System and method for detecting and measuring cracks in a structure using stereoscopic vision techniques

Abstract

The invention relates to a system coupled to a drone (2), for quantifying the width and depth of a crack (1.1) in a structure (building, civil engineering work, etc.), said system comprising, at least, anchors for installing, inter alia, elements such as: a frame (3), used as a support system; a single-board microcomputer (5); a camera in addition to the one incorporated in the drone (2); and two cameras (4) installed on the fuselage of the drone (2) to the right and left of the smaller side of the frame (3.1). The system according to the invention may include an additional power supply, which can also be external. The invention also relates to an inspection method comprising the following steps: a) obtaining a representative set of photographs to obtain the parameters of the camera (4) ; and b) dimensional adjustment, in order to be able to scale the images obtained of the crack (1.1) and thus monitor it over time.

Description

[0001] DESCRIPTION

[0002] SYSTEM AND METHOD FOR THE DETECTION AND MEASUREMENT OF CRACKS IN A STRUCTURE USING STEREOSCOPIC VISION TECHNIQUES

[0003] Technology sector

[0004] The present invention relates to the detection and control of damage in the civil and building sectors, and more specifically, to the monitoring of pathologies in buildings or structural elements of infrastructure and structures in general (bridges, dams, towers for installations, etc.). In particular, this invention relates to a system, a principal form of use thereof, and a method for monitoring the temporal development of certain anomalies or failures in structures, such as cracks, that may compromise the strength of the structural element in which they appear.

[0005] State of the art.

[0006] Upon completion of any structure, whether civil or building-related, its service life begins, during which fissures or cracks often appear and intensify over time. These can be due to various causes, including poor-quality materials, calculation or design errors, or even faulty workmanship in the structural element in question (columns, beams, vertical or horizontal surfaces, etc.).

[0007] The importance of monitoring the evolution of structural cracks is undeniable, given that they can even cause the collapse of the structural element where they appear. However, monitoring and tracking the evolution of these cracks is a complex and critical task for preserving the integrity of structures over time and preventing potential structural collapses, which, in extreme cases, can be catastrophic.

[0008] In general, assessment methods are costly and complex, requiring the participation of specialized technicians in hard-to-reach places, which would entail personal risk.

[0009] Traditionally, crack monitoring on the surfaces of structural elements has relied primarily on visual inspection methods using conventional instruments such as calipers, steel rules, and test specimens. However, measurements taken with these tools can be subjective, depending on the personnel involved, and their use requires skilled labor. Therefore, despite their simplicity, these methods cannot be considered reliable for comprehensive monitoring and control of crack development.

[0010] On the other hand, linear variable displacement transducers (or LVDTs), which could be called "contact gauges," provide more objective and reliable measurements. However, they have the disadvantage of being expensive and difficult to install when the location is not easily accessible, for example, due to height.

[0011] Currently, various image processing techniques and algorithms are known for crack identification and the efficient extraction of crack information from images. Publications exist that reveal detection techniques such as SURF (Speeded Up Robust Features), SIFT (Scale-Invariant Feature Transform), and the application of different edge detectors (Sobel, Canny, Prewitt, etc.) to define a boundary between a crack and the background surface. Among others, the following can be cited:

[0012] “Concrete Crack Identification Using a UAV Incorporating Hybrid Image Processing” (https: / / www.mdpi.eom / 1424-8220 / 17 / 9 / 2052);

[0013] “Inspection, Identification and Repair Monitoring of Cracked Concrete structure -An application of Image processing”

[0014] (https: / / www.researchgate.net / publication / 333497217_lnspection_ldentification_and_Repair_Monitorin g_of_Cracked_Concrete_structure_-An_application_of_lmage_processing), or

[0015] “Analysis of edge-detection techniques for crack identification in bridges”, (https: / / www.researchgate.net / publication / 245282260_Analysis_of_Edge- Detection_Techniques_for_Crack_ldentification_in_Bridges)

[0016] In addition to the above, various possibilities have been disclosed regarding the analysis and dimensional quantification of a crack using convolutional neural networks (e.g., U-Net), as well as a combination of neural networks and Support Vector Machines.

[0017] Se pueden citar los siguientes artículos científicos: “Vision and Deep Learning-Based Algorithms to Detect and Quantify Cracks on Concrete Surfaces from UA V Videos"

[0018] (https: / / www.researchgate.net / publication / 346730791_Vision_and_Deep_Learning-

[0019] Based_Algorithms_to_Detect_and_Quantify_Cracks_on_Concrete_Surfaces_from_UAV_Videos), o

[0020] “The Feasibility Assessment Study of Bridge Crack Width Recognition in Images Based on Special Inspection UAV” (https: / / www.researchgate.net / publication / 346749871_The_Feasibility_Assessment_Study_of_Bridge _Crack_Width_Recognition_in_lmages_Based_on_Special_lnspection_UAV)

[0021] All these solutions have the drawback that, despite providing an approximate value for the width of the crack under study, they do not have the necessary precision to optimally monitor the evolution of these structural deficiencies and thus avoid irreversible damage to the structural element being analyzed.

[0022] The use of machine vision is also known, as it provides a more accurate dimensional estimate of the crack by working with higher image resolution. However, some disadvantages are known, such as the need for highly qualified personnel or the complexity of programming in certain scenarios. The publication “A UAV-based machine vision method for bridge crack recognition and width quantification through hybrid feature learning” is a relevant example.

[0023] (https: / / www.researchgate.net / publication / 352462413_A_UAV-based_machine_vision_method_for_ bridge_crack_recognition_and_width_quantification_through_hybrid_feature_learning)

[0024] Also, in recent years, in addition to the analysis options mentioned, research has been carried out using methods based on the deformation of triangular meshes obtained from photographs of the crack and 2D and 3D Digital Image Correlation (DIC) to measure deformation in the crack's vicinity. Examples include the publication "Application of Digital Image Correlation in Structural Health Monitoring of Bridge Infrastructures: A Review." (https: / / www.researchgate.net / publication / 356988200_Application_of_Digital_lmage_Correlation_in_Structural_Health_Monitoring_of_Bridge_lnfrastructures_A_Review.)

[0025] It should be noted that the Digital Image Correlation technique is largely limited by the generation of the point pattern and may not provide reliable results for various reasons, such as the area being analyzed being too small, or the absence of sampling points in the vicinity of the crack. Therefore, it is not the most suitable technique for small cracks.

[0026] On the other hand, research is being conducted on the use of photogrammetric techniques, both airborne and non-airborne. These studies would require painting a regular grid of circular targets on the surface of the structure, or placing reflective targets around both sides of a crack, acting as measuring devices. These solutions present drawbacks related to imperfections or discontinuities on the surface of the damaged structure, which could lead to false crack detections. Furthermore, the placement of the targets on the structure being evaluated can sometimes be complex. Additionally, when the surface where the crack is located is irregular or lacks flatness (e.g., eroded concrete), the dimensional analysis becomes more complicated. See the following scientific articles:

[0027] “Automatic crack monitoring using photogrammetry and image processing." (https: / / www.sciencedirect.com / science / article / abs / pii / S0263224112002941)

[0028] “Improved digital photogrammetry technique for crack monitoring." (https: / / iranarze.ir / wp-content / uploads / 2016 / 09 / E70.pdf)

[0029] “Automated rotational and scaling invariant image-based crack width monitoring with sub-millimeter accuracy and self-numbering label."

[0030] (https: / / www.researchgate.net / publication / 339347583_Automated_rotational-_and_scaling- ¡nvariant_image-based_crack_width_monitoring_with_sub-millimeter_accuracy_and_self- numberingjabel)

[0031] Additionally, the use of an unmanned aerial system (UAS) has been considered to overcome certain difficulties in accessing the cracked area. Regarding the use of a remotely piloted mechanism, such as a conventional drone, for crack detection, US patent US2017122909A1, entitled “non-destructive system and method for detecting structural defects,” mentions the possibility of including non-contact, non-destructive detection sensors on a drone to assess defects in a structural element without risk to human life. However, while the invention for which protection is sought also uses a drone, the drone alone would not constitute its core.

[0032] In this regard, several studies have also been published, such as the following: “Development of Crack Detection System with Unmanned Aerial Vehicles and Digital Image

[0033] Processing. ”

[0034] (http: / / www.i-asem.Org / publication_conf / asem15 / 5.ICSSS15 / 1w / W3F.2.MS582_2545F1.pdf)

[0035] “Concrete Crack Identification Using a U AV Incorporating Hybrid Image Processing.’’ (https: / / www.mdpi.eom / 1424-8220 / 17 / 9 / 2052)

[0036] Long-Term Monitoring of Crack Patterns in Historic Structures Using UAVs and Planar Markers: A Preliminary Study.

[0037] (https: / / www.mdpi.eom / 2313-433X / 4 / 8 / 99)

[0038] “Assessment of the feasibility of detecting concrete cracks in images acquired by unmanned aerial vehicles.’’

[0039] (https: / / www.researchgate.net / publication / 324867497_Assessment of_the_feasibility_of_detecting_concrete_cracks_in_images_acquired_by_unmanned_aerial_vehicles)

[0040] “Robotic System for Inspection by Contact of Bridge Beams Using UAVs.’’ https: / / www.mdpi.eom / 1424-8220 / 19 / 2 / 305

[0041] “UAS-Based Crack Detection Using Stereo Cameras: A Comparative Study.” https: / / www.researchgate.net / publication / 335200709_UAS-Based_Crack_Detection_Using_Stereo_Cameras_a_Comparative_Study

[0042] “Image-based crack assessment of bridge piers using unmanned aerial vehicles and three-dimensional scene reconstruction.” https: / / onlinelibrary.wiley.eom / doi / full / 10.1111 / mice.12501

[0043] All these studies reveal shortcomings of the implemented system, such as the inability to quantify the width and evolution over time of the crack in question, the complexity of the hardware, in certain cases, the limitation to the study of only certain structures caused by the requirement of other distance measuring instruments (total station) or the fact of not focusing on the specific quantification of the dimension of the crack.

[0044] In light of all the disadvantages of existing solutions, as described above, a technical solution is needed that meets the requirements of being applicable to all types of infrastructure, eliminating risks to people, and eliminating the need to install monitoring devices or sensors in the crack. Furthermore, it must be low-cost and easy to use. Object of the invention

[0045] To achieve these objectives and solve the technical problems discussed so far, the present invention proposes a low-cost, user-friendly system for evaluating and monitoring structural defects, such as cracks. This system eliminates potential hazards to the operator by enabling 3D crack monitoring in infrastructure through remote measurement of crack width and depth using a series of cameras and other components installed on a UAS, thus eliminating the need for physical access. Furthermore, a calibration process for the images captured by the cameras has been developed using a series of targets designed for this purpose.

[0046] One of the objects of the present invention is to provide a system for inspecting a surface of a structure and dimensionally quantifying any structural defects present (such as cracks). To this end, the system comprises at least a frame mounted on a drone (conventional or of custom design), a series of imaging devices that focus on the crack when the frame is attached, as far as possible, to the cracked surface, and a conventional single-board microcomputer with a secondary board as an interface, where the drone can be moved adjacent to a structure to detect or dimensionally quantify a defect along it (crack).

[0047] Preferably, these image acquisition means are at least 2 stereoscopic vision cameras that will be calibrated in due course.

[0048] Since the frame can be positioned in different ways, it allows for the inspection of vertical or horizontal structural surfaces, including inclined surfaces. It is effective for all materials commonly used in construction, civil engineering, and infrastructure in general, such as concrete structures, ceramic elements, metal structures, etc. Therefore, one of the essential features of the present invention is its versatility, in addition to its ease of use, as it does not require any special knowledge or skills to pilot the drone.

[0049] Another object of the present invention is to provide a method for inspecting a structure, the method comprising the steps of:

[0050] Obtaining a representative set of photographs to obtain the camera parameters, and

[0051] - Dimensional adjustment, in order to scale the images obtained from the crack and thus monitor them over time.

[0052] In addition to the above, a main method of use is established, consisting of the following steps:

[0053] 1. Fixing the nearest drone take-off point (2) on the ground to the location of the crack.

[0054] 2. Drone (2) approach flight by elevation to the level of the crack (1.1) under analysis.

[0055] 3. Translation flight of the drone (2) to center the position of the frame and the crack (1.1) in order to have the best focus of the cameras (4).

[0056] 4. Support of the frame (3.1) on the structural element containing the crack (1.1), in order to prevent drone movements that impair the quality of the photographs, if the conditions under which the inspection is carried out and the characteristics of the cameras (4) used allow it.

[0057] 5. Taking photographs of the crack (1.1).

[0058] 6. Separation of the frame (3.1) from the structural element containing the crack (1.1).

[0059] 7. Descent back to the initial starting point, where the drone (2) will land.

[0060] Description of the figures.

[0061] For a better understanding of the above, some drawings are included which, schematically and as a non-limiting example, represent preferred cases of realization.

[0062] Figure 1 shows a schematic view of a possible system installation on a drone (2) for inspecting vertical, near-vertical, or even inclined surfaces. Figure 2 reveals another possible system installation on a drone (2) for inspecting horizontal or near-horizontal surfaces.

[0063] Figure 3 contains a perspective view of the frame (3).

[0064] Figure 4 is a schematic view of the exterior of the frame (3.1) where a linear lighting element (3.3) and a padded covering (3.4) are located, and where point A is also located, the intersection of the two diagonals of the window delimited by the frame (3.1).

[0065] Figure 5 shows the targets (3.5) required for calibration, placed on the inside of the frame (3.1).

[0066] Figure 6 shows the position of the cameras (4) according to the angle <t>vision supplied by the manufacturer.

[0067] Figure 7 shows the set of targets used for the initial calibration of the two cameras (4).

[0068] Detailed description of the invention

[0069] The present invention relates to a system for inspecting a surface of a structure and reliably, easily and economically monitoring existing defects therein, e.g. a crack, to a calibration method to be able to implement the claimed solution and a preferred form of use of the claimed system.

[0070] Thus, an object of the present invention is to provide a system for inspecting and quantifying defects in an element of a structure, comprising at least one frame (3) anchored on a drone, a series of means for obtaining images that focus on the crack when such frame (3) is attached to the cracked surface, and a conventional single-board type microcomputer next to a secondary board that functions as an interface, wherein the drone can be moved adjacent to a structure to detect or dimensionally quantify a defect along it (crack).

[0071] The use of the claimed system has the following advantages:

[0072] 1. The system can be applied to the inspection of structures that would otherwise be difficult or dangerous to inspect;

[0073] 2. With the proposed system it is possible to measure the width / depth of the cracks and, in addition, to monitor their evolution over time.

[0074] 3. The elements that can be inspected include structures in the field of building, civil engineering or infrastructure, in general;

[0075] 4. These elements can be made of different materials, the proposed method being applicable to concrete elements, metal structures and other materials that can be used in the execution of any structure.

[0076] 5. The different arrangements of the system elements on the drone allow the inspection of cracks on surfaces with different inclinations, ranging from vertical structures to horizontal structures.

[0077] 6. The use of the proposed system does not require prior preparation of the surface of the structure to be examined;

[0078] 7. The use of the proposed system does not require any special knowledge or skills for piloting the UAS used;

[0079] 8. The system's own lighting allows for inspection in adverse environmental conditions, including low light or even shade.

[0080] With regard to the system that is the subject of the present invention, it achieves a compromise between flight time and the "payload" associated with the drone (2) used, understood as the weight that the drone (2) can carry, including its own weight with its components such as, for example, the battery. Additionally, it may include cameras, sensors, microprocessors to control the cameras, etc.

[0081] In addition to the drone (2) itself, the proposed system comprises at least: a frame (3) anchored to a drone (2) with the corresponding fixings; a series of means for obtaining images that focus on the crack when such frame

[0082] (3) is supported, where possible, on the cracked surface (1), and a conventional single-board computer (5) (SBC); where the drone (2) can be moved adjacent to a surface (1) of a structure to detect or dimensionally quantify a crack (1.1) along it.

[0083] In addition to the drone's main battery (2), an additional power supply system, even an external one, will be included if necessary for the operation of the aforementioned devices plus the lighting system (3.3).

[0084] In the present invention, the frame (3) comprises a rectangular frame (3.1) with a minimum width of 20-25 mm, supported by a rod structure (3.2) with a maximum width that allows for proper handling of the drone (2). This frame (3.1) may contain, around its perimeter, a linear lighting element (3.3) made up of commercially available components, such as LEDs or similar lighting elements. This allows the system to be used in any environmental condition, including low-light or shaded conditions, in which case longer exposure times are required on the cameras (4) to avoid focusing problems.

[0085] The linear lighting element (3.3) must have sufficient physical characteristics (luminous intensity, luminance and illuminance) so that, in a daytime inspection, carried out under the most adverse environmental conditions, even with shade, and with a camera configuration of ISO 100 and a minimum shutter speed of 1 / 1000, the crack (1.1) can be photographed in a way that allows its analysis.

[0086] Furthermore, the outer part of the frame (3.1) will contain a padded element (3.4), of the foam type, to cushion the contact and improve the support between the frame (3.1) and the surface (1) of the structural element being examined. On the inner part of the frame (3.1) is attached a series of targets (3.5), formed by sectors of a circle, which will allow the dimensions of the crack (1.1) to be measured after appropriate calibration.

[0087] The frame (3) can be anchored in two positions, allowing inspection of surfaces of vertical or inclined (figure 1) and horizontal (figure 2) structures, from a lower position, such as the bottom of a bridge.

[0088] Regarding the means for obtaining images of the crack (1.1), regardless of the camera that the drone (2) may incorporate, for filming the crack (1.1), at least two cameras (4) of any commercial type will be installed, to the right and left of the shorter side of the rectangle formed by the frame (3.1). These cameras will have an appropriate field of view (FOV), depending on the distance from the crack (1.1) to the camera (4.4), and a suitable resolution, in order to obtain a wider view of the area surrounding the crack (1.1). In this way, the two cameras provide a stereoscopic vision system suitable for any aircraft, even small ones, with a low price and low maximum takeoff weight.

[0089] Both cameras (4) will have their own power supply or can be powered additionally (even externally) and will be high resolution (minimum of 4000x3000 pixels), allowing high shutter speeds, minimum of 1 / 1000 s. Thus, the higher the resolution of the cameras (4) and their shutter speed, the simpler the process of photographing the crack (1.1) and the more accurate the results will be.

[0090] Furthermore, the orientation of the cameras (4) must be as shown in Figure 6; that is, the lines originating from the center of the CMOS sensor plane of the two cameras (4), located behind the lens and perpendicular to it, must intersect at the central point (A), the intersection of the diagonals of the frame (3.1)

[0091] With the aim of contributing to the development of an invention that combines functionality, precision, versatility, and cost-effectiveness, using conventional, small-sized drones, the video transmission system is implemented by connecting the cameras to a single-board computer (5) ("SBC"), such as one of the widely available Raspberry Pi™ models. A secondary, also conventional, board, such as those sold by ArduCAM™, is incorporated into this SBC. This secondary board acts as an interface (adapter) and manages the two cameras (4) so ​​that they function as a single camera. The real-time DMA (Direct Memory Access) transfer between the camera and the microcomputer (5) is performed using this secondary board.

[0092] With the careful selection of all these elements, an optimized payload can be achieved, which does not reach 500 g, which means having a flight autonomy of the drone (2) of approximately fifteen minutes with this type of small drone.

[0093] A second object of the present invention is to provide a method applying known stereoscopic vision and "Structure from Motion, Multiview Stereo (SfM-MVS)" processing techniques that allows: obtaining three-dimensional models of a crack (1.1) on structures built of any type of material, using stereoscopic vision cameras, and

[0094] - to carry out the temporal monitoring and tracking of the evolution of said crack (1.1) by comparing the distances between pairs of points, on both sides of the crack (1.1), on said three-dimensional models.

[0095] The complementary SfM and MVS techniques will be applied to images taken on the crack (1.1) using two stereoscopic cameras (4) and have the advantage of being known automated photogrammetric methods that are low cost, high resolution, flexible and easy to apply.

[0096] The first action to be carried out, in order to implement the claimed solution, is the calibration of the cameras (4), whose focal length or field of view (in English, Field of View, "FOV", or angle It will depend on the dimensions of the frame (3.1) and the distance between the cameras (4) and the crack (1.1), as well as the size of the drone used (2), and in any case, it must be possible to photograph the frame (3.1) with its targets (3.5) completely, with the crack (1.1) located in the area delimited by the frame (3.1).

[0097] The calibration process is carried out in 2 phases:

[0098] 1. Obtaining a representative set of photographs.

[0099] The stereoscopic system consisting of two cameras (4) will be used on a set of targets similar to that shown in Fig. 7. The number of images must be sufficient for the SfM-MVS software to perform a correct alignment. For reference, it can be verified that 30 shots, i.e., 60 photos, guarantee high accuracy.

[0100] After obtaining the aforementioned images of the set of targets (fig. 7), the working guidelines of the chosen SfM-MVS software are followed and, once the photos are aligned, the following intrinsic parameters of each camera (4) and the extrinsic variables that define the spatial relationship between them, including their relative rotation and translation, will be obtained: Focal distance on the X and Y axes

[0101] Cx, Cy Coordinates of the point of intersection of the focal axis with the CMOS sensor.

[0102] Skew Tilt transformation coefficient k1, k2, k3, k4 Radial distortion coefficients p1, p2, p3, p4 Tangential distortion coefficients.

[0103] These coefficients will be stored permanently in the software, as an XML file, making it unnecessary to repeat the process.

[0104] With a fixed camera configuration (4) and the obtained parameters, it is possible to accurately measure distances and create accurate 3D models in any other image data set.

[0105] 2. Dimensional adjustment.

[0106] Since the objective of the procedure is to determine the actual dimensions of the crack (1.1), a dimensional adjustment is necessary, for which the distances between the reference points of the frame (3.1) will be used. Therefore, the distance between two of the targets (3.5) of the frame (3.1) will be measured, either manually or by any other method, and entered into the SfM-MVS software.

[0107] Subsequently, the selected targets and the measured distance between them will be marked on the 3D model, and from this moment on, the three-dimensional models obtained for the crack (1.1) will be properly scaled, so that their monitoring can now be carried out.

[0108] To analyze the evolution of the crack (1.1) between successive inspections and to determine both the crack width (1.1) and its periodic monitoring over time, an initial measurement must be taken as a reference. This will allow for the precise calculation of the dimensional variations that the crack (1.1) has undergone between the different inspections. To do this, at least two singular points on opposite sides of the crack (1.1) are selected, and a sufficient number of point pairs (15 or more) are marked on the photographs taken of the crack around these singular points on each side. The singular points of the crack should be surface features that can be easily identified in subsequent three-dimensional models used to monitor its evolution.

[0109] Finally, a preferred use of the claimed system and method is described, with at least the following steps:

[0110] 1. Fixing the drone's take-off point (2), on the ground, closest to the crack (1.1).

[0111] 2. Drone approach flight (2) to achieve elevation to the level of the crack (1.1) under analysis.

[0112] 3. Translation flight of the drone (2) to center the position of the frame and the crack (1.1) in order to achieve the best focus of the cameras (4).

[0113] 4. Support of the frame (3.1) on the structural element (1) that contains the crack (1.1), in order to prevent movements of the drone (2) that impair the quality of the photographs, if the conditions in which the inspection is carried out and the characteristics of the camera used allow it.

[0114] 5. With the help of devices included in a drone (2), such as GPS and, where appropriate, visual positioning systems, optical flow cameras or infrared beams, and ultrasonic sensors, the fixed position will be maintained for the time necessary to take photographs of the crack (1.1).

[0115] 6. Taking photographs of the crack (1.1).

[0116] 7. Separation flight of the drone (2) from the frame (3.1) of the structural element (1) containing the crack (1.1).

[0117] 8. Return descent flight of the drone (2) to the initial starting point, where it will land.

[0118] The system and method of use can be implemented on any type of drone, including commercial drones. Preferably, drones already on the market weighing less than 2 kg are selected to minimize costs. However, the system and method can be implemented on larger and heavier drones that are therefore capable of carrying more payload and / or have greater autonomy, with longer flight times. Whether or not the drone already has a built-in camera, usually of low resolution, a preferred embodiment of the present invention incorporates an additional (non-stereoscopic) camera to assist with navigation and positioning of the drone within the crevice.Furthermore, if necessary, the system includes an additional power source (such as a battery), separate from the one built into the drone itself, or even an external one, to provide extra power to the drone's functions (such as powering stereoscopic cameras and the microcomputer). This power supply system can be external.< / t>

Claims

CLAIMS 1 Drone-mounted system (2) for detecting and measuring the width and depth of a crack (1.1) in a structural element comprising: - an anchoring system that allows elements to be installed on the drone (2), - a structure (3), as a support system; - at least two cameras (4) installed on the fuselage or supporting structure of the drone (2); to the right and left of the shorter side of the rectangle formed by the frame (3.1); - a single-board computer (SBC) (5) with a secondary board acting as an interface for the two cameras (4); wherein the structure (3), as a support system, can be placed in different positions and consists of a frame (3.1) including a linear lighting element (3.3) that allows viewing of the crack (1.1) in adverse environmental conditions and a padded element (3.4) to improve contact with the surface of the element to be inspected (1) and which is supported by four clamping rods (3.2); characterized in that the frame (3.1) additionally contains a series of targets (3.5) attached to its inner part, for the calibration of the cameras (4). 2- System according to claim 1, characterized in that the linear lighting element (3.3), included in the frame (3.1) is composed of conventional LED elements or similar lighting elements. 3- System according to claim 2, characterized in that the padded element (3.4) arranged in the front part of the frame (3.1) is made of a foam rubber type material. 4.- System according to claim 1 further comprising an additional camera to the one that may already be incorporated in the drone (2), to help to steer the drone towards the crevice. 5.- System according to any one of the preceding claims, characterized in that it may also comprise an additional power supply device, even of an external nature. 6- Method that uses the system according to any of the preceding claims, which applies stereoscopic vision techniques and “Structure from Motion, Multiview Stereo (SfM-MVS)” type processing and which allows: - obtain 3D models of a crack (1.1) on structures built in any type of material using stereoscopic vision cameras (4), and - the temporal monitoring and tracking of the evolution of said crack (1.1) by comparing the distances between pairs of points, on both sides of the crack (1.1), on said 3D models, characterized in that, prior to positioning the system in the crack, a calibration of the cameras (4) is carried out, which is developed in 2 phases: 1 a Obtaining a representative set of photographs, to calculate the intrinsic parameters of each camera (4) and the extrinsic variables that define the spatial relationship between them, including their relative rotation and translation, and 2 a A dimensional adjustment. 7- Method according to the previous claim, characterized in that in the first phase of the pre-calibration of the cameras (4) the following intrinsic parameters of each camera (4) are calculated: fx, fy Focal distance on the X and Y axis Cx, Cy Coordinates of point A of intersection of the focal axis with the CMOS sensor of the cameras (4). Skew Tilt transformation coefficient k1, k2, k3, k4 Radial distortion coefficients p1, p2, p3, p4 Tangential distortion coefficients, and are stored permanently in the SfM-MVS software used, as an XML file. 8- Method according to the preceding claim, characterized in that in the second phase of the pre-calibration of the cameras (4) the distances between the reference points of the frame (3.1) are used, measuring, manually or in any other way, the distance between 2 of the targets (3.5) of the frame (3.1), to introduce it into the SfM-MVS software used and, Subsequently, the selected targets and the measured distance between them are marked on the 3D model, and from this moment on, the three-dimensional models obtained for the crack (1.1) are properly scaled, so that they can now be monitored. 9- Method according to any of claims 7 and 8 characterized in that, in order to know both the crack width (1.1) and its periodic monitoring over time, the initial measurement must be made as a reference and, thus, it will be possible to calculate, with precision, the dimensional variations suffered by the crack (1.1) between the different inspections, for which at least 2 singular points are selected on opposite sides of the crack (1.1) and a sufficient number of pairs of points, preferably 15 pairs of points or more, are marked on the photos taken of it around said singular points on each side of the crack, selecting as singular points of the crack, surface singularities of the same that can be easily identified in the next three-dimensional models of monitoring the evolution of said crack. 10- Use of a system or method according to any of claims 1 to 9, characterized in that it comprises at least the following steps: i. Fixing the drone's take-off point (2) on the ground, closest to the crack (1.1). i. Approach flight of the drone (2) to achieve its elevation to the level of the crack (1.1) under analysis. iii. Translation flight of the drone (2), preferably in a direction perpendicular to the structural element (1), to center the position of the frame (3.1) and the crack (1.1), in order to achieve the best focus of the cameras (4). iv. Support of the frame (3.1) on the structural element (1) containing the crack (1.1), in order to avoid movements of the drone (2) that could impair the quality of the photographs, if the conditions under which the inspection is carried out and the characteristics of the camera used allow it. v. With the help of the devices included in a drone (2), such as GPS and, where applicable, visual positioning systems, optical flow cameras or infrared beams, and ultrasonic sensors, maintain the fixed support position for the time necessary to take the photographs of the crack (1.1). vi. Taking photographs of the crack (1.1). vii. Separation flight of the drone from the frame (3.1) of the structural element (1) containing the crack (1.1), preferably in a direction perpendicular to said structural element (1). viii. Descent flight for the return of the drone (2) to the initial starting point, where it will land.

Images