FR3109236A1 - Element classification process - Google Patents

Abstract

The present invention relates to a method of classifying elements comprising: collecting a set of three-dimensional model of elements, the model of each element being labeled with a classification of the element, for each three-dimensional model collected, the generation of a series of images of the element, each image of the element being labeled with the classification of the element, the images generated for all the elements forming a base of '' training images, training of an element classification model according to a training technique applied to the training image base to obtain a trained classification model, exploitation of the trained classification model comprising: receiving an image of an item to be classified, and determining a classification of the item by the trained classification model. the implementation of an action relating to the element based on the classification obtained for the element. Figure for abstract: 3

Description

Element classification process

The present invention relates to a method of classifying elements. The present invention also relates to an associated computer program product.

In the industrial field, many elements, such as industrial parts or office elements, are handled by the various actors.

However, when one of these elements needs to be renewed, these players do not necessarily turn to a supplier approved by their company due to a lack of knowledge of the existence of such suppliers.

Also, a tool allowing a supplier to be associated with such elements would be of interest.

To this end, in the field of research, classification tools based on neural networks exist. Such neural networks are in particular trained from an element image base, known as a training image base.

However, such classification tools are not suitable for the industrial context. In particular, to obtain a reliable classification tool, current tools have been trained on a multitude of images of elements taken under different conditions (angles, distances, luminosity, textures).

However, in an industrial context, the realization of such a training image base with sufficient quality is very complex and time-consuming to implement, and is therefore irrelevant in terms of feasibility.

There is therefore a need for a method making it possible to classify elements reliably in an industrial context.

To this end, the present description relates to a method of classifying elements comprising:

1. a collection phase of a set of three-dimensional model of elements, the model of each element being labeled with a classification of the element, the collection phase being implemented by computer,
2. for each three-dimensional model collected, a phase of generating a series of images of the element, each image of the element being labeled with the classification of the element, the images generated for all the elements forming a training image base, the generation phase being implemented by computer,
3. a training phase of a classification model of elements according to a learning technique applied to the training image base to obtain a trained classification model, the training phase being implemented by computer,
4. an operating phase of the trained classification model, the operating phase being implemented by computer and comprising:
   1. a step of receiving an image of an element to be classified, and
   2. a step of determining a classification of the element by the trained classification model.
5. an implementation phase of an action relating to the element according to the classification obtained for the element.

According to particular embodiments, the control method comprises one or more of the following characteristics, taken in isolation or according to all the technically possible combinations:

- the method includes images of the element taken from different angles and / or at different distances from the element;

the method comprises, upstream of the generation phase, a phase of processing the three-dimensional model of each element, the processing phase advantageously comprising the filtering of the texture and / or color characteristics present on the model;

- the implementation phase includes the purchase of an item with the same classification as the classified item during the operation phase, the implementation phase being performed by computer, preferably automatically;

- the implementation phase includes the generation of an alert based on the classification obtained for the element, the implementation phase being performed by computer, preferably automatically;

- the implementation phase includes the launch of an application to control the element according to the classification obtained for the element, the implementation phase being carried out by computer, preferably automatically;

- during the exploitation phase, each classification is associated with a probability representative of a level of confidence in the classification;

- the image of each element to be classified during the exploitation phase is obtained via a sensor, for example a camera;

- the elements are chosen from the list made up of: industrial equipment, parts of industrial equipment, individual protection elements, parts of individual protection elements and office supplies.

The present description also relates to a computer program product comprising a readable information medium, on which is stored a computer program comprising program instructions, the computer program being loadable on a data processing unit. and adapted to cause the implementation of a method as described above when the computer program is implemented on the data processing unit.

The present description also relates to a readable information medium on which is stored a computer program product as described above.

Other features and advantages of the invention will become apparent on reading the following description of embodiments of the invention, given by way of example only and with reference to the drawings which are:

, Figure 1, a schematic view of an example of a computer allowing the implementation of an element classification method,

, Figure 2, a flowchart of an exemplary implementation of a method for classifying elements,

, Figure 3, a schematic representation of an example of implementation of the phase of generating a series of images of an element from its three-dimensional model, and

, Figure 4, a schematic representation of an example of series of images generated for each element during the generation phase.

A calculator 10 and a computer program product 12 are illustrated in Figure 1.

The calculator 10 is preferably a computer.

More generally, computer 10 is an electronic computer suitable for handling and / or transforming data represented as electronic or physical quantities in computer 10 registers and / or memories into other similar data corresponding to physical data in data sets. memories, registers or other types of display, transmission or storage devices.

The computer 10 interacts with the computer program product 12.

As illustrated in Figure 1, the computer 10 comprises a processor 14 comprising a data processing unit 16, memories 18 and a reader 20 for supporting information. In the example illustrated in Figure 1, the computer 10 comprises a keyboard 22 and a display unit 24.

The computer program product 12 includes an information carrier 26.

The information medium 26 is a medium readable by the computer 10, usually by the data processing unit 16. The readable information medium 26 is a medium suitable for storing electronic instructions and capable of being coupled to a device. bus of a computer system.

By way of example, the information medium 26 is a floppy disk or flexible disk (of the English name " Floppy disc "), an optical disc, a CD-ROM, a magneto-optical disc, a ROM memory, a memory. RAM, EPROM memory, EEPROM memory, magnetic card or optical card.

On the information carrier 26 is stored the computer program 12 including program instructions.

The computer program 12 can be loaded onto the data processing unit 16 and is suitable for causing the implementation of a method of classifying elements E1,…, En when the computer program 12 is put into operation. work on the processing unit 16 of the computer 10

The elements E1,…, En to be classified are, for example, chosen from the list made up of: industrial equipment, parts of industrial equipment, individual protection elements (helmets, vests, shoes, coveralls), parts of personal protective equipment and office supplies. The term "parts" is understood to mean the various components of the equipment.

The operation of the computer 10 will now be described with reference to FIG. 2, which schematically illustrates an example of implementation of a method for classifying elements and to FIGS. 3 and 4 which illustrate in more detail certain phases of this method. .

The classification process comprises a phase 100 of collecting a set of three-dimensional models M1, ..., Mn of elements E1, ..., En. Such models M1, ..., Mn are in particular computer models of the elements E1, ..., En. Such models M1, ..., Mn have, for example, been obtained by a technique or via computer-aided design software.

The three-dimensional model M1, ..., Mn of each element E1, ..., En is labeled with a classification of the element E1, ..., En. The classification of the element E1,…, En is a category to which the element E1,…, En is assigned. The category is, for example, the nature of the element E1,…, En, an approved supplier for the renewal of the element E1,…, En or the serial number of the element E1,…, En. The classification of the element E1, ..., En a, for example, was determined by an operator.

The collection phase 100 is implemented by the computer 10 in interaction with the computer program product 12, that is to say is implemented by computer.

Optionally, the classification method comprises a phase 110 for processing the three-dimensional model M1, ..., Mn of each element E1, ..., En. The processing phase makes it possible to simplify the model M1, ..., Mn by filtering out characteristics specific to an element E1, ..., In particular, but not representative of the class of the element E1, ..., En.

For example, the processing phase includes filtering the texture and / or color characteristics present on the model M1, ..., Mn of the element E1, ..., En. Thus, this makes it possible to subsequently generate a training database independent of the textures and / or colors of the elements E1, ..., Considering which are characteristics with a high variability (object under corrosion, object repainted).

The processing phase 110 is implemented by the computer 10 in interaction with the computer program product 12, that is to say is implemented by computer.

The classification method comprises, for each three-dimensional model M1, ..., Mn collected, a phase 120 of generating a series of images IM1, ..., IMn of the element E1, ..., En. Each image IM1, ..., IMn of element E1, ..., En is labeled with the classification of element E1, ..., En (the same as that of the associated three-dimensional model M1, ..., Mn).

Advantageously, the series of images IM1,…, IMn of each element E1,…, En comprises images IM1,…, IMn of element E1,…, Taken at different angles and / or at different distances from the element E1,…, En. The successive images IM1,…, IMn are, for example, acquired with a step of angle of rotation and / or a step of distance to the element E1,…, En.

The generation phase 120 is akin to a "shooting" of the element E1, ..., En. In particular, Figure 3 illustrates a model M1 of an element (clamp) which is rotated to obtain images IM1, IM2, ..., IMn of the element at different angles and at different distances.

The images IM1,…, IMn generated for all the elements E1,…, En form a base of training images. The example of Figure 4 illustrates in particular the series of images IM1, ..., IMn obtained for four elements E1, E2, E3, E4. The number of IM1, ..., IMn images taken for each element depends, for example, on the size, area and complexity of the element.

The training base obtained is, for example, stored in a memory 18 of the computer 10.

The generation phase 120 is implemented by the computer 10 in interaction with the computer program product 12, that is to say is implemented by computer.

The classification method comprises a phase 130 of training an element classification model E1, ..., En according to a training technique applied to the training image base to obtain a trained classification model.

The classification model is, for example, a neural network. The learning technique allows the neural network to be configured as it is learned on the basis of training images.

In one example, part of the training image base is used to configure the neural network and the other part to validate the configuration.

The training phase 130 is implemented by the computer 10 in interaction with the computer program product 12, that is to say is implemented by computer.

The configuration method includes an exploitation phase 140 of the trained classification model.

The operating phase 140 includes in particular the reception of an image IM1,…, IMn of an element E1,…, En to be classified, and a step of determining a classification of the element E1,…, En by the trained classification model.

Advantageously, the image IM1, ..., IMn of the element E1, ..., En to be classified is obtained via a measurement by a sensor, such as a camera.

Advantageously, each classification is associated with a probability representative of a level of confidence in the classification. This allows the operator to verify a classification associated with an unsatisfactory probability, typically less than a predetermined threshold (eg less than 85%).

The operating phase 140 is implemented by the computer 10 in interaction with the computer program product 12, that is to say is implemented by computer.

The classification method comprises a step 150 of implementing an action relating to the element E1, ..., depending on the classification obtained for the element E1, ..., En. The implementation phase is, for example, carried out by computer, preferably automatically. The action implemented depends in particular on the type of elements E1, ..., En and the intended application.

For example, the implementation phase includes the purchase of an element E1, ..., Having the same classification as element E1, ..., As classified during the operation phase. The purchase is, for example, carried out automatically according to the classification obtained. Alternatively, the purchase is validated by an operator before being made. Such an action is, for example, part of a "guided buying" process within a company, for example to renew parts of equipment or office supplies. In an example scenario, a machine breaks down and must be replaced. An operator in the field is unaware that a contract with local suppliers exists. Therefore, without the classification model, the operator will replace the machine with a non-contractual supplier. On the other hand, with the classification model, the operator provides a photo of the defective machine as input to the model, and the model provides information (number for example) as output, making it possible to establish a link with the contractual supplier.

In another example, the implementation phase includes generating an alert based on the classification obtained for the element E1, ..., En. Such an action is, for example, part of the surveillance of an industrial site or of people. The alert makes it possible, for example, to alert an operator when an E1 element,…, In unauthorized mode is identified on the site or when an individual is not equipped with an E1 element,…, In protection compulsory (helmet).

In yet another example, the implementation phase includes launching an application to control the element E1,…, depending on the classification obtained for the element E1,…, En. This allows, thus, to launch an automatic pilot of the element E1,…, En. Such an action is, for example, performed by an autonomous robot.

Thus, the present method enables elements to be reliably classified in an industrial context. Indeed, the classification model used was trained on a plurality of images of elements obtained from a three-dimensional model of labeled elements. The images obtained are synthetic images obtained during a "shooting" of the three-dimensional model. Obtaining the training image base therefore does not involve any actual image acquisition.

The method of obtaining the image base is, therefore, simple to implement and not very time-consuming compared to taking actual images. It is therefore suitable for an industrial context.

Those skilled in the art will understand that the embodiments and variants described above can be combined to form new embodiments provided they are technically compatible.

Claims (10)

Element classification process (E1, ..., En) comprising:

1. a collection phase of a set of model (M1, ..., Mn) in three dimensions of elements (E1,…, En), the model (M1, ..., Mn) of each element (E1, …, En) being labeled with a classification of the element (E1,…, En), the collection phase being implemented by computer,
2. for each three-dimensional model (M1, ..., Mn) collected, a phase of generation of a series of images (IM1,…, IMn) of the element (E1,…, En), each image ( IM1,…, IMn) of the element (E1,…, En) being labeled with the classification of the element (E1,…, En), the images (IM1,…, IMn) generated for all the elements (E1,…, En) forming a base of training images, the generation phase being implemented by computer,
3. a training phase of a classification model of elements (E1, ..., En) according to a learning technique applied to the base of training images to obtain a trained classification model, the training phase being implemented by computer,
4. an operating phase of the trained classification model, the operating phase being implemented by computer and comprising:
   1. a step of receiving an image (IM1,…, IMn) of an element (E1,…, En) to be classified, and
   2. a step of determining a classification of the element (E1,…, En) by the trained classification model.
5. a phase of implementation of an action relating to the element (E1,…, En) according to the classification obtained for the element (E1,…, En).

Method according to claim 1, wherein the series of images (IM1,…, IMn) of each element (E1,…, En) comprises images of the element (E1,…, En) taken from different angles and / or at different distances from the element (E1,…, En). Method according to Claim 1 or 2, in which the method comprises, upstream of the generation phase, a phase of processing the model (M1, ..., Mn) in three dimensions of each element (E1, ..., En) , the processing phase advantageously comprising the filtering of the texture and / or color characteristics present on the model (M1, ..., Mn) of the element (E1, ..., En). Method according to any one of claims 1 to 3, in which the implementation phase comprises the purchase of an element (E1,…, En) having the same classification as the element (E1,…, En) classified during the operational phase, the implementation phase being carried out by computer, preferably automatically. Method according to any one of claims 1 to 3, in which the implementation phase comprises the generation of an alert as a function of the classification obtained for the element (E1,…, En), the implementation phase work being produced by computer, preferably automatically. Method according to any one of claims 1 to 3, in which the implementation phase comprises launching an application for checking the element (E1,…, En) as a function of the classification obtained for the element (E1,…, En), the implementation phase being carried out by computer, preferably automatically. A method according to any one of claims 1 to 6, wherein during the exploitation phase, each classification is associated with a probability representative of a level of confidence in the classification. Method according to any one of Claims 1 to 7, in which the image (IM1,…, IMn) of each element (E1,…, En) to be classified during the operating phase is obtained via a sensor, by example a camera. Process according to any one of Claims 1 to 8, in which the elements (E1,…, En) are chosen from the list consisting of: industrial equipment, parts of industrial equipment, individual protection elements, parts personal protective equipment and office supplies. A computer program product comprising program instructions recorded on a computer readable medium, for performing a method according to any one of claims 1 to 9 when the computer program is executed on a computer.