EP1546936A1

EP1546936A1 - Computer system and method for comparing data sets of three- dimensional bodies

Info

Publication number: EP1546936A1
Application number: EP03790960A
Authority: EP
Inventors: Frank Epple; Markus Zajc
Original assignee: Cadenas Konstruktions- Softwareentwicklungs- und Vertriebs GmbH
Current assignee: Cadenas Konstruktions- Softwareentwicklungs- und Vertriebs GmbH
Priority date: 2002-09-02
Filing date: 2003-09-02
Publication date: 2005-06-29
Also published as: WO2004021216A1; DE10240940A1; AU2003264250A1

Abstract

A computer system (100) serves for making available technical information; the computer system (100) includes a data processing unit (130), a database (1), a data input unit (110), and a data display unit (120), whereby in the database (1), data of a quantity of three-dimensional bodies are saved. By means of the data input (110), data of any selected body are input and via the data display unit (120), as at least one body is displayed, which has the greatest similarity with the selected body in the quantity of the three-dimensional bodies.

Description

COMPUTER SYSTEM AND METHOD FOR COMPARING DATA SETS OF THREE-DIMENSIONAL BODIES

The invention relates to a computer system for making available technical information, 5 which includes a data processing unit, a database, a data input unit, and a data display unit, whereby in the database, data of a quantity of three-dimensional bodies can be stored.

Such a computer system is known from WO 01/09742 A2, in which a web portal for 0 engineering information is disclosed. The computer system includes data and drawings for technical elements, which are bought by a consumer. The drawings can be stored in a customer computer (client) and integrated in a CAD system.

The known system is limited to reproducing the parts themselves and simple 5 information about the parts, for example, their dimensions.

On the other hand, from the attachment "Matching 3D Models with Shape Distributions", by R. Osada, Th. Funkhouser, B. Chazelle, D. Dobkin, published by Princeton University, Princeton, NJ 08540, USA, 2001, a method is known, which 0 serves to determine the similarity between three-dimensional models, samples, or shapes, which represents a fundamental problem of the illustration with the computer in molecular biology, however, also with the comparison of technical components. According to the attachment, a method is proposed, which determines shape distributions or shape functions of three-dimensional images. The determination of 5 similarities of three-dimensional models serves for further recognition of the models, for locating, classifying, and grouping.

Data-containing three-dimensional models have characteristics, which are essential for the compilation of algorithms for describing the three-dimensional structure; in contrast 0 to pictorial representations or screen representations, three-dimensional models do not depend on the configuration of cameras, light sources, or surrounding objects, for example, mirroring elements. As a result, they also contain no reflections, shadows, inclusions, projections, and the like. Thus, it is possible with the assistance of algorithms to describe conformity between objects of the same type. 5 A method for calculating 3D model signatures and degrees of dissimilarity is described, which is suited for the desired object. The basic idea is that the signature of an object is to be represented by a shape distribution, which is obtained from a shape function, which describes the global geometric characteristics of the object. The basic idea of this approach is to transform any 3D model into a function, which can be easily compared with others.

The characteristic of a three-dimensional object is reproduced by a feasibility distribution, which was obtained from a shape function. The shape function dispenses with the geometric characteristics of the object. Such a shape distribution provides, for example, the distribution of Euclidean removal between pairs of points selected according to random criteria on the surface of a three-dimensional model again. The distribution describes the entire surface of the selected object. As soon as the shape distributions from two different objects are obtained, the dissimilarity between the two objects can be determined by means of metrics, which dispenses with the distances between the distributions, for example, the L_N-norm, whereby possibly a standardizing step is necessary for adjusting the dimensional scales. In this manner, each 3D model can be covered in a parameterized function, which can be easily compared with others. The computing method is based on the original polygons of a 3D model. With this method, also such objects can be separated from one another, which have different fine structures.

Shape distributions have the necessary invariance, for example, relative to rotations and scaling. They are robust against small disturbances. The description of the deviation between two bodies assumes the metrics of the used standards, for example, the L_N-norm. The building of the shape distributions for a data compilation of 3D models can be achieved quickly and efficiently. The shape distributions are independent from the type of their representation, topology, or the application area of the 3D model described with their assistance. A single data component can have also a plurality of different forms. The shape distributions are also suited to describe machine parts.

As functions for the obtainment of shape distributions, for example, the following are considered: functions for measuring the angle between three points selected according to the random principle on a surface of a 3D model, functions for measuring the distance between a determined point and a point selected according to the random principle on a surface, functions for measuring the distance between two points selected according to the random principle on the surface, functions for measuring the third root of the volume of the tetrahedron between four points selected according to the random principle on the surface.

These functions have the advantage that they are invariant relative to a tessellating of a 3D model, since the points on the surface are distributed equally. Likewise, they are insensitive against small disturbances, caused by minimal blurring, cracks, or by the use or removal of polygons. AH of these shape functions are suited as means for distinguishing between various bodies, since significant changes to the bodily structures of three-dimensional images have an effect on the model functions.

It is an object of the present invention to produce a computer system, by means of which the selection of a body from a predetermined quantity of bodies according to characteristics of the body is made possible.

According to the present invention, this object is solved, in that via the data input unit, data of any selected body can be input and that the data display unit can output at least one body, which in the quantity of the three-dimensional bodies, has the greatest similarity with the selected body.

With the invention, then, the possibility is created of selected bodies according to the degree of their similarity. Under the term "similarity", in particular, the geometric similarity is to be understood. In this manner, it is possible to associate technical components of any type, for example, pneumatic cylinders, piston rods for a combustion engine, bevel wheels for a bevel wheel gear, or any other objects of an animate or inanimate nature according to the similarity provided by their outer appearance and to determine a degree of similarity between two bodies associated with one another. It is not impossible to use the computer system of the present invention on objects of microbiology.

By means of the invention, the expensive searching of one component is eliminated, which first must be associated to a determined class of components, in order to arrange its geometry alone with the assistance of the human eye in a specific group. If this type of grouping with simple components, such as a screw or a plate can be performed still with justifiable expense, then with objects comprised of many components, such as a cylinder or an engine, this type of grouping can hardly be performed and requires much time. First, by means of the invention, an automatic recognition of components is made possible, because corresponding components are available in a database as reference parts. Via algorithms, the searched components of a group of reference parts are associated from the database, and therewith, the corresponding attributes are assigned, which relate to these reference parts. Starting from a mathematically comprehensible association with the aid of a computer system, which is programmed accordingly, the search of a desired component is realized in only a portion of the time that would be required with a search by eye and with purely mechanical arrangement means.

An industrial manufacturer of a highly complex product, for example, an air plane or a ship, can locate individual components, which are mounted or machined on many different points of the product and can determine similarities between these components, when they are combined in a common data line. Thus, it is possible, instead of the same or similar components - with a degree of similarity determined by the manufacturer - to limit different suppliers to an individual component of an individual supplier and to reduce the complexity of the entire product from its output elements. In this manner, the diversity of parts within an undertaking can be reduced. Redundancy and inconsistencies within the management of all outsourced items of an undertaking are thereby avoided or are eliminated.

The body that can be selected according to the invention is either already contained in the database or it is input externally and than pre-processed, broken down into its characteristics, and mathematically described by shape functions, shape distributions, or other scalar or vectoral dimensions, which together form a completely described data set of the body at least in view of its geometry. In this manner, an offline as well as an online reduction can be performed according to the present invention. In the latter case, the data of the body to be searched are input externally.

With reference to the degree of similarity between the parts contained in the database input by him, the user of the database can determine the breadth of the database, in which perhaps, with a sufficiently appearing similarity of the components contained in the database, he reduces this to a single component, or if the similarity distances between the parts contained in the database are too large, new parts can be added in the database, in order to increase the complexity and diversity. The computer system of the present invention can be used in particular also during the construction process itself. When the user in the construction processes a new component made up of individual component parts, similar components are searched from the database and are represented on the data display unit in the form of a ranking list of the similarities. In this manner, the user can interrupt the construction process early, if the components with sufficient similarities can be combined from the parts provided in the database. Or the construction process can be advanced in consideration of the components offered in the database. In the first case, the use of double parts is avoided; the second case builds on already provided resources and technology.

In this manner, many resources can be saved with the construction as well as with the warehousing; likewise, time is obtained, which is saved by the minimal complexity of the substock.

With the assistance of the computer system, also an existing nomenclature variety for designation of component is overcome: in an optimal environment, all contractors of components would provide a functionally and geometrically equivalent part of the same name. This does not exist in reality. There are systems, such as for example, the unitary name allocation according to a German industry standard (DIN) or according to a European standard; however, these cover only five percent of all technical components, so that a substantial need still remains for the recognition of agreement. By the present computer system, a classification based on a geometric, spatial determination of the product works for various vendors, without an alphanumeric nomenclature, which normally and by the human language is used with terms like "screw", "t-support". The user is displaced by the use of the computer system into the situation of determining components, without having to resort to standards or language. Language has the disadvantage that also spoken duplicates exists: for example, a "crimping" in a component also can be designated as a "groove" or as a "channel", which leads to difficulty in relocating the same part when it is provided with different designations.

In the data set, which is associated with each component stored in the database, not only geometric data on the dimensions of the body can be introduced, according to another embodiment of the invention; rather also data on its physical and chemical qualities can be introduced, for example, the material composition (metal, plastic, rubber, wood), the electrical conductivity, the elasticity, the surface qualities (roughness, coating), the treatment, which the component had during its manufacture (hardening, sintering), etc.

Advantageous further embodiments of the invention are provided in the dependent claims and from the description.

Particularly advantageous is a computer system, in which the selected body can be broken down in the data processing unit into individual components. In this manner, the complexity of a body made up of multiple individual components is reduced to the individual component parts themselves.

Alternatively to breaking-down of an assembly into individual parts, the assembly also can be merged, in order to make it comparable as a whole. In this connection, Boolean operations are used, such as for example, those known from W.C. Thibault, B.F. Naylor, "Set operations on polyhedra using binary space portioning tree". Comput. Graph., 21 (4): 153-162, 1987.

In the computer system, on each of the bodies in the database, a characteristic extraction is compiled, by means of which a description is produced, by which the body can be compared with the description of the quantity of bodies already provided in the database.

For characteristic extraction, in a preferred embodiment of the invention, a breakdown of the surface of the selected body into triangular surfaces or other polygonal surfaces is suited.

For describing the selected body, preferably shape distributions are used. With the obtainment of shape distributions, various methods can be used, such as those only cited by way of example in the previously cited attachment.

Alternatively or in parallel, additionally other forms for describing the geometry of a selected body are taken into consideration, for example, the determination of the ratio of its surface to its volume.

According to the type of description of the geometry of the body, also various forms or descriptions for "similarity" are created. If different descriptions are provided for each body contained in the database for its geometry, the user of the database also can select which similarity criterion he wants to search. Thus, it can be provided that the selected body is compared with the bodies in the database with reference to the shape distributions and/or the ratio of its surface to its volume.

Also particularly advantageous, a computer system proves to be one in which the characteristic data set of a selected body can be compared with the characteristic data sets of bodies in the database according to the dynamic time warping method, whereby the shortest distance between the data sets can be found. Dynamic time warping is a known method from voice recognition systems, which serves to determine the conformity between a spoken word and a word sample dependent therefrom, with which the speed of the word is spoken. In this manner, time standardization is necessary, in which the words are temporally expanded or compressed. It is particularly advantageous if only selected parts of the characteristic data set of the selected body are compared with the corresponding parts of the body from the database. Dynamic time warping, then, is applied when shape distributions are contained in the characteristic data set.

Generally, the characteristic data set can be as long as desired. It combines various criteria obtained by algorithms, so that parts of the characteristic data set can be as long as desired. Dynamic time warping is only applied on the parts, which comprise a function that can be expanded or compressed. It is advantageous if the function is low- noise. However, noise also can be eliminated with the aid of a filter. The expansion or compression of a function is computed with the assistance of dynamic time warping.

When also the ratio of the volume to the surface of the body is part of the characteristic data set, it should be noted that dynamic time warping cannot be used on this criterion. Preferably, dynamic time warping is applied on the shape distributions D1, D1 , N1 (with this method, the standard vector of a surface is considered).

In addition to the computer system itself, the present invention relates to a method for comparing data sets of three-dimensional bodies, which are stored in a database of a computer system with a data input unit and a data display unit.

It is the object of the present invention to improve such a method for comparing the bodies or components. According to the invention, the object is resolved, in that via the data input unit, data of any selected body can be input and that by means of the data display unit, at least one body is displayed, which has the greatest similarity with the selected body in the quantity of the three-dimensional bodies.

Further advantages are provided in the dependent claims.

With the determination of the similarity of a new body with bodies already provided in the database, it is advantageous if upon cycling of the data set of a body from the database, the results of previous cycles are taken into consideration. In this manner, a data cycle can be interrupted when it is determined that the similarity of the body to be compared with the new body is smaller than the similarity between an already, previously compared body and the new body.

Alternatively, a method is used, in which, according to a type of evaluation or weight system, points of the individual evaluation algorithms are provided. Each algorithm approaches a margin, within which it distributes points. In this manner, the margin is determined according to the reliability of the algorithm. After running of a plurality of algorithms with a similarity comparison of the selected body with the bodies contained in the database, the points are added up, and the body is determined with the lowest number of points, which proximate the selected body.

Each algorithm that evaluates provides points. According to the quality of the algorithm (which is determined by tests), the algorithm may allocate points. For example, an algorithm A may allocate between 0 and 10 points, an algorithm B may allocate between 0 and 3 points, and an algorithm C may allocate between 0 and 50 points. Each algorithm is used on each part to be compared. The comparison occurs in different ways. For example, with the shape distribution D2, the dynamic time warping is used, while with comparison of volumes to the surface, a simple comparison of both numbers is performed. The determined difference or similarity between the objects to be compared is mirrored further in the number of points allocated by the algorithm. The algorithm A, for example, is a shape distribution. This algorithm has a relatively high importance and allocates evaluations between 0 and 10. The algorithm for forming the ratio of volume and surface is, for example, algorithm B. This is a little different for the end result and may allocate as many as three points. The more points that are allocated, then the more marked is the deviation between the selected body and the respective body from the database. The algorithm C is a typical filter. As soon as a part is safely classified as wrong, the algorithm allocates 50 points. In this manner, this part is placed in the ranking list of the most similar parts at a value which signifies a large dissimilarity. This can be an algorithm, which compares the lengths and generally sorts out equally a part which is 1 m long, when the part to be compared has a length of 2 mm.

The evaluation margin, which the individual algorithms permit, is either determined by that made available by the database, or it remains for the user to abandon, after his own specifications determine a valuation of the algorithm.

Preferably, during the performance of the comparison or after termination of the comparison, a list of the most similar bodies is displayed on the data display unit. In addition, in a further preferred embodiment of the method, the most similar body as such is displayed in its geometrical appearance on the data display unit, for example, a display screen. In addition, on the display screen, also the other qualities of the body, in particular, its longitudinal dimensions, its material qualities, and the like are displayed. Preferably, the selected body and the most similar body are shown near one another in a spatial representation in perspective, in order to make possible an optical comparison of parts of the characteristic data set of the selected body with the corresponding parts of the most similar body. In a preferred method, the most similar body itself and the selected body are placed in one another, so that the differences in the volumes are visible. In this manner, differences and similarities of the bodies can be viewed and are made particularly well-recognizable.

Next, the invention in one embodiment will be described in greater detail with reference to the figures. In the figures:

Figure 1 shows a computer system;

Figure 2 shows a schematic representation of the path for building a database from components;

Figure 3 shows a flow diagram for achieving a pre-processed 3D-model from the

3D-model of a body; Figures 4a, b show a flow diagram for producing a graphical visualization of differences with a comparison between a selected 3D-model of a body and 3D-data in the database;

Figure 5 shows a diagram for illustration the achievement of a ranking list of the most similar models from the database;

Figures 6a-e show an assembly and its breakdown into individual components, respectively, in perspective view;

Figures 7a, b show a surface breakdown of the surface of a component, by way of example, a T-shaped tube connection;

Figures 8a, b show the shape distribution associated with the T-shaped tube connection;

Figure 9 shows a comparison between the shape distribution of the selected body and a body similar to the selected body; and

Figures 10a and 10b show a comparison between the shape distribution of the selected body and a body similar to the selected body, respectively, in perspective representation.

The computer system is either an individual computer 100 (Fig. 1) or a network of computers, which are connected to one another either via a LAN or via the Worldwide Web. The invention can be realized in hardware in the computer 100, which is programmed accordingly. The invention, however, also can be realized in a client- server environment, in which devices that are remote from one another are connected to each other via a communication network. Program modules can be provided in a local as well as in a remotely arranged storage device. The computer 100 is equipped with a data input unit 110, for example, a keyboard, a data output unit 120, for example, a data display unit, in particular, a display screen, and a data processing unit, that is, a processor (CPU) 150. A read-only memory (ROM) 140, a random access memory (RAM) 160, a bulk memory 170, and a communication unit 190 for connection of the computer to a network, for example, a local or internal network or the Worldwide Web, via an output 200 is provided. For connecting the various components of the computer 100, a data line 210 is provided. In the individual computer or in a computer connected to the network, a database 1 is made available, which serves as the comparison database for the comparison with a desired body.

The database 1 (Fig. 2) contains a plurality of 3D models 21, 22, 23, 24, 25,..., which are obtainable as individual technical components and which are available either as such or in connection with one another to a complex component. The models 21 , 2, 23, 24, 25,... are then, for example, available in commerce as purchased parts, or they are parts of such purchased parts. The number of the models 21, 22, 23, 24, 25... is as large as desjred.

The models 21 , 22, 23, 24, 25..., for example, are screw nuts of various types, generally square nuts with rounded portions between the corners, cap nuts 22, tube connection pieces 23, wing screws 24, or double-T supports 25.

With incorporation into the database 1, the models 21, 22, 23, 24, 25... are processed in a pre-processing step 3 all in the same manner, such that they all form a common output basis for a characteristic extraction from an object desired by a user of the computer system. Database 1 contains all data sets of all models. The pre-processing step 3 includes a breakdown of the models in individual components, in the event this is necessary. (In the present case, the models 21, 22, 23, 24, 25, however, already are formed as no longer able to be broken down into individual parts). For example, a breakdown of the surfaces of the models 21, 22, 23, 24, 25... into triangles or other polygons can be provided. This step is also designated as tessellating.

In a step following the pre-processing step 3, via a characteristic extraction 4 for each of the models 21, 22, 23, 24, 25... characteristic data sets are produced, which is saved in the database 1 associated with the respective model 21, 22, 23, 24, 25... The characteristic extraction 4 includes an application of at last one of the known shape functions, which were previously set forth. Each of the models 21, 22, 23, 24, 25..., however, can also be processed by a plurality of functions. Subsequently, the results of the various functions are associated with the respective models 21, 22, 23, 24, 25 and saved in a storage medium.

With the assistance, for example, of the functions described in the essay, perhaps the production of the distances between randomly selected points on the surface of the body, a data set is produced. The models 21 , 22, 23, 24, 25... can be associated also then with pictorial representations of the functions. Thus, the representations 51, 52, 53, 54, 55 of the respective models 21 , 22, 23, 24, 25 show the associated distributions of curves, which are related to a selected function; the representations 51 , 52, 53, 54, 55 may also show or represent any other algorithm. To each function, for example, multiple curves or other forms of representations are associated.

The pre-processing step 3 (Fig. 3) serves to provide a static analysis of the models 21, 22, 23, 24, 25.... The surfaces of the models 21 , 22, 23, 24, 25... , for example, are tessellated, that is, broken down into triangles or polygons. Alternatively, the models are shown in another, likewise useable method as scatter plots. Entire data are transferred over import interfaces into the data processing system. Multiple, different methods for processing of the three-dimensional data can also be used in parallel, so that different data sets for each three-dimensional object are provided in the database 1.

In a next step 31 , it is checked whether the models are combined from various components. If this is the case, they are broken down in a further step 32 into their individual components or merged, that is, combined to a single part. This makes possible a comparison of such objects, which are represented in individual components in a database, while they are broken down in the other database as total objects or in other individual components.

In the case of the separation, in step 31 , a separation of the parts for determining the volume parts (solids) occurs. For example, then, a screw which is screwed into a cube is separated from this. In this step, also information on the original spatial association of the individual parts is obtained and placed in an additional data set. Preferably, in an additional step 33, also a data set on the complex part made up of individual components is stipulated as a data set. Thus, for example, with a search inquiry of an individual component, either this or the entire component is offered.

As far as the models emerging already as individual parts in step 31, they are preferably tessellated in a step 34, that is, their surface is broken down into a plurality of polygonal partial surfaces, for example, in triangular surfaces. Subsequently, it is checked whether the component (model) is completely covered by the selected surfaces. If surfaces are missing, the defective model is repaired. Also, false orientations are repaired. By means of these corrections, the quality of the data is improved; however, a completely, correctly represented model is not necessary, since these errors are improved by the static method of the characteristic extraction. This is also true for the breakdown of the data.

In step 34, a static equal distribution of the individual surfaces is produced, while as the same time, only a minimal memory use is required. In addition, almost any other format for illustration of the model in this representation can be transferred.

Subsequently, the pre-processed data of a three-dimensional model are stored in a data structure (step 35), which permits recognition of the connection between individual models and of their combined components or assemblies. All data of the models are collected and the characteristic extraction 4 is applied, in order to take into consideration, for example, global connections from a data set. Alternatively, the data sets are processed separately.

The data of a 3-D model are transferred from the pre-processing step 35 (Fig. 3) to a specified interface on a computer unit for performing the characteristic extraction. Through these steps, further pre-processing steps can be performed later within the pre-processing unit 3 (Fig. 2), without requiring that the data obtained in the characteristic extraction be modified.

On the models, either a shape distribution cited in the attachment is applied in a step 41 , or it is used in parallel several times from it. For the purpose of evaluation, the obtained data is standardized, to which, for example, the L_N-norm of Minkowski or the Kolmogorow-Smirnow distance belongs. In addition, a method for noise suppression can be used, in order to free the data from disturbing noise signals.

In addition, for the data sets, an optimal ratio between measurement pairs and the model size is developed. The smaller the measuring points that are used, the greater the noise. Thus, a suitable, calculatable value must found, that is, one that is not in too long of a time. The more points that are used, the longer the calculation lasts. A further parallel used method of the extraction exists in the calculation of the ratio of surface to volume, which is determined in step 42 parallel to step 41. The result of this method is represented only by a single number, and thus is accommodated in a very space- saving manner in the database. By this simple method, in particular, with a search process, an optimal pre-selection of the data quantity can be achieved. A further method for determining a characteristic exists in the determination of the convex casing of the object. Therefore, the "inner life" and depressions, holes, bores and the like are disregarded. On the convex casings, then, the shape distribution or other algorithms are again used. Also, however, a ratio number can be produced, which provides the surface ratio of the convex surfaces to the concave surfaces.

With a 2D comparison with algorithms from the type recognition, only the outline lines of a body are represented and an image of all six side views is produced, that is, in +x-, -*-_> ⁺y-_> -y-_ι +z-, and -z-. In this manner, a method designated as an OCR method (optical character recognition) is used.

A further method involves the use of the VOXEL algorithm. The goal of this algorithm is to observe a body in local regions and from these characteristics, to extract. The body is separated into uniform cells equally distributed over the body. From this, one can extract characteristics, for example, over points or volume parts. Possible methods are the already used methods of the shape distributions (model distribution) and further methods. Upon comparison of the bodies, each cell of one body is compared with the equivalent cell of another body, and thereby a difference amount is formed. In a further step 43, a further extraction method is used in parallel, which, for example, exists in the formation of a ratio between the curved to the flat surfaces. The method for data obtainment can be expanded by a flexible data structure. In this manner, the expansion function is already provided in the base software or is used as an additional tool by the user. The data sets produced in steps 41 through 43 are either used for filing in the database for the use with a search inquiry of a desired body. With the compilation of the data for the database, the data is used cumulatively, in order to produce a database with high information content. With the search inquiry, the results of steps 41 through 43 are used successively or incrementally, that is, first, a search process with the similarly obtained from a first shape distribution is performed, and then a search process with another shape distribution.

From the data made available by steps 41 through 43, a characteristic data set is produced in a step 44 for each component. In a step 45, a characteristic combination is provided. Thus, the determined data sets are collected according to guidelines, such as, for example, accuracy or memory use and abstracted to the characteristic data set. The data sets (Fig. 4a, b) obtained from the pre-processing (Fig. 2) and by the subsequent characteristic extraction forms the content of the database 1 , which forms a comparison database with the search for a desired body.

With the use of a computer system for determining the characteristics of a desired body 6 which is to be compared with all models 21, 22, 23, 24, 25 ... (Fig. 4b), the characteristics of the body are pre-processed in the same manner as with building the database 1 with the characteristic sets of all models 21, 22, 23, 24, 25 ... in a preprocessing step 7, in order to break down the surface of the body 6 into partial surfaces. Subsequently, the step 8 of characteristic extraction takes place according to the functions, which are contained in the database 1. Finally, the body is compared with all of its geometric characteristics with the objects from the database 1 in a comparison step 9.

In this manner, with each comparison, a degree of similarity is produced. From all similarity degrees, a ranking list 10 of the similar models is produced, which is graphically represented on a data output unit 11, for example, a monitor, especially the differences between the body and each of the models 21, 22, 23, 24, 25 ... may be visualized.

The bodies (models) to be searched are associated with data sets in the same manner as the components contained in the database 1 , in order to enable a comparison of it with all of the models stored in the database 1 up to that point. In this manner, the models to be compared is traversed in a processing step 61 (Fig. 5) in the same manner as with the production of the database 1. That is, the pre-processing and the characteristic extraction are applied to the model to be compared. Accordingly, likewise a characteristic data set to the body exists.

In a processing step 62, each data set of the body to be compared is compared with each data set in the database 1. In this manner, shape distributions, which are obtained according to the same shape functions, are placed opposite one another. When the comparison provides that the data exceeds a specified, predetermined distance between the shape functions, that is, they are too different, the comparison is interrupted and with the data set of a further component, a new comparison is performed (step 621). Preferably, for increasing the quality of the data comparison, the method of "dynamic time warping" (step 620) is used, which is known in language processing, in order to recognize the distortion of the temporal course of a spoken word and compare with a word example. In the present case, this means that in the level representation of the spatial problem with reference to the shape distribution, two graphs, namely, the shape distribution of the desired body and the body contained in the database, are compared with one another, whereby distortions in partial regions of the distribution of the graphs are balanced. In the present case, two characteristic data sets are compared with one another. Both data sets are performed from the front to the back, and with each step, the "distance" of the two data sets is adjusted. In this manner, a ranking list 63 of the desired body can be compiled from the most similar models.

With the performance of the method of the present invention, then, a multiple-member, essentially T-shaped body 7(Fig. 6a), which comprises a plurality of individual parts, is broken down into individual components or assemblies, which can be procured by a manufacturer X, connected as such, and are contained as such in the database 1. Individual components, for example, are a bar-shaped component 71 (Fig. 6b), a flat component 72 (Fig. 6c), an angular element 73 (Fig. 6d), and a cubical component 74 (Fig. 6e).

On each of the components, then, as shown in Fig. 7a, b with reference to a T-shaped tube connection 80, a surface dispersion (tessellating) of all surfaces into polygons, for example, triangles 81, is provided. On this, then, as described above, a method for producing the shape functions is used. In Fig. 8b, the shape distribution 82 of the tessellated tube connection 80 (Fig. 8a) is shown, whereby the level axis is represented according to the selected shape distribution, that is, for example, A3, D1 , D2, D3, D4, as are known from the attachment "Matching 3D Models with Shape Distributions", or accordingly to other shape distributions, a value of the tube connection 80, that is, for example, in the case of D2, the length, is shown, and the perpendicular axis provides the frequency of occurrence of the length in this shape distribution.

With a comparison with a shape distribution 90 for a body from the database 1 (Fig. 9), a function for describing the deviation between the shape distributions 82 and 90 (both represented here stretched) is begun at 0, whereby the distances between the shape distributions 82 and 90 are added up according to specified criteria in an addition function 91. The criterion, for example, is the distance of the data sets to be compared. If the added-up value is greater than the value found to be greatest in the previous comparison, the comparison that is running is interrupted early. Such a summation method is known, for example, from the attachment "A Dynamic Programming Algorithm Optimization for Spoken Word Recognition", by H. Sakoe, S. Chiba IEEE Trans. On Acoustics, Speech, and Signal processing, ASSP-26(1): 43-49, Feb. 1978.

From the data base 1, bodies 101 (Fig. 10a) and 102 (Fig. 10b) arranged according to the degree of their geometrical similarities are found and represented on a display screen. The user of the database 1 now distinguishes if one of the components under the bodies, which has the greatest similarity with the desired component, comes into question for his purposes.

Since with an exhaustive comparison search, the comparison of each component stored in the database 1 can take a long time, the inquiry preferably is limited on all of the components with reference to a dimension that is easy to determine, such as the ratio between the surface and volume, in an upstream rough search, which is taken into consideration approximately for a comparison with the desired component. First, then, shape functions for the comparison are used. Also, other methods are suitable for accelerating the comparison between the desired body and the components contained in the database 1 , for example, parallel comparison strategies or the distribution of the computing steps on multiple computers running in parallel.

According to the invention, the comparison between bodies, which comprise only a single component, or those which combine a plurality of components, to bodies in a database is made possible. Thus, a vector for each body is positioned, whose elements provided all of the characteristics of the body, in particular, its geometrical characteristics, however, also other qualities, such as material characteristics, optical characteristics, magnetic characteristics, etc.

Claims

1. A computer system (100) for making available technical information, which includes a data processing unit (130), a database (1), a data input unit (110), and a data display unit (120), wherein in the database (1), data from a quantity of three dimensional bodies can be saved, characterized in that data of any selected body can be input via the data input unit (110) and that via the data output unit (120), at least one body (101, 102) can be output, which in the quantity of the three-dimensional bodies has the largest similarity to the selected body.

2. The computer system (100) according to claim 1, characterized in that the selected body (7) can be broken down into individual components (71 , 72, 73, 74,...) in the data processing unit (130).

3. The computer system (100) according to claim 1 or 2, characterized in that from the body (3; 21, 22, 23, 24, 25, ...), by means of a characteristic extraction, a description can be produced, by means of which the body (3; 21, 22, 23, 24, 25) can be compared with the description of the bodies in the quantity.

4. The computer system (100) according to claim 2, characterized in that the surfaces of the selected body (80) can be broken down into triangular surfaces (81) or other polygonal surfaces.

5. The computer system (100) according to one of claims 1 through 4, characterized in that from the selected body (80), shape distributions (82) can be produced.

6. The computer system (100) according to one of claims 1 through 5, characterized in that from the selected body (80), a ratio of its surface to its volume can be produced.

7. The computer system (100) according to one of claims 1 through 6, characterized in that the selected body (80) can be compared with the bodies in the database (1) with reference to the shape distributions and/or the ratio of its surface to its volume.

8. The computer system (100) according to one of claims 1 through 7, characterized in that the characteristic data set of a selected body (80) can be compared with the characteristic data sets of the bodies in the database (1) according to a method corresponding to dynamic time warping, wherein the shortest distance between the data sets can be found.

9. A method for comparing data sets of three-dimensional bodies, which are saved in a database (1) of a computer system (100) with a data input unit (110) and a data display unit (120), characterized in that via the data input unit (110), data of any selected body are input and that via the data output unit (120) at least one body is output, which in the quantity of the three-dimensional bodies, has the greatest similarity with the selected body.

10. The method according to claim 9, characterized in that the selected body (7) is broken down in the data processing unit (130) into individual components (71 , 72, 73, 74, ...).

11. The method according to claim 9 or 10, characterized in that from the body (3; 21 , 22, 23, 24, 25, ...), by means of a characteristic extraction, a description is produced, by means of which the body (3; 21 , 22, 23, 24, 25,...) can be compared with the description of the body in the quantity.

12. The method according to claim 11, characterized in that the surfaces of the selected body (80) are broken down into triangular surfaces (81) or other polygonal surfaces.

13. The method according to one of claims 9 through 12, characterized in that from the selected body (80), shape distributions (82) are produced.

14. The method according to one of claims 9 through 13, characterized in that from the selected body (80), a ratio of its surface to its volume is produced.

15. The method according to one of claims 9 through 14, characterized in that the selected body (80) is compared with the bodies in the database (1) with reference to the shape distributions and/or the ratio of its surface to its volume.

16. The method according to one of claims 9 through 15, characterized in that the selected bodies (80) are compared with the bodies of the database (1) with reference to the ratio of the curved surfaces to the level surfaces and/or with reference to the portion of the convex surfaces, with reference to the entire surfaces, and/or a two-dimensional comparison with algorithms from the type recognition and/or a voxel comparison.

17. The method according to one of claims 9 through 16, characterized in that the selected body is compared with the bodies in the database (1) according to a method corresponding to dynamic time warping, wherein the shortest distance between the data sets is found.

18. The method according to claim 17, characterized in that with cycling of the data set of a body from the database (1), the results of previous cyclings are taken into consideration.

19. The method according to claim 18, characterized in that the cycling is interrupted when it is determined that the similarity of the body to be directly checked with the selected body is lower than the smallest similarity from a list of that body from the database (1), which has the greatest similarity with the selected body.

20. The method according to one of claims 9 through 19, characterized in that a list of the most similar bodies (101 , 102) is displayed on a data display unit (120).

21. The method according to claim 20, characterized in that the most similar body (101 , 102) itself and the selected body are displayed on the data display unit (120).

22. The method according to claim 21, characterized in that the most similar body (101 , 102) and the selected body are added to one another and that the differences of their volumes are displayed.