The subject of this invention is a method of segregation of industrial articles. The invention applies to identification and segregation in pipelined inflow of products, including such in which successively inflowing products can be normally subordinated to the same or different categories of industrial articles. The invention also applies to segregation of products previously gathered in a group by means of another previously applied criterion. The invention combines technological operations on a set of industrial articles with technological operations performed by computer-assisted equipment.

Commonly known are methods of segregation of industrial products based on unequivocal assignment of indicated attribute to selected object being assessed. This assignment results in two possibilities: "true" or "false", which entails classifying an industrial article as fulfilling this condition or failing to fulfil this condition. It's of minor importance whether positive fulfilling this condition results in acceptance or rejection of a an ailicle under examination. Commonly known are also methods of segregation consisting in simultaneous examination of several concurrent attributes. The known methods, however, assess unequivocally each time whether a set of attributes assigned to conduct grouping actually occurred or not.

Known are also methods of identification and grouping or segregation of objects in which conformity of attributes of an object under examination with the pattern is assessed. Every method mentioned above can be effected in various ways that include implementation of equipment to use the described method. These implementations are often require analyses made by data-processing systems, predominantly analogue or analogue-to-digital converters. Such converters are preferably connected to analysers of received signals in the form of processing and logic units which are ideally supported by computer systems possessing memory chips or comparators to capture input data and compare it with model data previously saved permanently in a mass storage unit.

As an example may serve an invention described under number US5979240, which demonstrated a method and equipment to detect the type of materials of objects constituting waste, and these concern recycling. This method generates a sound wave of certain energy, which directs itself to an object under examination, and then analyses a resonant wave against a model sound wave emitted through resonance that was provoked as a model by a known waste material. Obtained results were compared in a sound wave analysis unit, and on that basis the object under examination was subordinated to a proper material group.

From another solution concerning invention revealed under number US5501344 is known a method of identification and sorting of individual formed objects or mixture of objects on the basis of a spectral analysis of materials from which these objects were made. In this method an electromagnetic wave is generated, particularly from UV spectrum range or visible band which is directed to the object under examination. Resonance return wave is registered by means of a sensor, and then saved as data set in the memory unit which also contains model resonant waves coming from known objects. In addition, saved data is compared, and the comparison generates a signal controlling the equipment that sorts out the objects under examination.

In the presented known methods, a classification is used that's related to statistical probability of receiving an identical signal. This is reasonable, because for the assessment of returned electromagnetic resonant wave there must exist an assumption that this result is influenced by a number of factors, such as e.g. arrangement of product, disturbances caused by the environment, variable conditions of measurement, etc. An example combining real data and measurement data with their digital representation as a data set is an international application with number PCT US 1999/015431, whose subject matter is computer assignment of values for a numerous set of attributes in a shortened way. This method demonstrates the supply of efficient, ordered and space- saving representations of multi-dimensional data. Data values for every attribute are saved in a way that's advantageous e.g. in terms of using space and/or quickness of gaining access, for instance in a compressed form and/or in the order of sorting.

A specific data value for a selected attribute has only one representation. This individual representation is associated with another one concerning a different attribute by means of unique assignment. A set of assignments identifies all data values for all attributes, which means that identical information is not reproduced, and only the distance among individual data values is indicated. In fact, such a solution may accelerate mathematical processing of a set of data, however, nothing beyond that.

What the last known solution shows is that it's useful to perform a series of assignments in the assessment of a numerous set of data, where data means the attribute and its corresponding value. Unfortunately, the last exemplary indication still remains a database record, and has no practical advantages for industrial processes consisting in a possibility of reliable change of the very method of sorting or grouping industrial articles.

It's possible that on the basis of the last indication of the known solution, I conjunction with those mentioned earlier, one can arrive at such conclusions that during industrial processes of sorting and grouping of articles it is possible to perform such assignment that one article having more than one attribute can be assigned to as many sets of data at a time as there are its attributes, or to as many sets of data at a time as there are all combinations of attributes and their values. Considering computer support of such processes, these assignments can be picked up and saved in the data mass memory for the purposes of comparing them as a model set later on to assess conformity with incoming single product that can be described by the attribute and possibly at the same time by a value for the attribute, where this attribute's values can be represented by a finite set of occurring events. It also seems possible to refer in a given method of grouping to previously completed operations, which through the probability calculus can be linked to events to come in the nearest or near future, as long as they are based on the same method of grouping as before.

Unfortunately, all these above-mentioned methods of segregating products restrict themselves to a specific number of compared attributes which are assumed as a base and finite set for selected products. Even if for a given attribute, in relation to which a product is to be assigned to by other products, there is a possibility of determining different values, it is restricted by their finite number that a given attribute can reach. This neither allows the appearance of a self-learning method, nor using a correlation among objects which seem to have no common attributes, or for which there is actually no possibility of applying a common attribute. The method according to this invention eliminates these inconveniences, as this invention's objective is the possibility of segregating industrial articles based on the most representative common attributes not known a priori, which can characterise articles subject to grouping. Another objective is through an applied method to improve its results also owing to the use of self-learning mode based on the main assumptions of the method according to the invention, It also aims at acquiring a possibility of automatic occurrence of an anticipatory step for the previously selected rules in the method according to the invention, as well as even a possibility of automatic determination of rules based on previous assignments.

According to this invention, the method of identification and segregation of industrial articles according to the invention consists in analysing them as a set of objects one by one, directly as a whole, and at the same time or sequentially by means of their digital representation formed thanks to a converter, preferably digital one, which combines the attributes of a given object with that object, and saves as a set of events in the memory unit. The set of events of other objects from the set of objects is saved in the memory unit, and the sets of events are compared using computer-assisted methods, and then the objects under examination are grouped as a result of the comparison. The invention is characterised by that at least one attribute of at least the first examined object is adopted as a representative attribute for a given group of objects that will be examined following the first object for conformity and appearance of the attribute in all objects of a given group. Then, the percentage of conformity with the adopted representative attribute is determined for every new examined object, and the examined object is located in the hierarchy of the set of objects the more farther on, the lesser its conformity with the representative attribute. With every pick-up of successive object for assessment of confonnity, successive attributes of an object regarded as representative for a given set of already analysed objects are checked against an object that's subject to checking for conformity, and a new set of objects conforming to the new rule is created in relation to the currently examined attribute by placing the examined object in the new set the more farther on, the lesser the conformity with the representative object for a given set of objects. It is assumed each time that the representative object for a given set of objects is one whose attributes demonstrate the highest convergence with the other objects of this set of objects, and all the objects have a measure included in their set of events, which measure is one of convergence with representative objects of different sets of objects. After assessment of the last picked-up object, the sets of events of representative objects are considered as rules of belonging to individual sets of objects.

All sets of events of individual objects, and at the same time or sequentially all grouped objects in sets of objects, and at the same time or sequentially all rules of belonging to individual sets of objects, are preferably saved in the memory unit, ideally in the mass memory unit.

Representative attributes, and at the same time or sequentially, representative objects, are preferably saved in the memory unit, ideally in the mass memory unit, ideally by combining them with sets of events of individual objects, and at the same time or sequentially with all grouped objects in sets of objects, and at the same time or sequentially with all rules of belonging to individual sets of objects. Consecutive object that's subject to conformity assessment is selected or rejected automatically from the memory unit, preferably from a mass memory unit, in which sets of events for this consecutive object were previously saved, and as the closest ones to representative attributes, and at the same time or sequentially representative objects, they are preferably called up based on the principles of probability.

Set of events is preferably defined, and at the same time or sequentially preferably introduced for the first object from the set of objects.

Attributes are preferably variable, and preferably change depending on objects in the set of objects.

Set of events preferably has also a value assigned to attribute, which value then becomes a fundamental element of comparisons. Attribute value can be a binary expression or an alphanumeric expression or a numeric expression, including a flat-point one.

Attribute value can be a complex expression, preferably as an alphanumeric or numeric vector of partial values. Attribute can be the colour, and at the same time or sequentially the size, and at the same time or sequentially the shape, and at the same time or sequentially the material property, and at the same time or sequentially the object name - customary or professional one, and at the same time or sequentially the product category of object name - customary or professional one.

Attributes are preferably assigned weights in the form of multiplier, and only after taking it into account a comparative assessment of object convergence is performed. Set of events can be formed automatically thanks to an analogue-to-digital converter connected with an analyser of spectrum, preferably visible light, and at the same time or sequentially set of objects can be formed by a pipelined inflow of products performed automatically or semi-automatically, also on operator's suggestion. An advantage of the solution according to the invention is high measure of reliability of acquired division of articles into groups of objects based not only on frequency of co- appearance of objects in an given set (cluster), but also on the strength of this co- appearance expressed in probability, i.e. conformity with the attributes of the majority of objects.

The solution according to the invention is presented in the below example of performance.

Exemplary method of identification and segregation of industrial articles consists in analysing them as a set of objects one by one by means of their digital representation formed thanks to a digital converter, which combines the attributes of a given object with that object, and saves as a set of events in the memory unit. The set of events of other objects from the set of objects is saved in the memory unit, and the sets of events are compared using computer-assisted methods by means of a software comparator, and then the objects under examination are grouped as a result of the comparison.

Set of objects is the following: silver watch with rectangular dial and brown strap, set of four off-road winter car tyres, grey SUV car with bike carrier, silver aluminium wheel rims 18", oval wooden bead necklace. The first item from the set of objects to be examined is the silver watch with rectangular dial. The first attribute determined for it is the dial shape. It was acknowledged that rectangle is representative as a feature of shape. Another attribute was recognised automatically, i.e. colour, and it was ascribed two values: silver (for the dial) and brown (for the strap). Third attribute, one of object category, was ascribed: jewellery. It was acknowledged that the watch was a representative object for the group of objects being analysed, as only one object was put to examination.

Second object from the pipelined stream of objects incoming for analysis was analysed, as previously using a digital image converter, which object appeared as the second in the sequence. It is a set of four off-road winter car tyres. Attributes were recognised in the form of shape: round, colour: black, and a category of object expressed by a group of features: spare part, rubber, set in motion. Attributes by type are successively identical, like for the first object, therefore after analysing the second object it can be assumed that the objects are located in one set of objects, however the value ascribed to individual attributes is totally different. Hence, the convergence of values for the attributes expressed in percentage terms equals zero, except convergence of black colour of the tyres against the brown watch strap, where this value is slightly higher and equals 20%.

In picking up another object in the form of a grey SUV car with bike carrier, attributes were recognised and assigned in the form of shape composed of its body and independent external elements in the form of four wheels and carrier, attribute of colour: grey, and attribute of category of object: means of transport. Out if it, on comparison convergence of colour was acquired at 70% with the colour of the watch dial, convergence of the colour of external elements at 100% with the colour of the set of winter tyres, and convergence of the shape of external elements (installed tyres) at 100% with the shape of the set of four off-road winter car tyres, and convergence of category of object in the form of a SUV car as a means of transport was assessed at 80% against component attribute in the form of the set of four tyres as the "set in motion" object. Thus, it was acknowledged that the SUV car would be closer to the set of four winter tyres, and because of that it would make up with them a two-element set of objects different from the one-element set of object in the form of the watch with strap. The representative object for the indicated two-element set will be the set of four tyres as consistent with both the SUV car and the watch, and the representative object for the one-element set will still be the watch with strap. All the objects in all the sets have their attributes and their values entered in the set of events, in which are also located elements in the form of percentage convergences of attributes and their values with one another.

The moment when the fourth object in the form of silver aluminium wheel rims 18" was recognised and assessed, attributes were automatically matched in the form of shape: round, colour: silver, category of object: metal spare part, set in motion. Convergence in this case, however, was assessed both with the set of four tyres and the SUV car on average at 90% and 80% respectively, while in a detailed breakdown into colour, shape and category of object, with minor departures from the indicated average and with the watch at 50% because of its dial colour. Therefore, the wheel rims were assigned to the previously two-element set which since then became a three-element set: tyres, SUV car, wheel rims 18". The wheel rims have become a representative object for the three- element set, because their convergence level reached the highest values in cross comparisons with the three-element set objects, and because the wheel rims are the closest in their convergence with the watch which stays in the one-element set, and still is the representative object for the one-element set.

In analysing the last, fifth object, oval wooden bead necklace, as previously were automatically recognised the following: shape: oval, colour: brown, and category of object: jewellery. Convergence with the watch was assessed at 100% for the attribute of category of object: jewellery, and at 90% for attribute "colour" by comparing the necklace's brown colour with partial brown colour of the watch taking into account its strap. The necklace, however, is oval, which gives convergence at 75% with the elements of the three-element set, i.e. set of tyres, SUV car and wheel rims 18". Bearing in mind the previous method of assignment based on convergence, the representative element was changed in the present two-element to which belongs the watch and the necklace. The necklace stayed the representative element as more convergent with the representative element of the three-element set, i.e. wheel rims, and its distance (convergence) in respect of the set of tyres is also very high, but much lesser is the convergence with the car. Currently, the sequence in the sets of objects is the following:

First set: necklace, watch;

Second set: wheel rims, set of tyres, SUV car. It must be assumed that the first element of each set of objects is the representative element for a given set of objects, however, the last one is the least convergent with it. The representative elements of all sets of objects are the most convergent, with one another, elements of various sets of objects, despite differences related to the general character of sets and differences of attributes saved in the sets of events.

Considering the presented rules of segregating and assignment of objects to groups, the below general rules of grouping and segregating industrial articles as objects analysed before segregation are genuine. At least one attribute of at least the first examined object is selected as the representative attribute for a given group of objects which as consecutive after the first object will be examined for conformity and appearance of attribute in all the objects of a given group. Then, the percentage of conformity with the adopted representative attribute is determined for every new examined object, and the examined object is located in the hierarchy of the set of objects the more farther on, the lesser its conformity with the representative attribute, and the set of events also has a value assigned to attribute, which value becomes a fundamental element of comparisons. With every pick-up of successive object for assessment of conformity, successive attributes, and their values, of an object regarded as representative for a given set of already analysed objects are checked against an object that's subject to checking for conformity, and a new set of objects conforming to the new rule is created in relation to the currently examined attribute by placing the examined object in the new set the more farther on, the lesser the conformity with the representative object for a given set of objects. It is assumed each time that the representative object for a given set of objects is one whose attributes demonstrate the highest convergence with the other objects of this set of objects, and all the objects have a measure included in their set of events, which measure is one of convergence with representative objects of different sets of objects. After assessment of the last picked-up object, the sets of events of representative objects are considered as rules of belonging to individual sets of objects. All sets of events of individual objects, and at the same time all grouped objects in sets of objects, and at the same time all rules of belonging to individual sets of objects, are saved in the memory unit, ideally in the mass memory unit, in this case a computer unit. Taking this into account, it should be noted that objects can be displayed on monitor screen according to the order of their appearance in the stream of objects incoming for analysis.

Representative attributes, and at the same time representative objects, are saved in the mass memory unit by combining them with sets of events of individual objects, and at the same time with all grouped objects in sets of objects, and at the same time with all rules of belonging to individual sets of objects.

Consecutive object that's subject to conformity assessment is selected automatically from the mass memory unit, in which sets of events for this consecutive object were previously saved, and as the closest ones to representative attributes, and at the same time representative objects, they are called up based on the principles of probability. Considering this for our example, it is therefore possible that what is going to be automatically assigned to the first set of objects is a set of round brown earrings, and to the second set of objects will be assigned a two-wheeled bicycle with silver wheel rims and black tyres as a transport vehicle which can be mounted on the roof of the SUV car. Then, both the bicycle and the set of earrings will become representative objects, thus dynamically changing the records in the mass memory.

Further analysing the example of performance, the below general rules of the exemplary methods are genuine.

Set of events is defined, and at the same time introduced, at least for the first object from the set of objects, and in the example of performance operations were carried out for all analysed objects.

Attributes are variable, and their value changes depending on objects in the set of objects. Attribute value is saved in the set of events and becomes a fundamental element of comparisons. Attribute value is an alphanumeric expression.

Attribute value is a complex expression as an alphanumeric or numeric vector of partial values.

Attributes in this example of performance were assigned each time a weight equalling one, in the form of multiplier, and only after talcing them into account a comparative assessment of object convergence is performed. Attribute can be the colour, and at the same time the size, and at the same time the shape, and at the same time the material property, and at the same time the product category of object name, customaiy one. Set of events is formed automatically thanks to a digital converter connected with a spectrum analyser of visible light, and at the same time it can be formed by a pipelined inflow of products performed automatically