CA1157154A - Method and apparatus for real time detection of faults in industrial objects - Google Patents

Abstract

ABSTRACT OF THE DISCLOSURE
Method and apparatus for the detection of faults in industrial objects and signaling such faults to handling apparatuses. According to the invention, the image is taken of a sample object by a television camera or other sensing device;
the image is digitized and stored in a frame memory. The digitized sample image is then divided into small areas, for each of which the square root of standard deviation for the luminance levels of dots belonging of each area is calculated and stored. The same operations and elaborations are followed for each object to be checked and the obtained values are compared with those Or the sample. When the differences thereof are above predetermined thresholds, an output signal is generated for distinguishing the fault and generally the faulty object and for controlling a handling apparatus for the objects.

Description

\

~7~

- "METHOD AND APPARATUS FOR REAL TI~E DETECTION OF ~AULTS IN INDUSTRrAL
o~Ecrs.~
This invention relat~ to a method and apparatus ~or industrial automatic quality control, i.e. ~or real time detecti~on and signallir~ of faults in objects or industrial products, either two-dimensional or three-dimensional, ~ni~orm or decorated, single or multicoloured. We define as fault those visi~le differences between the object and a sample thereo~ which cause a quality degradation Or said o~ject, 10. Many studies, researches and devices using ir,lage processing methods have been proposed for the detection of faults, such as spots, points, scores, chipped edges, irregularity in shape or in any case differences in the object to be checked as compared with the sample thereof. We can distinguish three r~ain trends in the image processing methods ror the investigation of faults or anomalies in objects:
1) Structural and formal analysis;

2) Methods for image processing (e.g., use Or Fourier transform);

3) Arithmetic methods.
As a matter Or fact, the ~irst method mainly deals with linear 20. shapes or two-dimensional ~eometries, such as in written character recognitions or in pattern making and planninng.
The second method has been hitherto successful only i~ laLoratory, as requiring on one hand great calculation p~wers and not be~ng sufficient on the other hand for an automati`c select;on ~ecause of ~571S4 r ;~u ir~ ~ c^ntinuol~s opeI-ator's inte~i~ltiol; a( ~sl y it aias in research, but carl not be applied to industria~ Jr~ctice-.
The third method does not solve at present th~ selectiorl problem becal~se Or not taking into acco~nt all or ~t least some o~ the big problems that may bè sul~marized as follows:
Illwninat _ n_unifo~mity: although accurate and thoro~ ly studied, it is extremely difficult to be obtained; in theory, from the standpoint of a positive industrial application, it is almost iml~ossible to assurè a uniform illumination of the object throu~hout the field 10. to be scanned. As a consequence, normally in any optical sensing system, each fault having a li~ht intensity difference (measured by its reflectance) lower than the 1um~nosity or brightness difference over the im~ge field, though provided by very good illuminators (thè best ones give differences not lo~er than 10%) r,ay not be detected.
As a matter of fact, besides illumination, also t~.e optical apparatus and detector (photodiodes, television tube, etc.) affect the illumination uniformity, often adding more differences. In fact, both do not have a homogeneous response throughout the field to be checked.
Detector Noise: every detector has its own noise, from the actual 20. detection to the di~ital conversion that nay be limited but not cut off; in thls noise the differences between dot and dt that should be detected as faults, "drown". This problem adds to the former problem with similar effect.
"Noise" of the object to ~e checked: the o~jects to be checked is for itself never fully homogeneous, but the surface has a "granularity"
of its ow~ with also high local differences which may be greater, between dot and dot, than the faults to be detected. I~is e~fect is less visible to the eye in some objects having a great apparent homogeneity, such as a unifo~m well made priht or a smoothed steel 3~ surface; on the other hand it is easier to detect this kind Or noise in more coarse industrial objects which have a rou~l or painted surface.

~L~57154 ~ io~ev~r t}~e pro~lem is actually the same becau~* in both c s~s one is lookir~ for faults having visibilit~ lir~ts only sli~;3tly greater than the granularity of the object to be checked, however reduced this granularity is.
Drawin~, edge and shape inaccuracies of the object to be checked:
the syste~s, for instance for industrial decoration, are always affected by inaccuracies that depend on the production system itself.
In some cases, such as for instance in the xero~raphy on tiles, silk or other objects, or in four-color process in a rotary machine~
10. the inaccuracy with respect to the sample may be dramatic.
Similarly, the dimension and shape of the object reveal differences from the sample of generally reduced tolerances, but again lar~er than the faults to be found on the "edGes".
The positive advantage of the method proposed ~y this invention is just to have solved these four problems and to have solved them in a completely automatic way, without the operator's intervention and in real time, that is, at such a speed that from 100 up to 200 objects per minute can be examined, according to the requirements of the production line.
20. Substantially, the method automatically provides the investigation process performed by hum3n eye, that rejects and does not detect differences due to illumination uniformity lack, noise of the object and inaccuracy of drawing and size, but it integrates the collected information, and distinguishes those particular differences that are referred to as faults.
To ~etter illustrate the characteristics of the method and apparatus that we claim~ we will no~ refer to the selection of ceranic tiles according to the presence and degree of faults; it is however understood that the method and appar2tus according to - 30. the invention may he used for the fault detecti`on on any o~ject ~571S4 or ~oduct, either ~;di~3~i~lal or tridi~i~~io~-~l, or fl^J'7 th~
~ault detectioll on ar~ i~a~e, that is positive or negative photo, print reproduction or the li`ke, for example on fabrics, metal plates and pieces, mechanical objects and the like.
Both the tiles referred to, and any other object or image to be checked may be composed of one or more colours, with regular or irregular drawings.
Hereinafter we will particularly describe the apparatus which, while used for multi-coloured objects, senses a black and white 10. image: it being understood that the same apparatus, with the addition of two further sensors and suitable colour filters (e.g.
a colour television camera) and the addition of as many systems for the conversion, storing and calculation, would sense the image of the object as resolved into its three basic colours and search for the faults thereof.
As a matter of fact the trichr~matic detection is only a multiplication by thnee of the monochromatic apparatus and will not be described in the following because of its substantial identity.
The basic problem in selecting ceramic tiles arises in that during 20. the varicus production steps (support, glazing and baking), differences may occur between the tile and its pattern, as visible to the nake eye, and degrading the quality of the tile, limiting its use in case of serious or high difference, and disrating it in case of slight difference. These differences or faults may, for instanc~, be Yery strong ar~ smQll spots, less strong but ~ide spots, chipped edges, cracks, clefts, gla æ lack and the like.
At present, tiles ~re exanined b~ skilled staff that, more or less quickly, divide them into first choice, other inferior choices and scrap.
- 3~- ~or the ceramic industr~, the problem Or the selection INade by ~71S4 the open.-lto~ is, the peI-~olmel's hi~ co~t, the t~1i~ng co~-.t c3~id the non~omogeneous selection: for the persor~lel there is the problem of a ~onotonous and dull work, with scarce human contents and requiring a constant physical and intellectual care w;thout possibility Or inattention. The problem of inattention, namely of he who, for one moment re ves his eyes from the point to be examined, is great as in all the sight checks and, toge~her with the different concentration capability, is the main cause of a wrong object classification.
10. Iherefore, it is the object of the present invention to provide a method and apparatus for the detection of faults in objects, faults that the human eye, with eventually the aid of an optical instrument, considers visible, classifyin them according to the producer's demand ar.d supply suitable signals to a mechanical system dividing said obiects into homogeneous groups as to quality thereof.
Generally, according to the claimed method, it is provided to effect the observation Or the sample objects through a television camera or any other optical observing means, capable Or outputting electrical signals (e.g. photodiode bar or pthotodiode matrix)j this 20. image is then converted from analogical signals, to corresponding luminance levels; these signals are then divided ard digitized according to a scale of luminance levels; these digital signals are then stored into a suitable frame memory, with biunivocal correspondan-ce with the signals. The digitized image is then divlded into small areas, for instance square areas, that in the enbodiment to be described in the following conprise 16 dots (e.g. 4 for each side), ~ut could vary according to applicati-ons, from areas comprising 4 dots (side 2) to any other reasona~le s-ize that does not exceed however, according to our experience, areas comprising 256 dots (side 16).
-- ~3- In each of these areas the square root of the standard deviation ~lS7~4 is calculated accordillg to th~ ronmula:

V n n2 wherein e~= square root of standard deviation of the involved area;
Y-i = luminance value of the individual dots pertaining to the area;
n = number of dots contained in the area.
The ~ ~values of each area are stored as characteristic data of the comparison sample object.
The same method, through the same apparatus, is followed for each 10. object to be compared and the~ values for each area of said object are directly compared with the stored sample values.
The ccæparison values which n~y be the difference compared with predetern~ned thresholds, are then used by a calculating device for the production of a signal suitable to control mechanical selection - and shunting apparatus.
The invention will now be described by way of example with reference to the accompanying drawings, in which:
Fig. 1 is a block diagram of the apparatus according to the invention;
Fig. 2 is a block diagram of t~e arithmetic unit for calculating 20. the square root of the st~ndard deviation for the sample and objects to be checked and comparison ~f the square roots of the standard deviation relat;ng to tfie same areai Yigs. 3A an~ 3~ are a grap~c exemplification Or the image transform~tion carried out on a decorated tile; and Fi~s. 4 and 5 are the detailed block diagrams of the arithn~tic ~ ~ ~7:~4 urit ou~ ~?d in Fig. 2.
In Fig. 1 the dashed block 1 designates a ~eneral purpose apparatus for the readir~ and division or subdividing the ir~ge into dots, for each of which a determined value of an analogic lurinance level signals is obtained; in the specific case the above mentioned reading apparatus is a black and white television camera 2, such as - a commercially available one according to CCIR standard with external synchronization, in front of which either the object 3 to be checked or the sample are placed.
10. m e television camera 2 takes the image of the object 3 located on a general han~ling apparatus ~' and translates the information into analogic signals that are supplied to block 4 for amplification and synchronism generation (synchronism generator commercially available at the CCIR standard with H, V, A, S signals), this generator, in addition to supplying all of the required synchronisms to the television camera, provides for signal amplification.
e analogic signal corresponding to the luminance level of each dot is supplied to an analogic-digital converter 5 (Datel mod. ADC
TV 8B) which converts the analogic signal into 6-bit digital values, - 20. supplying it to frame memory 6.
As above described, the reading device 2 is black and white television camera, but which in the application, may be replaced by any reading device more appropria~e to the handling system of the object to be checked, such as for example a sin~le dot image scanning system with a photo multl~lier ~or the sensing of a single dot or a sensing system with a si`licon photodi`ode bar or a solid state television camera.
~ or the sensing of the coloured image, the black and ~hite television camera 2 can be rePl~ced Ey a color televlsion c&mera or by a ~ 3- plurality of black and white television camelas provided with suitable i15~i4 fil-ter for the scanning of the image into its fundamental chromatic components: in this case the analogic signal may be divided and sequentially oonverted (one colour after the other) or in parallel for higher speeds by a multiplication of block 5.
m e digital signal representing the luminance value of each dot ocming out of block 5 is supplied to frame memDry 6 storing the values relating to the individual dots acoording to the sensing sequence thereof, which are therefore stored together with the ocordinates of the dot and then retrivable at any instant according to the 10. storing sequen oe or in a programmable ~anner.
A frame memDry 6 for oontaining, for example, a matrix of 512 x 512 dots is manufactured by TASCO S.p.A. - Milan, but it is available at other firms, such as Colorado Video Inc., Eoulter Colorado, U.S.A.
In Fig. 1, block 7 is a calculating device which is oonnected at one side to frame memory 6 and at the other to an arithmetic unit 8;
this calculating device 7 may be a ~ P for instance a ZYLOG Z 80*
manufactured by MY~EK Corp. Texas - U.S.A. mcdel OEM 80/4CPU with memDry RAM 80 A x 48 K bytes by the same firm.
20. The calculating device 7 controls the frame memDry 6 be reading the dots in the programmed sequen oe , supplying the same to the arithmetic unit 8 for the calculations to be hereinafter described in details, receiving the results therefrom and processing the signals relating to the object to be checked, for the operator or for the mechanical handling or selecting devi oe .
The calculating device 7 is programmed to read, in the frame memory, the dots of the stored image, of square matrixes of different sizes, in the example square areas of 4 dots per 4 dots, that is to say-16 dots per area for a total of 16.384 areas; the device 7 supplies 30. the numerical value of the first dot from the first area to the arith-metic unit 8, the block diagram of which is shcwn in Fig. 2, while *Tradem~rk ~571S4 ~, th ~ d ~ ipt~o~ of t}~ cks i~
~ a~ithn~?tic ullit 8 outlined in Fig. 2 calculatcs in r~ ll time the square root of the standard deviation of the lumin~ce values ~ccording to the fo m ula:
i V n ~2 and then compares the ~ values of said area between the sample and the object to be checked. The input datum from the calculating device 7 is a 6 bit numerical value xi Or the dot represented by the signals ~DP0~DP5> in the sequence of the 16 dots Or the area.

( ~ Xj ) On one hand to obtain 10- the luminance value xi Or each dot is supplied to the summing block 9 for the sum of the 16 dots: the result, which is represented by the signals <Sl S10 ~ , and is a 10 bit number as a sum Or 16 6-bit numbers, is supplied to block 10 which calculates the square thereof, dividing it by n (that is to say 256) and complement on 1, namely calculating the neg~tive value thereof.
The result of this operation at the output of block 10 is supplied as a signal ~QS0iQS15~ to summihg ~lock 11.
The input datu~ ~DP0~DP5~ is simultaneousl~ supplied to the two input groups x ~nd y of block 12 for the calculation of the other term ~ X2 ~7~S4 s ~ 3tt~ f ~a~t, ~c(i~ aC ~l ir~l~ut ~.;ic~ t~; s.~
n~lical va~ue of the dot, tile2-. multip~ s eaci~ Ot.!el', that iS tG
say calculates the square value thereof, sequentially adds the sou~ e values and divides the ~esult by the rlw ~er n o~ ~ots, r~mely 1~ in the present exa~,le.
The output datum ~ SQ0~SQ15~ is supplied to the summing block 11 in order to subtract therefrom the value ~QS0~QS15 ~ coming from block 10. The output of block 11 is the result of the difference, that is a 10 bit signal <D0.D9~ (the 10 bit difference coming 10- out of block 11 is the result Or the cut of the less significative

4 bits for next reduction of the calculation).
The sig1lal <D0 D9> is supplied to block 13 which estimates the last stage of the calculation for the square root of the standard deviation, that is to say the square root of the difference.
The result at the output of block 13, namely ~R0-.R7~ is supplied to the calculating device 7 and contemporaneous to the c~mparator 14 comparing the value of the square root of standard deviation ~ for the object to be checked with the square root of standard deviation CRC0~RC7~ from the calculatir~ device 7 and relating to the same 20. area of the sample object.
The comparator will output a positve signal <PLUS~ to the calculating device 7 when <R0~R7~ is greater than CRC0iRC7~
When the arithrrietic unit 8 is processin~ the dots of an area of the sample, the device 7 supplies signals ~RC0~RC7~ equal to 0 to the comparator; the output signal from the comparator will be always ~PLUS~ and the output result from block 13 ~Ræ~R7~ will be stored by calculatir~g device 7.
On the other hand, when thé arithmetic unit 8 is operating on the image of the object to be checked the calculating device 7 SUPP1ieS
- 30. through ~RC0.RC7~ the corresponding values Or the sample and the cor~arator 14 will provide positive differer-lces with si~lal <PLUS ~ only ~S71S4 ~ e~ t~ to b.~ cl~t'~ }~ t~r v~lu~;~s t~ t~1~? s'~,lc;
howev~r, t}~e si~ials <RCY~RC7~ supplie~ from device 7, n~y b--either the squ~ e roots of the standard deviation calculated over the s~nple, or the sar~e values added or multiplie~ by a predetel~irled factor.
The calculating device will utilize these positive values as fault signals or referred to an area to be checked.
We have schematically described the calculation process carried out by the arithmetic unit 8; and the result of such a transformation 10. of the sarnple in ge into the square root of standard deviation referred to small areas is stored into the calculating device 7: this new sarnple image which is transforrned into ~ values (and no longer into luminance values) can be defined as an image of the object variability: that is to say no longer the in ge in its normal shape or as commonly seen, but represented by its deviations.
To numerical values ~ = 0 a fully uniforrn area would correspond, that is to say without any differences from dot to dot, whatever may be the luminance value of said area; at maximum values of6' , an extremely contrasted area would instead correspond, that is to say 20. an "edge" area, which has half white dots and half black dots.
This transformation is schematically represented in Figs. 3A and 3B, where Fig. 3A represents a normally white tile with a round grey drawing and a square black drawing.
On the other hand Fig. 3B represents the same tile transformed according to the above formula, ~hich in this case will have a square root of standard deviation equal to 0 over the whole tile left white, and higher values of ~ on the three edges: tile edge, circum~erence of the round drawing and perimeter of the square.
The calculating device 7 stores the parameters of the sar~le 3o. image and is pr~grammed to multiply them, for instance by a multiplicative factor, for causing the values of the parameters to ~57~S4 , ~e i3~ `i';;~ Uli',il t~ ;'Siit`~ VGl~r~3nce iS ol~tairleù; that i."
~o ~r~vi ~ that the difference, as calculated by co~p~ator ]~, t~iit~
the object to be checkea, is positive only when the object to br checked actually has a contrast that inplies the presence of a fault.
T~is multiplicative factor (or adding factor) is experimentally determined according to the type of objects to be checked, the characteristic of the faults, and also according to the lighting type and readin~ apparatus.
The calculating device 7 is also programmed to add to one another 10. all Or the values of positive difference, or only the adjacent values, so as to eliminate the small casual differences and transmit different fault signals to the selecti~3 and handling device according to the fault magnitude (difference value) and fault size (value of the difference sum).
In Figs. 4 and 5 the detai]ed block diagr~m of the arithmetic unit 8 outlined in Fig. 2 is represented.
In Fig. 4, blocks 9 and 10 of Fig. 2 are shown: to calculate the sum schematically represented in block 9, the signals ~DP0~DP5~
of the first in-coming dot from calculating device 7 are transferred 20. to the inputs <ID~6D ~ Or register 17 (74 LS 374 circuit) and, to the leading edge of a signal <CI~X~ from calculating device 7 are transferred to the outputs ~IQ~5Q~ of said register, where they will remain until the leading ed8~ of the next signal ~CKX> .
Upon the appearance of this last signal, the data, present at outputs ~lQ.6Q~ of register 17, are trans~itted to inputs <Al.A4~ , more simply referred to as inputs A of summing devices 18, 19 and 20 (summing devices 7483); said summing devices 18, 19 and 20 add the values present at inputs A to the values present at inputs B which, at this first stage, are still equal to 0.
- 30. The result of this first sum, present at the outputs Q Or the ~1571S4 l, ahov men~ion~à SUl.~nil-~ device, is supplied to inputs D of reg~.ter 21 ~ld 22 (r*gisters 74.174) and, to t~e leading edge of signal ~CYP~ from calculating device 7 are then transferred to the outputs Q;
this datum, which is the result of the first sum, is supplied to inputs B of the summing devices 18, 19 and 20, as outlined in Figure 4.
The numerical information Or the second dot <DP0~DP5~ at the arrival of a new signal ~CKX~ will be again supplied to inputs A
of summing devices 18, 19 and 20 and added to the values of the 10. first dot now present at inputs B of said summing devices, which devices will compute the second sum which is again stored in registers 21 and 22 which, at the leading edge of the next signal ~CKP~ will allow the transfer thereof to inputs B of sunning devices 18, 19 and 20.
This adding operation is repeated 16 times for the 16 dots of the involved area (generally n times) and, at the end of the 16 additions, in the registers 21 and 22, the sommation of the 16 dots will appear ( ); this summation at the signal ~CLR? from calculating device 7 (which clears the summing devices in readiness for the next 16 dots) 20. is supplied to memories 23 and 24 (memories EPROM 2708) which are shown in Figure 2 by block 10.
These memories 23 and 24 are programmable and contain, in our case, some tables in which, at each address presented to their inputs A, corresponds at outputs D a value, equal to the square of said ~alue divided by n (256 in the case being examined) and conplemented to 1: the outputs D of memories 23 and 24 are then the second term of the formula under square root, already made negative, and is supplied as signal ~QS0~QS15~ to the summing device 11 to be described below.
Fig. 2 shows block 12 commprising an accumulator multiplier (TRW-TDC 1010 J).

~lS7~i4 r.l~ si~ Dr~P5 ~ reLatil-~ to the n do~s ol the irlvolved a~ea is supplied to ;nputs X and Y. In said multiplier 12, the data ;DP0~DP57 , contenporaneously presen~ at said inputs X and Y, are multiplied to one another at the arrival Or si~nal CCKX> from calculating device 7; the result is added to the dots which were already present at the arrival of signal ACC from calculating device 7 which, for the first dot, is equal to 0 and therefore does not cause any addition and, for the next 15 dots, is equal to 1 and causes the addition thereof to the preceding sum.
10. At he signal <CKP~ this sum is stored in the output register of multiplier 12 and, at the 16th or last signal, the result is presented to the 16 outputs P of said multiplier for supply as signal SQ0~SQ15~ to the inputs B of the set of four summing devices 25, 26, 27 and 28 (summing devices 7483), see Fig. 5, forming the summing block 11 of Fig. 2; as it will be seen from Fig. 5, each summing device supplies to the next summing device a CARRY signal which can be a logical level 0 or 1 depending on whether the input bits overflow or not the summing device capacity, with the exception of the last summing device in the set, in which the cARRY signal is 20. ~always at level 1, because of the required complement of the output sum.
In Fig. 5 the inputs A of the summing devices 25, 26, 27 and 28 receive the signal ~QS0~QS15 > from memories 23 and 24 of Fig. 4 and, as abo w e referred to, the inputs B of said summing devices receive the signal ~ S~0~S~15> from multiplier 12.
Taking into account that <~S0~S15~ is already a negative datum, the result of summ~n~ de~ices 25, 26, 27 and 28 is the difference under square root o~ the preceding formula. Only the most si~lificant outputs (10 in the case ~ei~g examined) of this result -- 30. are connected to block 13 (mem~ry ~PR0~ 2708).
Also this memory is a pr~gramnable memory w}~ich in this case ~S71S4 contairrl a table, so that upon pr~sentation of the ~ifierellce result to tle inputs A <I~D9> , the datum at the outputs ~ R0.-R7~ is the square root of said result. If desired, the table could also be pro~ ammed to provide the square root logarithm.
The comparator block 14 of Fig. 2 is shown in ~ eater detail in Fig. 5, in which the blocks 29 and 30 are comparators 74S85.
The inputs A of the two comparators 29 and 30 are presented with the results <R0.R7> of bloclc 13; to the inputs B of said comparators the calculating device 7 supplies the data (RC0;.RC7) relating 10. to the sample, for the same area on which the calculation has been - carried out.
The comparator 29 receiving at inputs A and at inputs B the most significants bits of the information, supplies to the second comparator 30 one of the following three signals:
Z if the number at inputs A is greater than that at inputs B;
W if the number at inputs B is greater than that at inputs A;
K if they are equal.
20. m e comparator 30, receiving the least significant data of the information and the above described signals will supply the sigpal C PLUS ~ if ~ R0~R7~ is g~eater than c RC0~RC7~ as initially referred to.

. .

Claims (9)

The embodiments of the invention in which an exclusive property or privilege is claimed are defined as follows:

1. A method for the detection and signalling, in real time, of faults in industrial objects, by comparing the objects to be checked with a sample, according to which the object image is sensed, and the analogue signals of the luminance level of a finite number of dots of the object image, are converted into corresponding digital signals according to a scale of luminance levels, said digital values being stored into a frame memory, developed characteristic parameters being provided for each image, which are compared with the corresponding parameters as sensed on a sample image, whereupon a signal is generated for distinguishing the scanned object, characterized in that the characteristic parameters are obtained by:
- dividing the digitized object image into predetermined number of areas;
- calculating the square root of standard deviation for the luminance values of the dots of each area into which the object image has been divided according to the formula:

wherein: Xi = luminance value of the individual dots pertaining to an area;
n = number of dots in the area;

- storing the square roots of standard deviation ? of the areas of the object image.

2. A method according to Claim 1, characterized in that the square roots of standard deviation for the areas of the sample are multiplied, prior to comparison by a predetermined multiplicative factor.

3. A method according to Claim 1, characterized in that the square roots of standard deviation for the areas of the sample are added, prior to comparison, to a predetermined factor.

4. A method according to Claim 1, characterized in that the square roots Or standard deviation are transformed into logarithms prior to comparison.

5. A method according to Claim 1, 2 or 3, characterized in that the calculation and comparison steps are simultaneously carried out as to the fundamental chromatic components into which the image of the object to be checked is divided.

6. A method according to Claim 1, 2 or 3, characterized in that the calculation and comparison steps are sequentially carried out, as to the fundamental chromatic components into which the image of the object is divided.

7. An apparatus for real time detection of faults in industrial objects, including a device for the detection of the object image, connected, through an analogic - digital converter of the signals corresponding to the luminance values of the dots of said image object, to a device for the sequential storing of the digital signals, characterized by comprising a calculating device connected to said storing device and acting to divide the digitized image of the object into a predetermined number of areas and in that the calculating device is connected to an arithmetic unit for calculating the square roots of standard deviation for the luminance values of a predetermined number of dots of each area of an object to be checked and for the comparison of the stored values of the square roots of standard deviation for the corresponding areas of a sample object, provided by said calculating device.

8. An apparatus according to Claim 7, characterized in that said arithmetic calculation unit includes a multiplier, to the two input groups of which the same signal corresponding to the luminance values of the dots of a same area is supplied, said multiplier being suitable to add the square values of said values dividing the sum thereof by the number of said dots:
- a first summing device, to the input of which said signal of the luminance values is supplied, connected to a suitable squaring device and complementing on 1 the sum of said values;
- a second summing device, to the inputs of which the output signals from the multiplier and squaring device are supplied;
- a programmable memory for calculating the square root of the digital output signal from the second summing device; and - a comparator device; the output of the programmable memory being supplied to a first input to the comparator and respectively to said calculating device which, on turn, supplies signals relating to the stored square roots of standard deviation of the sample, to a second input to the comparator; an operative connection being also provided between the comparator output and the calculating device, in order to enable the latter to the reception of the output signals from said programmable memory.

9. An apparatus according to Claim 7, further characterized by an operative connection between an output of the calculating device and an apparatus for the handling of the objects, operable by a signal on said output corresponding to a sensed fault.