Structure detection to optimize resource utilization in physical systems

Abstract

The present invention relates to a method for optimizing the use of volume and / or surface area in a generic physical system (e.g., development plans, logistics and storage systems, quantum computers, solid-state systems, electron gases, plasmas), comprising the following steps: identifying a physical system in an n-dimensional space, in particular a three-dimensional space; identifying an (n-1)-dimensional space, in particular a two-dimensional space, suitable for describing an (n-1)-dimensional projection of the physical system in the n-dimensional space; providing n-dimensional point data, in particular lidar data and / or LoD data; providing secondary data comprising (n-1)-dimensional data, in particular polygonal data, in particular a cadastre;identifying one or more subsystems in the physical system, in particular using secondary data, in particular also assigning secondary data to the subsystems;Determining subsystem data from at least two subsystems that describe variables of the respective subsystem, in particular comprising the secondary data of the respective subsystem and n-dimensional point data and / or LoD data relating to the corresponding subsystem, in whole or in part; selecting at least one subsystem, using a supervised machine learning model, in particular a trained artificial neural network and / or linear regression, to predict the properties of the selected subsystem from the secondary data of the selected subsystem and data from other unselected subsystems. The invention also relates to computers, computer programs, computer networks, data and computer-readable data media, as well as their relevant uses.

Description

Field of invention

[0001] The present invention relates generally to the field of physical systems and their optimization and optimal use. Many physical systems can be described by descriptions that exist, for example, in two-, three- or even higher-dimensional space.

[0002] For the progress and improvement of physical systems, it is often desirable to confidently identify the potential for improvement in the systems, to assess their feasibility, and to implement the proposed improvements in reality.

[0003] In one embodiment, the development or use of areas in a physical system can be optimized, leading to improved use of space and volume in the system.

[0004] According to the invention, a ground reference point can be determined more precisely, which increases the process accuracy. Technical background

[0005] Specifications for physical systems, such as laws like natural laws, but also artificial or man-made laws (e.g., building regulations), can potentially restrict the possible or permissible modifications to these systems. Another example is provided by quality assurance regulations, which define, for instance, quality and / or homogeneity requirements for an electronic or mechanical component.

[0006] Several techniques known from the prior art will be briefly outlined below.

[0007] Cadastral data is a well-known example of state-of-the-art technology. Such data, often fragmentary and specialized to specific geographical regions, can be accessed via electronic systems. An example of this is the Official Real Estate Cadastre Information System (ALKIS).

[0008] Such data are usually measured by surveying technicians or publicly appointed surveying engineers, who carry out corresponding geodetic measurements on site using appropriate mobile measuring equipment, which can be brought to the location to be surveyed, set up there and used successively.

[0009] Furthermore, the data are generally limited to highly simplified measurements, particularly given the complex structure of the actual building, and therefore provide only extremely rough insights into the true situation. Substructures such as individual details of individual buildings are frequently ignored. In the most common scenario, surveyors or publicly appointed surveyors only measure building outlines, usually only in the form of two-dimensional data. Sometimes, the number of stories in a building is also given, as determined by the surveyor or publicly appointed surveyor based on the building structure.

[0010] A method is also known from the prior art that essentially provides a survey of the Earth's surface, including the objects located on it. One example of this technique is Light Detection and Ranging (LIDAR). For instance, a LIDAR measurement can be performed as a grid-based laser scan on board an aircraft, measuring the structures located below the aircraft. However, this data often does not go beyond spatial point data ("point clouds"). This is partly due to the fact that LIDAR is essentially based on distance measurements. In particular, LIDAR as such only creates point clouds without interpretation, which significantly hinders further, insightful analysis of LIDAR data.

[0011] Furthermore, point clouds are frequently affected by individual measurement errors (e.g., outliers), which can lead to incorrect conclusions even after a thorough examination of the point cloud data. This results in significant consequential errors, especially in automated data processing.

[0012] Another technique from a different technical field involves artificial neural networks (ANNs). For example, the translation service DeepL is known to use such networks to translate texts from one natural language to another with high accuracy. This is possible, among other things, due to extensive databases, such as the Linguee database, which allow for the appropriate training of the ANN(s). The ANN(s) are precise, but only for the aforementioned purpose.

[0013] The current state of technology has the disadvantage that enormous data potential remains untapped. The fragmentation of data and data sources, along with differing formats, further complicates the identification of potential changes and improvements.

[0014] Similar problems exist in other physical systems, such as solid-state physics (e.g., two-dimensional electron gases) and quantum computers.

[0015] Document EP 3 376 160 A1 discloses a method and system for data processing. The technical result is the improvement of the accuracy of identifying urban objects when creating urban panorama models based on laser scanner and photographic data. The stated objective is the computer-based recognition of urban objects. This consists of processing primary data, which includes at least photographic and laser scanning data, including information on urban locations, and comprises the following steps: cleaning and verification of the obtained primary data containing urban features; preprocessing of the cleaned laser scanning data, including: constructing normals for each point in the cloud; constructing planes for point clouds; and creating reliefs of the terrain and individual objects (e.g., buildings, streets, and buildings).Building facades) using photogrammetric algorithms based on photos taken from different points in the room.

[0016] The present invention is therefore based on the technical problem of providing methods that overcome the disadvantages of the prior art and enable new solutions. In particular, a method for optimizing and identifying improvement potentials in physical systems, especially generic physical systems, is to be created. Exemplary applications of the invention are conceivable in the real estate sector, the logistics sector, control engineering, integrated circuits, and quantum computers.

[0017] The aim is to create more flexible, scalable, transferable, secure, and reliable solutions. Cost minimization also plays a crucial role. The methods should be able to operate on technical system data, particularly measured "raw data." Furthermore, the methods should not only be able to propose potential optimizations of physical systems, for example, through simulation, but also implement these optimizations in reality by modifying the physical systems accordingly. Generic, cross-field applicability, especially for diverse physical systems of various types, should be guaranteed.

[0018] The disadvantages of the prior art are overcome by the method according to claim 1, the computer, the computer program, the computer network and the computer-readable data carrier according to claim 15. Description of the invention

[0019] The present invention is described in the attached set of claims. It provides a method with the features of claim 1. Further advantageous embodiments are specified in the dependent claims.

[0020] A method for optimizing volume and / or area utilization in a physical system is provided, comprising the following steps: Identifying a physical system in an n-dimensional space, in particular three-dimensional space; identifying an (n-1)-dimensional space, in particular two-dimensional space, suitable for describing an (n-1)-dimensional projection of the physical system in the n-dimensional space; providing n-dimensional point data, in particular LiDAR data, and / or 3D models at different levels of detail (LoD data), which in particular describe one or more surfaces; providing secondary data comprising (n-1)-dimensional data, in particular polygon data, in particular from a cadastre, which in particular describe one or more levels; identifying one or more subsystems in the physical system, in particular by means of the secondary data.in particular also including the assignment of secondary data to the subsystems, determining subsystem data of at least two subsystems which describe quantities of the respective subsystem, in particular encompassing the secondary data of the respective subsystem and the n-dimensional point data and / or LoD data of the respective subsystem, in whole or in part, selecting at least one subsystem, using a machine learning model of supervised learning, in particular a trained artificial neural network and / or linear regression, to predict properties of the selected subsystem based on the secondary data of the selected subsystem and the subsystem data of other, unselected subsystems.

[0021] The method is applicable to generic physical systems.

[0022] Physical systems are generic physical systems for whose structure and properties measurement data, i.e., technical data, can be provided.

[0023] Aspects of the invention are partly explained within the framework of an exemplary embodiment for an application in proactive sustainable urban planning, but are not limited to this.

[0024] One step involves identifying a physical system in an n-dimensional space. This can be a three-dimensional space, or it can be a part or section of an n-dimensional space. n is a natural number. In a two-dimensional example with n=2, a two-dimensional electron gas is described. This description can be classical or quantum mechanical. In another example, it is a part or section of a three-dimensional space. In yet another example, this space comprises built-up areas or plots of land. Buildings, including their roofs, can be located within this space. A wide variety of sizes and scales for systems are conceivable. These could include neighborhoods, villages, cities, countries, continents, or even a world map.Other limited structures are also conceivable, such as grid squares or the sum of different properties that are subject to uniform building laws or development plans.

[0025] Other physical systems are also conceivable. It should be emphasized that these physical systems are by no means limited to descriptions in n=3 dimensions. In a three-dimensional example, for instance, various boxes or other structures for receiving and storing goods could be housed in large warehouses or the like. This example is also usually three-dimensional. There are also optimization requirements in this area where the present invention can be applied. For example, individual, very heavy boxes (boxes with high weight due to their contents) should not be placed high on top of other, very light boxes. Light boxes often contain light and also not very rigid contents, and are therefore more at risk than heavy boxes of being damaged or "crushed" or at least "dented" by other heavy boxes.

[0026] The invention can be applied in a wide variety of fields. It is particularly advantageous in areas where certain intangible and therefore not directly and precisely definable requirements are placed on individual elements, especially homogeneity requirements of the elements in relation to their surrounding environments. Such applications include systems engineering, integrated circuits, data storage, and quantum computing.

[0027] The next step involves identifying an (n-1)-dimensional space, in particular a two-dimensional space, which is suitable for describing an (n-1)-dimensional projection of the physical system in n-dimensional space.

[0028] For example, this is a two-dimensional surface like a piece of land, as it might be represented on a simple map. Buildings can be represented in this example by projection onto the two-dimensional map. Thus, buildings here typically have a two-dimensional outline. Information is often lost in this projection, especially regarding the building's height, roof shape, and so on.

[0029] In one example, we are dealing with a two-dimensional projection of a building, a property, a neighborhood, a village, a city, or a country. Several of these can also be contained within the two-dimensional space. For example, this could be the city of Münster, including its surrounding districts, or it could be the state of North Rhine-Westphalia, Germany, or Europe. However, it could also simply be a small new development area or the area of ​​a construction project.

[0030] For both n-dimensional and (n-1)-dimensional space, appropriate data can be obtained, for example, retrieved from one or more databases. In particular, this data can also be measured within the scope of the invention. Combining different data sources and data is also possible. For example, self-measured data can be combined with data available in databases in a way that is both advantageous and reduces errors.

[0031] In one example, n-dimensional point data can be used.

[0032] Three-dimensional point data is particularly suitable for a three-dimensional space with various properties and buildings. LiDAR (Light Detection and Ranging) data is especially well-suited. Therefore, a LiDAR measurement can also be used as an alternative or in combination within the scope of the invention.

[0033] Existing LiDAR point data can be obtained or retrieved from state authorities in Germany, for example. One suitable format for such data is LAS, but other point cloud formats can also be used. In one example, the data is compressed, e.g., LAZ, which saves technical resources, especially during transmission and / or direct retrieval. This also saves time and allows results to be made available more quickly. Point data can be partially assigned to specific subsystems of a physical system. Such an assignment is based on technical considerations and / or knowledge of the technical structure and substructure of the physical system. Various options and further training will be explained in more detail later.

[0034] In an alternative according to the invention, so-called LoD data is used. Level of Detail (LoD) describes 3D models in which data is available at different levels of detail. This has the advantage that the level of detail for a calculation is freely selectable and remains variable. Thus, a suitable LoD level can be chosen for each case, enabling adequate calculations or representations.

[0035] These are, simply put, easy-to-use "block models" that can adequately describe buildings, for example. The data is reduced to the essentials and is available at varying levels of detail. These models are often generated from LiDAR data and therefore generally have a lower level of accuracy than the LiDAR data itself.

[0036] LoD model-based data is particularly easy to handle and results in short query processing times, as these are less computationally intensive due to the LoD models.

[0037] As a further option, LoD data can be derived, i.e., generated, from the LiDAR data within the scope of the invention. This results in a further intermediate step. Subsequently, calculations can then be performed based on the original LiDAR data and / or the LoD data derived from it.

[0038] Secondary data, including (n-1)-dimensional data, particularly polygon data, especially from a cadastral register, is being provided. This includes, for example, two-dimensional data from a cadastral register. Such data can be obtained, for instance, from an official cadastral information system (ALKIS). For example, the shape and location of properties and / or parcels can be extracted from the ALKIS data. Often, only data on parcels are available, but the property data can sometimes be reconstructed through skillful recombination of the parcels. Alternatively, this recombination can be performed using user input. For the purposes of this application, "property" and "parcel" are essentially considered synonyms unless the distinction is explicitly emphasized. In many cases, a property corresponds to a parcel, i.e.,The property comprises only one parcel, for which ALKIS data provides information. Outlines of buildings, for example, can also be extracted from the ALKIS data. Furthermore, additional data or metadata (often, for example, information on the number of stories or the use or type of use) may be included.

[0039] Information on other sealed surfaces can, in principle, also be contained in such a dataset and then extracted from it. However, this is generally not the case, which can be remedied by various further developments of the present invention. This will be discussed in more detail later. Terraces and parking lots are examples of such examples.

[0040] In the case of a two-dimensional electron gas, the secondary data can include scalars.

[0041] In one example, the data, or a portion thereof, is based on a standards-based exchange interface for geoinformation (NAS). One example format is the GML (Geography Markup Language) format. Using such a standardized format increases interoperability with existing systems and solutions.

[0042] In the case of a system that exists on four dimensions (e.g., Minkowski spacetime), the secondary data comprise three-dimensional data. For example, this secondary data corresponds to the spatial part of an observer's spatiotemporal reference frame.

[0043] One or more subsystems within the physical system are identified, particularly using secondary data. For example, in one embodiment, the system can be divided into plots of land using ALKIS data. The secondary data corresponding to a plot of land and its buildings can then be assigned to the respective subsystem.

[0044] Subsystem data can then be generated for subsystems. Such subsystem data is assigned to the subsystems and can describe parameters of the respective subsystem. This subsystem data can include the aforementioned secondary data. However, it can also contain the corresponding n-dimensional point or LiDAR data, for example, over the subsystem's property (e.g., obtained by appropriate data "intersection"), or parts of this LiDAR data. Furthermore, the subsystem data can contain additional derived parameters from other subsystem data. A ground reference point is one example. Another example could be a ridge and / or eaves height. Within the scope of the invention, the subsystem data can thus be extended as desired by further calculated or derivative parameters for the subsystem. The subsystem data can be used to train and utilize an artificial neural network.This will be explained in more detail later.

[0045] Secondary data can also be intersected with point data, such as LiDAR point data. This is explained below. For example, cadastral data describes two-dimensional polygons for parcels, properties, and / or buildings on those properties (e.g., in the form of the outlines of the area covered by the building). To determine certain dimensions of a house as accurately as possible from the LiDAR data, LiDAR points can be selected that, when projected onto the two-dimensional plane, lie above the property ("intersecting" the data). This increases the accuracy of the derived measurements because sources of error can be eliminated and thus avoided. For example, a tree near the house will not be taken into account when determining the ridge and eaves height of the roof.

[0046] In another example, this increases accuracy from a legal perspective. In many areas, building regulations may define the eaves height of a roof in such a way that, even with an overhanging roof, it must be measured precisely at the edge of the building plot. Since the intersection process cuts off any roof overhanging the property line, the accuracy of determining the dimension, which is considered particularly relevant under building regulations, is increased. The relevant, intersected points can be specially marked in the data, thus eliminating the need for remeasurement. This technique is suitable for all "intersection processes" within the scope of the present invention and its further developments. This saves computing time, as the relevant data points are immediately available for further calculations thanks to the attached flags.

[0047] Important properties, measurements, and derivatives, as well as dimensions relevant to building regulations, will be discussed in detail later. At this point, we will only refer to the ridge and eaves heights of a building's roof as examples. The combined use of cadastral and LiDAR data allows for precise determination of any building. Cadastral and LiDAR data are readily available with good coverage and accuracy, some of it commercially available and some even free of charge.

[0048] The example described above works particularly well when the building outlines are known very precisely, for example from ALKIS data. This method is therefore the preferred method of choice for Germany, where building outlines have been measured very precisely using ALKIS data.

[0049] At this point, further alternative approaches within the scope of the present invention should be mentioned. The point data can be, but need not be, intersected with, for example, the building outline. All point data across the entire property can also be considered, appropriately in each case. Even point data from neighboring properties can be included.

[0050] In another example, a specially trained machine learning model is used to determine the buildings, building outlines, and / or derivative parameters such as ridge and / or eaves height as precisely as possible from the point data. This approach is particularly precise and also reliably handles outlier cases such as the peculiarities of unconventional buildings. For example, an artificial neural network is used as the machine learning model.

[0051] In one example, the machine learning model is trained using data from Germany. In Germany, building outline data from the ALKIS (Official Cadastral Surveying and Information System) is available with comparatively high precision. The trained machine learning model can then be effectively used in regions abroad where building outlines are not as precisely known or measured, and therefore not as accurately available in cadastral data. This not only increases overall precision, but also allows the machine learning model to be reliably extended to geographical regions where sufficiently precise data is unavailable.

[0052] Within the scope of the present invention, LIDAR measurements specifically adapted to the purpose of the present invention can also be carried out in a separate manner. This allows for specific adaptations such as a higher raster / scan resolution in cities or densely populated areas.

[0053] For the numerous subsystems in a suitable example, subsystem data should be generated. This includes, for example, the important properties, measured variables, and other derivative properties mentioned above. The subsystem data should contain technically interesting and / or, for example, legally relevant data about the subsystems. In another example concerning quality assurance of integrated circuits, this could include specific technical requirements and parameters that a component must meet to pass the quality assurance process.

[0054] A subsystem is then selected. There could also be several, but for the sake of simplicity, the following discussion will focus on the case of a single selected subsystem.

[0055] This subsystem is now to be examined for its potential for optimization and / or expansion / improved usability. For this purpose, a specially trained machine learning model, in particular an artificial neural network, is used according to the invention. This is done on the basis of the secondary data of the selected subsystem on the one hand and the subsystem data of, in particular, other, especially unselected, subsystems on the other. This data can, for example, represent the input variables that are applied to corresponding input neurons of the artificial neural network. Deep neural networks are particularly suitable as artificial neural networks, but other techniques in the field of artificial intelligence and machine learning can also be used.

[0056] The secondary data of the selected subsystem includes, for example, cadastral data for the land parcel and / or plot. It may also contain information on buildings. In one example, derived quantities may also be included, such as a floor area. In this example, the secondary data mentioned here do not include any cadastral data on a building that may already exist on a property. This is particularly beneficial for a potential analysis of the subsystem, as property characteristics are difficult to change, whereas building characteristics can be modified when potential for improvement is identified. Therefore, for example, it is advantageous to include property characteristics when predicting the characteristics of buildings that could potentially be constructed on a property, but the analysis should not be based on the "status quo," i.e., the existing state of the property.to distort the current use of the area by, for example, buildings.

[0057] In one example, a plot of land in an area is occupied by a vacant, dilapidated building (an extreme example). This building could, however, be demolished, i.e., removed from the property, if necessary or beneficial to progress, in order to make room for something new. If the machine learning model, particularly the artificial neural network, makes a suggestion or prediction that, for example, indicates a high utilization rate or possibly other advantageous characteristics, implementing this suggestion or prediction can be considered quite sensible and structurally beneficial.

[0058] The subsystem data of other subsystems, particularly those not selected, are relevant for the artificial neural network to correctly capture specifications, especially environmental or contextual requirements. Specifically, the artificial neural network uses this subsystem data as features / influencing factors to capture various relevance and requirements that are otherwise difficult to capture, such as those related to homogeneity. For example, building regulations might require a "homogeneous appearance," or a new building might need to match the height of the surrounding buildings—at least approximately.

[0059] In another example, various transistors are incorporated as subsystems into an integrated circuit. Certain parameters, such as voltages, etc., may be subject to certain homogeneity requirements for ideally functioning circuits.

[0060] Furthermore, quantum computers comprising multiple qubits are in advanced stages of development. In many examples, these qubits interact both with each other through quantum entanglement and with external sources such as fields. Therefore, for an ideally functioning and maximally utilized quantum computer, it is essential that the qubits, which here represent the subsystems or are incorporated within the subsystems, and especially neighboring qubits, whether directly adjacent or more distantly adjacent, are designed to be as homogeneous as possible. According to the invention, this technical effect can be achieved by using the subsystem data of the surrounding qubits when predicting a selected qubit.

[0061] Feature engineering and the associated definition of the potentially relevant features with which the artificial neural network operates have a significant impact on the quality of the results obtained. These aspects will be discussed in detail in further developments of the present invention.

[0062] In many applications of the invention, secondary data includes features that describe environmental and / or spatial factors which are difficult to change. These are often factors determined by external circumstances or are fixed. Other factors may also be included that can be considered highly homogeneous to a reasonable approximation, such as temperature, air pressure, or humidity. Furthermore, in many cases, the secondary data does not include features that are treated as variables in the analysis of improvement potential. Excluding these features has the advantage that they can be predicted more precisely by the artificial neural network without the influence of interfering factors.

[0063] In many applications of the invention, subsystem data includes features that are treated as variables in the analysis of improvement potential. Considering these features within the subsystem data has the advantage that it enables the artificial neural network to precisely predict these or similar features for the selected subsystem.

[0064] The invention combines, in particular, secondary data of the selected subsystem(s) with subsystem data from other subsystems. Specifically, the invention combines secondary data of the selected subsystem(s) with subsystem data from other subsystems that are not selected. This deliberate choice for feature engineering enables the achievement of precise and reliable results that identify potential improvements to the system and facilitate their implementation.

[0065] The combination of the aforementioned system data thus unlocks new, previously untouched potentials for system improvement, while respecting external system requirements, in particular local or semi-local homogeneity requirements.

[0066] The machine learning model itself provides lossy compression of the input and training data.

[0067] According to further training, the machine learning model or the artificial neural network is and / or was trained using a selected subset of existing subsystems, in particular by means of supervised learning of the artificial neural network.

[0068] In one example, this selection is made manually, for instance by an employee or user. For example, appropriate [options / methods] are / were used for this purpose.

[0069] Subsystems, for example, within a specific region, are selected that already exhibit certain high levels of building utilization (and are therefore good examples of efficient land and volume use). In this example, such conditions are desirable. Therefore, such desirable or advanced cases of subsystems should be preferentially selected. By training the artificial neural network through supervised learning with these selected subsystems, the network's suggestions / predictions also tend to be oriented towards high building utilization. This makes the suggestions particularly useful for improving the physical system, and the network converges more quickly to viable solutions. Unnecessary network training, as well as "incorrect" or unusable training, is avoided. This pre-selection gives the overall system an additional "offset."a "drive" that translates the dynamics towards continuous improvement. This avoids simply replacing subsystems with very similar subsystems without significant improvement.

[0070] According to further training, subsystems are selected in particular which show high usage in the secondary and / or subsystem data, especially volume and area usage, characterized in particular by exceeding and / or falling below threshold values ​​with regard to the secondary and / or subsystem data.

[0071] Such thresholds are an easy and inexpensive way to implement the "drive" that triggers the desired time-dynamic system development. They can be absolute or relative thresholds. A relative threshold, for example, can be defined relative to a mean value such as the arithmetic mean or the median.

[0072] More complex systems such as decision trees can also be used to define the relevant subsystems and select them.

[0073] According to further training, the procedure also includes a step of providing photogrammetry data relating to the n- and / or (n-1)-dimensional space, and deriving at least one quantity from the photogrammetry data, in particular in combination with the secondary data and / or subsystem data.

[0074] The inclusion of photogrammetry data makes it possible to determine the actual conditions of a subsystem, especially based on LIDAR point cloud data, in a more precise manner.

[0075] For various building dimensions, such as the ridge and eaves height of a roof, a precise determination of a ground reference point for the subsystem (e.g., a "height above sea level") may be necessary or useful. In a simple implementation, cadastral data, which includes a property and one or more buildings on it, are overlaid with LiDAR data. For example, a ground reference point can then be determined by calculating an average value over the reference points not located within the building's footprint. Alternatively, an existing pre-classification in the LiDAR data can be used, e.g., based on all points with a "Ground" flag in the LiDAR data. However, it has been shown that a particularly precise determination is only achieved through combination with photogrammetry.Based on photogrammetry and corresponding selection and processing of the LIDAR points, a higher precision is achieved than with other methods for determining the ground reference point.

[0076] The determination of other reference points, reference values, derivative quantities, and dimensions of building use can also be accomplished using photogrammetry data. For example, the size or actual proportion of sealed surfaces to the total area can be determined more precisely, or further structural statements about the nature, use, type of use, etc., of the subsystem or parts thereof can be made. Correspondingly, further classification of subsystems based on these defined characteristics is enabled. The use of photogrammetry data can also influence the calculation of complex quantities. For example, it can be determined whether an attic has been converted or is in use.

[0077] The combination of data with photogrammetry data also proves highly advantageous for other system types. For example, in micro- and nanoelectronics, this yields further valuable insights into the existing structures, particularly through combined, synergistic analysis together with the other data already mentioned.

[0078] According to the present invention, the method further comprises a step of detecting at least one sealed surface and its dimensions based on photogrammetry data. The result is further refined by detecting and successively taking into account existing sealed surfaces. In building regulations and development plans, the proportion of sealed surfaces to the total area is often of great importance. This ensures compliance with legal requirements. It should be noted that these legal requirements are not purely abstract or legal in nature, but rather serve to enable technical effects (including the avoidance of undesirable technical effects). The proportion of sealed surfaces is relevant for the overall system, for example, in the case of heavy rainfall, which should flow into the ground ("seep into the ground") without disturbing, destroying, or negatively impacting the overall system.Flooding poses a major problem for people and obviously severely disrupts the sustainability of urban development. According to the training, this approach allows the potential of building use to be further exploited without endangering the physical system and its intended function and use. For example, an impermissible exceedance of critical limits is thus avoided in advance.

[0079] According to the present invention, the method further comprises a step of determining a ground reference point by recognizing an object in the photogrammetry data that is suitable to serve as a ground reference point, in particular manhole and sewer covers. There can also be several objects.

[0080] This allows for a more precise determination of the ground reference point, one that more closely reflects the actual conditions within the physical system and is to be determined from the measurement data. A more precise determination of the ground reference point can also significantly impact derivative dimensions such as ridge height, eaves height, and number of stories. Furthermore, it enables compliance with different definitions of ground reference points (e.g., those relevant to building regulations and requiring consideration). For example, if a ground reference point is defined as a manhole cover for a sewer, this can be measured or determined from the measurement data. Similarly, if it is the height of a curb or curb edge, this dimension can also be determined using the photogrammetry data.

[0081] According to further training, the procedure also includes a step of intersecting the dimensions of the detected object with the n-dimensional point data to generate intersection point data. For example, only LIDAR point data in the relevant area are used to determine a reference point, such as the points above or within the area of ​​the manhole or sewer cover (a "gully").

[0082] According to further training, the procedure also includes a step of averaging the intersection point data to determine the ground reference point. Averaging a well-chosen set of data points further reduces the error and leads to a highly precise determination of the ground reference point and all derivative quantities dependent on it.

[0083] According to further training, the procedure also includes a step of merging n-dimensional point data with a profile defined by the secondary data.

[0084] The building component of a subsystem for generating intersection point data includes a step of marking the points in the intersection point data that fall within the area of ​​the building component, within the n-dimensional point data. This marks all points that lie within or above the building, as defined in the cadastral map. This technique has numerous advantages.

[0085] To determine specific dimensions of a house as accurately as possible from LiDAR data, LiDAR points can be selected that, when projected onto a two-dimensional plane, lie above the property. This increases the accuracy of the derived measurements because sources of error can be filtered out and thus avoided. For example, a tree near the house is not taken into account when determining the ridge and eaves height of the roof.

[0086] In another example, this increases accuracy from a legal perspective. In many areas, building regulations may define the eaves height of a roof in such a way that, even with an overhanging roof, it must be measured precisely at the edge of the building plot. Since the intersecting process cuts off any roof overhanging the property line, the accuracy of determining the dimension, which is considered particularly relevant under building regulations, is increased. The relevant, intersected points can be specially marked in the data, thus eliminating the need for remeasurement. This saves processing time, as the relevant data points are immediately available for further calculations thanks to the attached flags.

[0087] From the point data, especially the intersected point data, normal vectors and curvature can be determined. These are helpful parameters for a multitude of subsequent calculations, which is why having these quantities readily available saves computing time and resources. Furthermore, normal directions and angles, whether in discrete or continuous form, as in the case of a curved roof, are architecturally relevant parameters.

[0088] The example described above works particularly well when the building outlines are known very precisely, for example from ALKIS data. This method is therefore the preferred method of choice for Germany, where building outlines have been measured very precisely using ALKIS data.

[0089] Further applications of the invention's development should be discussed at this point. As already discussed above, another example involves using a specially trained machine learning model to determine, as precisely as possible, the buildings, building outlines, and / or derivative parameters such as ridge and / or eaves height from the point data. This approach is particularly precise and also reliably covers outlier cases such as the peculiarities of unconventional buildings. For example, an artificial neural network is used as the machine learning model.

[0090] In one example, the machine learning model is trained using data from Germany. In Germany, building outline data from the ALKIS (Official Cadastral Surveying and Information System) is available with comparatively high precision. The trained machine learning model can then be effectively used in regions abroad where building outlines are not as precisely known or measured, and therefore not as accurately available in cadastral data. This not only increases overall precision, but also allows the machine learning model to be reliably extended to geographical regions where sufficiently precise data is unavailable.

[0091] The described training creates valuable training data that enables the training of the aforementioned machine learning model and thus makes it possible to achieve the aforementioned technical effects.

[0092] According to further training, a classification and marking of points as roof points takes place on the set of intersection point data, in particular clustering and / or unsupervised learning, especially on the basis of the Euclidean distance of the points to each other as a relevant measure and / or on the basis of the vertical distance, especially the vertical or z-component, of the points to each other as a relevant measure.

[0093] This approach reduces the influence of parts of the point cloud that do not belong to the actual roof. For example, a chimney or something above the roof can be excluded.

[0094] The overhanging crown of a tree is "clustered away" and thus rendered harmless for determining the roof. In one example, dormer windows are also "clustered away," while in another example, such dormers are deliberately retained as part of the roof.

[0095] For example, the Euclidean distance (or its square, or another equivalent quantity dependent on the Euclidean distance) between two points in the point cloud is used as a measure for clustering. This is a spatially isotropic measure and thus has the advantage that the results are rotationally invariant. Normal vectors are also well suited for clustering.

[0096] It has been shown that, with regard to a clear separation of the roof from components of the point cloud that do not belong to the roof in the strict sense, the vertical distance provides a particularly suitable measure to consider. Clustering is therefore performed according to the z-component, in the vertical direction, or according to the distance from the ground (or an equivalent quantity). The results obtained in this way are particularly successful and precise. Due to this surprising technical effect, clustering according to the z-component represents a preferred embodiment of the invention.

[0097] Clustering by normal vector, Euclidean distance, and vertical distance is even more precise when these parameters are used simultaneously in a single clustering procedure. However, they can also complement each other by employing several subsequent clustering procedures, for example, first by Euclidean distance (possibly taking the normal vectors into account) and then by vertical distance (also possibly taking the normal vectors into account). This results in particularly precise results, as smaller structures (e.g., a chimney, roof shafts, air conditioning units, satellite dishes, etc.) may still remain in the dataset identified as a roof after the first clustering process, but can then be reliably removed in a subsequent clustering step (e.g., by z-component).

[0098] The roof identified in this way allows precise dimensions to be determined in relation to the roof, the building and its use.

[0099] For example, the volume of a building can be calculated. Technically, this is achieved particularly advantageously through the numerical integration of the z-components of the point cloud over the relevant area. This result is especially precise and simultaneously avoids the need to make unnecessary model assumptions, such as about the roof shape. Furthermore, this method benefits from a previously determined ground reference point that is exceptionally precise and accurate. This synergistic effect significantly reduces errors and error propagation, and the determination of quantities, such as the volume, is performed with even greater precision.

[0100] Another quantity that can be precisely calculated in this way is the floor space, also known as gross floor area. This is also helpful for correctly calculating the gross floor area (GFA). A particularly advantageous numerical method for calculating the floor space involves numerical integration over the built-up area, divided by the story height. For example, the story height may be known or estimated, especially depending on the type of use (e.g., residential or commercial).

[0101] The result obtained through numerical integration is particularly precise and simultaneously avoids the need to make unnecessary model assumptions, e.g., about roof shape. Furthermore, this method benefits from a previously determined ground reference point that is exceptionally precise and accurate.

[0102] It is also possible to determine the number of usable floors. A method that has proven particularly suitable involves creating cross-sections through the building or its point cloud. The spacing of these cross-sections is defined, for example, by a floor height. A floor height may be known or estimated, especially depending on the type of use (e.g., residential or commercial). Cross-sections with a cross-sectional area above a certain threshold (minimum floor area) are then counted as floors when determining the number of usable floors. For example, such a threshold could be set at five square meters. This helps to avoid miscalculations in borderline areas (just before the next floor).Furthermore, it further enhances the avoidance of errors caused by structures such as chimneys, unless these have already been successfully removed from the roof or roof point cloud, i.e., "clustered away." Such attic spaces also only become economically viable above a certain size (e.g., when renting them out as rooms or apartments is feasible).

[0103] The "roof detection" method discussed here can also take normal vectors and curvatures into account. This makes the result even more precise. For example, roofs are generally only rarely curved.

[0104] According to further training, the procedure also includes a step of classifying and marking points as roof points based on a trained second machine learning model, in particular a second artificial neural network.

[0105] It has been shown that using a machine learning model can be superior to other strategies for this step as well. This further increases precision.

[0106] As explained elsewhere, this can be particularly beneficial if, for example, the outlines of buildings in a geographical region of interest have not been accurately measured and are therefore not precisely known.

[0107] According to further training in this context, the second machine learning model is and / or was trained by a procedure comprising a step of intersecting n-dimensional point data with a building part of a subsystem defined by secondary data to generate intersection point data.

[0108] This increases precision. For example, data from a geographical region where building outlines have been precisely measured (e.g., Germany) can be used with the overlay technique to generate intersection point data. This data can then be used to train the (second) machine learning model so that it can independently perform the classification according to the invention in other regions as well, without relying on data of precise building outlines, which are not always available accurately across many relevant geographical regions.

[0109] According to further training, a classification and marking of points as roof points takes place on the set of intersection point data, in particular clustering and / or unsupervised learning, especially on the basis of the Euclidean distance of the points to each other as a relevant measure and / or on the basis of the vertical distance, in particular the vertical or z-component, of the points to each other as a relevant measure, and / or on the basis of normal vectors.

[0110] For the advantages of these additional features, please refer to the discussion above. The use of this technique for generating high-quality training data to train the second machine learning model has also proven highly beneficial in practice. The improved and more accurate identification of roofs using this clustering technique, whose advantages have already been described above, can thus successively improve the quality of the training data, in turn the quality of the training of the second machine learning model, and ultimately the classification results for roof points when the second machine learning model is used as intended.

[0111] The "roof detection" method discussed here can also take curvatures into account. This makes the result even more precise. For example, roofs are generally only rarely curved.

[0112] According to a training course, an incoming feature vector of the artificial neural network comprises, on the one hand, property-related secondary data of the selected subsystem, and on the other hand, subsystem data from other, unselected subsystems. Subsystems in the immediate vicinity are regularly of particular relevance here.

[0113] For example, but not exclusively, such property-related secondary data includes the area, location, shape, dimensions or use (e.g. according to the land use plan) of a property.

[0114] For example, but not exclusively, such subsystem data from other, unselected subsystems include data such as site coverage ratio, floor area ratio, ridge height, eaves height, volume, number of usable floors, roof angle, area, location, shape or dimensions of a property, number of full stories, type of use (e.g. according to the land use plan) and much more.

[0115] All dimensions relating to land and / or buildings mentioned in the present patent document may potentially be included in the subsystem data.

[0116] This allows the artificial neural network to be trained with appropriate data to make suitable predictions or suggestions.

[0117] Put simply, the building on the property under consideration (see selected subsystem) is simply not included in the calculation. The artificial neural network can then be used to generate a suggestion or prediction. In doing so, it takes into account all relevant "environmental factors." These include, on the one hand, properties of the property (see secondary data of the selected subsystem). On the other hand, they include the properties and buildings in the surrounding area, particularly the immediate vicinity (subsystem data of the other subsystems). This structuring with respect to input and output variables allows for particularly efficient and targeted use of the artificial neural network and especially effective control of the dynamic improvement that can ultimately be implemented in the system according to the invention.For example, through repeated use of the invention, volume and area resources in a settlement or agglomeration are successively and increasingly better utilized. System homogeneity is maintained in the process – and in the case of building regulations, these can be automatically taken into account and thus complied with.

[0118] According to a training course, building-related secondary data of the selected subsystem are predicted by the artificial neural network.

[0119] It has been shown that the artificial neural network is suitable for making appropriate suggestions that exploit the potential – in real-world cases – significantly better than other methods. The artificial neural network can "think" in extremely high-dimensional feature spaces, thereby finding and taking into account correlations within these spaces that are difficult to capture through manual or otherwise automated approaches and therefore remain unaddressed. Comparative studies as well as studies on real-world samples, such as existing and especially recently constructed buildings, have been conducted.For these buildings, for example, enormous investments were made and numerous professionals / experts were employed for extended periods to ensure that a solution was found that met external requirements and specifications while simultaneously identifying and fully utilizing existing potential. The artificial neural network was able to reproduce many of the results achieved by these professionals. In some cases, the AI's invention significantly surpassed the results of the professionals. It can therefore be concluded that, in important cases, AI outperforms both manual system optimization and other computer-automated system improvements.

[0120] According to further training, the procedure also includes a step of measuring at least one point datum using LIDAR, in particular by means of an airworthy measuring device.

[0121] The data can therefore also be measured and recorded electronically within the scope of the invention itself. This has the advantage that recording parameters – for example, resolutions or repetition rates – can be adapted to the intended purpose. Self-measured data can also be combined with existing data, for example, to increase accuracy. The newly measured data is also more current and therefore represents the actual state of the physical system more accurately compared to, for example, official data, which may be several years old.

[0122] According to further training, the procedure also includes a step of identifying improvement potentials by comparing key figures of the selected subsystem in its existing form with key figures of the selected subsystem in the form proposed by the Artificial Neural Network, in particular by comparing measures of building use.

[0123] This allows potential to be clearly identified and even "searched for." Comparing actual and target values ​​enables a meaningful quantification of the potential hidden behind the optimization of subsystems. Therefore, the invention can also be used or considered as an invention in control engineering.

[0124] According to further training, the procedure also includes a step of graphically representing a map, which graphically indicates (visually) perceptible improvement potentials for two or more subsystems to the user.

[0125] Such a presentation is very helpful and clear for the user. The effect that a graphical representation of the improvement potential on a map or system overview influences users' perception, processing, and reaction to the results has been confirmed by a large number of our users from diverse user groups.

[0126] Various concrete types of graphical representations of the subsystems and their associated improvement potentials are conceivable. The invention does not intend to be limited to a single specific type of graphical representation. It is irrelevant whether bar charts, pie charts, or other types of representation are used. For example, a user can choose from various possible types of graphical representation one that suits their personal taste or is particularly intuitive and easy to understand.

[0127] According to further training, the procedure also includes a step of adapting the physical system by changing the selected subsystem in the form proposed by the Artificial Neural Network, in particular by constructing, converting or rebuilding one or more building structures in the selected subsystem.

[0128] This allows the system to be gradually improved and, while adhering to requirements, its further development to be sustainably managed. Recognizing potential improvements and implementing them gives our users enormous competitive advantages and ultimately financial benefits compared to parties who lack the technical analysis and information for system improvement.

[0129] According to further training, the procedure also includes a step of clustering subsystems into related clusters and marking the subsystems accordingly with regard to their cluster membership, in particular using an AI technique of unsupervised learning.

[0130] Within subsystem clusters, the latter must be given particular consideration in systems with homogeneity requirements. Clusters may therefore, for example,

[0131] Differences may exist between clusters, while the application of a homogeneity rule within a cluster can be more pronounced without jeopardizing the correct function and order of the system. Often, certain clusters of buildings, such as groups of buildings, are subject to the same or similar building regulations, and the inclusion of an individual building in its corresponding cluster, but not in other clusters, is required or desired. This can be particularly relevant at cluster boundaries. For example, a residential area may border a commercial area. For houses in this border area, the immediate neighbors are not necessarily relevant if they belong to a different cluster. Cluster affiliation is therefore particularly important here. A similar principle applies, for example, to components or parts of components in micro- or nanoelectronics.

[0132] According to further training, the subsystem data includes one or more of the following: ridge height, eaves height, roof height, roof pitch, site coverage ratio (GRZ), site coverage ratio GRZ1, site coverage ratio GRZ2, floor area ratio (GFZ), floor area, volume, number of usable stories, number of full stories, presence and extent of sealed surfaces, ground reference point, region, city and / or district in which the subsystem is located, building density of the nearest neighbor, building density of the next nearest (second nearest, third nearest, etc.) neighbor, a Boolean statement as to whether the nearest neighbor is located on the same connection, in particular the same street, as the selected subsystem, a Boolean statement as to whether the next nearest neighbor is located on the same connection, in particular the same street, as the selected subsystem, a Boolean statement as to whetherwhether the nearest neighbor has an open or closed building style, a Boolean statement regarding whether the nearest neighbor has an open or closed building style, a Boolean statement regarding whether the nearest neighbor has a building in the yard or on the street, a Boolean statement regarding whether the nearest neighbor has a building in the yard or on the street, a length specification that determines the distance of the nearest neighbor's building to the street, a length and / or depth specification that determines how far a house is built into the property in relation to the street, relevant building area, relevant development plans, land use plan, usage regulations according to legal requirements, in particular the Building Use Ordinance, cluster affiliation of the subsystem to a cluster of subsystems.

[0133] These are merely examples, but ones that have proven particularly relevant for the application of the present invention to buildings. This allows for a particularly precise prediction of the properties of buildings to be constructed.

[0134] m denotes a natural number. It is otherwise freely selectable. Thus, all neighbors, for m of any size in principle, can be considered (each in combinations of all their properties or parts thereof), or not.

[0135] For example, individual subsets of the following properties can be considered for a finite number of neighbors: ridge height, eaves height, roof height, roof angle, ground coverage ratio (GRZ), ground coverage ratio GRZ1, ground coverage ratio GRZ2, floor area ratio (GFZ), floor area, volume, number of usable floors, number of full floors, presence and extent of sealed surfaces, ground reference point, region, city and / or district, and much more.

[0136] One measure of building utilization could be, for example: GRZ or GFZ.

[0137] Statements concerning relational data (which include at least one property describing a relation between the selected subsystem and other subsystems, especially unselected subsystems) can be considered for a finite number of neighbors, or not. In practice, these relations can be arbitrarily complex. As a simple example of such a relation, consider the statement describing whether the selected subsystem and a specific m-nearest neighbor are "on the same street" (yes or no).

[0138] It should be emphasized once again that arbitrarily complex derivative quantities are possible here based on the relation between selected and non-selected subsystems.

[0139] The relations can also refer to property-related (i.e., secondary data-related) properties of the subsystems.

[0140] According to further training, the step of using the machine learning model, in particular the artificial neural network, is further based on relational data, which includes at least one property that describes a relation between the selected subsystem and other subsystems, in particular non-selected subsystems.

[0141] Including relational data makes the calculation even more precise, as homogeneity requirements can be better addressed. For example, it's possible to consider whether a selected subsystem and another subsystem are located on the same street or belong to the same housing estate. For detailed descriptions, please refer to the information above.

[0142] According to further training, the procedure also includes a step of generating LoD data based on the n-dimensional point data, in particular LIDAR data, wherein the step of determining subsystem data is based on the generated LoD data.

[0143] This allows for the efficient generation of Level of Detail (LoD) data. This reduces data complexity and simplifies further data processing. The different levels of detail enable easy access to macroscopic dimensions at lower levels, while higher levels ensure greater detail.

[0144] One example uses the CityGML format. This is flexible and increases compatibility and interoperability between program modules as well as with external interfaces.

[0145] The invention enables the efficient densification of living space (as well as other used areas or volumes). This also avoids the need for new land sealing, since existing areas can be densified and / or repurposed using the present invention before new sealing (e.g., by authorities) is considered.

[0146] Since densification usually takes place on already sealed or constructed buildings, this technology counteracts new sealing and thus makes an important contribution to ecological urban development. The invention also reduces climate-damaging CO2 emissions. The climate can therefore be positively and sustainably influenced, especially if enough stakeholders utilize this invention.

[0147] Furthermore, lightweight construction materials such as solid wood structures can be used in all types of building extensions to generate more living space on a smaller footprint. This invention thus promotes a responsible approach to urban densification.

[0148] The invention also provides computers, computer programs, computer networks, data and computer-readable data carriers, as well as a method for training an artificial neural network according to the invention.

[0149] All features disclosed in connection with the corresponding processes can be used in connection with the devices and computer programs, and vice versa.

[0150] Although some aspects have been described within the context of a device, it is clear that these aspects also constitute a description of the corresponding process, where a block or device corresponds to a process step or a function of a process step. Similarly, aspects described within the context of a process step also constitute a description of a corresponding block or element or property of a corresponding device.

[0151] Exemplary embodiments of the invention can be implemented in a computer system. The computer system can be a local computing device (e.g., a personal computer, laptop, tablet computer, or mobile phone) with one or more processors and one or more memory devices, or it can be a distributed computing system (e.g., a cloud computing system with one or more processors or one or more memory devices distributed at different locations, for example, at a local client and / or one or more remote server farms and / or data centers). The computer system can comprise any circuit or combination of circuits. In one exemplary embodiment, the computer system can comprise one or more processors, which can be of any type.In our usage, "processor" can mean any type of computing circuit, such as, but not limited to, a microprocessor, a microcontroller, a complex instruction set (CISC) microprocessor, a reduced instruction set (RISC) microprocessor, a very long instruction word (VLIW) microprocessor, a graphics processor, a digital signal processor (DSP), a multi-core processor, a field-programmable gate array (FPGA), or any other type of processor or processing circuit. Other types of circuitry that may be included in the computer system may be a custom-designed circuit, an application-specific integrated circuit (ASIC), or similar, such as one or more circuits (e.g., a communications circuit) for use in wireless devices, such as...The computer system may include mobile phones, tablet computers, laptop computers, two-way radios, and similar electronic systems. It may comprise one or more storage devices, which may include one or more memory elements suitable for the application, such as main memory in the form of random access memory (RAM), one or more hard disk drives, and / or one or more drives that handle removable media, such as CDs, flash memory cards, DVDs, and the like. The computer system may also include a display device, one or more speakers, and a keyboard and / or controller, which may include a mouse, trackball, touchscreen, voice recognition device, or any other device that allows a system user to input information into and receive information from the computer system.

[0152] Some or all of the process steps can be performed by (or using) a hardware device, such as a processor, a microprocessor, a programmable computer, or an electronic circuit. In some embodiments, one or more of the key process steps can be performed by such a device.

[0153] Depending on specific implementation requirements, embodiments of the invention can be implemented in hardware or software. Implementation can be carried out using a non-volatile storage medium such as a digital storage medium, for example, a floppy disk, DVD, Blu-ray disc, CD, ROM, PROM, EPROM, EEPROM, or FLASH memory, on which electronically readable control signals are stored. These signals interact (or can interact) with a programmable computer system to execute the respective method. Therefore, the digital storage medium can be computer-readable.

[0154] Some embodiments according to the invention include a data carrier with electronically readable control signals that can interact with a programmable computer system so that one of the methods described herein is carried out.

[0155] In general, embodiments of the present invention can be implemented as a computer program product with program code, wherein the program code is effective for executing one of the methods when the computer program product runs on a computer. The program code can, for example, be stored on a machine-readable medium.

[0156] Further embodiments include the computer program for carrying out one of the methods described herein, which is stored on a machine-readable medium.

[0157] In other words, an embodiment of the present invention is therefore a computer program with program code for carrying out one of the methods described herein when the computer program is running on a computer.

[0158] Another embodiment of the present invention is therefore a storage medium (or a data carrier or a computer-readable medium) comprising a computer program stored thereon for executing one of the methods described herein when executed by a processor. The data carrier, the digital storage medium, or the recorded medium is generally tangible and / or not seamless. Another embodiment of the present invention is a device as described herein comprising a processor and the storage medium.

[0159] Another embodiment of the invention is therefore a data stream or signal sequence that represents the computer program for carrying out one of the methods described herein. The data stream or signal sequence can, for example, be configured to be transmitted via a data communication connection, such as the Internet.

[0160] Another embodiment includes a processing means, for example a computer or a programmable logic device, which is configured or adapted to perform one of the methods described herein.

[0161] Another embodiment comprises a computer on which the computer program for performing one of the methods described herein is installed.

[0162] Another embodiment of the invention comprises a device or system configured to transmit (for example, electronically or optically) a computer program for executing one of the methods described herein to a receiver. The receiver may be, for example, a computer, a mobile device, a storage device, or the like. The device or system may, for example, include a file server for transmitting the computer program to the receiver.

[0163] In some embodiments, a programmable logic device (e.g., a field-programmable gate array, FPGA) can be used to perform some or all of the functionalities of the methods described herein. In some embodiments, a field-programmable gate array can cooperate with a microprocessor to perform one of the methods described herein. In general, the methods are preferably performed by any hardware device.

[0164] Examples of implementation can be based on the use of a machine learning model or machine learning algorithm. Machine learning can refer to algorithms and statistical models that computer systems can use to perform a specific task without the use of explicit instructions, instead of relying on models and inference. For example, instead of a rule-based transformation of data, machine learning can use a transformation of data derived from an analysis of historical and / or training data. For instance, the content of images can be analyzed using a machine learning model or a machine learning algorithm.To enable the machine learning model to analyze the content of an image, it can be trained using training images as input and training content information as output. By training the machine learning model with a large number of training images and / or training sequences (e.g., words or sentences) and associated training content information (e.g., labels or annotations), the machine learning model "learns" to recognize the content of the images, so that the content of images not included in the training data can be recognized using the machine learning model.The same principle can be used for other types of sensor data: By training a machine learning model using training sensor data and a desired output, the machine learning model "learns" a conversion between the sensor data and the output. This can be used to provide an output based on non-training sensor data provided to the machine learning model. The provided data (e.g., sensor data, metadata, and / or image data) can be preprocessed to obtain a feature vector, which is used as input for the machine learning model.

[0165] Machine learning models can be trained using training input data. The examples above use a training method called supervised learning. In supervised learning, the machine learning model is trained using a plurality of training samples, where each sample can include a plurality of input data values ​​and a plurality of desired output values; that is, each training sample is associated with a desired output value. By specifying both training samples and desired output values, the machine learning model "learns" which output value to provide based on an input sample that is similar to the samples provided during training. In addition to supervised learning, semi-supervised learning can also be used. In semi-supervised learning, some of the training samples lack a desired output value.Supervised learning can be based on a supervised learning algorithm (e.g., a classification algorithm, a regression algorithm, or a similarity learning algorithm). Classification algorithms can be used when the outputs are limited to a finite set of values ​​(categorical variables), i.e., the input is classified as one from the limited set of values. Regression algorithms can be used when the outputs are any numerical value (within a range). Similarity learning algorithms can be similar to both classification and regression algorithms but are based on learning from examples using a similarity function that measures how similar or related two objects are. In addition to supervised or semi-supervised learning, unsupervised learning can be used to train the machine learning model.In unsupervised learning, only input data may be provided, and an unsupervised learning algorithm can be used to find a structure in the input data (e.g., by grouping or clustering the input data, finding commonalities in the data). Clustering is the assignment of input data, which comprises a plurality of input values, into subsets (clusters) such that input values ​​within the same cluster are similar according to one or more (predefined) similarity criteria, while they are dissimilar to input values ​​included in other clusters.

[0166] Reinforcement learning is a third group of machine learning algorithms. In other words, reinforcement learning can be used to train the machine learning model. In reinforcement learning, one or more software actors (called "software agents") are trained to perform actions in an environment. A reward is calculated based on the actions performed. Reinforcement learning relies on training the one or more software agents to select actions in such a way that the cumulative reward increases, resulting in software agents that become better at the task they are given (as evidenced by increasing rewards).

[0167] Furthermore, some techniques can be applied to certain machine learning algorithms. For example, feature learning can be used. In other words, the machine learning model can be trained, at least partially, using feature learning, and / or the machine learning algorithm can include a feature learning component. Feature learning algorithms, also known as representation learning algorithms, can preserve the information in their input but transform it in a way that makes it useful, often as a preprocessing stage before performing classification or prediction. Feature learning can be based, for example, on principal component analysis or cluster analysis.

[0168] In some examples, anomaly detection (i.e., outlier detection) can be used to identify input values ​​that raise suspicion because they differ significantly from the majority of input and training data. In other words, the machine learning model can be trained, at least partially, using anomaly detection, and / or the machine learning algorithm can include an anomaly detection component.

[0169] In some examples, the machine learning algorithm can use a decision tree as a predictive model. In other words, the machine learning model can be based on a decision tree. In a decision tree, observations about an object (e.g., a set of input values) can be represented by the branches of the decision tree, and an output value corresponding to the object can be represented by the leaves of the decision tree. Decision trees can support both discrete and continuous values ​​as output values. When discrete values ​​are used, the decision tree can be called a classification tree; when continuous values ​​are used, the decision tree can be called a regression tree.

[0170] Association rules are another technique that can be used in machine learning algorithms. In other words, the machine learning model can be based on one or more association rules. Association rules are created by identifying relationships between variables in large datasets. The machine learning algorithm can identify and / or utilize one or more relationship rules that represent the knowledge derived from the data. The rules can be used, for example, to store, manipulate, or apply this knowledge. Machine learning algorithms are typically based on a machine learning model. In other words, the term "machine learning algorithm" can refer to a set of instructions that can be used to create, train, or use a machine learning model.The term "machine learning model" can refer to a data structure and / or a set of rules that represents the learned knowledge (e.g., based on training performed by the machine learning algorithm). In some implementations, the use of a machine learning algorithm may imply the use of an underlying machine learning model (or multiple underlying machine learning models). The use of a machine learning model may also imply that the machine learning model and / or the data structure / set of rules that constitutes the machine learning model are trained by a machine learning algorithm.

[0171] For example, the machine learning model could be an artificial neural network (ANN). ANNs are systems inspired by biological neural networks, such as those found in a retina or brain. ANNs comprise a plurality of interconnected nodes and a plurality of connections, called edges, between the nodes. There are typically three types of nodes: input nodes, which receive input values; hidden nodes, which are connected (only) to other nodes; and output nodes, which provide output values. Each node can represent an artificial neuron. Each edge can transmit information from one node to another. The output of a node can be defined as a (nonlinear) function of its inputs (e.g., the sum of its inputs). The inputs of a node can be used in the function based on a "weight" assigned to the edge or the node providing the input.The weight of nodes and / or edges can be adjusted during the learning process. In other words, training an artificial neural network can involve adjusting the weights of the nodes and / or edges of the artificial neural network, i.e., to achieve a desired output for a given input.

[0172] Alternatively, the machine learning model can be a support vector machine, a random forest model, or a gradient boosting model. Support vector machines (i.e., support vector networks) are supervised learning models with associated learning algorithms that can be used to analyze data (e.g., in classification or regression analysis). Support vector machines can be trained by providing an input with a plurality of training input values ​​belonging to one of two categories. The support vector machine can be trained to assign a new input value to one of the two categories. Alternatively, the machine learning model can be a Bayesian network, which is a probabilistic directed acyclic graphical model. A Bayesian network can represent a set of random variables and their conditional dependencies using a directed acyclic graph.Alternatively, the machine learning model can be based on a genetic algorithm, which is a search algorithm and heuristic technique that imitates the process of natural selection. List of characters

[0173] The present invention will be explained in more detail below with reference to the exemplary embodiments shown in the schematic figures of the drawings. These show: Fig. 1 a schematic representation of a LIDAR survey using an aircraft, Fig. 2 a schematic representation of an embodiment of the present invention, Fig. 3a-d an exemplary property with house based on ALKIS and LIDAR data, Fig. 4a-d a second exemplary property with a house based on ALKIS and LIDAR data, Fig. 5 a schematic representation of a small settlement for the purpose of illustrating the present invention, Fig. 6a schematic overview of a data flow according to the invention in an embodiment of the present invention, Fig. 7 a schematic representation of a deep neural network for use as a machine learning model within the scope of the present invention, Fig. 8 a schematic representation of a data relationship and flow according to an embodiment of the present invention.

[0174] In all figures, identical or functionally equivalent elements and devices – unless otherwise specified – have been provided with the same reference numerals. Description of the drawings

[0175] The Figure 1Figure 1 shows a schematic representation of a LiDAR survey of a physical system 100, symbolized here by various buildings 100 arranged on the Earth's surface. The aircraft 101 flies and carries a LiDAR surveying device. During the flight of the aircraft 101, the LiDAR surveying device emits signals 102, which are reflected by the buildings 100 and by the surface of the physical system 100 facing the aircraft.

[0176] The signals can generally be emitted and received in all directions. Signal 102 is shown in the figure as perpendicular to the Earth's surface, but only as an example and schematic representation.

[0177] Using appropriate reconstruction algorithms, a raster-resolved map ("LIDAR point clouds") can be created that attempts to reproduce the surface of the physical system.

[0178] An arrow indicates that the aircraft is moving during flight. The resulting "LIDAR map" is thus gradually expanded.

[0179] Figures 3a and 3b show a house on a plot of land. Figures 3c and d show this from a different angle perspective.

[0180] Property 100 is defined by a boundary 301. A hatched area 310 shows the part of the property on which a house stands. The house has a boundary 311 on the property. There is also an area 312 in which an ancillary structure is located. This data on the property and the buildings is derived, for example, from data in an official cadastral information system (ALKIS).

[0181] LiDAR point data 320, 321, and 322 are also shown. Data 320 shows the roof of the house, data 322 shows a chimney located on it, and data 321 shows the relatively flat roof of the exterior structure (e.g., garage).

[0182] The LIDAR data are intersection point data as defined in the invention and its further developments. In this embodiment, raw LIDAR data in the area of ​​property 300 were "intersected" with the developed parts of the property 310 and 312. Thus, the LIDAR points shown are those located above the developed parts of the property 310 and 312. In other words, for example, the points of the roof 320, when projected onto the plane of property 300, lie within the perimeter 311 of the house.

[0183] In the Figure 3bA numerical integration of the house and the outdoor area is graphically represented. As already described in detail, such a method can be used, for example, to calculate the usable volume and floor space. The column model shown here can also be used as a basis for other determinations, such as determining the number of usable floors. In the 3D figure Additionally, extension 332 is clearly shown.

[0184] In the Figures 4a-d Similarly, a second exemplary property with a house is shown, based on ALKIS and LiDAR data. In this case, an ancillary building 433 is directly attached to the house. This fact is not apparent from the roof 420 alone. However, through the correct evaluation of secondary data (here: ALKIS data), the volume of the (main) house 430 is correctly integrated numerically (see in particular). Fig. 4b and 4dOther relevant quantities are also correctly determined within the scope of the invention. Therefore, the invention can also correctly handle such a case and is not "misled" by the roof, i.e., it does not lead to incorrect conclusions.

[0185] The Figure 5 Figure 1 shows a schematic representation of a small settlement to illustrate the present invention. Schematically, five houses are arranged on opposite sides of a street. The houses are, for example, each situated on its own plot of land, with each plot comprising a parcel of land.

[0186] The houses and their plots of land represent subsystems. One subsystem, 501, has been selected. The invention is now able to identify and implement potential improvements in the depicted settlement.

[0187] In a simple, schematic example using the selected subsystem 501, a prediction is made, based on the surrounding houses, about how the property of subsystem 501 could be developed or better utilized. This is done based on the properties of the property (secondary data of the selected subsystem) as well as on the nine surrounding properties and their houses (subsystem data of the unselected subsystems). Relational data describing the relationship between subsystem 501 and the nine surrounding houses and properties can also be considered. In one simple example, it is taken into account that all 10 houses are located on the same street. In another example, it is taken into account that four houses are on the same side of the street and / or that five houses are on the opposite side. An evaluation of "how" relevant the properties of the four houses are, for example, is then made.In comparison to the properties of the five houses, the present invention achieves this, in particular through the use of machine learning (based on training the machine learning model).

[0188] In one example, the already existing house of subsystem 501 is disregarded. It is considered a variable within the context of system improvement. As a result, however, such a house could be modified, have additional stories added, or be rebuilt.

[0189] The Figure 6Figure 1 shows a schematic overview of a data flow according to an embodiment of the present invention. LiDAR or 3D model data / LoD data 601, ALKIS data 602, and photogrammetry data 603 are incorporated into an analysis 604 using a machine learning model 605. By combining these data sources, the invention can profitably utilize and evaluate synergies between the datasets. A suggestion or prediction 606 is generated for one or more selected subsystems, which is then implemented in reality. For example, a building is constructed, extended, or adapted; a process is adjusted; or a circuit element or qubit is modified.

[0190] The depicted data flow can be considered a control and / or regulation loop for the physical system.

[0191] The Figure 7Figure 1 shows a schematic representation of a deep neural network for use as a machine learning model within the scope of the present invention. The deep neural network (artificial neural network) has at least one input neuron 701, usually several, in particular numerous (the designation 701a-z is merely symbolic and does not limit the number in any way). There is at least one output neuron 703. Various structures can be arranged between these, in particular one or more hidden layers 702.

[0192] For example, secondary data of a selected subsystem and subsystem data of non-selected subsystems, as well as relational data between selected and non-selected subsystems, can be entered at input neurons 701a-z.

[0193] For example, at output neuron 703, a prediction is made available as to how a selected subsystem could be modified or rebuilt.

[0194] The Figure 8 Figure 1 shows a schematic representation of a data relationship and flow according to an embodiment of the present invention. Reference is made to the general part of the invention discussion.

[0195] The subsystem data 803 can include the secondary data and the LiDAR / LoD data 801 and / or be derived from them. A prediction 805 is made possible by a machine learning model. The secondary data 802 are considered for the selected subsystem 810x. The subsystem data 803 are considered, in particular, for non-selected subsystems 810. In addition, relational data 804 can be created or derived and successively considered. Reference symbol list

[0196] 100 Physical System 101 Aircraft 102 Measurement Signal 300 Property (here identical to parcel) 301 Edge of the Property (Edge of the parcel) 310 Developed part of the property (house) 311 Edge of the developed part of the property (i.e., of the house) 312 Developed part of the property (outdoor facility) 320 Roof of a house / "point cloud" of the intersection point data 321 Roof of an outbuilding / "point cloud" of the intersection point data 322 Chimney / Fireplace 330 Numerically integrated house / volume 331 Numerically integrated outbuilding / volume 332 Numerically integrated extension / volume 400 Property (here identical to parcel) 410 Developed part of the property (house) 420 Roof of a house / "point cloud" of the intersection point data (shown here including outbuilding) 422 Chimney / Fireplace 430 Numerically Integrated house / volume (here without outbuildings directly attached to the house) 433 Developed part of the property (here, however, belonging to the outbuildings) 501 Selected subsystem / selected property or building 601 LiDAR / LoD data602 ALKIS data 603 Photogrammetry data 604 Analysis 605 Machine learning model, in particular artificial neural network 606 Proposal / Prediction 607 Implementation (construction of a building, adaptation of a process, adaptation of a circuit, modification of a qubit, etc.) 700 Artificial neural network (schematic) 701-(az) Exemplary input neurons 702 Exemplary hidden layers (schematic) 703 Exemplary output neuron 801 LiDAR / LoD data 802 Secondary data 803 Subsystem data 804 Relational data 805 Prediction / Potential / Outcome 810 Non-selected subsystems 810 Selected subsystem(s) (x denotes selected)

Claims

1. A computer-implemented method for determining a ground reference point, comprising the following steps: - identifying a physical system (100) in a three-dimensional space, - providing three-dimensional LIDAR data (601, 801) in the three-dimensional space, - identifying a two-dimensional space suitable for a digital description of a two-dimensional projection of the physical system (100) in the three-dimensional space, in particular a two-dimensional projection of the LIDAR data (601, 801) - providing photogrammetric data (603) relating to the three-dimensional space - Detecting at least one sealed surface and its dimensions in the two-dimensional space based on the photogrammetric data (603) characterized by - determining a ground reference point by detecting an object in the photogrammetric data that is suitable for serving as a ground reference point, in particular manhole and sewer covers; - intersecting the dimensions of the detected object with the three-dimensional point data to generate intersection point data - calculating an average of the intersection point data to determine the ground reference point.

2. A method according to claim 1, further comprising a step of measuring at least one point datum using a LIDAR measuring device, in particular an airborne LIDAR measuring device (101).

3. A method according to claim 1 or 2, further comprising providing secondary data comprising two-dimensional polygon data, in particular from a cadastral survey.

4. A method according to any of the preceding claims, further comprising a step of classifying and marking points as roof points based on a second machine learning model, in particular a second artificial neural network.

5. A method according to claim 4, wherein the second machine-learning model is trained and / or was trained by a method comprising a step of intersecting three-dimensional point data with a building component of a subsystem of the physical system defined by secondary data to generate intersection point data.

6. A method according to any one of the preceding claims, wherein, furthermore, the points are classified and marked as roof points based on the set of intersection point data, in particular by clustering and / or through unsupervised learning, in particular based on the Euclidean distance between the points as a relevant measure and / or based on the vertical distance, in particular the vertical or z-component, of the points relative to one another as a relevant measure, and / or based on normal vectors.

7. A method according to any one of claims 3-6, further comprising a step of intersecting three-dimensional point data with a building component of a subsystem of the physical system defined by the secondary data to generate intersection point data, as well as a step of marking the points of the intersection point data that fall within the area of the building component in the three-dimensional point data.

8. A method according to any of the preceding claims, further comprising the use of a supervised machine learning model, in particular an artificial neural network and / or linear regression, to predict characteristics of a selected subsystem of the physical system.

9. A method according to claim 8, further comprising a step of adapting the physical system by modifying the selected subsystem in the manner proposed by the machine learning model, in particular the artificial neural network, specifically by constructing, renovating, or rebuilding one or more building structures within the selected subsystem.

10. A method according to any of the preceding claims, further comprising a step of clustering subsystems of the physical system into related clusters and correspondingly labeling the subsystems with respect to their cluster membership, in particular using an AI technique of unsupervised learning.

11. A method according to any of the preceding claims, wherein data relating to subsystems of the physical system comprises one or more of the following: - ridge height, eave height, roof height, roof pitch, floor area ratio GRZ, floor area ratio GRZ1, floor area ratio GRZ2, floor space index GFZ, floor area, volume, number of usable stories, number of full stories - Presence and extent of impervious surfaces - ground reference point - Region, city, and / or county in which the subsystem is located - Extent of building use by the nearest neighbor - Extent of built-up area of an m-nearest (second-nearest, third-nearest, etc.) neighbor - A Boolean statement indicating whether the nearest neighbor is located on the same route, specifically the same street, as the selected subsystem - A Boolean statement indicating whether the m-next neighbor is located on the same route-specifically, the same street-as the selected subsystem - A Boolean statement indicating whether the next neighbor has an open or closed structure - a Boolean statement indicating whether the m-th nearest neighbor has an open or closed design - a Boolean statement indicating whether the next neighbor has a building in the courtyard or on the street - a Boolean statement indicating whether the m-next neighbor has a building in the courtyard or on the street - a length specification that determines the distance of the nearest neighbor's building from the street - a length and / or depth value that determines how far a house is set back from the street - Relevant Development Area - Applicable zoning plans - Land use regulations in accordance with legal requirements, in particular the Building Land Use Ordinance - Cluster affiliation of the subsystem to a cluster of subsystems.

12. A method according to any one of claims 8-11, wherein the step of using the machine learning model, in particular the artificial neural network, is further based on relational data comprising at least one property that describes a relationship between the selected subsystem and other subsystems, in particular non-selected subsystems.

13. A method according to any of the preceding claims, further comprising a step of generating LoD data based on the three-dimensional point data, in particular LIDAR data.

14. A method according to claim 13, wherein data relating to subsystems of the physical system is generated and / or has been generated based on the generated LoD data.

15. A computer, computer program, computer network, or computer-readable medium configured to carry out the method according to any of the preceding claims 1 or 3-14, or a computer, computer program, computer network, or computer-readable medium configured to carry out the method according to claim 2, comprising at least one instruction suitable for causing a LIDAR measuring device to perform the step of measuring at least one point data.

Images