Method and system for detecting building objects installed in a building

Abstract

The present invention relates to a building object detection system for detecting building objects installed in a building, the system comprising: at least one scanner adapted to scan a space within a building to generate scan data of the corresponding scanned space; and at least one artificial intelligence module implemented on a processing unit of the building object detection system, adapted to process the scan data generated by the at least one scanner to automatically detect and classify building objects installed in the scanned space, wherein the artificial intelligence module is trained using labeled training data, the labeled training data comprising labels derived from scan data of previously scanned building objects provided with associated object identifiers, the object identifiers being located in the vicinity of the corresponding installed building objects.

Description

[0001] The present invention relates to a computer-implemented method and a building object detection system for detecting building objects installed at fixed positions in a building, in particular building-installed equipment providing building services.

[0002] US2016 / 328660 A1 describes a machine learning arrangement for a point of sale system including a scanner component. The scanner component scans a barcode, QR code, RFID or another code. The arrangement further includes a camera component that acquires an image or picture of an object, and a computing component having a predictive algorithm for classifying the object.

[0003] WO 2015 / 017796 A2 describes a learning system that can be trained using a plurality of training images.

[0004] WO 2018 / 005369 A1 describes a camera system for inventory monitoring. The system includes a movable base supporting multiple cameras. The multiple cameras are directed toward the inventory to take a series of photos along the aisles of a retail store or warehouse.

[0005] In many use cases, it is necessary to detect building objects or static objects installed in public or private buildings. For example, it may be desirable to provide an inventory of building objects located in a building such as a factory. In addition, it may be necessary to check whether security requirements have been met, such as whether there are a sufficient number of smoke detectors in the rooms of a building. However, the detection and annotation of equipment in buildings using conventional scanners is time-consuming and cumbersome, as it is based on manual visual identification of building objects performed by an operator. Furthermore, the visual identification of building objects by an operator based on the image or scan data is prone to errors, as the operator may overlook building objects within the generated image or scan data.

[0006] It is therefore an object of the present invention to provide a method and a system for effectively detecting building objects installed in a building.

[0007] This object is achieved by a building object detection system, a method for detecting a building object and a method for training an artificial intelligence module comprising the features of the independent claims. The subject matter of the dependent claims defines further advantageous embodiments.

[0008] According to a first aspect of the present invention, the present invention provides a building object detection system for detecting a building object installed in a building, wherein the building object detection system comprises:

[0009] at least one scanner adapted to scan a space within a building to generate scan data of the corresponding scanned space, and at least one artificial intelligence module AIM implemented on a processing unit of the building object detection system adapted to process the scan data generated by the at least one scanner to automatically detect and classify building objects installed in the scanned space,

[0010] Wherein the artificial intelligence module AIM is trained using labeled training data, the labeled training data comprising labels derived from scan data of previously scanned building objects provided with associated object identifiers located in the vicinity of the respective installed building objects.

[0011] The building object detection system according to the first aspect of the present invention provides the advantage of quickly and reliably detecting a building object installed in a building.

[0012] In a possible embodiment of the building object detection system according to the first aspect of the present invention, the associated object identifier comprises a visible object identifier. In a possible implementation, the visible object identifier comprises a one-dimensional barcode, a two-dimensional QR code and / or a three-dimensional code.

[0013] In a further possible alternative embodiment, the associated object identifier comprises an invisible object identifier.In a possible embodiment, the invisible object identifier comprises an electromagnetic signal object identifier adapted to provide an electromagnetic identification signal in an invisible frequency range.

[0014] In a further possible embodiment of the building object detection system according to the first aspect of the invention, said at least one scanner is mounted on a remote controlled or autonomously driven robot of said building object detection system.

[0015] The driving robot may move within a building to provide scan data of a scanned space within the building to a trained artificial intelligence module AIM of the building object detection system.

[0016] In a possible embodiment of the building object detection system according to the first aspect of the invention, the artificial intelligence module AIM is implemented on a local processor of said driving robot.

[0017] In a further possible alternative embodiment of the building object detection system according to the first aspect of the invention, the artificial intelligence module AIM is implemented on a local controller of the building's network.

[0018] In still another possible alternative embodiment of the building object detection system according to the first aspect of the present invention, the artificial intelligence module AIM is implemented on a remote server of the cloud platform to detect and classify building objects within a scanning space installed in a building.

[0019] In a possible embodiment of the building object detection system according to the first aspect of the invention, the scanning data generated by the at least one scanner of the driving robot equipped to the building object detection system comprises point cloud data and / or image data, radar data and / or acoustic data.

[0020] In still further possible embodiments of the building object detection system according to the first aspect of the invention, the trained artificial intelligence module AIM comprises a trained deep neural network.

[0021] In still another possible embodiment of the building object detection system according to the first aspect of the present invention, the trained artificial intelligence module AIM receives scanning data from different scanners, wherein the classification of building objects based on the scanning data provided by one scanner is verified by a verification unit of the building object detection system based on the scanning data provided by another scanner.

[0022] The classification of the building object is verified by a verification unit of the building object detection system based on predetermined installation rules for the installation of the building object within the building.

[0023] In still another possible embodiment of the building object detection system according to the first aspect of the present invention, the detected and classified building objects installed within the scanning space of the building are stored in a database of the building object detection system and processed by a processing unit of the building object detection system to generate a digital twin of the scanned physical building, wherein the digital twin forms a real-time digital copy of the scanned physical building.

[0024] In another possible embodiment of the building object detection system according to the first aspect of the invention, the generated digital twin of the scanned physical building is compared with a digital planning model of the same building stored in a database to automatically detect deviations between the scanned physical building and the planned building.

[0025] In case a deviation is detected, corresponding measures may be automatically triggered by a controller of the building object detection system.

[0026] In a further possible embodiment of the building object detection system according to the first aspect of the invention, the scan data generated by the at least one scanner comprises georeferenced scan data, which is processed by a processor of the building object detection system to determine the position of the detected building object.

[0027] In a further possible embodiment of the building object detection system according to the first aspect of the invention, the determined positions of the detected and classified building objects are compared with building floor plan data of the building to automatically verify the classification of the respective building objects performed by the trained artificial intelligence module AIM.

[0028] In a still further possible embodiment of the building object detection system according to the first aspect of the invention, building objects detected and classified by the artificial intelligence module AIM are provided with annotation data and stored in a database of the system.

[0029] In still another possible embodiment of the building object detection system according to the first aspect of the present invention, the scanner mounted to the mobile driving robot is automatically moved to a position within the building indicated by the coordinates of the building object specified in the digital twin of the building to scan the surrounding environment of the indicated position.

[0030] In still another possible embodiment of the building object detection system according to the first aspect of the invention, the building objects installed in the building and detected by the building object detection system include: security and alarm installation equipment, in particular temperature sensors, fire detection equipment, fire protection equipment, lightning protection equipment, communication installation equipment, in particular telephone sockets and / or radio sockets and communication lines, artificial lighting equipment and lamps, electrical switches, electrical sockets and power lines, energy supply installation equipment, heating and cooling installation equipment, ventilation installation equipment, water supply and drainage installation equipment, escalator and elevator installation equipment, windows and doors installed in the building.

[0031] In still further possible embodiments of the building object detection system according to the first aspect of the invention, the object identifiers used for providing labeled training data for the artificial intelligence module AIM comprise hierarchical classifications and / or object attributes.

[0032] By using hierarchical classification, the training of the artificial intelligence module AIM can be improved and accelerated.

[0033] In still further possible embodiments of the building object detection system according to the first aspect of the invention, object identifiers placed in the vicinity of the mounted building object and used to provide labeled training data for the artificial intelligence module AIM provide absolute or relative position information.

[0034] According to a further aspect, the invention provides a computer-implemented method for detecting a building object installed within a building.

[0035] According to a second aspect, the invention provides a computer-implemented method for detecting a building object installed in a building,

[0036] The method comprises the following steps:

[0037] Scanning a space within a building to generate scan data corresponding to the scanned space; and

[0038] The resulting scan data is processed by the Artificial Intelligence Module AIM to automatically detect and classify building objects installed in the scanned space,

[0039] Wherein the artificial intelligence module AIM is trained using labeled training data, the labeled training data comprising labels derived from scan data of previously scanned building objects provided with associated object identifiers located in the vicinity of the respective installed building objects.

[0040] According to a further aspect, the present invention provides a method for training an artificial intelligence module AIM implemented in a processing unit of a building object detection system according to the first aspect of the present invention, wherein the training method comprises the following steps:

[0041] physically attaching the object identifier near the building object installed in the building,

[0042] Scanning a space within a building to generate scan data of the corresponding scanned space,

[0043] processing the generated scan data to derive labels from object identifiers included in the scan data,

[0044] automatically annotating the scan data using the derived labels to provide labeled training data; and

[0045] The artificial intelligence module AIM is trained using the labeled training data.

[0046] According to still another aspect, the present invention provides a driving robot comprising: at least one scanner adapted to generate scanning data of a scanning space within a building, wherein the generated scanning data is provided to an artificial intelligence module AIM implemented on a processing unit of a building object detection system according to the first aspect of the present invention, the system being used to detect building objects installed in a building.

[0047] In the following, possible embodiments of different aspects of the invention are described in more detail with reference to the accompanying drawings.

[0048] Figure 1 A block diagram showing a possible exemplary embodiment of a building object detection system for detecting building objects installed in a building according to aspects of the present invention;

[0049] Figure 2 a flow chart showing a possible exemplary embodiment of a computer-implemented method for detecting a building object installed in a building according to a further aspect of the invention;

[0050] Figure 3 shows a schematic diagram for illustrating generation of labeled training data for training an artificial intelligence module for detecting objects installed within a building;

[0051] Figure 4 A block diagram showing a possible exemplary embodiment of a system for detecting a building object installed within a building;

[0052] Figure 5 A flow chart showing a possible embodiment of a training process for training an artificial intelligence module of a building object detection system according to the first aspect of the present invention;

[0053] Figure 6 An embodiment of a driving robot used in a building object detection system according to the first aspect of the present invention is shown.

[0054] As in Figure 1 As can be seen in the schematic block diagram of , in the illustrated embodiment, the system 1 for detecting a building object OBJ installed in a building according to aspects of the present invention comprises: at least one scanner 2 and at least one artificial intelligence module (AIM) 3. Figure 1 As shown in the schematic block diagram of , in a possible exemplary embodiment, the scanner (SC) 2 can be mounted to the robot (ROB) 4. Figure 1As shown in , in a possible exemplary embodiment, the artificial intelligence module 3 can be implemented on a processing unit (PU) 5. A building object detection system 1 for detecting a building object OBJ installed in a building such as an office building or a factory may include: one or more scanners 2, which are adapted to generate scan data SD, which are provided by the scanner 2 to an artificial intelligence module 3 implemented on the processing unit 5. The building object detection system 1 may include one or more scanners 2 or scanning devices 2 that provide scan data SD to the artificial intelligence module 3. One or more scanners 2 may be mounted on a robot 4 that moves within a room of a building. In a possible embodiment, the robot 4 may be remotely controlled. In an alternative embodiment, the robot 4 moves autonomously within the scanned building. The scanner 2 is adapted to scan the surrounding space within the building to generate scan data SD of the scanned space, and provide the scan data SD to an artificial intelligence module 3 implemented in a processor or processing unit 5 of the building object detection system 1. The artificial intelligence module 3 is adapted to process the received scan data SD to automatically detect and classify the building object OBJ installed in the scanned space. The artificial intelligence module 3 is trained using labeled training data LTD, which includes labels L automatically derived from scan data SD of previously scanned building objects OBJ provided with associated object identifiers OBJ-ID. The scanned building objects OBJ may be, for example, smoke detectors with associated object identifiers OBJ-ID, such as visible codes. In a possible embodiment, the trained artificial intelligence module 3 is implemented in a local processor of the mobile robot 4. In an alternative embodiment, the artificial intelligence module 3 is implemented on a local controller of the network of the building scanned by the scanner 2. In still other possible embodiments, the artificial intelligence module 3 is implemented on a remote server of a cloud platform.

[0055] Figure 2 A flow chart of a possible exemplary embodiment of a computer-implemented method for detecting a building object OBJ installed in a building is shown.

[0056] The building object OBJ installed in the building and detectable by the building object detection system 1 according to the present invention may include a variety of different installed building objects OBJ. The building object OBJ is located at a predefined fixed position in the room of the building. The building object OBJ may include security and alarm installation equipment, such as temperature sensors, fire detection equipment, fire protection equipment or lightning protection equipment. The fire detection equipment may include, for example, a smoke detector. The building object OBJ may also include communication installation equipment, such as telephone sockets and / or radio sockets and communication lines. The building object OBJ installed in the room of the building may include lamps or lighting equipment that provide artificial light in the room of the building. In addition, the building object OBJ installed in the building may include switches and electrical sockets installed in the building and power cords. In addition, the building object OBJ may include energy supply installation equipment. In addition, the detectable building object OBJ may include heating and / or cooling installation equipment and ventilation installation equipment, such as fans. The building object OBJ may further include water supply and / or drainage installation equipment. Other examples of building objects OBJ include escalator and elevator installation equipment. Moreover, windows and doors installed in a building form building objects OBJ installed at predefined positions in the building according to the building plan of the building. Some building objects OBJ form an integral part of the building, i.e. they cannot be easily removed or replaced. Other building objects OBJ can be removed and / or replaced, but are still located at fixed predefined positions of the building. For example, a fire extinguisher device is located at a predefined position in the building, however, after a certain life, the fire extinguisher device is replaced with a new fire extinguisher device.

[0057] Most of the building objects OBJ installed in a building are installed according to predefined installation rules IR depending on the type of building object. The installation rules IR may be very strict for safety-critical installation equipment, while they may be more general for less critical building objects OBJ. In a possible embodiment, different types of building objects OBJ and the associated installation rules IR may be stored in a database 8 of the building object detection system 1. The installation rules IR may be used to support the classification of building objects OBJ within the computer-implemented method according to the present invention.

[0058] In the illustrated embodiment, the computer-implemented method comprises two main steps. The detection of building objects OBJ can be performed in the reasoning phase IPH by the trained artificial intelligence module 3. The trained artificial intelligence module 3 is also able to classify building objects OBJ that do not include an associated object identifier OBJ-ID.

[0059] In the first step S1 IPHIn the present invention, at least one scanner 2 of the system 1 scans a space in a building to generate scanning data SD of the corresponding scanning space.

[0060] In another step S2 IPH The generated scan data SD are processed by an artificial intelligence module 3 for automatic detection and classification of building objects OBJ installed in the scanning space. The artificial intelligence module 3 has been trained in a training phase TPH and / or during operation of the building object detection system 1 using labeled training data LTD comprising labels L derived from scan data SD of previously scanned building objects OBJ provided with associated object identifiers OBJ-ID. Figure 5 The diagram shows the training of the artificial intelligence module 3 in the training phase TPH.

[0061] In a possible embodiment, the associated object identifier OBJ-ID includes: a visible object identifier OBJ-ID, which can be attached to the shell of the building object, or it can be provided near the corresponding building object. For example, the associated object identifier OBJ-ID may include a visible code, which is attached to the shell of the building object, or is provided within a predetermined range around the corresponding building object. For example, if the building object is a smoke detector mounted to the ceiling of a room in a building, the visible code forming the object identifier OBJ-ID can be attached to the shell of the smoke detector, or fixed to the ceiling of the room near the smoke detector, for example, by means of a sticker with a predetermined radius r of, for example, 30 cm around the smoke detector. In a possible embodiment, the visible code can be a one-dimensional barcode. In a preferred embodiment, the visible code can include a two-dimensional QR code. In still another alternative embodiment, the visible code forming the object identifier OBJ-ID can also include a three-dimensional code.

[0062] In a possible embodiment, in step S1 IPHAt least one scanner 2 providing scan data SD may be a camera adapted to provide image data. The image data forms scan data SD that can be processed by the artificial intelligence module 3. In a possible embodiment, the image data may be generated in different frequency ranges. In a possible embodiment, the image data generated by the scanner 2 may include visible image data in a visible frequency range. In another possible embodiment, the generated image data may include data generated in an invisible frequency range, i.e. data that is invisible to the human eye, such as infrared data. The scanner 2 for generating scan data SD may also include a two-dimensional or three-dimensional scanner 2, in particular a three-dimensional laser scanning device that generates point cloud data. The point cloud includes a set of data points in a predetermined scanning space. The point cloud data is generated by a scanner 2 that measures a large number of points on the outer surface of an object surrounding it. As an output of the 3D scanning process, the scan data SD including the point cloud data may be provided to the artificial intelligence module 3 for further processing. The following processing steps S2 may be performed in real time IPH .

[0063] In still further possible embodiments, the scanner 2 may also be used to generate acoustic data as scan data SD. In still further possible embodiments, the scan data SD may also include radar data of radar signals reflected by objects in the surroundings of the scanner 2 .

[0064] exist Figure 2 Step S1 shown in IPHIn the invention, the space around the mobile robot 4 in the building can be scanned by different kinds of scanning devices 2, in particular cameras, laser scanners, radar scanners and / or acoustic scanning devices. The scanning data SD generated by one or more scanning devices 2 is passed to at least one trained artificial intelligence module 3. In a preferred embodiment, the artificial intelligence module 3 includes a trained deep neural network DNN. In a possible implementation, the deep neural network DNN is a trained convolutional neural network CNN for automatically detecting the corresponding object. In a possible embodiment, the trained artificial intelligence module 3 receives scanning data SD from different scanners 2 or scanning devices 2 (such as cameras, laser scanners 2, etc.). In a possible embodiment, the classification of building objects based on the scanning data SD provided by one scanner 2 (such as a camera) is verified in another data processing step based on the scanning data SD provided by another scanner 2 (such as a laser scanning device). In still another possible embodiment, the classification of the object can be further verified based on a predetermined installation rule IR for the installation of the object in the building. The installation rule IR can, for example, indicate where a specific type of object must be installed in the building and in what way it must be installed in the building. The installation rules IR may, for example, comprise a rule indicating that a smoke detector must be mounted at a minimum distance from another building object, such as a window of a building. The installation rules IR may be stored in a database 8 accessible by a processing unit 5 of the system 1. For different kinds of building objects OBJ, such as smoke detectors or switchgear, there may be a plurality of installation rules IR stored in the database 8 of the system 1, which may be used to verify a specified building object OBJ. In a possible embodiment, the scan data SD generated by at least one scanner 2 may comprise geo-referenced scan data SD, which is processed to determine the orientation of the detected building objects within the building. In a possible embodiment, the determined orientation of the detected and classified building objects is compared with the building floor plan data of the corresponding building to verify the classification of the corresponding building objects formed by the trained artificial intelligence module 3. In a possible embodiment, the building objects detected and classified by the artificial intelligence module 3 may be provided with annotation data and stored in the database 8 for further processing.

[0065] In a possible embodiment, Figure 2 The scanning data SD generation step S shown in the flowchart IPH1 and object classification step S IPH2 This is followed by a validation step using different scan data sources 2 and / or installation rules IR and / or building floor plan data of the building. This validation increases the reliability of the classification results and the robustness of the building object detection against external influences.

[0066] Figure 3 A schematic diagram is shown for illustrating the generation of training data for training the artificial intelligence module 3 of the detection system 1 implemented on the processing unit 5. Figure 3 As can be seen in the schematic diagram of , a building object such as a smoke detector or any other building object installed in a building comprises an associated object identifier OBJ-ID, such as a visible object code. The object code forming the object identifier OBJ-ID may be, for example, a QR code attached to the housing of the building object or attached within a predetermined range around the installed building object. The scanning device 2 may be used to scan the building object and the associated object identifier OBJ-ID to generate scan data SD, which is provided to a code detection algorithm installed in the object identifier OBJ-ID detection unit 6, such as Figure 3 . The object identifier OBJ-ID detection unit 6, running, for example, a QR code detection algorithm, can generate labels L derived from the received scan data SD to generate labeled training data LTD for training the artificial intelligence module 3 in a training phase TPH. The object identifier OBJ-ID detection unit 6 can have access to a database 8 to generate labels L and extract and / or retrieve associated labels L. The trained artificial intelligence module 3 can later be used in an inference phase IPH to perform detection and classification of other installed physical building objects OBJ by means of at least one scanning device 2, the at least one scanning device 2 being a Figure 1 The robot 4 illustrated in the block diagram of can be moved, or mounted on a trolley that is pushed through a building by an operator.

[0067] Figure 4 A block diagram of a further possible exemplary embodiment of a detection system 1 according to the invention is shown. In the illustrated exemplary embodiment, the building detection system 1 comprises more than one scanning device or scanner 2-i. A first scanning device 2-1 may be an image source, such as a camera providing an image of the surroundings as scan data SD1, which is provided to a first trained artificial intelligence module 3-1 for classifying building objects OBJ in the surroundings.

[0068] In the illustrated exemplary embodiment, the building detection system 1 includes a second scanning device 2-2, which can be used as an image source to generate a two-dimensional point cloud as scan data SD2. In the illustrated exemplary embodiment, the two-dimensional point cloud is provided to a second artificial intelligence module 3-2 for classifying building objects OBJ in the surrounding environment.

[0069] In addition, in the illustrated embodiment, the building detection system 1 includes a third scanning device 2-3, such as a laser scanner that generates a three-dimensional point cloud as scan data SD3, which is provided to a third trained artificial intelligence module 3-3 for classifying building objects OBJ in the surrounding environment of the scanner 2-3.

[0070] exist Figure 4 In the illustrated embodiment, the building detection system 1 includes a verification unit 7, which is used to verify the correct classification of the building object OBJ. For example, the verification unit 7 can use the classified building object OBJ output by different artificial intelligence modules 3-i to perform verification as to whether the classified building object OBJ has been correctly classified. For example, if the classified building object OBJ output by different artificial intelligence modules 3-i belong to the same category, the verification unit 7 can verify that the classification is successful. In contrast, if the building objects OBJ classified by different artificial intelligence modules 3-i are different, the verification unit 7 cannot successfully perform the verification. In still another possible embodiment, the verification unit 7 can verify the classification of the building object based on a predetermined installation rule IR for the installation of the building object OBJ within the building, as also shown. Figure 4 As illustrated in the block diagram of . After successful verification by the verification unit 7, the detected and classified building objects OBJ can be stored in the database 8 of the detection system 1. In a possible embodiment, the processing unit 9 can automatically generate a digital twin DT of the scanned building based on multiple detected and classified building objects OBJ and building data such as floor plans of the corresponding buildings. The digital twin DT includes a real-time digital copy of the scanned physical building. The digital twin DT can be updated in real time by the processing unit 9. In a possible embodiment, the digital twin DT of the scanned building can be compared with a digital planning model of the same building to detect deviations between the scanned physical building and the planned building. Depending on the detected deviation, the following actions can be triggered: for example, installing missing objects, in particular building objects OBJ (such as smoke detectors). In addition, the processor can compare the determined orientation of the detected and classified building objects OBJ with the building floor plan data of the planned building to check whether the classification of the corresponding building objects by one or more trained artificial intelligence modules 3-i is correct. The scanner 2, the artificial intelligence module 3 and the verification unit 7, as well as the database 8 and the processing unit 9 can be integrated in Figure 1 , 6 In the movable robot 4 shown in the figure.

[0071] Further embodiments of the computer-implemented method and building detection system 1 according to the present invention are possible. In a possible implementation, a point of interest POI may be generated using an indoor viewer. The point of interest POI comprises a point defined within the three-dimensional coordinate system 1 of the indoor viewer instance with additional information, which may include GPS coordinates indicating a specific orientation. Further point of interest POIs may include types, type groups, and orientations. Point of interest POIs may be associated with objects in the three-dimensional scan.

[0072] In a possible embodiment, the artificial intelligence module 3 may include a deep learning artificial neural network DNN. In another possible embodiment, the artificial intelligence module 3 may be used for other kinds of machine learning algorithms. In a possible embodiment, the artificial intelligence module 3 may be implemented in a graphics processing unit GPU. The artificial intelligence module 3 including an artificial neural network ANN for implementing a machine learning ML algorithm uses labeled training data LTD, wherein the labeled training data LTD includes a label L automatically derived from the scan data SD of a previously scanned object with an associated object identifier OBJ-ID. In a possible embodiment, the previously scanned building object OBJ may include a reference building object OBJ scanned in a training environment and / or scanned on site (i.e., in a building). In a possible embodiment, the label L of the training data is derived from a two-dimensional QR code. The QR code includes: a quadratic matrix consisting of black and white squares, which represents the encoded data in binary form. The QR code can be protected by an error correction code. The QR code may include a direct description of the corresponding object, such as a smoke detector "Sinteso FDOOT241-9", or may include a network link. The scanner 2 may focus on the surroundings of the QR code and may assign a description to the detected building object based on the information encoded by the QR code.

[0073] Since building objects OBJ within a building are assembled and installed according to functional directions and installation rules IR, it is also possible to provide a filtered association of object selection. For example, conventional standard smoke detectors are installed on the ceilings of rooms in office buildings. In many cases, such specifications or standards or even laws include instructions for maintaining installation distances and similar information. These installation rules IR can be used for building object detection performed by the building object detection system 1 according to the present invention, as well as Figure 4A possible installation rule IR could be, for example: "If the detected building object is a smoke detector, then there must be another smoke detector within x square meters and y meters of surroundings". This allows excluding or including objects on a logical semantic level.

[0074] In a possible embodiment, the installation rule IR can be used by the detection system 1 to increase the detection quality, for example by implementing Figure 4 The verification unit 7 illustrated in the embodiment of . Another example of the installation rule IR may be that in addition to the passage, the switch of the light must be installed on the wall of the room at a height of 1 to 1.4 meters.

[0075] In a possible embodiment of the detection system 1 according to the invention, a point cloud forming the scan data SD is evaluated for detecting and classifying objects. The evaluation of the point cloud can increase the detection quality when compared to pure image-based detection. The point cloud data includes depth information, i.e. the points include not only x and y coordinates, but also z coordinates. This allows the form and geometry of building objects OBJ to be recognized in a more reliable way. Image recognition algorithms can use both point cloud data and / or two-dimensional photo / video data for object detection to further improve qualified analysis results. The number and type of scanning devices 2-i can vary depending on the use case.

[0076] With the geo-referencing of the scanning device, it is also possible to determine WGS 84 coordinates (GPS) or data of an alternative coordinate system 1 (such as UTM), which makes it possible to annotate building objects in a two-dimensional or three-dimensional floor plan of a building. In a two-dimensional floor plan of a building, it is also possible to place corresponding symbols. In addition, it is possible to transfer interface coordinates via another system 1 to the building object OBJ. The detected building object OBJ may include a wide variety of different building objects OBJ that may be provided in the building, such as sensors or sensor devices, in particular smoke sensors or temperature sensors. Other building objects OBJ may include light switch units or other hardware components that are installed in the rooms of the building.

[0077] In still further possible embodiments of the detection system 1 according to the invention, switching between different scanning devices 2-i may be performed depending on the environment and / or the operating mode of the system 1. The scanning device 2-i may comprise a camera that generates an image of the surroundings as scan data SD. In further possible embodiments, the scanning device 2-i may comprise a camera that generates a video stream consisting of a plurality of images of the scanned space. In a possible embodiment, the scanning device 2-i may comprise a fisheye camera that provides a 360° image of the surroundings.

[0078] In a preferred embodiment, a visible object identifier OBJ-ID (such as a QR code) can be attached to the associated building object, for example, directly on its housing or in its surroundings or in its vicinity. In an alternative embodiment, the attached object identifier OBJ-ID can also include an identifier that is particularly invisible to the human eye.

[0079] The invisible object identifier OBJ-ID may include an object identifier OBJ-ID that provides an electromagnetic signal invisible to the human eye, such as a radio signal, in particular a Bluetooth or RFID signal, or provides an optical signal in an invisible frequency range, such as an infrared signal.

[0080] In some countries, smoke detectors must be equipped with associated object identifiers OBJ-ID. These mandatory marking identifiers can be used as associated object identifiers OBJ-ID by the detection system 1 according to the present invention. The object identifier OBJ-ID can be a mandatory marking, which is formed by a text or character sequence attached to the shell of an object (especially a smoke detector) or provided near a building object. The characters may include a coded address. If such an address is detected and identified, there is a high probability that a physical building object (such as a smoke detector) exists in the direct surroundings of the coded address. The associated object identifier OBJ-ID may also include a three-dimensional code, such as the form of the shell of the corresponding object. For example, if the smoke detector has different shell forms, the form of the corresponding shell can form an associated object identifier OBJ-ID, which is used by the detection system 1 according to the present invention. In contrast, if different building objects OBJ have the same shell, an additional QR code may be used to identify the corresponding building object.

[0081] In a possible embodiment, a three-dimensional (3D) code may extend a two-dimensional QR code with an additional dimension (i.e., color). Each pixel or dot in the two-dimensional code may include a color (wavelength λ) that encodes a value. The 3D code may be provided by LEDs. In this embodiment, the object identifier OBJ-ID includes an LED matrix to provide a three-dimensional code.

[0082] In a possible embodiment, if a building object is included in the generated scan data SD, the object can be annotated at the determined position of the building object using the indoor viewer. The position of the building object can be referenced in the digital twin DT. For this purpose, WGS 84 position or room position (distance from wall, ceiling, floor) can be used. The 360° image can include pixels that can be enriched by point cloud data generated by laser scanning.

[0083] In a possible embodiment, additional labeled training data LTD are generated in the inference phase IPH during the operation of the detection system 1 and stored in the database 8 of the system 1. The detection system 1 according to the present invention can be used to generate an inventory of buildings, including all detected and classified building objects OBJ, and their orientation within the building. The generated digital twin DT of the scanned building can be used for different purposes. In a possible embodiment, the generated digital twin DT of the scanned building can be compared with the building data to identify deviations from the building floor plan, thereby triggering corresponding actions. In another possible embodiment, the digital twin DT can also be used to support service technicians performing services within the building. The digital twin DT forms machine-readable data stored in the database 8 of the system 1.

[0084] In addition to the scan data SD provided by the scanning device 2-i of the detection system 1, a floor plan with spatial details and equipment specified in a legend can also be used as an additional input to the system 1. Therefore, in a possible embodiment, a scanning device 2-i can be used to scan a printed floor plan of a building and a corresponding legend of building objects OBJ installed in the building. Building objects may include, for example, smoke detectors, alarm devices, connectable switches or lamps. The generated images of the scanned floor plan can be uploaded. The recognition algorithm can make these images searchable, making it possible to discover building objects OBJ and the scenes that appear in them. A lambda function can be created that automatically adds the newly detected image label L directly to the search index when a new image has been uploaded. The recognition algorithm retrieves the image and returns the label L of the detected image. The returned image can be pushed together with the confidence score provided with the search index to provide an output digital twin DT. The scanning device 2-i of the detection system 1 can capture an image as scan data SD, which contains building objects OBJ, scenes and / or, for example, text data. In a possible embodiment, the image provided by the scanning device 2-i (such as a camera) can be uploaded to a remote server of a cloud platform including at least one trained artificial intelligence module 3-i. A number or function can be triggered to call a recognition algorithm that retrieves an image and returns a label L of the detected image. The retrieved label L can be used to generate training data applied to the artificial intelligence module 3-i during the training phase TPH and / or the operational reasoning phase IPH of the system 1. In a possible embodiment, an associated search index can be provided to the derived label L. In a possible embodiment, the scan data SD includes images that can be captured by a NavVis M6 cart that forms a mobile robot 4 to provide a realistic point cloud, and includes a 360° image provided by a fisheye camera. In a possible embodiment, the QR code can contain any details and / or information of the scanned building object next to it. The scan data SD may include the orientation and location of the corresponding building object at the correct coordinates within the digital twin DT.

[0085] In a possible embodiment, the scanned detected and classified objects may include static objects mounted at fixed locations in a building.

[0086] Therefore, the computer-implemented method and system 1 according to the present invention can be used for a wide range of different use cases for any kind of building to detect and classify a plurality of building objects OBJ installed within the respective building.

[0087] In a possible embodiment, the building object OBJ can be detected based on the coordinates of a predetermined space or room within the building. For example, if in the digital twin DT BIM data model, the building object OBJ is provided with associated coordinates. These building object OBJs can be selected as possible candidate building object OBJs so that at least one scanner 2-i of the detection system 1 can focus on the area indicated by these coordinates to detect the physical building object at the indicated position. In a possible implementation, the scanner 2-i can be automatically moved to the indicated position, for example, by a mobile robot 4 to scan the surrounding environment of the indicated position. Then, it can be determined whether the possible candidate building object indicated by the coordinates of the possible candidate building object of the digital twin DT does exist as a physical building object at the indicated position. This allows the detection system 1 to check whether the various building objects OBJ of the planned building have been installed in the constructed physical building. After successfully verifying that the building object has been installed at the planned position (coordinates), the scanner 2-i can pass the generated scan data SD to the artificial intelligence module 3 in real time.

[0088] According to a further aspect, the invention provides a method for training an artificial intelligence module 3 implemented in a processing unit 5 of a building object detection system 1 according to the first aspect of the invention.

[0089] Figure 5 A flow chart of a possible embodiment of such a training method performed in a training phase TPH is shown.

[0090] In the first step S1 TPH In the embodiment, the object identifier OBJ-ID is physically attached near the building object OBJ installed in the corresponding building. The object identifier OBJ-ID may, for example, comprise a QR code printed on a sticker, which is attached in the surroundings of the building object (such as a smoke detector). In a possible embodiment, the QR code printed on the sticker may be attached to the ceiling of the room within a predefined radius r around the corresponding physical building object (i.e., the smoke detector). Depending on the form of the building object, the object identifier OBJ-ID may also be attached to the housing of the corresponding building object.

[0091] In another step S2 TPH In the embodiment, the space in the building is scanned by the scanner 2 to generate scanning data of the corresponding scanning space. The scanner 2 can be mounted on a movable driving robot 4. The driving robot 4 moves in the room of the building and is adapted to scan the surrounding rooms or spaces.

[0092] In another step S3 TPHIn the embodiment of the present invention, the generated scan data SD is processed to automatically derive a label L from an object identifier OBJ-ID included in the scan data SD. The scan data SD may, for example, include an image of an area in which the object identifier OBJ-ID is located near a corresponding building object. The generated image captures the object identifier OBJ-ID and the corresponding building object. By processing the image data, it is possible to detect an object identifier OBJ-ID such as a two-dimensional QR barcode, and automatically derive a label L associated with the corresponding building object included in the same image from the object identifier.

[0093] In another step S4 TPH In the method, the scan data SD (such as images) are automatically annotated with the derived labels L to provide labeled training data, which can be stored in a database 8 of the system 1.

[0094] In another step S5 TPH In the example, the labeled training data LTD is used to train the artificial intelligence module 3. The artificial intelligence module 3 trained in this way can then be used in the reasoning phase IPH to detect and classify building objects OBJ that do not have an associated object identifier OBJ-ID, such as Figure 2As illustrated in the flowchart of . Thus, the trained method allows the training of the artificial intelligence module 3 based on the building objects OBJ provided on site with the attached object identifier OBJ-ID. This has the following advantages: namely, only a part of the building objects OBJ installed in the building must be provided with the attached physical object identifier OBJ-ID (such as, QR code stickers) to provide a trained artificial intelligence module 3, which has the capacity and ability to detect and classify multiple other physical building objects OBJ, which are installed in the same building or another building, but do not have the attached physical object identifier OBJ-ID in their vicinity. Therefore, the training process becomes more efficient. During the operation of the system 1, the labeled training database 8 can still be increased. The building object detection system 1 according to the present invention is labeled to train itself and is automatically optimized. For example, in a building with several floors, the building object OBJ installed in the first floor can be attached with the object identifier OBJ-ID by the operator, and the object identifier OBJ-ID is placed near the labeled building object OBJ. The building object OBJ and the attached object identifier OBJ-ID can be used to provide labeled training data LTD for training the artificial intelligence module 3 of the system 1. Then, the mobile driving robot 4 with the scanner 2 can be placed in the next floor of the building to process the generated scan data SD so as to detect and classify the building object OBJ installed in the next floor of the building. The training data provided by the building object OBJ in the first floor can be sufficient to train the artificial intelligence module 3, so that it is not necessary to attach the object identifier OBJ-ID to the building object OBJ in the next floor (i.e., the second floor) of the building to provide classification results.

[0095] In a possible embodiment, the object identifier OBJ-ID used to provide labeled training data LTD for the artificial intelligence module 3 may include hierarchical classifiers and / or object attributes. For example, the hierarchical object identifier OBJ-ID of a smoke detector may include several hierarchical classifiers, as shown below. The first classifier may indicate that the building object (e.g., a smoke detector) is a building object installed at the ceiling of a room. The next classifier of the hierarchical object identifier OBJ-ID may indicate that the building object is a fire detection element. The next classifier of the hierarchical object identifier OBJ-ID may indicate that the building object is a smoke detector. Further, the next classifier of the hierarchical object identifier OBJ-ID may indicate the type of the corresponding smoke detector. By using the hierarchical object identifier OBJ-ID, it is possible to facilitate the training of the artificial intelligence module 3 and improve the performance of the training process.

[0096] In another possible embodiment, an object identifier OBJ-ID placed near the installed building object OBJ and used to provide labeled training data LTD for the artificial intelligence module 3 can provide absolute or relative orientation information. For example, a QR code forming the object identifier OBJ-ID of a smoke detector attached near the smoke detector can indicate the relative orientation of the corresponding smoke detector. For example, the object identifier OBJ-ID or the QR code can indicate that the corresponding smoke detector is located to the left of the object identifier OBJ-ID. The absolute or relative orientation encoded in the object identifier OBJ-ID can also facilitate the training process and increase the efficiency of the training process. Using the training method according to the present invention, the time-consuming and cumbersome learning process can be significantly alleviated, because the object identifier OBJ-ID such as a QR code can include an exact description of the corresponding building object (e.g., a hierarchical classifier).

[0097] The building object description can also be provided by a link indicated by an object identifier OBJ-ID (such as a QR code). Therefore, in a possible implementation, the object identifier OBJ-ID can also include a code for indicating a link to a data source, which includes the exact description of the associated building object. Additionally, the object identifier OBJ-ID (such as a QR code) must be provided to a considerable number of building objects OBJ. As the number of scans provided by the scanner 2 increases, the number of objects to which the object identifier OBJ-ID must be provided can be significantly reduced, because the building objects to be detected and characterized have been stored in the database 8 of the system 1. Therefore, when executing a computer-implemented method for detecting building objects OBJ installed in a building, the system 1 can automatically expand its training database 8. By learning on the fly, the building object detection system 1 continuously improves its performance. Each scan generates multiple training data for the same building object OBJ associated with the same object identifier OBJ-ID at different locations and under different environmental conditions. This provides an advantage for training the artificial intelligence module 3. Once a building object has been detected, it can be annotated at the detected position. The new position of the detected building object can be georeferenced. The position can be transferred to the BIM data model. This allows the building object to be annotated in a two-dimensional or three-dimensional building plan. The detected object can be integrated into the digital twin DT of the building. In addition, the detected building object can be linked to an adjacent building object OBJ or element, for example, located at the same ceiling or wall of the building. In a possible embodiment, it can be verified whether the installed building object satisfies the predefined installation rule IR. It is possible to automatically detect whether the classified building object located at a specific determined position satisfies the associated installation rule IR. In the case of a violation of the installation rule IR, the operator can be guided to the building object, for example to change its position at the wall or ceiling. Multiple different installation rule IRs for different types of building objects OBJ can be stored in the database 8. These installation rule IRs can vary depending on the location of the building (for example, the country in which the building is located). Depending on the exact location of the building, different sets of installation rule IRs can be loaded from the database 8 of the building object detection system 1. For example, if the building is located in Munich, the installation rule IR for Germany can be downloaded from the database 8 of the building object detection system 1. In a possible embodiment, if there is an installation rule IR for a building object, the user U can use a corresponding editor to set or define the corresponding installation rule IR. Therefore, in a possible embodiment, the installation rule IR can be defined according to the best practice experience of the user U. The stored installation rule IR can improve the quality of object detection and classification.For example, if according to the installation rule IR, a specific building object (such as a smoke detector) should have been installed at the ceiling of a building room, but has not yet been detected at the requested position within the scanning area of ​​the scanner 2 of the building object detection system 1, a message can be automatically generated. Therefore, if the necessary building object (such as a smoke detector) has not been detected in the building room, a corresponding alarm message can be automatically generated, and the alarm message can trigger countermeasures performed by the operator. In addition, if a building object is detected that should not be in a scanning area that may be safety-critical, a corresponding alarm message can be automatically triggered, and the operator can be guided to the position of the detected unwanted building object to remove the building object from the specified position. Therefore, the building object detection system 1 according to the present invention is able to detect the wrong placement of a building object OBJ within a building and automatically generate a corresponding message to the user U of the building object detection system 1. The building may include an office or a residential apartment. In addition, the building may also include an industrial facility, such as a factory.

[0098] Figure 6 A possible exemplary embodiment of a driving robot 4 is shown, which can also be used as Figure 1 The driving robot 4 may include a scanner 2 for automatically capturing point clouds, 360° immersive virtual data, and sensor data while moving through a building at walking speed. In the illustrated exemplary embodiment, the driving robot 4 includes a platform 4A for carrying an operator or user U, such as Figure 6 As shown in . The platform 4A is mounted on wheels 4B, 4C to move on the ground floor of the room in the building. Different kinds of scanners 2-1 to 2-7 can be mounted on the top of the driving robot 4 to automatically generate scan data SD of the scanned space around the driving robot 4. The scanner 2 can provide point cloud data, image data, radar data and / or acoustic data. In a possible embodiment, the processing unit 5 can also be integrated in the driving robot 4, such as Figure 6As shown in . The processing unit 5 may, for example, comprise a laptop or tablet device attached to the structure of the driving robot 4. In a possible embodiment, the processing unit 5 may comprise a graphical user interface U GUI which may be operated by the user U. The driving robot 4 may also comprise a handle 4D to provide stability to the user U during the movement of the driving robot 4 through the building, and / or to manipulate the driving robot 4. The driving robot 4 may comprise a motor, in particular an electric motor 4E, for operating the wheels. The processing unit 5 may receive scan data SD from the scanner 2-i. In a possible embodiment, the processing unit 5 of the driving robot 4 may comprise a communication interface to forward the scan data SD to another processing unit with an integrated artificial intelligence module 3. In an alternative embodiment, the artificial intelligence module 3 may also be integrated in a Figure 6 In the processing unit 5 of the driving robot 4 shown in . In a possible embodiment, the scan data SD received from the scanner 2 can be processed in real time to detect and classify the building objects OBJ in the scanning space installed around the driving robot 4. The movable driving robot 4 allows faster and easier capture of indoor building data. Real-time feedback on the screen of the graphical user interface U GUI allows the operator or user U to keep track of the area that has been scanned. The movable driving robot 4 can keep scanning on uneven ground, up ramps, down ramps, and narrow corridors or other spaces of the building. The 6DSLAM is characterized by allowing the scanning robot device 4 to be moved in six degrees of freedom, thereby allowing a multi-faceted scan (versatilescan) of the surrounding environment. The scan or process data SD can be stored on a removable SSD drive. The stored data can be transferred to other workstations or servers of the cloud platform via a communication interface. Several laser scanners can be provided to capture high-density point clouds, while multiple cameras take high-resolution pictures, which provide immersive 360 ​​° demonstrations (walkthrough). The mobile driving robot 4 will provide a survey-grade point cloud even when moving. Accuracy can even be improved by including SLAM anchor survey markers. SLAM anchors include ground control points that can increase accuracy and can be automatically geo-referenced during data processing. In a possible implementation, the mobile driving robot 4 can be designed for easy transportation in a car trunk or in a luggage compartment.

[0099] In a possible embodiment of the system 1, scan data SD provided by the scanner 2 of the mobile driving robot 4 is processed to automatically detect a building object OBJ. In a possible embodiment, first, an object identifier OBJ-ID is detected. Starting from the detected object identifier OBJ-ID, a scan can be initiated around the detected object identifier OBJ-ID to provide scan data SD, which is processed to detect a building object (such as a smoke detector) near the detected object identifier OBJ-ID.

[0100] The building object detection system 1 and the associated computer-implemented method according to the present invention can be used for a wide range of applications. For example, the building object detection system 1 can be used to verify whether a building includes mandatory building objects OBJ, especially safety-related building objects OBJ or elements (such as smoke detectors). In addition, the difference between the constructed physical building and the building plan of the corresponding building can be automatically detected to trigger corrections or changes in the constructed physical building. In addition, the building object detection system 1 can be used to verify whether the constructed physical building has met all the required installation rules IR at the location of the building. This makes it possible to perform an automatic approval process for approving the constructed physical building in a specific state or district. In addition, in the event of an unwanted event (such as a fire that damages a building), the building operator can prove that all required installation rules IR have been met. In addition, by verifying whether all installation rules IR have been met in the planning and construction of the building, the probability of such unwanted events (such as fires) is significantly reduced. The building object detection system 1 according to the present invention can be used for any building in which building objects OBJ are installed according to the installation rules IR. These buildings can include office buildings, hotels, cruise ships, apartment buildings or houses. Buildings may also include public buildings, in particular transportation buildings such as airports, ship ports or train stations. Buildings may also include tunnels or other infrastructure buildings for transporting goods or personnel. In a possible embodiment, the driving robot 4 may also include a navigation unit linked to a navigation platform, which is adapted to automatically navigate the driving robot 4 through the studied building of interest. The mobile driving robot 4 may be manually controlled by the transported user U and / or remotely controlled by the navigation system 1. In a possible embodiment, the processing of the scan data SD may be performed in real time. For example, if the building object of interest has been detected or classified, the area around the building object of interest may be studied more thoroughly. For example, if the building object of interest has been detected and classified by the building object detection system 1, the mobile driving robot 4 may be navigated to the area of ​​the detected building object OBJ to perform additional scanning by the scanner 2 of the robot 4. Additional scanning may be performed by the same or different scanner devices 2 of the mobile robot 4. After the scan data SD of the detected building object has been analyzed and processed, additional actions may be triggered, such as removing the detected building object from the building. In a possible embodiment, the scan data SD may also be processed to automatically determine the operational status of the detected building objects. For example, an additional scanner 2 may be activated to determine the current operational status of the detected building objects.For example, in the original example, if a lamp mounted on the ceiling of a room in a building is detected, the sensor may provide sensor data indicating whether the detected lamp is turned on or off. In addition, the scan data SD may also indicate whether the installed lamp is flashing. The detected operating state of the detected building object may be notified to the platform for triggering corresponding measures. For example, if the scan data SD indicates that the detected building object (such as a lamp) is faulty or in an undesirable operating state, the detected building object may be repaired or replaced. Therefore, the building object detection system 1 according to the present invention may be linked to a maintenance system 1, thereby improving the maintenance of a plurality of building objects OBJ installed in a building.

[0101] Reference Symbols List

[0102] 1 Object Detection System

[0103] 2 Scanner

[0104] 3 Artificial Intelligence Module

[0105] 4 Driving the Robot

[0106] 4A Platform

[0107] 4B, 4C wheels

[0108] 4D Handle

[0109] 4E Electric Motor

[0110] 5 Processing Unit

[0111] 6 Object Identifier Detection Unit

[0112] 7 Verification Unit

[0113] 8 Database

[0114] 9 Processing Units

Claims

1. A building object detection system (1) for detecting building objects (OBJ) installed at fixed predefined positions in a building, the system (1) comprising: at least one scanner (2) adapted to scan a space within the building to generate scan data (SD) of the corresponding scanned space; at least one artificial intelligence module (3) implemented on a processing unit (5) of the building object detection system (1), adapted to process scan data (SD) generated by the at least one scanner (2) to automatically detect and classify building objects (OBJ) installed in the scanning space, wherein the artificial intelligence module (3) is trained by a processing unit (5) using labeled training data (LTD), the labeled training data (LTD) comprising labels (L) derived from scan data (SD) of previously scanned building objects (OBJ) provided with associated object identifiers (OBJ-ID), the object identifiers (OBJ-ID) being in the vicinity of the respective installed building object (OBJ); a database (8) adapted to store different types of building objects (OBJ) and associated installation rules (IR) for the installation of the building objects (OBJ) within a building; as well as a verification unit (7) adapted to verify the classification of the building object (OBJ) based on installation rules (IR) for the installation of the building object (OBJ) within the building stored in said database (8), Depending on the location of the building, different sets of installation rules (IR) are loaded from the database (8).

2. A system (1) according to claim 1, wherein the at least one scanner (2) is mounted on a remote-controlled or autonomous driving robot (4) of the system (1), wherein the driving robot (4) can move within a building to provide scanning data (SD) of a scanning space within the building to a trained artificial intelligence module (3) implemented on a local processor of the driving robot (4), or on a local controller of a network of the building, or on a remote server of a cloud platform, to detect and classify building objects (OBJ) installed in the scanning space of the building.

3. The system (1) according to claim 2, wherein the scan data (SD) generated by the at least one scanner (2) mounted to the driving robot (4) comprises: Point cloud data, and / or image data, radar data and / or acoustic data.

4. The system (1) according to claim 1 or 2, wherein the trained artificial intelligence module (3) comprises a trained deep neural network.

5. A system (1) according to claim 1 or 2, wherein the trained artificial intelligence module (3) receives scan data (SD) from different scanners (2), wherein the classification of a building object (OBJ) based on the scan data (SD) provided by one scanner (2) is verified by a verification unit (7) of the system (1) based on the scan data (SD) provided by another scanner (2).

6. A system (1) according to claim 1 or 2, wherein the detected and classified building objects (OBJ) installed in the scanned space of the building are stored in a database (8) of the system (1) and processed by a processing unit (5) of the system (1) to generate a digital twin (DT) of the scanned physical building, wherein the digital twin (DT) forms a real-time digital copy of the scanned physical building.

7. A system (1) according to claim 6, wherein the generated digital twin (DT) of the scanned physical building is compared with a digital planned model of the same building stored in the database (8) to automatically detect deviations between the scanned physical building and the planned building.

8. A system (1) according to claim 1 or 2, wherein the scan data (SD) generated by the at least one scanner (2) includes geo-referenced scan data (SD), and the geo-referenced scan data (SD) is processed by a processor of the system (1) to determine the orientation of the detected building object (OBJ).

9. A system (1) according to claim 8, wherein the determined orientation of the detected and classified building objects (OBJ) is compared with building floor plan data of the building to verify the classification of the corresponding building object (OBJ) performed by the trained artificial intelligence module (3).

10. A system according to claim 1 or 2, wherein the building objects (OBJ) detected and classified by the artificial intelligence module (3) are provided with annotation data and stored in a database (8) of the system (1).

11. A system (1) according to claim 2, wherein the scanner (2) mounted on the driving robot (4) is automatically moved to a position within a building indicated by the coordinates of a building object (OBJ) specified in a digital twin (DT) of the building to scan the surrounding environment of the indicated position.

12. The system (1) according to claim 1 or 2, wherein the building object (OBJ) installed in the building and detected by the building object detection system (1) comprises: Security and alarm installation equipment, Communications installation equipment, Artificial lighting, electrical switches and sockets, and power cords, Energy supply installations, heating and cooling installations, ventilation installations, Water supply and drainage installation equipment, Escalator and elevator installation equipment, windows and doors.

13. The system (1) according to claim 1 or 2, wherein the object identifier (OBJ-ID) used to provide labeled training data (LTD) for the artificial intelligence module (3) comprises a hierarchical classification and / or object attributes.

14. A system (1) according to claim 1 or 2, wherein an object identifier (OBJ-ID) placed in the vicinity of the installed building object (OBJ) and used to provide labeled training data (LTD) for the artificial intelligence module (3) provides absolute or relative position information.

15. The system (1) according to claim 12, wherein the safety and alarm installation equipment comprises at least one of a temperature sensor, a fire detection device, a fire protection device, a lightning protection device.

16. The system (1) of claim 12, wherein the communication installation equipment comprises at least one of a telephone socket, a radio socket, and a communication line.

17. A computer-implemented method for detecting building objects (OBJs) installed at fixed predefined locations within a building, the method comprising the steps of: Scan(S1 IPH ) spaces within a building to generate scan data (SD) of the corresponding scanned spaces; and Processed by the artificial intelligence module (3) (S2 IPH ) to automatically detect and classify building objects (OBJ) installed in the scanned space, wherein the artificial intelligence module (3) is trained using labeled training data (LTD) comprising labels (L) derived from scan data (SD) of previously scanned building objects (OBJ) provided with associated object identifiers (OBJ-ID) in the vicinity of the respective installed building objects (OBJ), wherein different types of building objects (OBJ) and associated installation rules (IR) for the installation of the building objects (OBJ) within a building are stored in a database (8), wherein depending on the orientation of the location of the building, different sets of installation rules (IR) are loaded from said database (8), and Therein, the classification of the building object (OBJ) is verified based on installation rules (IR) for the installation of the building object (OBJ) within the building loaded from the database.

18. A method for training an artificial intelligence module (3) implemented in a processing unit (5) of a building object detection system (1) according to any one of the preceding claims 1 to 16, The training method comprises the following steps: Physically attaching an object identifier (OBJ-ID) to (S1 TPH ) in the vicinity of a building object (OBJ) installed in a building; Scan (S2 TPH ) spaces within a building to generate scan data (SD) of the corresponding scanned space; Processing (S3 TPH ) generated by the scan data (SD) to derive a label (L) from an object identifier (OBJ-ID) included in the scan data (SD); The derived labels (L) are used to automatically annotate (S4 TPH ) the scan data (SD) to provide labeled training data (LTD); as well as Using labeled training data (LTD) to train (S5 TPH ) said artificial intelligence module (3).

19. A driving robot (4), comprising: at least one scanner (2) adapted to generate scan data (SD) of a scanned space within a building, The generated scan data (SD) is provided to an artificial intelligence module (3) implemented on a processing unit (5) of a building object detection system (1) according to any one of the preceding claims 1 to 16, for detecting building objects (OBJ) installed in a building.

Images