DE202023000938U1 - Arrangement for photogrammetric object recording - Google Patents

Abstract

Arrangement for photogrammetric object recording,
characterized in that
an object stands, lies, or is fixed to a slide and that the slide has both an irregular, non-repeating geometric pattern of varying delicacy with a matte finish,
as well as at least three flat, geometrically unambiguously coded targets with a matt surface and known spatial position.

Description

Field of application and technical background

The invention relates to an arrangement for photogrammetric recording of objects for the purpose of 3D reconstruction. An object is imaged (camera) from several positions, therefore from different perspectives, with the same object areas being imaged in several recordings. Distinctive points (feature points) in the recorded images are extracted, corresponded and their position is registered in each image. With the help of this information, each camera location as well as the imaging geometry can be determined (resection) and in the next step a 3D point cloud of the object can be calculated. Using the color information of the recorded images, it is also possible to color the point cloud in real colors.

State of the art

• - Durch Projektion eines ortsfesten Musters (so genannte visuelle Hinweise) auf das Objekt werden extrahierbare Featurepunkte erzeugt [ EP 3 257 026 ].
• - Das Objekt wird auf einem Objektträger oder in einem Objektraum positioniert, der ein Muster aufweist, welches eine Extraktion von Featurepunkten ermöglicht. Dieses Muster kann asymmetrisch [ EP 3 477 254 ], gleichmäßig geometrisch [ DE 10 2018 105 589 ] oder natürlich wiederholungsarm [ EP 3 640 583 ] ausgestaltet sein. Objektträger bzw. Objekthintergrund werden zusammen mit dem Objekt und mit ausreichendem Bildanteil aufgenommen.

Arrangements are known that ensure extraction of a larger number of unique feature points, even though the object itself has only a few extractable features:

• - Extractable feature points are generated by projecting a stationary pattern (called visual cues) onto the object [ EP 3 257 026 ].
• - The object is positioned on a slide or in an object space that has a pattern that allows extraction of feature points. This pattern can be asymmetrical [ EP 3 477 254 ], regular geometric [ DE 10 2018 105 589 ] or naturally low in repetition [ EP 3 640 583 ] be configured. Object carriers or object backgrounds are recorded together with the object and with sufficient image content.

These arrangements have the disadvantage that no spatial scale can be assigned to the model. This is only possible by processing additional information, e.g. B. by evaluating individually determined dimensions.

• - Neben dem Objekt wird ein Referenzobjekt bekannter Geometrie und Abmessung positioniert, mit aufgenommen und rekonstruiert [ EP 3 640 583 ]. Die erzeugte 3D-Punktwolke wird danach so skaliert, dass die Geometrie des rekonstruierten Referenzobjektes dem Original entspricht, wodurch auch das rekonstruierte Objekt maßhaltig ist.
• - Der Objektträger oder das Objekt selbst besitzen definierte Zielmarken, deren Position in einem definierten räumlichen Koordinatensystem bekannt sind [ DE 10 2016 102 412 ], [ EP 3 436 773 ], [ DE 10 2013 211 342 ], [ DE 10 2021 111 417 ]. Diese Zielmarken werden bei der bildgebenden Aufnahme des Objekts mit erfasst und lassen sich durch ihre geometrische Form und/oder Farbgebung besonders gut und eindeutig extrahieren. Dabei werden sowohl flächige (2D) als auch räumliche Zielmarken (3D, kugelförmig) [ DE 10 2013 211 342 ] eingesetzt. Die Zielmarken sind durch ihre geometrische Gestalt so codiert, dass eine eindeutige Identifikation jeder Zielmarke mit Hilfe automatischer Bildauswertung möglich ist [ DE 10 2016 102 412 ]. Die Zielmarken werden mit dem Objekt aufgenommen und rekonstruiert. Die erzeugte 3D-Punktwolke wird danach so skaliert und transformiert, dass die Positionen der Zielmarken in der Punktwolke den originalen entsprechen, wodurch auch das rekonstruierte Objekt maßhaltig ist.

Furthermore, arrangements are known that provide a dimensional reference for the 3D reconstruction:

• - A reference object of known geometry and dimensions is positioned next to the object, recorded and reconstructed [ EP 3 640 583 ]. The generated 3D point cloud is then scaled in such a way that the geometry of the reconstructed reference object corresponds to the original, which means that the reconstructed object is also true to size.
• - The slide or the object itself has defined targets whose position in a defined spatial coordinate system is known [ DE 10 2016 102 412 ], [ EP 3 436 773 ], [ DE 10 2013 211 342 ], [ DE 10 2021 111 417 ]. These target marks are also recorded when the object is imaged and can be extracted particularly well and unambiguously due to their geometric shape and/or coloring. Both flat (2D) and spatial targets (3D, spherical) [ DE 10 2013 211 342 ] used. The target marks are coded by their geometric shape in such a way that a clear identification of each target mark is possible with the help of automatic image analysis [ DE 10 2016 102 412 ]. The target marks are recorded with the object and reconstructed. The generated 3D point cloud is then scaled and transformed in such a way that the positions of the target marks in the point cloud correspond to the original ones, which means that the reconstructed object is also true to size.

For objects with few prominent surface features, only an insufficient number of unique feature points can be extracted. The consequences are defects or incorrectly interpreted geometries in the calculated point cloud.

Arrangements are also proposed for the methods mentioned, in which the object recording is realized from different camera locations with the help of a rotary platform [ DE 100 11 241 [, [ DE 201 18 522 ], [ DE 20 2011 103 512 ], whereby the camera can also assume several vertical positions in addition to the horizontal rotation of the object [ DE 10 2011 104 216 ].

Technical task

The invention is based on the technical task of enabling a process-reliable and dimensionally accurate photogrammetric 3D reconstruction with a suitable arrangement that does not have the disadvantages mentioned. In particular, a sufficient number of feature points should be extracted for objects with little distinctive surface structure.

Solution to the problem, description of the invention

The problem is solved with the invention characterized in the main claim. Advantageous configurations are specified in the further claims.

The slide, in the simplest case a dimensionally stable plate, is covered with a pattern ensure that this can be captured as completely as possible from all camera positions. It is also sufficient for this if partial areas of the slide bear the pattern. The pattern is designed in such a way that it has a matt surface and high contrast, which means it can be easily recorded and evaluated. It is irregular, has no repetitions and has both coarse and fine structures in all areas, so that the automatic image evaluation in different geometric frequency ranges can reliably extract a large number of unique feature points in each image and independently of the object. The uniqueness refers to the correspondence of the same feature points in different images, which is necessary for the photogrammetric evaluation.

The object carrier also has a large number of planar target marks, which are arranged uniformly around the object in such a way that at least three target marks from a number of camera positions are recorded with the object. The targets are uniquely coded by their geometric shape, have a matte surface and high contrast, which enables automatic extraction and registration in the recorded images. For each target, the spatial coordinates, related to a meaningfully defined coordinate origin, are known and thus provide a dimensional reference. The photogrammetrically generated 3D point cloud is thus scaled and transformed in such a way that the spatial coordinates of the target marks in the point cloud correspond to the original ones, which means that the reconstructed object is dimensionally accurate.

Both the pattern and the target marks are permanently attached to the slide, so that all dimensional references remain constant throughout the image acquisition. The same applies to the object, whose position in relation to the object carrier also remains constant during the entire image recording.

An advantageous embodiment of the slide provides for one or more polygonal or round structures, the top and side surfaces of which are equipped both with target marks and with a pattern with the features already described. Not every surface has to have all of these characteristics. A configuration that is particularly advantageous for small objects provides for a rhomboidal shape of the pedestal, for example in the form of a cube placed on its tip, for the surfaces of which the described combination of pattern and target marks also applies. High platforms, which resemble a multi-sided polygon or round column, are also advantageous for certain types of objects. The pedestal or several pedestals can rest on a base plate, which also has target marks and/or patterns.

A configuration of the arrangement that is advantageous for object recording with a plurality of camera locations provides a rotating platform which is either itself designed as an object carrier according to the invention or carries it as a support structure. Instead of moving the camera around the object for a number of different recording positions or arranging a large number of cameras around the object, the object is moved on the rotary platform in front of the camera. Vertical camera positioning or the vertical arrangement of several cameras is also possible when using such a rotary platform.

example

• The 1 shows a recording arrangement in accordance with the invention.

A base plate 1 designed as a rotating platform carries an object carrier 4 in the form of an 8-sided pedestal. The object to be digitized is positioned on the slide for image recording and rotates with it around the central vertical axis of rotation.

The surface of the base plate has an irregular pattern 3 which is not repeated at any point and which is interrupted by concentrically arranged planar targets 2 . Both the pattern and the target marks are designed as high-contrast black and white prints with a matt surface.

The slide is firmly mounted on the base plate and has a similar pattern on the top surface as the base plate, but this does not match the base plate. The lateral surfaces are provided with target marks of the same type as the base plate. The object is either placed on the slide or fixed by means of a holder mounted on the slide. In any case, the object must not change in shape and relative position during the rotation of the rotating platform or during the image recording.

Each target in the system has its own unique code. The planar targets are circularly encoded in such a way that their position is irrelevant for registration. In order to avoid disturbing reflections, the target marks are provided with a very thin, matt top layer. The positions of the target center points on the surface of the mark are determined with high accuracy in a separate process and relate to a zero point - in this case the center of rotation of the top surface of the slide.

2 shows a rhomboidal slide in accordance with the invention. A cube 6 standing on its tip is firmly attached to a pedestal with a foot 7 . At the end of the pillar there is a small platform 5 or an adequate holder for fixing small objects. A flat target of the described design is attached to all six faces of the cube. The target positions are determined in a separate process and are referenced to the center of the small platform surface. According to the invention, all planar surfaces of the slide are provided with a pattern of the type described.

3 shows an exemplary pattern in accordance with the invention. It is irregular, does not repeat itself anywhere and has maximum edge contrast. The generated structures alternate between coarse and fine geometries in all areas, which is particularly advantageous for the extraction of unambiguous feature points using known methods. In order to avoid disturbing reflections, the pattern has a very thin, matt top layer

QUOTES INCLUDED IN DESCRIPTION

This list of documents cited by the applicant was generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Patent Literature Cited

• EP 3257026 [0002]
• EP 3477254 [0002]
• DE 102018105589 [0002]
• EP 3640583 [0002, 0004]
• DE 102016102412 [0004]
• EP 3436773 [0004]
• EN 102013211342 [0004]
• EN 102021111417 [0004]
• DE 10011241 [0006]
• DE 20118522 [0006]
• DE 202011103512 [0006]
• EN 102011104216 [0006]

Claims (4)

Arrangement for photogrammetric recording of objects, characterized in that an object stands, lies or is fixed to a slide and that the slide has both an irregular, non-repeating geometric pattern of varying fineness with a matt surface and at least three flat, geometrically unambiguous has coded targets with a matt surface and known spatial position. arrangement according to claim 1 , characterized in that the slide has a three-dimensional shape or has such partial areas. arrangement according to claim 1 , characterized in that the object carrier or partial areas thereof is at least one rhomboid arrangement, advantageously a cube standing on its tip. arrangement according to claim 1 , characterized in that the object carrier is a rotatable unit, for example in the form of a rotating platform.

Images