensors Ci, Cik, so as to acquire, for each object 2 during its movement, a set of one-dimensional radiographic images processing of the object, in which each one-dimensional radiographic processing image comprises a projection of a section of the object along a section plane Pk containing the base line B. More precisely, this set of images comprises:

such processing radiographic one-dimensional images for a number NK of distinct section planes Pk containing the base line;

for each distinct section plane Pk, a number NP of such one-dimensional radiographic processing images Spk, Sp'k, Sp ”k, of the region to be inspected, obtained according to at least three different projection directions Djik in the section plane Pk .

This set of images therefore comprises, for each section plane Pk in a plurality of section planes containing the base line, at least three one-dimensional radiographic processing images Spk of the region to be inspected of the object, which are each obtained according to a different direction of projection Djik in the section plane Pk.

[0142] In FIG. 2, we have illustrated the trace of several section planes Pk, Pk ’

distinct, containing the base line B and therefore containing the focus Fj. It is noted that, in embodiments comprising several distinct foci Fj, since these foci are placed on the same base line B, a given section plane Pk, Pk ’, Pk”. contains all Fj outbreaks. We note that the planes of distinct sections Pk, Pk ', ..., are by definition not parallel to each other, but arranged in a fan around the base line B. The base line B is therefore the intersection of all the section planes Pk, together constituting a family of planes.

[0143] A section plane Pk is therefore defined by the base line B, and by at least one sensitive element Cik, Ci'k, Ci ”k, ..., of an image sensor, contained in this section plane Pk. and / or, as in the example of FIG. 9, by a linear image sensor Cik, Ci'k, Ci ”k, ..., contained in this section plane Pk. In the case of a linear sensor, it can be oriented parallel to the base line B, or be oriented in a direction that intersects the base line B. The base line B being fixed for installation, the plane of section Pk is therefore associated with this sensitive element Cik contained in the section plane Pk, or with this linear image sensor Cik contained in the section plane Pk. For at least some section planes Pk, the section plane is therefore likely to cut the object

2.

[0144] The intersection of a section plane Pk with the object 2 defines a section of the object 2. As seen above, each section plane Pk defines, at its intersection with the object, a delineation of object 2 in the section plane, the delineation being formed by one or more two-dimensional curves, considered together in the section plane.

[0145] Note that there is only one section plane Pk which is parallel to the conveying plane PC. We have seen above that the conveying plane is considered, at least in certain embodiments, to be horizontal. In some applications of the invention, the series of objects may be a series of containers, including bottles, especially glass bottles. For these containers, it is generally observed that they have a central axis along which their general shape is elongated, this central axis possibly being, for some bottles, an axis of symmetry, or even an axis of symmetry of revolution. Usually, such objects are transported with their central axis in a vertical position. It follows that the different section planes Pk each intersect the object at a different angle with the central axis of the object. In the context of an object which would have mainly an envelope or external surface of cylindrical shape of revolution, the section planes Pk then intersect the object so that the external contour of the object appears in the form of a curve elliptical.

The section plane Pk also defines a projection of this section of the object on an associated sensitive element Cik contained in the plane Pk, or on a linear image sensor Cik oriented parallel to the base line B and contained in the Pk plan.

[0147] In operation, the sensitive element Cik which is intersected by the plane Pk, or the linear image sensor Cik which is intersected by the plane Pk, makes it possible to obtain a one-dimensional radiographic image Spk of a projection of this. section according to this section plane.

[0148] In both cases, whether it is acquired directly by a linear image sensor, or acquired indirectly by sampling with a single element sensitive, this one-dimensional radiographic image will be used for the treatment according to the invention and will therefore be referred to as a one-dimensional radiographic treatment image.

[0149] In the case illustrated in the examples of FIGS. 2, 3 and 5 where a unique

Cik sensitive element is implemented to acquire this image

unidimensional radiographic treatment, it is necessary to form it by sampling a point image acquired using a sensitive element Cik, for a scanning period corresponding to the duration of the movement of the object between the focus and the 'Cik sensitive element, during its movement, therefore by juxtaposition of several images acquired

successively in time, during the scanning time, by the Cik sensitive element used, which forms a point image sensor ,. It is noted that then, each image acquired successively over time by the point image sensor Cik used is a pixel of the one-dimensional radiographic processing image. The time between two point images, acquired successively by the implemented Cik point image sensor, corresponds to an incremental movement of the object along the movement path. The scanning time therefore corresponds to the time required for

the entire section of the object by the section plane Pk has passed through the line which supports the direction of projection. We also note that, in this case, each pixel is obtained with a direction of projection

strictly identical, corresponding to the direction of the vector joining the focus Fj to the sensitive element Cik, since all the pixels of the image

radiographic treatment are obtained by the same sensitive Cik element, with only a time lag.

[0150] In the case illustrated in FIG. 9 where a Cik linear image sensor is used to acquire this one-dimensional radiographic image of

processing, it can be acquired in a single acquisition time or sensor integration time. The one-dimensional linear radiographic processing image then corresponds to the simultaneous acquisition of several point images each delivered by a sensitive element belonging to a set of sensitive elements aligned consecutively on a straight line contained in the section plane Pk. We then note that, in this case, each pixel of the processing radiographic image corresponds to a different sensitive element, and it follows that each pixel is obtained with a projection corresponding to the direction of the vector joining the focus Fj to the sensitive element specific to this pixel. However, by convention, it has been considered that the radiographic projection direction for this processing radiographic one-dimensional image is that which joins the focus Fj to the center of the linear image sensor is implemented to acquire this radiographic one-dimensional image of

treatment. We therefore consider here a direction of projection Djik which is made an average direction of projection for the one-dimensional processing radiographic image.

[0151] Each pixel of the one-dimensional Spk processing radiographic image consists of a value representative of the signal collected by the corresponding sensitive element during the integration time of this pixel. The value of this signal therefore depends on the intensity of the X-ray received, therefore depends on the cumulative attenuation undergone by the X-ray between the focus Fj and the sensitive element

corresponding. This depends on the thickness and the coefficient of the material or of each of the layers of materials which have been traversed by the X-ray between the focus Fj and the corresponding sensitive element. Each radiographic one-dimensional image of Spk processing of the region to be inspected can therefore be represented by a set of these values, digital or analog, for the entire section of the object by the corresponding section plane, this set of values having been acquired in a single acquisition time of a linear sensor or in successive acquisition times of a point image sensor.

The invention provides that, for each object to be measured, and for each distinct section plane Pk, the delineation of the object in the section plane Pk considered is determined. This determination, in each section plane, is carried out on the basis of a number of one-dimensional radiographic treatment images of the region to be inspected, obtained according to at least three different direction of projection Djik in the section plane, therefore on the base of at least three one-dimensional radiographic images of Spk processing of this section of the region to be inspected of the object, preferably between 3 and 40, more preferably between 8 and 15, which are each obtained according to a different direction of projection Djik in the section plane. [0153] Note that this must be repeated for each section plane which intercepts the region to be inspected of the object. So this has to be repeated for NK section plans Pk.

[0154] Also, in the examples of FIGS. 1 to 3, it is intended to use physical sensor components which are linear (only two of these linear physical sensor component sensors are shown in Fig. 2). In these examples, each physical sensor component CC1, CC2, ..., CCnmax comprises a linear array of elements sensitive to X-rays, distributed along a support line Ln defining with the associated focus Fj, a projection plane PPji (Fig. . 2). In the example, the linear physical sensor components have their support lines Ln parallel to each other and orthogonal to the conveying plane PC. In the example of FIG. 1, there are eight linear physical sensor components. These physical sensor components CCi are arranged so that at least m sensitive elements Cik of each of these physical sensor components receive a radiographic projection of the region to be inspected of the object by the X-ray beam coming from the focus Fj during the displacement of the object between the focus Fj and the sensor. In each section plane Pk, each linear physical sensor component CCi comprises a single sensitive element Cik, The sensitive elements Cik of each linear physical sensor component CCi are therefore each capable of acquiring, by sampling during a scanning period corresponding to the duration the displacement of the object between the focus and the sensitive element, therefore by juxtaposition of several images acquired successively over time, a one-dimensional radiographic image with Spk processing. It is understood that, in a given section plane Pk, each sensitive element present in this plane belongs to a different linear physical sensor component, and defines, with the focus Fj, a different direction of projection, and that one-dimensional images are obtained radiographic processing Spk according to different projection directions Djik, in this case according to as many different projection directions as there are sensitive elements in the section plane Pk. As indicated above, the linear physical sensor components are preferably arranged such that they define different directions of projection Dji, Djik forming between them, taken two by two, an angle minimum at least equal to 3 degrees of angle, preferably at least equal to 5 degrees of angle in the section plane Pk, this in each section plane Pk implemented.

[0155] In the example of FIG. 5, we find the same principle, but with two focal points F1 and F2, aligned on the base line B, each of which is associated with several, in this case five linear physical sensor components having their support lines Li parallel to each other and orthogonal to the PC conveyor plane. Thus, we obtain for each plane of section Pk, by

sampling, for each linear physical sensor component, of the point image acquired by the sensitive element Cik of the linear physical sensor component which is included in the section plane considered, 10 images

unidimensional radiographic processing Spk according to 10 different projection directions Dji.

[0156] In the example of Figs. 9 to 10, we plan to use components

physical sensors that are matrix, or two-dimensional. In this example, there are three of them. In this example, each physical sensor component CC1, CC2, CC3 comprises a matrix network of elements sensitive to X-rays, distributed according to a support plane. In the example, the matrix physical sensor components CC1, CC2, CC3 have their support planes which are not parallel to each other. Indeed, in this example, each matrix physical sensor component CC1, CC2, CC3 is arranged so that its support plane is orthogonal to the direction of projection defined by the focus Fj and the center of the matrix physical sensor component CC1, CC2, CC3. However, the matrix physical sensor components CC1, CC2, CC3, or at least some of them, could be parallel to each other. In this example, their respective support planes are perpendicular to the conveying plane PC.

For each matrix physical sensor component CC1, CC2, CC3, at least m lines of sensitive elements Cik receive a radiographic projection of the region to be inspected of the object by the X-ray beam coming from the focus Fj when the object is located between the focus Fj and the sensor. In each section plane Pk, each physical matrix sensor component CCi comprises a line

of sensitive elements which form a linear image sensor Cik. The sensitive elements of this linear image sensor Cik are therefore susceptible, together, to acquire a one-dimensional radiographic image of Spk processing in a single acquisition time or integration time of the sensor. Typically, the sensitive elements on a matrix sensor component are arranged in a matrix arrangement, with vertical (columns) and horizontal (rows) alignment of the sensitive elements. In the case where a matrix sensor component CCi constitutes a plane not parallel to the base line B, the intersection of the planes Pk with the plane of the matrix sensor component CCi is a line not aligned with the physical arrangement of the pixels. In this case, the linear image sensor Cik is a subset of sensitive elements which does not follow

the horizontal or vertical alignment of the array sensor component arrangement. The linear image corresponding to the linear image sensor Cik, which is therefore part of the matrix sensor component CCi, can be obtained by combining the values delivered by the sensitive elements cut by a virtual straight line, representing the linear image sensor Cik, possibly also taking into account the pixels supplied by neighboring sensitive elements. For example, it is known to those skilled in the art to use interpolations and resamples in the combination of pixels.

It is understood that, in a given section plane Pk, each line of sensitive element forms a linear image sensor Cik in a matrix physical sensor component and defines, with the focus Fj, a different direction of projection, and that one-dimensional radiographic processing images Spk are obtained according to different projection directions Djik, in this case according to as many different projection directions as there are matrix physical sensor components CCi.

From the delineations of the object in each distinct section plane, themselves each obtained from radiographic processing images according to at least three different projection directions in this section plane, it is possible to determine, for the object to be measured, at least one linear dimension measure of the region to be inspected of the object to be measured. For example such a linear dimension measure of the region to be inspected of the object to be measured can be determined as the distance between at least two points

three-dimensional objects each belonging to a border surface of the region to be inspected. For this, the acquisition system is linked to a computer system, a nonlimiting example of which is represented symbolically in FIG. 3, but which can be of any type known per se. According to an advantageous embodiment characteristic, the computer system records with the aid of the image sensors Ci, Cik, for each object of the series during its movement, radiographic images originating from a determined number of radiographic projections of the region to be inspected from different projection directions.

[0159] We have seen that the delineation of the object comprises a curve or a set of curves which represent the intersection, with the section plane Pk, of the boundary surfaces of the object. In such a delineation, the curve or each curve of the delineation is a plane curve which can be modeled by a parametric system, in particular a system of one or more parametric equations. The curve or each of the curves of the delineation is preferably modeled by a fixed number of parameters. Such a delineation curve can for example be a polygonal curve. In this case, the coordinates of the vertices of the polygon can act as parameters.

[0160] Among the possible methods for determining a delineation of the object in the section plane, it is possible to implement a curve fitting algorithm starting from an a priori delineation of the object in the section plane. the section plane.

[0161] Such an algorithm can be an iterative algorithm, in particular a nonlinear regression algorithm.

[0162] As illustrated in FIG. 12, an iterative algorithm 100 capable of being used can thus iteratively implement a simulation step 110, a comparison step 120 and a step 130 of error reduction by adjusting one or more parameters in the parametric system .

[0163] Such an algorithm can advantageously, as an initial step 101,

taking into account 101 an a priori delineation DLk1 of the object in the section plane Pk as a calculated delineation of the first iteration rank. Such a delineation a priori DLk1 amounts to defining, in a given section plane, an initial curve or a set of initial curves, preferably sufficiently close (s) to the expected delineation, which is however not known. Of preferably, an initial delineation will make it possible to determine the number and the order of the materials which are cut by a number of X-rays emitted by the focus Fj, contained in the section plane considered, and collected by the image sensor after having crossed the object. This a priori delineation DLk1 can be taken from an a priori geometric model of the object, and / or from the measurement of one or more objects of the same series by a measuring device, and / or from values entered and / or drawings produced and / or shapes selected by an operator on a man-machine interface of a computer system. In the case of a series of hollow cylindrical objects comprising only one material and having a theoretical central axis perpendicular to the conveying plane, the delineation in a section plane may consist of an initial inner curve and of an external initial curve, preferably closed curves, for example of circle or ellipse type, or of polygonal curve type. Such curves will be particularly effective for objects of revolution around the theoretical central axis. They may however also be used with satisfactory results for objects whose section through a plane perpendicular to the theoretical central axis is of the prismatic type.

[0164] An iterative algorithm can then iteratively implement the following steps.

An iterated step may be a simulation step 110 which involves the

SIM calculation, from the calculated delineation DLkr of a given iteration rank r of the object in the section plane, of a number NP at least equal to three simulated radiographic one-dimensional images SSpkr, SSp'kr , SSp ”kr of the region to be inspected, each calculated in the section plane Pk according to one of the different projection directions Djik which were used for the acquisition of the one-dimensional radiographic processing images Spk, Sp'k, Sp” k in the section plane.

[0166] Here, therefore, we consider one of the one-dimensional radiographic processing images Spk, Sp’k, Sp ”k, therefore an image for which real values of the image signal are known, as collected by the image sensor.

This image necessarily corresponds to a given direction of projection Dijk, therefore to a given focus Fj and a given linear image sensor Cik. The principle is to calculate, for each given iteration rank of the iteration, a estimate SSpkr, SSp'kr, SSp ”kr of the values representative of the signal which would be collected by this same sensor for an X-ray beam emitted by the same focal point Fj but after having passed through an object whose delineation would have the calculated delineation DLkr of rank iteration r given. During the first iteration, it is possible to use the calculated delineation of the first rank of iteration DLk1, namely the initial delineation. For the following iterations, we use the delineation calculated during the previous iteration.

[0167] At each iteration, this calculation is performed for the number of images

unidimensional radiographic processing which are taken into account by the method in this section plane, namely at least three, for example between three and forty, within the meaning of three to forty, limits included, plus

preferably between four and fifteen, within the meaning of four to fifteen, limits included.

[0168] In this way, it is then possible to perform, at each iteration, the

120 comparison of the simulated radiographic one-dimensional images

SSpkr, SSp’kr, SSp ”kr to one-dimensional processing radiographic images Spk, Sp’k, Sp” k. This comparison may be performed, e.g. for each simulated radiographic one-dimensional image as a comparison function COMP, e.g. as a function of the differences of the signal values of the simulated radiographic one-dimensional image SSpk from the values of the Spk processing radiographic linear image corresponding to the same radiographic projection. This function of the differences can be a function of pixel-to-pixel difference. This

comparison can be made, for example, by considering together several or all the simulated radiographic one-dimensional images SSpkr, SSp'kr, SSp ”kr with the associated one-dimensional treatment radiographic images Spk, Sp'k, Sp” k, This comparison can for example understand the calculation of a COMPVAL comparison value, which may be an error value, for example a quadratic error value.

[0169] We note that this iterative algorithm works on linear images

radiographic processing in a given section plane Pk, therefore on images belonging to the same plane. The calculations implemented in the iterative algorithm therefore relate to data belonging to the same plane determined Pk, corresponding to plane two-dimensional entities that are the delineations. This considerably simplifies the calculations compared to an iterative algorithm which would relate to data corresponding to three-dimensional entities.

[0170] Depending on the comparison, the iterative algorithm can then predict, for example at each iteration before the last, the modification 130 of the

calculated delineation having the iteration rank r considered in a new calculated delineation DLk (r + 1) of higher iteration rank (r + 1), which will be used for the next iteration, therefore of higher iteration rank. The modification implemented can be perhaps a MOD modification function, which can take into account the current delineation Dlkr, the comparison value COMPVAL calculated by the comparison function COMP at the comparison step 120, and / or possibly of other calculations made in the preceding iterations in order to reduce the value of this COMPVAL function at the following iteration according to the principle of the descent methods in optimization. A modification function that can be used is, for example, a least squares method, for example a linear least squares method.

[0171] The above steps can in fact be repeated until, for a last iteration, the comparison reaches a predefined CRIT optimization criterion. The achievement of this predefined optimization criterion can be verified during a verification step 125. It is possible, for example, to verify whether a verification function returns a defined value. One or more criteria can be checked, such as the value of the COMP function with respect to a threshold value, the number r of iterations with respect to a maximum threshold number, etc. This verification can be based on the value of COMPVAL comparison, for example by checking that it has reached or exceeded a predefined squared error value.

[0172] Typically, for each section plane, the number of directions of

different projection Dji, Djik is between three and forty, within the meaning of three to forty, limits included, and preferably between four and fifteen, within the meaning of four to fifteen, limits included. Also, according to a variant

advantageous embodiment, the installation 1 comprises, in a given section plane, between three and forty image sensors Ci, in the sense of three to forty, terminals included. According to a preferred variant embodiment, the installation 1 comprises, in a given section plane, between four and fifteen image sensors Ci, in the sense of four to fifteen, terminals included.

[0173] As will be explained in detail in the remainder of the description, the

computer system is programmed to analyze, for each object, the at least three one-dimensional radiographic images obtained, in each section plane from at least three radiographic projections of different directions, so as to know the three-dimensional geometry of the object.

In some cases, we can thus choose to build a model

digital geometric of each measured object. This digital geometric model can be produced in any suitable manner, with a degree of precision depending on the precision required for the desired distance measurement. Thus, the digital geometric model can be formed by at least two three-dimensional points each belonging to a border surface of the region to be inspected of the object and not located in a plane orthogonal to a direction of projection Dji, Djik, and not located in a parallel to the direction of displacement T. The at least two points can belong to two different border surfaces, for example to measure a thickness or an air gap.

[0175] The digital geometric model can also consist of one or more preferably several sections of the region to be inspected, each section being in a different plane from a plane orthogonal to a direction of projection Dji, Djik. This section plane can be one of the Pk section plans implemented for image acquisition or be an even different plane. Furthermore, the digital geometric model can consist of at least one three-dimensional surface of the region to be inspected, different from a plane orthogonal to a direction of projection Dji, Djik and different from a plane parallel to the direction of displacement T.

According to the methods described above, one thus obtains, for each measured object, the delineation of the object in a whole series of section plane Pk containing the base line B, therefore in fan-shaped planes around this basic line. Each delineation can be rendered as a parametric system, e.g. rendered as a set of points and / or segments, in particular a set of points and / or segments belonging to external boundary surfaces of the object. The set of delineations thus obtained can be considered as a geometric model of the object, obtained by measurement. As a variant, a geometric model of the object can be constructed from this set of delineations, for example by interpolation methods. Thus, the delineations in the Pk planes can be merged into 3D curves (borders), for example to obtain an STL type model. Then, it is possible to cut the 3D surface model thus obtained by section planes which correspond to planes in which the measurements are made.

In summary, the above methods allow the construction, for an object to be measured, using the computer system and from the delineations of the object in each of the separate section planes Pk, of a model three-dimensional digital geometric of the region to be inspected comprising:

- three-dimensional points in space each belonging to a boundary surface of the region to be inspected of the object;

- and / or at least one three-dimensional surface of the region to be inspected.

In cases where the determination of a three-dimensional geometric model is implemented, it is possible to determine a measure of linear dimension of the region to be inspected of the object to be measured by determining the distance between at least two three-dimensional points of the three-dimensional digital geometric model of the area to be inspected.

[0179] Of course, the invention makes it possible to construct a geometric model

digital with a large number of three-dimensional points or three-dimensional point clouds.

[0180] The digital geometric model is constructed using the attenuation coefficient of the material (s) of the objects in the series.

[0181] In certain embodiments of the invention, it has been seen that the

delineations in each section plane, and therefore possibly a digital geometric model, can be constructed using a model

geometric a priori of the region to be inspected for the series of objects. In other words, in such a case, the computer system uses to build the digital geometric model of each object, on the one hand, a geometric model a priori of the region to be inspected for the series of objects, and, on the other hand, the attenuation coefficient of the material or the different attenuation coefficients of the different parts of each object in the series.

[0182] Thus, the computer system takes into account the coefficient (s)

attenuation of the material (s) of the objects being inspected for this calculation operation. Advantageously, the installation 1 comprises a device for making available to the computer system, the attenuation coefficient (s) of the material (s) of the objects in a series.

[0183] This provisioning device can be produced by a memory of

mass, a man-machine interface or by a wired or wireless computer network.

[0184] Likewise, in certain embodiments, the computer system

has a so-called a priori geometric model of the region to be inspected to perform this calculation operation. Thus, the installation 1 may include a device for making available to the computer system, an a priori geometric model of the region to be inspected for the series of objects.

[0185] The device for making an a priori geometric model of the region to be inspected available to the computer system is, for example, a mass memory, a wired or wireless computer network or a man-machine interface.

As indicated in the definition part, the a priori geometric model is a digital model of the series of objects, which can be positioned in a frame of reference linked to the device, and which can serve as an initialization for a determination method. delineations of the object in each section plane Pk.

[0187] In the absence of knowledge of the a priori geometric model, the

reconstruction can be extremely computationally expensive, because for each point in 3D space its attenuation must be calculated. The implementation of an a priori geometric model thus makes it possible to carry out measurements of linear dimensions on objects, with good precision, in a very short time and at low cost. [0188] According to a first variant, the a priori geometric model is obtained by the digital computer design model of the objects of the series, produced during the design (3D CAD) of the objects. In this case, it is made available to the computer system by various possible means, such as a connection through a computer network, to a database containing several CAD models corresponding to the various series of objects capable of being measured in production, selection by the operator from a database internal to the installation, etc.

According to a second variant, the a priori geometric model is obtained from a geometric digital model built from the measurement of one or more objects of the same series by a measuring device, for example by a machine. measuring by feeler or an axial tomography device, the slowness of which compared to the invention is recalled. The a priori geometric model can be built by a fusion of measurements of several manufactured objects of the same series.

[0190] According to a third variant, the a priori geometric model is a geometric digital model generated by the computer system from values entered and / or drawings made and / or shapes selected by an operator on the man-machine interface. of the system.

[0191] For example, to provide the a priori geometric model in the case of a standard type M13 external hexagon nut, with a threaded hole, the following is sufficient. The operator enters on a keypad the number and height of the flats, the diameter and the thread pitch, the system being configured to inspect metric nuts. No additional precise dimension is indicated. In another example, for the inspection of a single-material container made of glass or synthetic polymer material such as polyethylene or polyester, the operator only gives as information that the object is a closed cylinder at the bottom, surmounted of a cone, two diameters, two heights and one thickness are sufficient for the computer system to know an a priori geometric model of the object to be inspected. According to another example, the computer system can, through its interfaces, receive technical descriptions of the a priori model such as a number, diameters, depths and positions of various bores present in a surface which would be part of the region to be inspected of a larger object. The description can be geometric for example if the computer system receives the number and the general shape of the boundary surfaces making it possible to describe it, the number of cavities, the number of faces or sides of a polyhedron. In summary, it should be understood that the a priori geometric model must at least contain sufficient technical, geometric, topological and / or numerical information, to inform the computer system on the 3D structure of the object, the degree of detail and accuracy of this information can be very low without penalizing the accuracy sought for linear measurements.

One of the advantages offered by a determination of a geometric model is that it is possible to determine, by the same computer system or by another system to which this model would be supplied, for each object in the series, from the digital geometric model of the region to be inspected corresponding to said object of the series, at least one linear measurement of the region to be inspected in any direction, therefore not necessarily contained in a plane orthogonal to a direction of projection, nor necessarily contained in a plane parallel to the direction of travel.

[0193] At least one dimension, and generally several dimensions are

controlled on objects 2. The objective is generally to compare the measurements obtained on objects with required values, for example defined by a quality department. These dimension measurements or the deviations of these measurements from the required values can be displayed, saved, etc. They can also be used to make object compliance decisions that can be sorted automatically.

[0194] The measurements can come from the measurements of the geometric model

number of the inspected region established for each object. For example, the inspected region may have a bore. In the digital geometric model, it is possible to determine measurements of diameter or depth of the bore, by calculating on the digital geometric model the distances between diametrically opposed surface elements. When the object is mono-material, the determination of the position of the surface elements can be more precise with a minimum of calculations. Another means of determining measurements of the diameter or depth of the bore is by comparing the digital geometric model of the inspected region with a reference or theoretical geometric model.

[0196] The geometric reference model is an ideal model from the series of objects inspected. To carry out a dimensional check, we can compare the digital geometric model of the inspected region with the reference geometric model, by an algorithm comprising the setting

correspondence of the models, then the measurement of the differences between the models. The geometric reference model can be obtained from CAD.

[0197] It is thus possible to carry out a matching operation of the digital geometric model of the inspected region with the reference geometric model, then to determine dimensional deviations by measuring distances between surface elements belonging to the model. reference and surface elements belonging to the digital geometric model. In the example of the bore measurement, it is possible to virtually position a cylinder of maximum diameter inscribed in the modeled internal surface of the bore, and likewise a cylinder of minimum diameter containing said surface internal modeled, and consider as measurements of the diameter of the bore in the inspected region, the diameter of one and / or the other of the cylinders inscribed and excreted. This type of analysis is also possible to verify internal diameters in the neck of glass bottles obtained by the press-blown or blown-blown processes, or in plastic.

[0198] According to a variant of the invention, the reference geometric model and the a priori geometric model are the same geometric model.

[0199] According to another variant of the invention, the a priori geometric model is less precise, less complete and / or is different from the reference geometric model.

[0200] To carry out such measurements, the installation advantageously comprises a device for making available to the computer system, values of linear dimensions, and / or tolerances on these dimensions, and / or geometric reference models. . [0201] According to an advantageous embodiment characteristic, the computer system is connected to a device for displaying the linear measurement values of the region to be inspected and / or dimensional deviations from reference values, and / or deviations. between the digital geometric model of the inspected region and a reference geometric model. For example, for a nut are displayed measurements such as a thread depth, an average thread pitch, an average thread root radius, a height, a minimum or maximum internal diameter, a flatness of one or more of its external faces. For a glass or plastic container, the system will display the total height and for example the minimum diameter and the maximum diameter of the cylindrical part at a height predefined by the adjustment of the dimensions to be checked. The ribs can be displayed with different colors depending on their conformity or not.

[0202] According to an advantageous embodiment characteristic, the system

computer is connected to a device for sorting objects according to the linear measurement of the region to be inspected. Thus, this sorting device can, for example using an ejector, eject from the transport device the objects considered to be defective in consideration of the measured linear dimensions.

[0203] According to an advantageous embodiment characteristic, the system

computer can be connected to a device for marking objects according to the linear measurement of the region to be inspected. This marking device can record, for example, the measured linear dimensions or the compliant or defective condition of the object.

[0204] The relative positions of the foci Fj and the sensors Ci, Cik in a fixed reference frame C, U, Z of the installation are known to the computer system. This position can be obtained by hypothesis or by calibration. Calibration consists, for example, of placing or conveying a precision machined gauge into the installation.

[0205] Of course, the relative positions of the foci Fj and of the image sensors Ci, Cik are diverse, being reminded that the foci Fj and the image sensors Ci, Cik are positioned outside the conveying volume Vt.

[0206] According to an alternative embodiment, the installation 1 comprises a single hearth Fj = F1 arranged on one side of the conveying volume Vt and a series of

linear physical sensor components CC1, CC2,, CCnmax, arranged with their support line perpendicular to the conveying plane and each comprising image sensors for a multitude of section planes, arranged on the opposite side of the conveying volume Vt to receive the rays coming from the focus F1 and having passed through the region to be inspected. In this example, the focus has an opening Of which is measured in at least one arbitrary plane, such as for example the X, Y plane in FIG. 1, which is greater than or equal to 120 °. This opening Of is considered at the outlet of the focal point, in the case where the installation comprises between the focal point and the volume Vt, or between the volume Vt and the image sensors, screens limiting the beams to only useful beams, in the goal of reducing the broadcast.

[0207] According to another variant embodiment, at least two focal points Fj (F1 and F2) for producing X-rays are positioned separately in two distinct positions along a base line B parallel to the rectilinear trajectory of the objects, and at least three physical sensor components each comprising, for a multitude of section planes, X-ray sensitive image sensors, are placed so that each focus is associated, in each section plane with at least one image sensor, and that each image sensor is associated with a focal point and receives the X-rays coming from said focal point and passing through the region to be inspected. In this example, each hearth has an opening greater than or equal to 60 ° so that the sum of the openings of the two hearths is greater than or equal to 120 °.

[0208] In the exemplary embodiment illustrated in FIGS. 5 and 6, the installation 1 comprises two foci F1, F2, aligned on a base line B parallel to the path of the objects 2. The two foci F1, F2 are each associated with a separate generator tube 7. Installation 1 also includes five linear physical sensor components CC11, CC12, CC13, CC14 and CC15 which are arranged with their support line perpendicular to the conveying plane and which each include, for a multitude of section planes, image sensors each sensitive to X-rays from the first associated focus F1. Installation 1 also includes five linear physical sensor components CC21, CC22, CC23, CC24 and CC25 which are arranged with their support line perpendicular to the conveying plane and which each include, for a multitude of section planes, image sensors each sensitive to X-rays from the associated second focal point F2. According to this exemplary embodiment, it should be noted that a focus (F1 and F2 in the example) from which a diverging X-ray beam emerges is positioned on one side of the intersecting plane PS so that its beam crosses the intersecting plane PS and the region to be inspected, while at least one image sensor Ci associated with said focus Fj to receive the X-rays coming from said focus Fj is disposed on the opposite side with respect to the intersecting plane PS. (In the example, these are the five image sensors C11, C12, C13, C14 and C15 each sensitive to X-rays from the associated focus F1 and the five image sensors C21, C22, C23, C24 and C25 sensitive each with X-rays from the associated focus F2).

[0210] According to an advantageous variant embodiment, the hearth (s) Fj is / are arranged in the conveying plane PC. Preferably, these foci cooperate with associated image sensors located opposite them from the intersecting plane PS. Thus, in the case of transporting objects arranged on a flat conveyor, this arrangement allows that in the radiographic images, the projections of the objects are not superimposed on the projection of the conveyor. Thus, in the digital geometric model of objects, the part of the object in contact with the conveyor can be precisely determined.

[0211] According to an advantageous embodiment characteristic, the arrangement of

Ci image sensors and focal points is such that the X-rays from the focal point (s) Fj and reaching the Ci image sensors only pass through one region to be inspected at a time. In other words, X-rays only pass through one object at a time. It should be noted that the installation may include a system for controlling the spacing between successive scrolling objects.

[0212] An object of the invention is to obtain a method which is not only fast, but also inexpensive, able to calculate a three-dimensional geometry of each object transported on the line with the precision necessary for dimensional control. The invention aims to reduce the number of images required for reconstruction to the minimum number capable of achieving precision

desired dimensional. For example, the invention allows with nine projections and a limited number of images of the inspected region, to measure with an accuracy of +/- 0.05 mm the internal diameter of a cylinder between 10 and 120 mm.

Advantageously, the installation according to the invention comprises between one and four hearths Fj and preferably one or two hearths Fj and preferably one. number of image sensors making it possible to acquire, in each section plane, radiographic projections according to between four and fifteen different projection directions.

[0213] According to the invention, the image sensors and the focal point (s) should be arranged so that the combination of at least three projection directions optimizes the determination of the delineation of the inspected region of the objects transported on the line, in each of a multitude of section planes comprising a base line B parallel to the rectilinear trajectory T of the objects, considering that the volume Vt traversed must be left free for the movement of the objects. The following rules are advantageously implemented in the context of the invention, these rules being valid for linear or matrix image sensors.

[0214] In what follows, an angle is an absolute value. Figs. 7 and 8 illustrate two directions of projection Dji and D’ji which are also vectors. We consider here two directions of projection Dji and D’ji which are contained in the same plane of section Pk. These Figures show the angle a between these two directions of projection either a = (DIJ, D'IJ) and s the angle complementary to the angle a, or s = 180 ° -a. By definition, the useful angle a between two different directions of projection Dji and Dji in their section plane Pk is the smallest of the angles a and s, that is to say a = Min (a, s). Thus, the useful angle a is the smaller of the angles formed by the two lines bearing the directions of projection Dji, Dji.

[0215] According to an advantageous variant of the invention, one acquires for each object and for each section plane Pk, among the at least three images

radiographic images from radiographic projections along at least three different projection directions, at least two images from two radiographic projections along two different directions Dji and D'ji making between them a useful angle a greater than or equal to 45 ° and less than or equal to 90 °. According to an advantageous variant embodiment, for each object and for each section plane, among the at least three radiographic images obtained from radiographic projections in different directions, at least two images originating from two radiographic projections in two directions are acquired. different forming between them a useful angle a greater than or equal to 60 ° and less than or equal to 90 °.

[0216] To do this, the installation 1 according to the invention comprises at least one focal point and two image sensors arranged so that the directions of projection of the inspected region which they receive have between them a useful angle a

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 °.

[0217] For example, as illustrated in FIG. 5, the useful angle a between directions D15 and D11, and between directions D13 and D25 are greater than 45 °. Well

obviously it must be understood that at least one useful angle is greater than or equal to 45 ° and less than or equal to 90 ° and advantageously that at least one useful angle is greater than or equal to 60 ° and less than or equal to 90 ° and the other useful angles between two directions Dji, D'ji are arbitrary. Those skilled in the art using this rule will know how to look for an arrangement which offers the most complete possible distribution of the directions of projections of the inspected region.

[0218] In a particular case, provision can be made for the foci to be in the conveying plane PC, and for the physical sensor components to be positioned so that the conveying plane PC is one of the section planes Pk. Then, in the section plane Pk corresponding to the conveying plane PC, this condition applies, namely that we acquire for each object and for the horizontal section plane Pk coincident with the conveying plane PC, among the at least three radiographic images from radiographic projections in different directions, at least two images from two radiographic projections in two different directions forming between them a useful angle a greater than or equal to 60 ° and less than or equal to 90 °.

[0219] Preferably, the useful angle has between two directions of projection Dijk

consecutive in a given section plane Pk is identical for all the consecutive Dijk projection directions implemented to acquire the one-dimensional processing images in the given section plane Pk. In other words, the Dijk projection directions implemented to acquire the at least three one-dimensional processing images in a given section plane Pk are angularly spaced in a regular manner in the plane of Pk section given. Preferably, the directions of projection implemented to acquire the at least three one-dimensional processing images in a given section plane Pk are angularly distributed in the section plane Pk so as to cover the angular amplitude, measured in this plane of section Pk, covered by the opening Of of the focal point (s) Fj, preferably so as to cover at least 50% of this angular amplitude, preferably at least 75% of this amplitude.

[0220] According to another advantageous characteristic, for each object, the computer system acquires at least one radiographic image of the inspected region corresponding to a projection direction forming, in orthogonal projection on the conveying plane PC, an opening angle b determined with the direction of travel T.

[0221] As illustrated in FIG. 9, the angle p between a direction of projection (vector Dji), brought into orthogonal projection on the conveying plane PC, and the trajectory of the objects (vector T) is considered, i.e. the angle p = (Dji, T ) i.e. p = (D11, T) and p = (D12, T) in the example illustrated in Fig. 9. The angle q complementary to the angle p is such that q = 180 ° -p. By definition, the opening angle b between a projection direction Dji, brought into orthogonal projection on the conveying plane PC, and the trajectory T is the smallest of the angles p and q, namely b = Min (p, q). Thus, the opening angle b is the smallest of the angles formed by the two straight lines, one carrying the direction of projection Dji brought into orthogonal projection on the conveying plane PC, and the other the trajectory T.

[0222] According to another advantageous characteristic, for each object, the computer system acquires at least one radiographic image of the inspected region corresponding to a projection direction Dji, Djik having with the direction of movement T, an opening angle b included. between 10 ° and 60 °. In other words, the installation according to the invention comprises at least one focus and one image sensor Ci arranged so that, when an object passes through the field of the image sensors, the direction of projection Dji, Djik of the region inspected on the image sensor Ci makes an opening angle b with the direction of

displacement T between 10 ° and 60 °. In other words, the configuration of the installation 1 is optimized to reduce its bulk in the direction of movement while maintaining a traversed volume Vt suitable for the objects and good reconstruction quality.

[0224] Due to the volume Vt traversed, the installation does not produce a projection around the direction of movement T. The volume traversed Vt imposes a minimum opening angle b min. For example, the minimum opening angle b min = 10 °. There is no sensor arranged to provide an opening angle projection b of less than 10 °.

[0225] It must be deduced from the above that the distribution of the projection angles for each object is not necessarily uniform.

[0226] As illustrated in FIG. 9, the projection angle distribution can

have a gap, called a blind spot region, of twice 2 x 10 ° or 20 °, instead of having full 180 ° coverage.

[0227] For example, as illustrated in FIG. 9, an installation according to the invention

comprises at least one focus F1 and two physical sensor components CC1, CC2, here among three and for example two-dimensional, which each comprise, for a multitude of section planes Pk, image sensors C1k, C2k, C3k, for which the directions of projections D11, D12, brought into orthogonal projection on the conveying plane PC, define with the direction of displacement T, an opening angle b between 10 ° and 60 ° corresponding respectively to the angles p and q. Likewise, the installation illustrated in FIG. 5, comprises an image sensor CC11 associated with the focus F1 and whose direction of projection D11 forms an opening angle b of between 10 ° and 60 ° with respect to the direction of movement T.

[0228] The physical sensor components which form the image sensors Ci are generally of the matrix or linear type.

[0229] According to a preferred variant embodiment, the installation 1 comprises

linear physical sensor components. According to this preferred variant, each physical sensor component CCi, CCi ’comprises a linear network

elements sensitive to X-rays, distributed along a support line Li, Li 'defining with the associated focus Fj, a projection plane PPji, PPji' containing the direction of projection Dji (Fig. 2). These physical sensor components Ci are arranged so that at least m sensitive elements of each of these image sensors receive the radiographic projection of the region to be inspected by the X-ray beam coming from the associated focus Fj, with the projection planes PPji for the different sensor components physical which are distinct from each other and not parallel to the conveying plane PC. The number m of sensitive elements of each linear physical sensor component is greater than 128, preferably greater than 512. The distance between neighboring sensitive elements (called “pitch” or “pitch”) and / or the dimension of the sensitive elements is preferably less than 800 µm. The reading frequency of the image lines is preferably greater than 100 Hz, advantageously greater than 1 kHz. Of course, these parameters are adapted according to the size of the objects, the required precision and the scrolling speed. Thus, it is possible to obtain, by suitably arranging the physical sensor components, planes Pk spaced vertically by the value of the pitch, the spacing being measured at the level of the sensors. It is possible that the number of section planes Pk reaches 128, or even 512, or even more.

[0230] According to an advantageous embodiment characteristic, at least three

Ci linear physical sensor components have their supporting lines Li parallel to each other.

[0231] According to another advantageous embodiment characteristic, at least three linear physical sensor components Ci have their support lines Li orthogonal to the conveying plane PC.

[0232] According to a variant, a focus Fj is positioned so that its beam passes through the inspected region then the conveying plane PC. In addition, at least one associated linear physical sensor component Ci is positioned opposite the focal point Fj relative to the conveying plane PC and so that its straight support Li is parallel to the conveying plane PC.

[0233] According to these variant embodiments with linear physical sensor components, the acquisition system acquires with the aid of each of the at least three physical sensor components Ci, at each incremental displacement of each object on the trajectory, linear images x-rays of the region to be inspected according to a number chosen so that for each object, the entire region to be inspected is completely represented in the set of linear radiographic images. Thus, during the movement of an object, each image sensor is able to acquire linear radiographic images so that the whole of the region to be inspected of the object is completely represented in the set of linear radiographic images. obtained from said image sensor. Thus, for each object, at least three sets of radiographic linear images of the region to be inspected are obtained which are then analyzed. It is possible to constitute matrix radiographic images of the inspected region, by juxtaposition of the sets of linear radiographic images. But the reconstruction of the geometric model and the measurement do not necessarily impose it.

[0234] It should be noted that taking into account the volume Vt traversed, no projection

radiographic is only acquired in the blind spot region (b <± 10 °) located on either side of the direction of movement T. The method according to the invention allows, despite the absence of radiographic projections in this range of angles, to reconstruct, for example using the a priori geometric model, a precise and complete digital geometric model of the object. It is thus possible to carry out measurements of linear dimension over the entire digital geometric model and in particular in directions not orthogonal to the possible projection directions, including measurements of linear dimension according to directions of measurement orthogonal to the corresponding directions of missing projections. to the blind spot region located on either side of the direction of movement T. In fact, without the method according to the invention, for example with the methods intended for traditional "complete" axial tomography, in the case of where no radiographic projection is acquired according to the directions of a blind angle, then the reconstructed model also presents in an angular sector orthogonal to the blind angle, reconstruction errors making it impossible to determine a surface precisely and therefore making everything impossible linear dimension measurement of an object, even mono-material.

[0235] Thus, as illustrated in FIGS. 10 and 11, according to the invention, no projection is possible in a dead angle of for example 20 ° (b min = 10 °). According to the prior art, no precise measurement could be made in the direction A, which is not orthogonal to any of the directions of projection. Direction A is close to the orthogonal of any of the directions of projection to at least 10 °, in the sense that it presents, in orthogonal projection on the conveying plane PC, an angular deviation of at least 10 degrees of angle with n any of the projection directions. With the reconstruction from a priori models of the series of objects and the constant and uniform attenuation, the measurement of the internal diameter in direction A (distance a1) and the measurement of the distance between the two external faces perpendicular to direction A (distance a2) are correct and precise. In other words, the three-dimensional geometry determined for the region to be inspected does not present any missing or blurred boundaries in the direction A.

[0236] Of course, the number of focal points, the number of image sensors associated with each focal point, and their relative arrangements are chosen in any suitable manner depending on the degree of measurement precision desired, the shape of the objects and the size thereof. their spacing on the conveyor.

[0237] In summary, as illustrated schematically in FIG. 13, the process of

the invention provides:

- the transport of objects (TRANS OBJ),

- during this transport, the acquisition (ACQUIMAGE), using the image sensor (s) Ci, Cik, for each object during its movement, of a set of one-dimensional radiographic processing images, comprising:

one-dimensional processing radiographic images for a number NK of distinct section planes Pk containing the base line;

for each distinct section plane Pk, a NP number of images

unidimensional radiographic processing Spk of the region to be inspected, obtained according to at least three different directions of Dijk projection in the section plane;

- for each object to be measured, and for each distinct section plane Pk, the determination (CALC DLk), using the computer system, of a delineation of the object DLk in the section plane Pk considered, from at least three one-dimensional radiographic processing images Spk of the region to be inspected, obtained according to at least three different directions of Dijk projection in the section plane, - And determining (MES), for an object to be measured, from the delineations of the object in each distinct section plane, of at least one linear dimension measurement of the region to be inspected of the object to be measured.

[0238] The method allows real-time computer processing of the images to create a three-dimensional model in order to measure fast moving objects in production. The arrangement and the translational movement make it possible to work in the section planes Pk. It is possible to use reconstruction algorithms in the section planes Pk which computationally manipulate two-dimensional geometries, which is computationally much faster than manipulating three-dimensional data. It is thus possible to manipulate parametric curves by computer, which is much faster than manipulating elementary volume elements of a solid model. The method can take into account a priori information from the objects, in order to further limit the computer computing power necessary for the implementation. The proposed method allows reconstructions without the so-called "missing edge" error zone in the direction orthogonal to the displacement, despite the absence of an image of projection directions close to the direction of displacement.

[0239] Thanks to the acquisition geometry proposed by the method, the problem of the 3D reconstruction of the object from the acquired radiographic data is decoupled into NK problems of reconstructions of oblique 2D sections of the object, which limits the computing power necessary for the implementation. Indeed, the problem of 3D reconstruction from A acquisitions on C linear sensors produces AxCxNK data and would allow estimation of the order of AxCxNK voxels in an algebraic approach, by solving a linear system of the order of AxCxNK by AxCxNK. The cost of such an operation is of the order of (AxCxNK) ^A 3 by a Gauss type method and

iter3Dx (A + C + NK) xAxCxNK for an iterative method exploiting the hollow character of the matrix, with iter3D iterations.

[0240] In a decoupled approach in NK, problems of reconstructions of

oblique 2D sections of the object, we should solve NK linear systems of the order of AxC by AxC i.e. a cost of the order of NKx (AxC) ^A 3 by a Gauss-type method (we gain a factor NK ^A 2 ) and iter2D * (A + C) xAxCxNK with here in general iter2D <iter3D and of course (A + C) <(A + C + NK). The method therefore requires much less calculations, and therefore less computing power. This gain in complexity is found in the fact that it is possible to reconstruct curves in each 2D section which is much simpler than to reconstruct a surface in 3D: the volume of data in local memory is lower, and the complexity of the management of curves is much less than that of surfaces.

[0241] It is noted that, because the trajectory of the objects is a translation, and because each of the section planes contains the base line and is by

therefore parallel to the trajectory of the objects, each of the section planes Pk is invariant with respect to the object during translation. Thus, all the one-dimensional radiographic processing images Spk of the region to be inspected obtained in the same section plane Pk are one-dimensional radiographic images of the same section of the object, therefore of the same delineation of the object. It can therefore be seen that the acquisition geometry proposed by the method makes it possible to work directly on data belonging to the same plane, which considerably simplifies the calculations.

[0242] It should be noted that in industrial production in series, it is possible that several series are present at the same time on the same production or control line. In this case, the installation includes an indication system to the computer system of the series to which each of the objects belongs in order to implement the method of the invention for all the objects of the same series. Indeed, the installation according to the invention can be used to inspect a flow of manufactured objects composed of several series of different objects, for example a first series and a second series. The series may differ in the shape of the objects or in their own attenuation coefficient, or both. In this case, the installation must be provided with a means to provide the computer system with an a priori geometry model of each series of objects, an attenuation coefficient for each series of objects and it is necessary to provide a means of associating in the computer system, the images

x-rays of each object with the series to which it belongs.

Claims

Claims

[Claim 1] A method of automatically measuring linear dimensions of manufactured objects (2) of a series comprising:

- the choice of a series of manufactured objects (2) in which each of said objects consists of one or more distinct parts, the number of parts being known and each part being made of a material with an attenuation coefficient known and uniform at all points of the part of the object;

- the transport, by means of a transport device, of objects in a direction (T) of displacement along a rectilinear path in a conveying plane (PC), these objects generating a conveying volume (Vt) during their displacement;

- the arrangement, outside the conveying volume (Vt),

at least one focus (Fj) of an X-ray generator tube, each focus being arranged on the same basic straight line parallel to the direction (T) of displacement along the rectilinear path and

one or more image sensors (Ci) each exposed and sensitive to X-rays from an associated focal point (Fj), these X-rays having passed through at least the region to be inspected producing on each image sensor a radiographic projection of the region to be inspected in a direction of projection (Dji, Djik);

- the acquisition, using the image sensor (s) (Ci, Cik), for each object during its movement, of a set of one-dimensional radiographic processing images, each image

unidimensional radiographic treatment comprising a projection of a section of the object according to a section plane (Pk) containing the base line, the assembly comprising:

said processing radiographic one-dimensional images for a number (NK) of distinct section planes (Pk) containing the base line;

for each distinct section plane (Pk), a number (NP) of said one-dimensional treatment radiographic images (Spk) of the region to be inspected, obtained according to at least three different projection directions (Dijk) in the section plane;

- for each object to be measured, and for each distinct section plane (Pk), the determination, using the computer system, of a delineation of the object in the section plane (Pk) considered, from the one-dimensional processing radiographic images (Spk) of the region to be inspected, obtained according to at least three different projection directions (Dijk) in the section plane,

- And determining, for an object to be measured, from the delineations of the object in each distinct section plane, of at least one linear dimension measurement of the region to be inspected of the object to be measured.

[Claim 2] A method according to claim 1, characterized in that

that a delineation of the object comprises a curve or a set of curves which represent the intersection, with the section plane, of the boundary surfaces of the object.

[Claim 3] A method according to claim 2, characterized in that the or each curve of the delineation of the object is a plane curve modeled by a parametric system.

[Claim 4] Method according to one of the preceding claims, characterized in that the determination of a delineation of the object in the section plane comprises a curve fitting algorithm starting from an a priori delineation of the object in the section plane.

[Claim 5] Method according to one of the preceding claims, characterized in that the determination of a delineation of the object in the section plane comprises a nonlinear regression type curve fitting algorithm.

[Claim 6] Method according to one of the preceding claims, characterized in that the determination of a delineation of the object in the section plane comprises an iterative curve fitting algorithm comprising:

- taking into account an a priori delineation of the object in the section plane as a calculated delineation of the first iteration rank; - then iteratively,

the calculation, from the calculated delineation of a given iteration rank of the object in the section plane, of a number (NP) at least equal to three simulated radiographic one-dimensional images (SSpk) of the region to be inspected, calculated in the section plane according to the at least three different projection directions (Dijk) which used for the acquisition of the one-dimensional processing radiographic images (Spk) in the section plane,

comparing simulated radiographic one-dimensional images (SSpk) to processing radiographic one-dimensional images (Spk),

according to the comparison, the modification of the calculated delineation into a calculated delineation of higher iteration rank,

until the comparison of one-dimensional images

simulated radiographic (SSpk) to one-dimensional images

Radiographic processing (Spk) reaches a predefined optimization criterion.

[Claim 7] Method according to one of the preceding claims, characterized in that it comprises:

- acquisition, using the image sensors (Ci, Cik), for each object during its movement, of a number (NP) at least equal to three two-dimensional radiographic images (Ri) of the region to be inspected, each obtained according to a different direction of projection (Dji),

- Extracting, in the two-dimensional radiographic images (Ri), the processing one-dimensional radiographic images (Spk) to form the set of one-dimensional radiographic images.

[Claim 8] Method according to one of the preceding claims, characterized in that a one-dimensional processing radiographic image (Spk) of an object is formed by sampling a point image acquired using a d-sensor. 'point images (Cik), for a scanning time corresponding to the duration of the movement of the object between the focus and the point image sensor (Cik).

[Claim 9] Method according to one of the preceding claims, characterized in that it comprises the construction, for an object to be measured, using the computer system and from the delineations of the object in each of the planes of separate section (Pk), of a three-dimensional digital geometric model of the region to be inspected comprising:

- three-dimensional points of space each belonging to a border surface of the region to be inspected of the object;

- and / or at least one three-dimensional surface of the region to be inspected.

[Claim 10] A method according to claim 9, characterized in that determining, for an object to be measured, from the delineations of the object in each distinct section plane, at least one linear dimension measurement of the region. to be inspected of the object to be measured includes the

determining the distance between at least two three-dimensional points of the three-dimensional digital geometric model of the region to be inspected.

[Claim 11] Method according to one of the preceding claims, characterized in that it comprises providing the computer system, for each section plane, with an a priori delineation of the object in the section plane.

[Claim 12] A method according to claim 11, characterized in that the a priori delineations are obtained by:

- a digital computer design model of the objects in the series;

- and / or from the measurement of one or more objects of the same series by a measuring device;

- and / or from values entered and / or drawings produced and / or shapes selected by an operator on a man-machine interface of a computer system.

[Claim 13] Method according to one of the preceding claims, characterized in that it comprises supplying the computer system with an a priori three-dimensional geometric model of the region to be inspected of the series, which is obtained by:

- a digital computer design model of the objects in the series; - and / or a geometric digital model obtained from the measurement of a or several objects of the same series by a measuring device;

- and / or a geometric digital model generated by a system

data processing from values entered and / or drawings produced and / or shapes selected by an operator on a man-machine interface of the computer system.

[Claim 14] Method according to one of the preceding claims, characterized in that it comprises the arrangement of the hearth (s) in the conveying plane (PC).

[Claim 15] Method according to one of the preceding claims, characterized in that it comprises the acquisition, using the image sensor (s) (Ci, Cik), for an object of the series during its displacement, and for each considered section plane (Pk) of the object, of at least two one-dimensional radiographic processing images of the inspected region corresponding to projection directions (Djik) defining, in the considered section plane, a useful angle (a) 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 °.

[Claim 16] Method according to one of the preceding claims, characterized in that it comprises the acquisition, using the image sensor (s) (Ci, Cik), for an object of the series during its displacement, and for each section plane (Pk) considered of the object, of at least one radiographic image of the inspected region corresponding to a projection direction (Djik) having, brought into projection in the conveying plane (PC) , an opening angle (b) with the direction of movement (T) of between 10 ° and 60 °.

[Claim 17] Method according to one of the preceding claims, characterized in that it does not include the acquisition, using the image sensor (s) (Ci, Cik), for each object of the series during his

displacement, of any radiographic image of the inspected region corresponding to a direction of projection (Dji, Djik) having an opening angle (b) with the direction of displacement (T) of less than 10 °.

[Claim 18] Method according to one of the preceding claims, characterized in that it comprises making and acquiring radiographic projections of the inspected region of an object so that the X-rays coming from the focal point (s) and reaching the image sensors (Ci) do

cross no other object.

[Claim 19] Method according to one of the preceding claims,

characterized in that it comprises acquiring, for each object of the series during its movement and for each section plane, one-dimensional processing radiographic images from between three and forty radiographic projections of the region to be inspected from directions of different projections, preferably from between four and fifteen radiographic projections of the region to be inspected from different directions of projection.

[Claim 20] Method according to one of the preceding claims,

characterized in that:

- the image sensors (Ci) form part of at least three physical sensor components (CCi) which are each of linear type, each comprising a linear array of elements sensitive to X-rays distributed along a support line (Li) , which defines with a focus (Fj) a projection plane (PPji) containing the direction of projection (Dji, Djik), these image sensors being arranged so that:

at least m sensitive elements of each of these physical sensor components receive the radiographic projection of the region to be inspected by the X-ray beam coming from a focal point (Fj);

the projection planes (PPji) for the various physical sensor components are distinct from one another and not parallel to the conveying plane (PC);

- using each of the at least three linear physical sensor components (Ci) is acquired, at each incremental displacement of each object along the trajectory (T), one-dimensional radiographic images of the region to be inspected according to a chosen number so that for each object, the entire region to be inspected is represented completely in the set of one-dimensional radiographic images;

the at least three sets of one-dimensional radiographic images of the region to be inspected are analyzed for each object.

[Claim 21] Installation for automatic measurement of dimensions

lines of at least one region to be inspected of manufactured objects in a series, the installation comprising:

- a device for transporting objects in a direction materialized by a displacement vector (T), along a rectilinear path in a conveying plane (PC), the objects traversing a conveying volume (Vt) extended in the direction of displacement ( T);

- at least one focus (Fj) of an X-ray generator tube located outside the volume crossed (Vt), and creating a divergent beam of X-rays directed to pass through at least one region to be inspected of the object, each focus being arranged on the same basic straight line parallel to the direction (T) of movement along a rectilinear path;

- image sensors (Ci, Cik), located outside the conveying volume (Vt), so as to receive X-rays from an associated focus (Fj), the focus (s) (Fj) and the sensors of images (Ci) being arranged so that each image sensor receives the radiographic projection of the region to be inspected by the rays coming from the focus (Fj) when the object crosses these rays, the directions of projection 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 its movement, a set of one-dimensional radiographic processing images, the set comprising:

said processing radiographic one-dimensional images for a number (NK) of distinct section planes (Pk) containing the base line;

for each distinct section plane (Pk), a number (NP) of said one-dimensional treatment radiographic images (Spk) of the region to be inspected, obtained according to at least three different projection directions (Dijk) in the section plane;

- a computer system configured for:

for each distinct section plane (Pk), determine a delineation of the object in the section plane (Pk) considered, from said one-dimensional radiographic processing images (Spk) obtained according to the at least three directions of projection (Dijk ) different in the section plane.

[Claim 22] Installation according to claim 21, characterized in that it comprises at least two foci (Fl, F2) for producing X-rays, positioned separately in two distinct positions on the same basic line parallel to the direction (T ) of displacement along the rectilinear path, and at least three image sensors (Ci), sensitive to X-rays and

positioned so that:

- each focus emits its beam through at least the region to be inspected to reach at least one associated sensor (Ci, Cik);

- Each sensor (Ci) is associated with a focus and receives the X-rays from said focus after passing through the region to be inspected.

[Claim 23] Installation according to one of claims 21 or 22, characterized in that it comprises at least one focal point from which a divergent beam of X-rays emerges with an opening greater than or equal to 90 ° or at least two focal points from which are derived from divergent X-ray beams, the sum of the openings of which is greater than or equal to 90 °.

[Claim 24] Installation according to one of claims 21 to 23,

characterized in that it comprises at least one hearth arranged in the conveying plane (PC).

[Claim 25] Installation according to one of claims 22,

possibly taken in combination with one and / or the other of

claims 23 and 24, characterized in that at least one focus and two image sensors are arranged so that the directions of projection of the inspected region which they receive have between them a useful angle (a) 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 °.

[Claim 26] Installation according to one of claims 21 to 25,

characterized in that at least one focal point and one image sensor (Ci) are arranged so that, when an object passes through the field of the sensors, the direction of projection (Dji, Djik) of the inspected region on the sensor images (Cik, Ci) makes an opening angle (b) with the direction of

displacement (T) between 10 ° and 60 °.

[Claim 27] Installation according to one of claims 21 to 26,

characterized in that, no focal point (Fj) of an X-ray generator tube being located in the traversed volume (Vt), and no image sensors (Ci) being located in the conveying volume (Vt) ), the direction of projection (Dji, Djik) of the region inspected on the image sensor (Ci) never makes an opening angle (b) with the direction of displacement (T) less than 10 °.

[Claim 28] Installation according to one of claims 21 to 27,

characterized in that the image sensors (Ci) and the focal points (Fj) are arranged so that the X-rays from the focal point (s) and reaching the image sensors (Cik, Ci) and passing through the region of a object do not pass through another object at the same time.

[Claim 29] Installation according to one of claims 21 to 28,

characterized in that it comprises between one and four foci (Fj), originating from one or more tubes generating X-rays.

[Claim 30] Installation according to one of claims 21 to 29,

characterized in that the number and arrangement of the image sensors (Cik) and of the associated foci, are such that, for each object of the series during its movement, the radiographic projections of the region to be inspected on the sensors of images present between three and forty different directions of projection, preferably between four and fifteen different directions of projection.

[Claim 31] Installation according to one of claims 21 to 30,

characterized in that the image sensors (Ci) form part of physical sensor components (CCi) of linear type each comprising an array linear of X-ray sensitive elements distributed along a support line (Li), which defines with an associated focus (Fj) a projection plane (PPji) containing the direction of projection (Dji, Djik), these image sensors being arranged so that:

- at least m sensitive elements of each of these physical sensor components (CCi) receive the radiographic projection of the region to be inspected by the X-ray beam coming from the associated focus (Fj);

- the projection planes (PPji) for the different sensors are distinct from one another and not parallel to the conveying plane (PC).

[Claim 32] Installation according to claim 31, characterized in that at least three linear image sensors (Ci) have their support lines (Li) parallel to each other.

[Claim 33] Installation according to one of claims 31 and 32, characterized in that at least three linear physical sensor components (CCi) have their support lines (Li) orthogonal to the conveying plane

(PC).