Equipment layout on a georeferenced plan

Abstract

Method for detecting equipment of a structure built during the generation of a georeferenced plan of the structure built and comprising: a) an acquisition of georeferenced images of a portion of the equipment; b) a processing of the acquired images comprising: b1) a three-dimensional modeling of the structure built by photogrammetry, b2) a georeferencing of the structure built, b3) a segmentation of the acquired images so as to obtain at least one contour of the portion of the equipment, implemented on each acquired image rotated by an angle of rotation determined from an angle formed between an axis associated with the portion of the equipment and a reference axis, b4) a projection of the contours onto the three-dimensional modeling of the structure built, and b5) a grouping of the projected contours forming a trace of the equipment;the process further comprising: c) generation of the georeferenced plan from the results of the automated processing and including the layout of the equipment. Figure from the abstract: Figure 1;

Description

Title of the invention: Tracing of equipment in a geo-referenced plan

[0001] The present disclosure relates to the field of generating plans of constructed works and more particularly of the layout of equipment in such georeferenced plans. Prior art

[0002] The cartographic update of works on structures - particularly underground networks - is done by drawing plans of the topographic survey of the sites / structures concerned. Such plans correspond to Georeferenced Plans of Constructed Works (or PGOC). These plans make it possible in particular to obtain a map of the construction sites carried out and to identify their main characteristics, for example once the construction site has been filled in and the positioning of the structures is no longer visible. The reliability of such plans makes it possible, for example, to limit and better plan subsequent construction sites and interventions on underground networks.

[0003] The generation of PGOC generally requires three-dimensional modeling and spatial identification of the structures by a land survey carried out by a surveyor on site, in order to obtain a georeferenced plan. This typically involves the intervention and coordination of several actors and trades on the site both for the execution of the works (including clearing, installation of the structures, backfilling, etc. for example) and the taking of measurements (for example, of the positions of reference points) on the site (by the intervention of the surveyor for example). Existing methods may then involve leaving a site in suspense for several days (for example, by leaving an excavation site cleared) while a survey of the position of the constructed structure is carried out, which can disrupt the progress of work on a site.Furthermore, the evolution of regulations relating to work near networks - concerning for example the Declaration of Project of Works (or DT) or the Declaration of Intention to Commence Work (DICT) in France - requires, in most cases, the establishment of PGOC in precision class A.

[0004] Regulatory and technical obligations in the field of construction thus lead to an increased need to map construction sites precisely, faithfully to the works constructed, by reducing the uncertainty of measurements, by providing specific characteristics of the works, and this without compromising the progress of work on the sites concerned.

[0005] Furthermore, there is a need to provide traceability of the material and equipment implemented on such construction sites, such as electrical cables or electrical boxes. Such traceability requires in particular the recognition and identification of this equipment on georeferenced plans. Due to the generally discontinuous arrangement of this equipment (which may, for example, be partially buried or covered by ducts) on construction sites, such traceability is generally not systematic or automated and often lacks precision. Summary

[0006] The present disclosure provides a solution to such a problem.

[0007] According to one aspect, a method is proposed for detecting equipment of a constructed structure during generation of a georeferenced plan of the constructed structure implemented by a computer acquisition and processing system, the method comprising: (a) an acquisition of images, using a mobile camera, said acquired images being georeferenced according to a predefined geographic coordinate system and presenting at least two distinct viewing angles of at least a portion of the equipment of the constructed structure; (b) automated processing of the acquired images by at least one processing software and comprising: bl) a three-dimensional model of the structure constructed by photogrammetry, b2) georeferencing of the constructed structure in the predefined geographic coordinate system, b3) a segmentation of each of the acquired images so as to obtain at least one outline of the portion of the equipment, said segmentation being implemented on each acquired image rotated by a rotation angle determined from an angle formed between an axis associated with the portion of the equipment on the acquired image and a reference axis of the acquired image, b4) a projection of the contours of the portion of the equipment respectively obtained on the images acquired on the three-dimensional modeling of the constructed structure, (b5) a grouping of projected contours, said grouping of projected contours forming a layout of the equipment; the method further comprising: (c) a generation of the georeferenced plan of the structure constructed from the results of the automated processing, said georeferenced plan including at least the layout of the equipment.

[0008] Consequently, the proposed method advantageously makes it possible to generate plans georeferenced constructed structures (or PGOC), or more generally georeferenced plans, including detection, recognition and precise tracing of equipment and elements included in the structures. The proposed automated processing of acquired images makes it possible to obtain georeferenced plans including equipment tracings while respecting a level of precision known as class A. Such a class A can in particular designate a cartographic precision class in which the maximum uncertainty of location of a structure is less than or equal to 40, 50 or 80 centimeters depending on the characteristics of the structure (for example, its rigidity, its flexibility or its operating context).

[0009] The proposed method addresses in particular the difficulties classically encountered during the generation of PGOC, when one or more pieces of equipment (or other equipment present on construction sites) are captured sparsely in the plurality of acquired images. The reconstruction of equipment in its entirety and in its precise arrangement in a structure is then difficult. The proposed method makes it possible to include the layout of equipment in the georeferenced plans in a systematic, automated and precise manner.

[0010] According to another aspect, there is provided a computer program comprising instructions for implementing the method when this program is executed by a processor.

[0011] According to another aspect, there is provided a non-transitory recording medium readable by a computer on which is recorded a program for implementing the method when this program is executed by a processor.

[0012] The features set out in the following paragraphs may, optionally, be implemented, independently of one another or in combination with one another:

[0013] In one embodiment, the segmentation of each of the acquired images is a segmentation of instances.

[0014] Consequently, the proposed image segmentation not only makes it possible to detect several segmented elements (or polygons or contours) in the images potentially corresponding to the equipment, but also to link the segmented elements as belonging to the same equipment. Indeed, most constructed structures include cleared, backfilled elements, obstacles (at least visual), walkways, metal plates, tarpaulins, ducts, earth or other elements obstructing visibility and at least partially masking equipment on the images. For example, an electrical cable laid in a trench may be partially covered with earth or may pass under a walkway. Furthermore, a perspective effect linked to the shooting or a partial shooting of the structure may lead to a portion of the equipment being visible in a different way. continuous on an image. As a result, instance segmentation advantageously allows the presence of such a discontinuity to be detected by linking the segmented elements as belonging to a single instance (for example, to a single portion of the same cable) on each image.

[0015] In one embodiment, the segmentation of each of the acquired images comprises: b31) a first segmentation of the portion of the equipment on the acquired image so as to obtain at least one polygon associated with the portion of the equipment, b32) a determination of the axis associated with the portion of the equipment on the acquired image, including: - determine a center of gravity of the polygon, - obtain two sub-polygons resulting from a section of the polygon along an axis of the acquired image crossing the center of gravity, - determine a barycenter of each sub-polygon, - determine the axis associated with the portion of the equipment from an axis associated with the polygon obtained by connecting the barycenters of the sub-polygons, b33) a determination of the angle formed between the axis associated with the portion of the equipment and the reference axis, b34) a second segmentation on the acquired image rotated by a rotation angle calculated from said angle, so as to obtain a rotated polygon associated with the portion of the equipment.

[0016] Consequently, the proposed method advantageously makes it possible to improve the detection of the equipment in the structure as captured on the acquired images, by implementing several segmentations. Indeed, the method optimizes the image recognition allowing the detection of the portion of the equipment in a systematic manner by automating a mutual realignment of the images. The segmentation is advantageously improved by an arrangement of the image facilitating the recognition of contours associated with the portion of the equipment. Such an intermediate manipulation of the images is neutral on the result of the method but facilitates automatic processing.

[0017] In one embodiment, at least two polygons associated with the portion of the equipment are obtained at the end of the first segmentation and each acquired image is rotated by a rotation angle determined from the average of the angles formed between each axis associated with each polygon and a common reference axis of the acquired image.

[0018] Consequently, the method advantageously makes it possible to improve the segmentation of the equipment portions on the acquired images even when several disjoint contours are detected on the equipment portion (for example, when the first segmentation implemented on the initial image before rotation is imprecise).

[0019] In one embodiment, the equipment comprises a marking composed of a set of characters and the method further comprises, before the generation of the georeferenced plan of the constructed structure: (b) identification of the equipment by optical character recognition implemented on at least one of the acquired images including the marking, and the generated georeferenced plan further comprises the identification of the equipment.

[0020] Consequently, the proposed method advantageously makes it possible to identify the equipment detected on the acquired images. The method therefore makes it possible to provide information that is as exhaustive as possible on the situation of the construction site and the structure. In situations where the marking is visible and recognizable, at least in part, on a plurality of acquired images, the identifier can also be used to make the link between two images more reliable, i.e. the automatic detection of the fact that the two images correspond to the same physical element captured from two points of view.

[0021] In one embodiment, the identification of the equipment comprises: b'1) an initial detection of an identification zone of the acquired image, said identification zone comprising the marking, b'2) optical recognition of several groups of patterns making up the set of characters of the marking via detection by a sliding window on said identification zone, a size associated with the sliding window being less than a size of the identification zone, two groups of patterns successively detected by the sliding window overlapping in at least one common pattern, b'3) a concatenation with each other of the groups of patterns successively detected by means of the at least one common pattern, b'4) identification of the equipment by association of the equipment marking with the concatenated pattern groups.

[0022] Consequently, the proposed method advantageously makes it possible to identify the equipment present on the constructed works and having long markings (of the reference number or equipment identification number type). The method is particularly advantageous in that it makes it possible in particular to identify the equipment even when the marking associated with the equipment is captured on several separate shots, and for example partially (part of the characters is invisible or not recognized).

[0023] Thus, the method makes it possible to extract exhaustive information without imposing constraints in the capture of the acquired images. The images of the structure can therefore be captured by any individual on the site, without requiring particular expertise or even knowing the precise elements to be isolated or identified on the images. The method makes it possible to obtain convincing results by filling in any possible defects in taking view. Once the images have been acquired, the proposed method advantageously allows the equipment to be identified automatically.

[0024] In one embodiment, the projection of the contours of the portion of the equipment onto the three-dimensional model of the constructed structure comprises a projection of coordinates associated with said contours in the predefined geographic coordinate system, said coordinates associated with said contours being defined according to a two-dimensional image reference associated with the acquired images.

[0025] Consequently, the proposed method advantageously makes it possible to reconstruct the entirety of the equipment in the constructed structure, in three dimensions. The proposed projection then makes it possible to reconstruct the depth of the structure and the arrangement of the equipment in such a context.

[0026] In one embodiment, the grouping of the projected contours comprises: b51) filtering the projected contours so as to select intersection zones corresponding to an intersection of at least two projected contours, b52) a grouping of non-disjoint intersection zones, so as to obtain a set of georeferenced polygons associated with the equipment.

[0027] Consequently, the proposed method advantageously makes it possible to transform a set of contours (or polygons) potentially isolated from each other in the acquired images into grouped polygons approximating a layout of the equipment in its entirety in the structure. In pictorial terms, the method makes it possible to “guess” the position of certain portions of the physical elements which are invisible or masked in the images.

[0028] In one embodiment, the layout of the equipment is obtained by: cil) an extraction of at least one direction curve associated with the equipment from the grouping of the projected contours, cl2) a modeling of said direction curve in a graph, c 13) tracing of the equipment via tracing by graph traversal of the vertices of said graph, said graph traversal being constrained by at least a predefined minimum distance and a predefined maximum angle between two successive vertices of the graph.

[0029] Consequently, the proposed method advantageously makes it possible to generate, in an automated manner, a continuous and precise layout of the equipment despite the practical constraints of the arrangement of the equipment in the structure. The method then makes it possible to obtain a georeferenced plan including a precise, realistic layout of equipment generated in an automated and systematic manner from acquired images without the level of expertise required to capture such images being particularly high. Brief description of the drawings

[0030] Other characteristics, details and advantages will appear on reading the detailed description below, and on analyzing the attached drawings, in which: Fig. 1

[0031] [Fig-1] shows a diagram of an acquisition and processing system according to a mode of realization. Fig. 2

[0032] [Fig.2] shows steps of a method for generating a georeferenced plan according to one embodiment. Fig. 3

[0033] [Fig.3] shows images acquired according to one embodiment. Fig. 4

[0034] [Fig.4] shows a segmentation of equipment according to one embodiment. Fig. 5

[0035] [Fig.5] shows a segmentation of equipment according to one embodiment. Fig. 6

[0036] [Fig.6] shows a segmentation of equipment according to one embodiment. Fig. 7

[0037] [Fig.7] shows a projection of equipment according to one embodiment. Fig. 8

[0038] [Fig.8] shows filtering of equipment according to one embodiment. Fig. 9

[0039] [Fig.9] shows a layout of equipment according to one embodiment. Fig. 10

[0040] [Fig. 10] shows a georeferenced plan generated according to one embodiment. Description of the embodiments

[0041] Reference is now made to [Fig.l]. [Fig.l] illustrates a computer system 1,2 allowing the acquisition and processing of images of a construction site, so as to generate a georeferenced plan of constructed work (or PGOC, or more simply georeferenced plan) of the site.

[0042] The construction site, or construction site, can refer to any type of land comprising one or more works or installations, of an electrical nature for example, in the context of works such as the installation or connection of underground networks. A work can for example correspond to a trench dug by clearing the ground on a road in which network connections are installed. The construction site generally extends over three dimensions and the works on such a construction site are characterized by a dimensioning according to variable depths, curvatures and directions relative to a re such as the ground surface. For example, in the case of a work corresponding to a trench, the trench can be characterized by an earthwork depth, an earthwork volume, a total length, etc. The work then consists of an excavation in a ground with a specific leveling and characteristics. Such information, relating to the transformation of the land and the work carried out on the construction site, requires a cartographic update of the land concerned and the production of plans faithful to the reality of the site.

[0043] It is specified here that the physical elements that one wishes to georeference can be of any nature and of any shape. The tests carried out by the applicant are particularly convincing for elements of elongated shape, which is why many examples refer to cables, pipes or elements of similar shape, without being limited to them. The site may in particular include one or more EQ equipment such as for example pipes, electrical cables, ducts, sheaths, electrical fittings, electrical boxes, underground accessories (junctions, tangents, etc.), etc. Each EQ equipment has specific dimensions and is integrated into the site according to a given positioning and arrangement adapted to the environmental constraints. The determination of such a positioning will be detailed later.In addition, each EQ equipment may be associated with a marking comprising a reference or an identifier enabling the equipment to be identified, uniquely or in batches. The marking associated with an EQ equipment may, for example, take the form of a succession of typed characters (composed of letters, numbers and / or symbols) placed on one side of the EQ equipment (for example via a sticker, an inscription, an engraving on the EQ equipment) and visible when the EQ equipment is positioned on the construction site. Such a marking may also correspond to a machine-readable marking of the barcode or QR code type. Alternatively, the marking may take the form of an electronic chip of the NFC (or “Near Field Communication”) type. Such a marking may also be slightly offset from the equipment itself, for example carried by a label itself attached to the equipment.Such marking may in particular make it possible to associate each piece of equipment with information stored in a database or any other type of digital register on a server 21, a website or a remote digital platform for example. Such information may include intrinsic characteristics of the equipment such as the type of equipment, its dimensional and / or functional characteristics, its serial number, etc.

[0044] The acquisition and processing system 1, 2 may comprise a mobile device 1 and a processing device 2. The mobile device 1 is a device for taking pictures of the construction site using a lens and may refer to a smartphone for example. Such a mobile device 1 can be handled by an operator on the construction site. The use of a device known and common to the general public, such as a smartphone, for taking pictures allows use by an operator without any particular skills, unlike devices whose handling is more complex. The mobile device 1 may include a memory unit (not shown in [Fig.l]) and allowing the captured images to be stored at a minimum. In addition, a single mobile device 1, single-piece and with a single lens, may be sufficient for taking pictures. In other words, the parallax necessary for photogrammetry can be obtained by manually moving the single mobile device 1 and taking two successive shots. The images may come from photos or extracted from one or more videos.The mobile device 1 may be associated with a sensor 11 enabling the reading of the marking (an optical sensor 11 for visual markings and / or an NFC sensor 11 for reading NFC chips, for example). Such a sensor 11 may correspond to an optical sensor. Such an optical sensor may be directly integrated into the mobile terminal 1 as shown in [Fig. 1]. The mobile device 1 may also be associated with a geographic coordinate acquisition device 12. The geographic coordinate acquisition device 12 makes it possible to georeference the images acquired by the mobile device 1 and to assign geographic coordinates to the pixels representing elements of the structure captured in the images. Such a geographic coordinate acquisition device 12 may, for example, include a satellite antenna.The geographic coordinate acquisition device 12 may also be confused with the mobile device 1 in the case of mobile devices 1 having an integrated geolocation or georeferencing functionality for example. The geographic coordinate acquisition device 12 may also take the form of a peripheral paired with the mobile device 1.

[0045] The mobile device 1 may further be connectable to the device 2 and / or to a server 21 integrated or not to the device 2 via a communication network, the device 2 being able to receive data comprising the images captured by the mobile device 1 via an input interface 23. In addition to the images acquired by the mobile terminal 1, the device 2 is further capable of receiving and storing data acquired by the optical sensor 11 and / or the geographic coordinate acquisition apparatus 12. Such transmission of data from the mobile terminal 1 (respectively from the optical sensor 11 and / or the geographic coordinate acquisition apparatus 12) to the device 2 may be carried out in real time (or in near real time) during the capture of the data, for example when the mobile terminal 1 is connected to the device 2.Transmission of captured data in real time or almost in real time allows progress on the work to be continued. on the captured area (for example to backfill the captured area and more generally to make it less accessible and / or visible) or to move to another area of ​​the construction site, without having to keep such an area accessible for a new subsequent measurement. Indeed, a real-time data transmission can allow the device 2 to carry out subsequent processing immediately and therefore to signal in return a possible need to repeat or complete the data capture of the captured area at a time when the user has easy access to the captured area. Alternatively, the data transmission can be subsequent to the data capture, in particular when the mobile terminal 1 is located in an area not covered by the communication network. In the remainder of the description, it will be considered that the optical sensor 11 and the geographic coordinate acquisition device 12 are integrated into the mobile device 1 as illustrated in [Fig.l] and that the mobile device 1 is connected to a communication network, so that all of the acquired data is transmitted by the mobile device 1 to the device 2 in real time or near real time. In another embodiment, the device 2 is capable of receiving different data streams from different mobile devices 1, 11, 12. In other words, the device 2 can be centralized and cooperate with a plurality of mobile devices 1 so that the hardware and software resources of the device 2 are shared.

[0046] The processing device 2 comprises at least one processing unit 22 capable of processing the data received, directly from the mobile device 1 and / or indirectly via a server 21 for example. In one embodiment, the processing unit 22 may comprise several functional sub-units forming a processing circuit, each processing sub-unit being able to include a processor and a memory unit. The processing unit 22 may for example include functional sub-units making it possible, by processing the images acquired by the mobile device 1, to: - implement a three-dimensional reconstruction of the work, for example by projection or by photogrammetry; - to geo-reference the elements of the work in a predefined coordinate system, - to segment and filter the acquired images so as to enable tracing of the building's equipment, - to identify the equipment of the structure from markings present on the equipment, or in the immediate vicinity.

[0047] The device 2 makes it possible to process images acquired by taking pictures of a structure, so as to generate a georeferenced plan. In particular, the device 2 is configured to generate a georeferenced plan comprising at least one trace of equipment belonging to the structure. In one embodiment, the device 2 is configured to generate a georeferenced plan further comprising an identification of the equipment belonging to the work. Such identification of the equipment may for example correspond to an association of the equipment, on the georeferenced plane with a reference or a marking making it possible to identify the equipment in a database. Such a database may for example be stored on the server 21.

[0048] The acquisition and processing system 1, 2 may be connected to the server 21. In one embodiment, the server may be integrated into the processing device 2, as shown in [Fig.l]. Alternatively, the server 21 may be a remote server (for example on a “cloud”). The server 21 may be configured to store the data transmitted by the mobile device 1 and / or to store data associated with the equipment belonging to the work.

[0049] The acquisition and processing system 1,2 is configured to implement the method described subsequently in [Fig.2].

[0050] Reference is now made to [Fig. 2]. [Fig. 2] represents a succession of steps for generating a Georeferenced Plan of Constructed Work (or PGOC), also referred to as a georeferenced plan in the context of this description. The generated georeferenced plan may be associated with a constructed work or an installation (e.g., a trench dug in the ground) belonging to a construction site or a construction site (e.g., an underground network construction site). The work considered notably comprises at least one EQ equipment (e.g., an electrical cable) arranged in the work. Such a method may be implemented by the system represented in [Fig. 1].

[0051] In a first step 200, images of the structure are captured using the mobile camera 1. Such image acquisition consists of capturing several portions of the structure via shots (in two dimensions) from at least two distinct viewing angles. The acquisition of the images of the structure in step 200 thus makes it possible to have two-dimensional shots of all the elements constituting the structure, from different viewing angles. In particular, the images have a priori identical or similar characteristics (dimensions, number of pixels, focal length, etc.).

[0052] The images can be acquired at different times or different stages of the progress of the work on the construction site. The images can optionally be acquired by different mobile devices 1. Thus, the step 200 of acquiring the images of the structure can be implemented at selected times so as not to disrupt the progress of the work: a portion of the structure can be photographed by an individual working on this portion of the site (for example when the cables arranged in the structure are visible) and then backfilled without delay. In the example described here, the images are photos. Alternatively, several images 1 can be extracts from the same video.

[0053] With reference to [Fig.3], six images acquired in step 200 are shown as an example.

[0054] In one embodiment, the images acquired in step 200 are taken so as to capture one or more portions (or sections) of PEQ equipment included in the portion(s) of the structure. For example, the structure considered may be a trench excavated in the ground so as to have several curvatures and at least one EQ equipment corresponding to an electrical cable may be arranged in the trench. The images acquired in [Fig. 3] then capture several portions of the structure, each acquired image comprising a capture of a portion of PEQ equipment.

[0055] In one embodiment, the images acquired in step 200 are georeferenced according to a predefined coordinate system. In other words, each pixel (or set of pixels) of each of the acquired images can be associated with so-called GPS (Global Positioning System) coordinates, or coordinates of a positioning system other than GPS. For example, each pixel of each acquired image can be associated with a longitude and a latitude, expressed in degrees, minutes, seconds (or DMS), as well as optionally a height or depth (vertical coordinate). The georeferencing of each image can in particular be implemented by the geographic coordinate acquisition apparatus 12.Each image acquired in step 200 is georeferenced and can then be associated with a set of geographic coordinates, for example in the form of metadata, which can be stored on the server 21 and / or by the device 2 in correspondence with each acquired image.

[0056] In one embodiment, at least one of the images acquired in step 200 may comprise a marking associated with a reference or an identifier of the EQ equipment captured on the acquired image, such a marking being able to be captured by the sensor 11. The capture, during the shots, of such a marking may allow the identification of the captured EQ equipment, which will be detailed later in a step 240.

[0057] At the end of step 200, or concomitantly, the acquired images can be transmitted from the mobile device 1 (or the different mobile devices 1) to the processing device 2 and / or the server 21. Such a transmission can be done automatically, the images being captured and transmitted to be stored remotely on a memory unit of the device 2 and / or the server 21 for example.

[0058] In a step 210, a three-dimensional (or 3D) modeling of the structure can be implemented from the images acquired in step 200. For this, a photogrammetry process can be implemented by the processing unit 22 of the device 2 from a plurality of images acquired in step 200. The three-dimensional modeling of the structure makes it possible in particular to restore a depth of the structure, for example the depth of the dug trench. Thus, at the end of step 210, a three-dimensional model of the structure can be obtained.

[0059] At a step 220, georeferencing of the three-dimensional model of the structure can be implemented. Such georeferencing can in particular be based on the georeferencing of each of the acquired images, obtained by the geographic coordinate acquisition device 12 during step 200. The georeferencing of the model then makes it possible to position the structure in space and to estimate the occupation and arrangement of the various elements making up the structure. Thus, at the end of step 220, a georeferenced and three-dimensional model of the structure can be obtained.

[0060] Steps 230, 240, 250 and 260 can then be implemented. In the context of steps 230, 240, 250 and 260, the images considered, acquired in step 200, have at least a portion of PEQ equipment, as shown in [Fig. 3],

[0061] In a step 230, a segmentation is implemented on each of the acquired images. The image segmentation step 230 makes it possible in particular to obtain at least one contour associated with the portion of the PEQ equipment visible in the image. In one embodiment, a segmentation of instances is implemented. For this, a convolutional neural network of the “Mask R-CNN” type can for example be implemented. Each acquired image can be segmented so as to detect one or more contours (forming polygons) associated with elements of the image, in particular with portions of PEQ equipment. The segmentation implemented in step 230 makes it possible, for example, to detect instances in the acquired image, for example via labeled rectangular detection zones, or “bounding boxes”, as shown in Figures 5 and 6. In [Fig. 5], the rectangular detection zone makes it possible to segment an element corresponding to a portion of a cable SEQ while in [Fig.6], the rectangular detection area makes it possible to segment an element corresponding to a cable sheath. The segmentation step 230 makes it possible to obtain segmented images, comprising segmented elements detected on the image. In particular, the equipment portions PEQ form, at the end of step 230, segmented instances SEQ. An equipment section SEQ will subsequently refer to the equipment portion after segmentation of the image in step 230.

[0062] With reference to [Fig.4], the images of [Fig.3] after the segmentation step are illustrated by way of example. The equipment section SEQ is detected on each of the acquired images via at least one contour. In particular, on an image IMG2 of the segmented images visible in [Fig.4], two distinct contours (or equivalently, two polygons), designated by SEQ1, SEQ2, are associated with the equipment section detected in the image IMG2. These two contours SEQ1, SEQ2 are disjointed by the presence of a central reservation present on the acquired image IMG2. The segmentation of instances implemented in step 230 advantageously makes it possible to link the polygons SEQ1, SEQ2 as forming a single instance (in other words, as app- connected to the same EQ equipment, for example the same cable).

[0063] In one embodiment, the segmentation step 230 can be implemented on each acquired image rotated by a rotation angle determined from an angle 0 formed between an axis Xo associated with the section of the equipment SEQ on the acquired image and a reference axis of the acquired image. The reference axis can for example be one of the axes (horizontal or vertical) forming the reference frame of the acquired image on [Fig. 5] (the units of the reference axes being expressed in pixels). Such an embodiment is illustrated in Figures 5 and 6. In such an embodiment, the segmentation step 230 of each of the acquired images comprises: - a first segmentation of the portion of the PEQ equipment on the acquired image so as to obtain at least one SEQ polygon associated with the portion of the PEQ equipment, as shown in [Fig.5], - a determination of an Xo axis associated with the SEQ polygon obtained on the acquired image - a determination (see details in the following paragraph) of the angle 0 formed between the Xo axis associated with the portion of the SEQ equipment and the reference axis, the reference axis being for example chosen as being the horizontal axis of the image reference frame on [Fig.5], - a second segmentation on the acquired image rotated by a rotation angle calculated from said angle 0, so as to obtain a rotated polygon SEQ* associated with the portion of the equipment, as shown in [Fig.6].

[0064] The determination of an axis Xo associated with the polygon SEQ obtained on the acquired image, includes for example: — determine a center of gravity G of the polygon SEQ, — obtain two sub-polygons resulting from a section of the polygon SEQ along an XG axis of the acquired image crossing the center of gravity G, — determine a barycenter Bl, B2 of each sub-polygon, — determining the axis Xo associated with the portion of the PEQ equipment from an axis associated with the polygon SEQ obtained by connecting the barycenters Bl, B2 of the sub-polygons,

[0065] Such an implementation then makes it possible to obtain a better segmentation of the image, the polygon(s) on the image having a more precise envelope (delimiting the polygon) when the portion of the PEQ equipment to be detected is arranged vertically or horizontally relative to the reference axes (or reference axes) of the image.

[0066] In one embodiment, the first segmentation implemented on the image (original, i.e. before rotation) may comprise several polygons, for example when the first segmentation lacks precision so that the cable section is detected discontinuously (i.e., by disjoint envelopes), or when the section of the SEQ equipment has two disjoint contours. associated with the same instance on the image, as on the image IMG2 of [Fig.4]. In such an embodiment, each acquired image is rotated by a rotation angle determined from the average of the angles formed between each axis associated with each polygon and a common reference axis of the acquired image, for example the horizontal axis of the acquired image reference frame, as shown in [Fig.5].

[0067] At the end of step 230, the method then makes it possible to obtain a plurality of segmented images, comprising several sections of equipment SEQ detected and delimited on each of the images, as shown in [Fig.4].

[0068] In an optional step 240, an identification of the EQ equipment is implemented. In the context of step 240, the image (or images) considered have(s) at least one visible marking associated with the EQ equipment and arranged on one or more portions of the PEQ equipment captured on the acquired images (or in the immediate vicinity). The step of identifying the EQ equipment can then be implemented by detecting the marking. According to embodiments, the identification step 240 can be implemented before or after the segmentation step 230. Preferably, the identification step 240 can be implemented after the segmentation step 230, so as to limit the identification zone of the marking to the zones occupied by the polygon(s) SEQ already detected on the acquired image, which reduces the area of ​​the image to be processed in order to detect the marking.

[0069] In one embodiment of step 240, the marking corresponds to a sequence (or set) of typed characters. The identification of the EQ equipment in step 240 can then be implemented by optical character recognition (OCR) on at least one acquired image comprising the marking. For this, an area associated with the marking (for example, a bounding box) can first be detected (for example, by detecting the presence of characters in the image, without detecting the nature or type of these characters at this stage). A size associated with the marking can also be estimated (for example in pixels), so as to delimit the area associated with the marking. Step 240 can then include: - optical recognition of several groups of patterns (eg, characters) composing the set of characters of the marking via detection by a sliding window on the area associated with the marking, a size associated with the sliding window being smaller than the size associated with the marking, two groups of patterns successively detected by the sliding window overlapping in at least one pattern, - a concatenation of the groups of patterns successively detected from the at least one pattern covering two groups of successive patterns, so as to obtain a set of character strings resulting from the concatenation (for example, if the pattern(s) covering two groups of successive patterns comprise two sequences of identical characters, several character strings resulting from the concatenation. can be obtained), and finally - an identification of the EQ equipment by associating the marking of the EQ equipment with the concatenated pattern group by selecting the character string having the minimum average Levenshtein distance with the other character strings (i.e., the character string for which the cost to move from one string to another is minimal).

[0070] In one embodiment, the concatenated group of patterns associated with the marking of the EQ equipment can be interpreted as a reference or an identifier of the EQ equipment, and can then be matched in a database, for example stored on the server 21 and / or the device 2, so as to associate the identified marking (and therefore the EQ equipment detected on the image) with a set of information stored in the database.

[0071] Alternatively, the identification of the EQ equipment can also be implemented by digital recognition, in the case where the marking corresponds for example to a QR code or a digital tag. The digital capture of the marking can then make it possible to associate the EQ equipment with a set of information stored in the database corresponding to the captured marking.

[0072] Thus, at the end of step 240, the EQ equipment as captured on one or more of the acquired images can be identified by detection of the marking.

[0073] In other embodiments, step 240 may be implemented earlier or later in the method, as long as step 200 of acquiring the images is implemented upstream of step 240.

[0074] At a step 250, a projection of the equipment sections SEQ detected at the end of step 230 (and possibly identified at the end of step 240) can be implemented, so as to model the equipment EQ in the three-dimensional georeferenced model of the structure as obtained at the end of step 220.

[0075] For this, step 250 may comprise a projection of the contours associated with the SEQ equipment sections, detected on each of the acquired images, onto the three-dimensional model of the structure by projection of the coordinates associated with the SEQ contours into the predefined geographic coordinate system. Such a projection at step 250 then makes it possible to obtain a three-dimensional and georeferenced model of the structure comprising a precise arrangement of the EQ equipment. Such a projection implemented at step 250 is shown in [Fig.7] as an example. As illustrated in [Fig.7], at the end of step 250, the EQ equipment can then be reconstructed by projection of the different SEQ equipment sections and by three-dimensional modeling integrated into the model of the structure.

[0076] At step 250, the different projected contours associated with the sections of the equipment SEQ may have disjoint sections D, as shown in [Fig.7].

[0077] At a step 260, a grouping of the projected contours may be implemented. Such a grouping of the projected contours may comprise: - filtering of the projected contours so as to select intersection zones corresponding to an intersection of at least two projected contours, - a grouping of non-disjoint intersection zones, so as to obtain a set of georeferenced polygons associated with the equipment.

[0078] Thus, in step 260, a set of polygons formed by an aggregation (or “clustering”) of the projected contours can be obtained. Such a grouping is for example illustrated in [Fig.8]. With reference to [Fig.8], three projected contours C1, C2, C3 are considered as an example. The georeferencing associated with each contour C1, C2, C3 allows an arrangement of the contours as shown in [Fig.8]. The contours C1 and C2 have a non-zero intersection zone, as do the contours C1 and C3, as well as the contours C2 and C3. Such intersection zones are then grouped together, so as to obtain a polygon formed by a single contour C. The zones covered by a single contour (for example C1 only or C3 only) are not grouped together, as represented by the element D in [Fig.8].

[0079] At the end of step 260, a first representation (potentially discontinuous) of the EQ equipment can be obtained and modeled in the georeferenced model of the structure.

[0080] At a step 270, a georeferenced plan can then be generated, such a georeferenced plan notably comprising a layout of the EQ equipment. Such a layout can be obtained at step 270 by: - an extraction of at least one guide curve associated with the EQ equipment from the grouping of the projected contours (i.e. here from the polygons obtained in step 260), - a modeling of the direction curve in a graph, then - a tracing of the equipment via a graph traversal tracing of the vertices of the graph, such a graph traversal being constrained by at least a predefined minimum distance and a predefined maximum angle between two successive vertices of the graph.

[0081] With reference to [Fig.9], such a tracing of the EQ equipment is represented. The extraction of the guide curve associated with the EQ equipment can be implemented by connecting the curvatures associated with the grouped contours under an assumption of continuity of the tracing of the EQ equipment. The curvatures associated with the grouped contours can also have branches (and not a single guideline), depending on the shapes of the contour, a graph modeling of the guide curve makes it possible to connect the different sections of the SEQ equipment under the assumption of continuity of the tracing of the EQ equipment. Indeed, each grouped contour (i.e., each polygon) can be represented as a connected component. The graph traversal can then be interpreted as a connection of the different connected components corresponding to the different disjoint polygons, so as to obtain a single trace of the EQ equipment. The tracing of the EQ equipment can also be subject to parameter constraints defining the EQ equipment in order to improve the accuracy of the tracing: a minimum distance separating two polygons can be predefined or a maximum angle can be defined. Such a maximum angle can for example be defined by the maximum curvature presented by an electrical cable laid in the structure.

[0082] Thus, during step 270, a continuous trace of the EQ equipment can then be obtained on a georeferenced plan of constructed structure (or PGOC) as generated. In addition, the EQ equipment thus traced can also be identified. Consequently, the proposed method makes it possible to generate a georeferenced plan of the structures including a precise, systematic trace and an identification of the EQ equipment present in the structures. [Fig. 10] represents such a georeferenced plan including a trace of the EQ equipment. Such a georeferenced plan is furthermore generated according to a level of precision known as class A, that is to say with a modeling error of less than 40 centimeters. The tracing of the equipment according to the proposed method then makes it possible to enrich the information offered by the georeferenced plan in an automated manner while respecting the level of precision required by construction regulations.

[0083] The georeferenced plan obtained can then be used as a reliable source of information for any action relating to the area of ​​the construction site thus georeferenced. In other words, the method can be completed by a final step using the georeferenced plan obtained or the information it carries, implemented or not by a computer system, having any technical effect within the meaning of the European Patent Office.

Claims

Claims

1. Method for detecting equipment (EQ) of a constructed structure during generation of a georeferenced plan of the constructed structure implemented by a computer acquisition and processing system, the method comprising: a) an acquisition of images (200), using a mobile camera, said acquired images being georeferenced according to a predefined geographic coordinate system and presenting at least two distinct viewing angles of at least one portion (PEQ) of the equipment (EQ) of the constructed structure; b) automated processing of the acquired images by at least one processing software and comprising: bl) three-dimensional modeling (210) of the structure constructed by photogrammetry, b2) georeferencing (220) of the structure constructed in the predefined geographic coordinate system, b3) segmentation (230) of each of the acquired images so as to obtain at least one outline of the portion (PEQ) of the equipment (EQ),said segmentation (230) being implemented on each acquired image rotated by a rotation angle determined from an angle (0) formed between an axis associated with the portion (PEQ) of the equipment (Xo) on the acquired image and a reference axis of the acquired image, b4) a projection (250) of the contours of the portion of the equipment (EQ) respectively obtained from the acquired images onto the three-dimensional modeling of the constructed structure, and b5) a grouping (260) of the projected contours (Cl, C2, C3), said grouping (260) of projected contours (Cl, C2, C3) forming a layout of the equipment (EQ); the method further comprising: c) a generation (270) of the georeferenced plan of the constructed structure from the results of the automated processing, said georeferenced plan comprising at least the layout of the equipment (EQ).,

2. The method of claim 1, wherein the segmentation (230) of each of the acquired images is an instance segmentation.

3. Method according to one of the preceding claims, in which the segmentation (230) of each of the acquired images comprises: b31) a first segmentation of the portion (PEQ) of the equipment (EQ) on the acquired image so as to obtain at least one polygon associated with the portion (PEQ) of the equipment (EQ), b32) a determination of the axis associated with the portion (PEQ) of the equipment (Xo) on the acquired image, including: - determining a center of gravity (G) of the polygon, - obtaining two sub-polygons resulting from a section of the polygon along an axis of the acquired image crossing the center of gravity (G), - determining a barycenter (Bl, B2) of each sub-polygon, - determining the axis associated with the portion (PEQ) of the equipment (Xo) from an axis associated with the polygon obtained by connecting the barycenters (Bl, B2) of the sub-polygons, b33) a determination of the angle (0) formed between the axis associated with the portion (PEQ) of the equipment (Xo) and the reference axis, b34) a second segmentation on the acquired image rotated by a rotation angle calculated from said angle (0), so as to obtain a rotated polygon associated with the portion (PEQ) of the equipment (EQ).

4. Method according to claim 3, in which at least two polygons associated with the portion (PEQ) of the equipment (EQ) are obtained at the end of the first segmentation, and in which each acquired image is rotated by a rotation angle determined from the average of the angles formed between each axis associated with each polygon and a common reference axis of the acquired image.

5. Method according to one of the preceding claims, the equipment (EQ) comprising a marking composed of a set of characters, the method further comprising, before the generation of the georeferenced plan of the constructed structure: b') an identification (240) of the equipment (EQ) by optical character recognition implemented on at least one of the acquired images comprising the marking, and in which the generated georeferenced plan further comprises the identification of the equipment (EQ).

6. Method according to claim 5, in which the identification of the equipment comprises: b' 1) an initial detection of an identification zone of the acquired image, said identification zone comprising the marking, b'2) an optical recognition of several groups of patterns composing the set of characters of the marking via a detection by a sliding window on said identification zone, an associated size to the sliding window being less than a size of the identification zone, two groups of patterns successively detected by the sliding window overlapping in at least one common pattern, b'3) a concatenation with each other of the groups of patterns successively detected by means of the at least one common pattern, b'4) the identification of the equipment by association of the marking of the equipment (EQ) with the concatenated groups of patterns.

7. Method according to one of the preceding claims, in which the projection (250) of the contours of the portion (PEQ) of the equipment (EQ) onto the three-dimensional model of the constructed structure comprises a projection of coordinates associated with said contours in the predefined geographic coordinate system, said coordinates associated with said contours being defined according to a two-dimensional image reference associated with the acquired images.

8. Method according to one of the preceding claims, in which the grouping (260) of the projected contours (Cl, C2, C3) comprises: b51) filtering the projected contours (Cl, C2, C3) so as to select intersection zones corresponding to an intersection of at least two projected contours (Cl, C2, C3), b52) grouping the non-disjoint intersection zones, so as to obtain a set of georeferenced polygons associated with the equipment.

9. Method according to one of the preceding claims, in which the layout of the equipment is obtained by: c11) an extraction of at least one direction curve associated with the equipment from the grouping of the projected contours (Cl, C2, C3), cl2) a modeling of said direction curve in a graph, c 13) a tracing of the equipment via a tracing by graph traversal of the vertices of said graph, said graph traversal being constrained by at least a predefined minimum distance and a predefined maximum angle between two successive vertices of the graph.

10. Computer program comprising instructions for implementing the method according to one of claims 1 to 9 when this program is executed by a processor.

11. Non-transitory recording medium readable by a computer on which is recorded a program for implementing the method according to one of claims 1 to 9 when this program is executed by a processor.

Images