Computer-implemented method and system for assigning a numerical value to an annotation of an object

Description

[0001] The present invention relates to a computer-implemented method for assigning a numerical value to an annotation of at least one object identified in image, video and / or point cloud data.

[0002] The present invention further relates to a system for assigning a numerical value to an annotation of at least one object identified in image, video and / or point cloud data.

[0003] The present invention also relates to a computer program. State of the art

[0004] Computer vision models, i.e., algorithms for object recognition in image, video, and / or point cloud data, are trained using training data. To reliably recognize objects in the image, video, and / or point cloud data, the objects in question are conventionally annotated visually and / or conceptually.

[0005] The classification of the aforementioned objects is traditionally performed manually by annotators using appropriate software tools. In the field of computer vision models for autonomous driving, image annotation is generally carried out using so-called bounding boxes. These allow, for example, vehicles, traffic signs, and other environmental objects to be marked or annotated.

[0006] The CVPR publication of the Computer Vision Foundation, entitled "Interactive full image segmentation by considering all regions jointly", discloses a software application which, by annotating extreme points of certain image objects, enables a prediction for full image segmentation, i.e., a division of the full image into the objects it contains.

[0007] Furthermore, the software application has the feature that, in the event of an incorrect prediction of certain areas of the image segmentation, the annotator can make changes to the image segmentation using a graphical user tool, which can then be implemented by the software application.

[0008] Further methods for automatically identifying and annotating objects in received image, video and / or point cloud data are known from the following papers: NS Manikandan, K. Ganesan, "Deep Learning based Automatic Video Annotation Tool for Self-Driving Car", arXiv:1904.12618v1[cs.CV], Cornell University, April 19, 2019. Aleksa Corovic et al. "The Real-Time Detection of Traffic Participants Using YOLO Algorithm", 26th Telecommunication Forum (TELFOR), IEEE, Belgrade, Serbia, November 20, 2018. Joseph Redmon et al., "You Only Look Once: Unified, Real-Time Object Detection," arXiv:1506.02640v5[cs.CV], Cornell University, May 9, 2016.

[0009] However, the aforementioned methods all have in common that a high level of annotation effort is always required, since a very large amount of training data and its annotation are necessary for the effective training of computer vision models, which results in high personnel and cost costs.

[0010] Consequently, there is a need to improve existing methods and systems for annotating objects in image, video, and / or point cloud data to enable simplified, more efficient, and more cost-effective annotation of specified objects. Therefore, the object of the invention is to provide a computer-implemented method, a system, and a computer program that enables simplified, more efficient, and more cost-effective annotation of specified objects in image, video, and / or point cloud data. Disclosure of the invention

[0011] The problem is solved according to the invention by a computer-implemented method for assigning a numerical value to an annotation of at least one object identified in image, video and / or point cloud data according to claim 1, a system for assigning a numerical value to an annotation of at least one object identified in image, video and / or point cloud data according to claim 11 and a computer program according to claim 12.

[0012] The invention relates to a computer-implemented method for assigning a numerical value to an annotation of at least one object identified in image, video and / or point cloud data.

[0013] The procedure involves identifying and annotating the at least one object in received image, video and / or point cloud data, whereby the identification and / or annotation is performed at least partially automatically.

[0014] The procedure further includes calculating the numerical value of the annotation of the at least one object, wherein the numerical value depends at least partially on a degree of correspondence between a dimension of a visual annotation and a dimension of the at least one object and / or a correspondence between a conceptual label of the at least one object and the at least one object.

[0015] The procedure also includes assigning the calculated numerical value to at least one object.

[0016] The invention further relates to a system for assigning a numerical value to an annotation of at least one object identified in image, video and / or point cloud data.

[0017] The system includes means for identifying and annotating the at least one object in received image, video and / or point cloud data, whereby the identification and / or annotation can be performed at least partially automatically.

[0018] The system further includes means for calculating the numerical value of the annotation of the at least one object, wherein the numerical value depends at least partially on a degree of correspondence between a dimension of a visual annotation and a dimension of the at least one object and / or a correspondence between a conceptual label of the at least one object and the at least one object.

[0019] The system also includes means for assigning the calculated numerical value to at least one object.

[0020] The invention further relates to a computer program with program code to carry out the method according to the invention when the computer program is executed on a computer.

[0021] One aspect of the present invention is to enable at least partially automated identification and annotation of objects in image, video, and / or point cloud data. This can significantly reduce the time and processing effort previously required for tasks performed manually, particularly in the area of ​​computer vision models for autonomous processes.

[0022] Another idea of ​​the present invention is to calculate billing or pricing based on the achievement of predetermined technical parameters, namely on the accuracy of a visual annotation and / or the correct assignment of at least one conceptual annotation, in contrast to the pay-per-use billing models common in the field of cloud computing.

[0023] Further embodiments of the present invention are the subject of the further dependent claims and the following description with reference to the figures.

[0024] The procedure further includes the requirement that the conceptual designation of the at least one object includes at least one property of the object.

[0025] In addition to visual annotation, this allows for a more precise classification of the object in question by assigning one or more properties to the object.

[0026] According to a further aspect of the invention, the method also includes the visual annotation comprising the automatic positioning and drawing of a boundary element surrounding the object, which is formed by a 2D boundary frame or, in particular in the case of LiDAR and / or radar image data, by a 3D boundary frame.

[0027] Due to the automatic positioning and drawing of the corresponding bounding box around the object in question, a precise and efficient, i.e., processing time-reduced, annotation of the objects contained in the image, video and / or point cloud data can be carried out.

[0028] According to a further aspect of the invention, the method further comprises that the at least one property assigned to the object has at least one object class, wherein a first object class comprises motor vehicles and a first object subclass comprises passenger cars, trucks, vans, buses, construction vehicles, rail vehicles and / or trailers, and wherein a second object class comprises persons and a second object subclass comprises a gender, a size and / or an age of the person.

[0029] Furthermore, other objects relevant to the computer vision model, such as traffic signs, buildings, etc., can also be classified. By classifying objects into object classes and object subclasses, both a precise classification of the objects in question and a prediction of their expected behavior can be made.

[0030] According to a further aspect of the invention, the method further comprises the step of assigning at least one property to the sensor acquiring the image, video and / or point cloud data, wherein the assigned property comprises at least one sensor class, wherein a first sensor class comprises an image sensor and a first sensor subclass comprises a position and orientation of the image sensor on a carrier device, in particular on a detection vehicle, and wherein a second sensor class comprises a LiDAR sensor and a third sensor class comprises a radar sensor.

[0031] Knowing the position and orientation of the image sensor relative to the detected object also makes it advantageous to enable a more accurate classification of the object.

[0032] According to a further aspect of the invention, the method further comprises that the first sensor subclass includes a wide-angle camera arranged centrally at the front of the carrier device, in particular at the front of the detection vehicle, a narrow-angle camera arranged centrally at the front, a camera arranged at the front left, a camera arranged at the front right, a camera arranged at the rear left, a camera arranged at the rear right and / or a wide-angle camera arranged centrally at the rear.

[0033] This advantageously allows for a 360° recording of the surrounding traffic situation, including moving and stationary objects.

[0034] According to a further aspect of the invention, the method additionally comprises that the identification and visual annotation of the at least one object in the image, video and / or point cloud data is carried out manually, in particular by a user, and wherein the assignment of the at least one property of the object and / or the assignment of the at least one property of the at least one sensor detecting the at least one object is carried out automatically.

[0035] The method according to the invention can therefore also be advantageously applied when the step of identifying and visually annotating the at least one object in the image, video and / or point cloud data is carried out manually by a user and, based on this, properties are automatically assigned to the object and / or the sensor capturing the image, video and / or point cloud data is automatically labelled.

[0036] According to the invention, the method further comprises that the degree of correspondence of the dimension of the visual annotation in relation to the dimension of the at least one object and / or the correspondence of the conceptual designation of the at least one object with the at least one object is checked by a user.

[0037] This advantageously enables the highest possible accuracy of visual annotation and / or correctness of conceptual annotation. With the high efficiency and effectiveness of the computer-implemented method of automatic visual and conceptual annotation, only a relatively small additional effort is required for post-processing by the user.

[0038] According to a further aspect of the invention, the method also includes the fact that the degree of correspondence between the dimensions of the visual annotation and the dimensions of the at least one object is deemed sufficient if the dimensions of the boundary element essentially correspond to the dimensions, in particular the outer dimensions, of the annotated object.

[0039] Thus, an objective evaluation criterion can be advantageously provided to determine the accuracy of the visual annotation.

[0040] According to a further aspect of the invention, the method also includes the creation and storage of a transaction data record containing information required for calculating the numerical value of the annotation, in particular at least one automatically performed action, when performing the at least one visual annotation and / or the assignment of the at least one property of the object and / or the at least one property of the sensor.

[0041] The transaction data record thus advantageously stores each individual action performed, i.e., it stores whether it is a visual and / or conceptual annotation and whether the conceptual annotation includes the assignment of object properties and / or sensor properties.

[0042] According to the invention, the method further comprises that a change made by the user to the annotation of the object, in particular a change to the visual annotation and / or a change to at least one property of the object, is included in the transaction record of the object or in a transaction record linked to the transaction record of the object and stored in the transaction data store.

[0043] Thus, in addition to the annotation steps automatically performed by the computer-implemented process, the transaction record can also capture whether and to what extent changes were made to the annotation of the object in question. Such a transaction record then forms the basis for pricing the actions performed.

[0044] According to a further aspect of the invention, the method additionally comprises the numerical value of the annotation forming a price for the annotation, wherein each entry in the transaction data record relating to a performed action is priced by a valuation module using a pricing plan. Thus, an exact price for the performed actions can advantageously be achieved.According to a further aspect of the invention, the method further comprises the evaluation module forming a first sum from the numerical value of the at least one entry of the at least one, in particular automatically performed, annotation, and if the transaction record has at least one entry of a change made by the user, the evaluation module forming a second sum from the numerical value of the at least one entry of the change made by the user, and wherein the second sum is subtracted from the first sum to calculate the numerical value, in particular the price, of the annotation.

[0045] The inventive method thus advantageously prices the annotation of the object in question depending on the specified technical parameters, namely the accuracy of the visual annotation and the correct assignment of the conceptual annotation, and enables a corresponding, success-based billing of the services rendered.

[0046] The features of the method described herein are also applicable to scenarios other than computer vision models, such as person detection in various environments. Brief description of the drawings

[0047] For a better understanding of the present invention and its advantages, reference is made to the following description in conjunction with the accompanying drawings. The invention is explained in more detail below with reference to exemplary embodiments shown in the schematic illustrations of the drawings.

[0048] They show: Fig. 1 a flowchart of a method for assigning a numerical value to an annotation of at least one object identified in image, video, and / or point cloud data according to a preferred embodiment of the invention; Fig. 2 a schematic diagram of an annotation of objects performed in image, video, and / or point cloud data according to a preferred embodiment of the invention; Fig. 3 a schematic diagram of an annotation of objects performed in image, video, and / or point cloud data according to a preferred embodiment of the invention; Fig. 4 a block diagram of a plurality of object properties according to a preferred embodiment of the invention; Fig. 5 a block diagram of a plurality of sensor properties according to a preferred embodiment of the invention; and Fig.6. A schematic diagram of a system for assigning a numerical value to an annotation of at least one object identified in image, video and / or point cloud data according to the preferred embodiment of the invention.

[0049] Unless otherwise specified, identical reference numerals denote identical elements of the drawings. Detailed description of the embodiments

[0050] Fig. 1 Figure 1 shows a flowchart of a method for assigning a numerical value to an annotation of at least one object identified in image, video and / or point cloud data according to a preferred embodiment of the invention.

[0051] The procedure includes identifying S1 and annotating S2 the at least one object 14a, 14b in received image, video and / or point cloud data 12. The identification S1 and / or annotation S2 is performed at least partially automatically.

[0052] Alternatively, the identification S1 and visual annotation 10a, S3 of at least one object 14a, 14b in the image, video and / or point cloud data can be carried out manually, i.e. by a user.

[0053] The procedure further includes calculating S3 the numerical value of the annotation 10a, 10b of at least one object 14a, 14b. The numerical value corresponds to the price to be charged for the annotation 10a, 10b.

[0054] The numerical value is calculated at least partially depending on the degree of correspondence between a dimension 34a, 34b of a visual annotation 10a and a dimension 36a, 36b of the at least one object 14a, 14b and / or a correspondence between a conceptual designation 10b of the at least one object 14a, 14b and the at least one object 14a, 14b and / or a conceptual designation 10c of an at least one sensor 16a, 16b detecting the at least one object 14a, 14b and the at least one sensor 16a, 16b.

[0055] The degree of correspondence between dimension 34a, 34b of the visual annotation 10a in relation to dimension 36a, 36b of at least one object means that the visual annotation, for example a bounding frame, is correctly dimensioned in relation to the object in its dimension and position.

[0056] The bounding box is therefore neither too small nor too large relative to the object and is also correctly positioned and / or aligned to the object.

[0057] The correspondence of the conceptual designation 10b of the at least one object 14a, 14b with the at least one object 14a, 14b means that the pictorial content corresponds to the conceptual content, i.e., a car captured in the image, video and / or point cloud data is also correctly described as such in a conceptual sense.

[0058] The correspondence of the conceptual designation 10c of the at least one sensor 16a, 16b capturing the at least one object 14a, 14b with the at least one sensor 16a, 16b means that the sensor or sensors in question, with which the image, video and / or point cloud data were acquired, are correctly designated.

[0059] If, for example, the image, video and / or point cloud data were acquired using a front-central wide-angle camera 32a and a front-left camera 32c, these cameras should therefore be correctly named or correctly assigned to the image, video and / or point cloud data.

[0060] Furthermore, the procedure includes assigning S4 of the calculated numerical value to at least one object 14a, 14b.

[0061] Furthermore, it is intended that the identification S1 and the annotation S2 of the at least one object are carried out automatically, preferably using a machine learning algorithm such as an artificial neural network.

[0062] The annotation S2 includes the visual annotation 10a and the assignment of a predetermined number of properties 10b1 to the at least one object 14a, 14b and / or the designation 10b2 of at least one sensor 16a, 16b capturing the image, video and / or point cloud data 12 to the object 14a, 14b.

[0063] Fig. 2 shows a schematic diagram of an annotation of objects performed in image, video and / or point cloud data according to the preferred embodiment of the invention.

[0064] The visual annotation 10a includes the automatic positioning and drawing of a boundary element 18a surrounding the object 14a, 14b. In the present representation, the boundary element 18a is formed by a 2D boundary frame 18a.

[0065] The accuracy of visual annotation 10a is verified by a user.

[0066] The visual annotation 10a is considered exact if the visual annotation 10a meets user-defined, dimension-related requirements, in particular if a dimension 34a, 34b of the boundary element 18a, 18b essentially corresponds to an outer dimension 36a, 36b of the annotated object 14a, 14b.

[0067] The aim of automatically positioning and drawing the corresponding bounding box 18a around the object 14a, 14b in question is to fully automate the process of annotating objects 14a, 14b in image, video and / or point cloud data 12, thus eliminating the need for post-processing by the user.

[0068] This enables precise, efficient and cost-reduced annotation of the objects contained in the image, video and / or point cloud data.

[0069] Fig. 3shows a schematic diagram of an annotation of objects performed in image, video and / or point cloud data according to the preferred embodiment of the invention.

[0070] Visual annotation 10a includes automatic positioning and drawing of a boundary element 18b surrounding the object 14a, 14b.

[0071] In the present example, the boundary element 18b is formed by a 3D boundary frame 18b. The present representation consists of image and / or video data 12. 3D boundary frames are also particularly suitable for LiDAR and / or radar image data, i.e., point cloud data.

[0072] Fig. 4 shows a block diagram of a plurality of object properties according to the preferred embodiment of the invention.

[0073] The (in Fig. 4 (Not shown) object assigned a predetermined number of properties 10b1 has at least one object class.

[0074] A first object class 22a comprises motor vehicles 22a1. A first object subclass 22b comprises passenger cars 22b1, trucks 22b2, vans 22b3, buses 22b4, construction vehicles 22b5, rail vehicles 22b6 and / or trailer hitches 22b7.

[0075] A second object class 24a contains persons 24a1. A second object subclass 24b contains a gender 24b1, a height 24b2, and / or an age 24b3 of person 24a1. The correct assignment of at least one property 10b1 to the object is verified by a user.

[0076] Fig. 5 shows a block diagram of a plurality of sensor properties according to the preferred embodiment of the invention.

[0077] The conceptual designation 10c relates to the sensor 16a, 16b that detects at least one object.

[0078] The predetermined number of properties 10b2 assigned to the sensor 16a, 16b, which captures the image, video and / or point cloud data 12, includes at least one sensor class. A first sensor class 26a includes an image sensor 16a.

[0079] A first sensor subclass 26b specifies the position and orientation of the image sensor 16a on a detection vehicle 28. A second sensor class 30 specifies a LiDAR sensor and a third sensor class 31 specifies a radar sensor 16c.

[0080] As an alternative to the detection vehicle 28, the sensor 16a, 16b can, for example, be arranged on a stationary support device, such as on a building and / or a traffic sign.

[0081] Alternatively, the sensor 16a, 16b can be arranged, for example, on a rail vehicle and / or an aircraft.

[0082] When the sensor 16a, 16b is installed on a building, for example in a parking garage, the sensor can detect parked, entering and / or exiting motor vehicles.

[0083] When the sensor 16a, 16b is arranged on a traffic sign, for example on a traffic light system and / or a display board of a traffic guidance system, passing motor vehicles can be detected by the sensor.

[0084] The first sensor subclass 26b has a wide-angle camera 32a, a narrow-angle camera 32b, a camera 32c, a camera 32d, a camera 32d, a camera 32e, a camera 32f, a camera 32f, and / or a wide-angle camera 32g, arranged centrally at the front of the detection vehicle 28.

[0085] The computer-implemented method according to the invention can alternatively, for example, assign a numerical value to an annotation of at least one object identified in audio data, in particular speech data, and / or structured data.

[0086] Fig. 6 Figure 1 shows a schematic diagram of a system for assigning a numerical value to an annotation of at least one object identified in image, video and / or point cloud data according to the preferred embodiment of the invention.

[0087] The system includes means 52, 54 for identifying and annotating the at least one object 14a, 14b in received image, video and / or point cloud data 12, wherein the identification and / or annotation can be performed at least partially automatically.

[0088] The system also includes means 56 for calculating the numerical value of the annotation 10a, 10b of at least one object 14a, 14b.

[0089] The numerical value can be calculated at least partially depending on the degree of correspondence between a dimension 34a, 34b of a visual annotation 10a and a dimension 36a, 36b of the at least one object 14a, 14b and / or a correspondence between a conceptual designation 10b of the at least one object 14a, 14b and the at least one object 14a, 14b and / or a conceptual designation 10c of an at least one sensor 16a, 16b detecting the at least one object and the at least one sensor 16a, 16b.

[0090] The system also includes means 58 for assigning S4 of the calculated numerical value to at least one object 14a, 14b.

[0091] When performing at least one visual annotation and / or assigning at least one property of the object and / or designating at least one sensor capturing image, video and / or point cloud data, a transaction data record 38 containing information required for calculating the price of the annotation, in particular at least one automatically performed action, is created and stored in a transaction data store 40.

[0092] Each time an object is newly annotated, a corresponding transaction record 38a is created and sent via a push message P to a transaction gateway 39, from which the transaction record 38a is forwarded to the transaction data store 40 and stored therein.

[0093] A change 42a, 42b made by the user to the annotation of the object, in particular a change 42a of the visual annotation and / or a change 42b of the at least one property of the object and / or a property of the at least one sensor capturing the image, video and / or point cloud data, is included in the transaction data record 38 of the object or alternatively in a transaction data record 38 linked to the transaction data record 38 of the object 14a, 14b and stored in the transaction data store 40.

[0094] Each entry 38a, 38b contained in the transaction data record 38, relating to a performed action, is priced by a valuation module 44 using a pricing plan 46. The valuation module 44 calculates an initial sum 48 from the price of at least one entry 38a or at least one automatically performed annotation.

[0095] If the transaction record 38 contains at least one entry 38b of a user-made change 42a, 42b, the valuation module 44 calculates a second sum 50 from the price of the at least one entry 38b of the user-made change 42a, 42b. The second sum 50 is then subtracted from the first sum 48 to calculate the price of the annotation.

[0096] The entry 38a and the entry 38b can alternatively be stored in two separate transaction records 38, wherein the transaction records 38 are linked in such a way that a price calculation of the annotation is possible using the entries 38a, 38b of both transaction records 38.

[0097] If no changes are made to the transaction data record, the first sum 48 is decisive for price determination. Another influencing factor in the pricing of the actions performed is a subscription module 45, which contains conditions stored for the customer in question, such as a discount in the pricing plan 46.

[0098] Although specific embodiments have been illustrated and described herein, it is understandable to those skilled in the art that a multitude of alternative and / or equivalent implementations exist. It should be noted that the exemplary embodiment(s) are merely examples and are not intended to limit the scope, applicability, or configuration in any way.

[0099] Rather, the above summary and detailed description provides the person skilled in the art with convenient guidance for implementing at least one exemplary embodiment, whereby it is understood that various changes in the scope of functions and the arrangement of the elements can be made without deviating from the scope of the attached claims and their legal equivalents.

[0100] In general, this application is intended to cover changes, adaptations or variations of the embodiments set forth herein.

Claims

1. Computer-implemented method for assigning a numerical value to an annotation (10a, 10b) of at least one object (14a, 14b) identified in image, video and / or point cloud data (12), comprising the steps of: Identifying (S1) and annotating (S2) the at least one object (14a, 14b) in received image, video and / or point cloud data (12), the identifying (S2) and annotating (S3) being performed at least partially automatically; Calculating (S3) the numerical value of the annotation (10a, 10b) of the at least one object (14a, 14b), wherein the numerical value is calculated at least partly depending on a degree of a match of a dimension (34a, 34b) of a visual annotation (10a) in relation to a dimension (36a, 36b) of the at least one object (14a, 14b) and / or a correspondence of a conceptual label (10b) of the at least one object (14a, 14b) with the at least one object (14a, 14b), wherein the conceptual label (10b) of the at least one object (14a, 14b) comprises at least one property (10b1) of the object (14a, 14b), wherein the degree of correspondence of the dimension (34a, 34b) of the visual annotation (10a) in relation to the dimension (36a, 36b) of the at least one object (14a, 14b) and / or the correspondence of the conceptual labeling (10b) of the at least one object (14a, 14b) with the at least one object (14a, 14b) is checked by a user, and wherein a change (42a, 42b) to the annotation (10a, 10b) of the object (14a, 14b) by the user is recorded in the transaction data record (38) of the object (14a, 14b) or in a transaction data record (38) linked to the transaction data record (38) of the object (14a, 14b) and stored in the transaction data memory (40); and assigning (S4) the calculated numerical value to the at least one object (14a, 14b).

2. Computer-implemented method according to one of the preceding claims, characterized in that the visual annotation (10a) comprises automatically positioning and drawing a bounding element (18a, 18b) surrounding the object (14a, 14b), which is formed by a 2D bounding box (18a) or, in particular in the case of LiDAR and / or radar image data, by a 3D bounding box (18b).

3. Computer-implemented method according to claim 2 or 3, characterized in that the at least one property (10b1) assigned to the object (14a, 14b) has at least one object class, a first object class (22a) being motor vehicles (22a1) and a first object subclass (22b) being passenger cars (22b1), trucks (22b2), vans (22b3), buses (22b4), construction site vehicles (22b5), rail vehicles (22b6) and / or trailers (22b7), and wherein a second object class (24a) comprises persons (24a1) and a second object subclass (24b) comprises a gender (24b1), a height (24b2) and / or an age (24b3) of the person (24a1).

4. Computer-implemented method according to one of the preceding claims, characterized in that at least one property (10b1) is assigned to the sensor (16a, 16b) detecting the image, video and / or point cloud data (12), the at least one property having at least one sensor class, wherein a first sensor class (26a) comprises an image sensor (16a) and a first sensor subclass (26b) comprises a position and orientation of the image sensor (16a) on a carrier device, in particular on a detection vehicle (28), and wherein a second sensor class (30) comprises a LiDAR sensor (16b) and a third sensor class (31) comprises a radar sensor (16c).

5. Computer-implemented method according to claim 4, characterized in that the first sensor subclass (26b) has a wide-angle camera (32a) arranged centrally at the front on the carrier device, in particular on the detection vehicle (28), a narrow-angle camera (32b) arranged centrally at the front, a camera (32c) arranged at the front left, a camera (32d) arranged at the front right, a camera (32e) arranged at the rear left, a camera (32f) arranged at the rear right and / or a wide-angle camera (32g) arranged centrally at the rear.

6. Computer-implemented method according to one of the preceding claims, characterized in that the identification (S1) and the visual annotation (10a, S2) of the at least one object (14a, 14b) in the image, video and / or point cloud data (12) is performed manually, in particular by a user, and wherein the assignment of the at least one property (10b1) of the object (14a, 14b) and / or the assignment of the at least one property (10b2) of the at least one sensor (16a, 16b) detecting the at least one object (14a, 14b) is performed automatically.

7. Computer-implemented method according to claim 1, characterized in that the degree of correspondence of the dimension (34a, 34b) of the visual annotation (10a) in relation to the dimension (36a, 36b) of the at least one object (14a, 14b) is assessed as sufficient if the dimension (34a, 34b) of the boundary element (18a, 18b) essentially corresponds to the dimension (36a, 36b), in particular the external dimension, of the annotated object (14a, 14b).

8. Computer-implemented method according to one of the preceding claims, characterized in that when performing the at least one visual annotation (10a) and / or assigning the at least one property (10b1) of the object (14a, 14b) and / or the at least one property (10b2) of the sensor (16a, 16b), a transaction data record (38) comprising information required for calculating the numerical value of the annotation (10a, 10b), in particular at least one automatically performed action, is created and stored in a transaction data memory (40).

9. Computer-implemented method according to claim 1 or 8, characterized in that the numerical value of the annotation (10a, 10b) forms a price of the annotation (10a, 10b), each entry (38a, 38b) contained in the transaction data record (38) and relating to an action carried out being priced by a valuation module (44) using a pricing plan (46).

10. Computer-implemented method according to claim 9, characterized in that the valuation module (44) forms a first sum (48) from the numerical value of the at least one entry (38a) of the at least one annotation (10a, 10b), in particular an annotation performed automatically, and if the transaction data record (38) has at least one entry (38b) of a change (42a, 42b) made by the user, the evaluation module (44) forms a second sum (50) from the numerical value of the at least one entry (38b) of the change (42a, 42b) made by the user, and wherein the second sum (50) is subtracted from the first sum (48) to calculate the numerical value, in particular the price, of the annotation (10a, 10b).

11. System for assigning a numerical value to an annotation (10a, 10b) of at least one object (14a, 14b) identified in image, video and / or point cloud data (12), comprising: means (52, 54) for identifying and annotating the at least one object (14a, 14b) in received image, video and / or point cloud data (12), wherein the identification and annotation can be carried out at least partially automatically; means (56) for calculating (S3) the numerical value of the annotation (10a, 10b) of the at least one object (14a, 14b), wherein the numerical value can be at least partly calculated based on a degree of correspondence of a dimension (34a, 34b) of a visual annotation (10a) in relation to a dimension (36a, 36b) of the at least one object (14a, 14b) and / or a correspondence of a conceptual label (10b) of the at least one object (14a, 14b) with the at least one object (14a, 14b), wherein the degree of correspondence of the dimension (34a, 34b) of the visual annotation (10a) in relation to the dimension (36a, 36b) of the at least one object (14a, 14b) and / or the correspondence of the conceptual labeling (10b) of the at least one object (14a, 14b) with the at least one object (14a, 14b) can be checked by a user, and wherein a change (42a, 42b) to an annotation (10a, 10b) of the object (14a, 14b) made by the user can be recorded in the transaction data record (38) of the object (14a, 14b) or in a transaction data record (38) linked to the transaction data record (38) of the object (14a, 14b) and can be stored in the transaction data memory (40); and means (58) for assigning (S4) the calculated numerical value to the at least one object (14a, 14b).

12. Computer program comprising program code for performing the method according to any one of claims 1 to 10 when the computer program is executed on a computer.