DK170845B1 - Installation, apparatus and method for on-line determination of quality characteristics of pieces of meat - Google Patents

Description

DK 170845 B1 i

Plant, apparatus and method for on-line determination of quality characteristics of meat cuts

The present invention relates to an installation for on-line determination of quality characteristics of meat pieces for quality control or sorting, which system comprises an S conveyor with a belt for conveying the meat pieces, a lighting arrangement arranged over the belt for illuminating the meat pieces, and in connection with the arrangement provided a color camera for recording three frames of the cuts in the red, green and blue spectral range, and a data equipment for storing the frames in the form of RGB light values, analyzing the stored frames and determining the quality characteristics of the cuts on the basis of the analysis.

Boiled hams are made from several smaller muscles that are filled in shape, intestine or can and cooked to a coherent product that can be cut into thin slices for use in sandwiches, for example. In the main, four different types of pork cuts are used: hip, inner thighs, outer thighs and lump. The muscles are delivered to the production site in vessels, which typically contain 500-1,000 kg of meat. Muscle checks are performed randomly, with some samples of each batch being evaluated visually at the production site for possible defects, such as insufficient fat depletion of the muscle surface, bleeding and blood stains in the muscle, PSE spots in the flesh and other unwanted occurrences (PSE means "Pale, Soft" , Exudative "). In addition, inspection visits are made at the suppliers.

20

It is important to keep the aforementioned defects at a low level, as they may result in a substantial waste in slicing the ham at the user or may even result in the buyer of a batch of ham returning the delivery due to quality dissatisfaction.

25 Proper trimming of fat on the surface of the muscles is important for the client's visual impression, and in addition, too much fat can make it difficult to comply with a 98% fat-free, lean meat declaration. Bleeds in the meat appear very clearly in the finished product, as the areas darken during the heat treatment and give the product a uniform appearance. PSE instances in the muscles cause problems because the meat after the cooking 170 17045 B1 2 is so loose that it cannot be sliced properly. This significantly reduces the yield of the slicing unit. In addition, PSE meat is responsible for greater boiling winds in the production of cooked hams.

A study has shown that over half the muscles in a production can have trimming errors in fat. Therefore, there is a need for improved control of the raw materials, S used in production. However, this can only be achieved by examining every one of the pieces of meat used for common errors, such as insufficient fat trimming, bleeding and PSE occurrences.

Manual inspection of each piece of meat seems to be extremely labor intensive and is subject to the fluctuating ability of the inspectors. It is tiring to examine 10 pieces of meat for hours every day, and different controllers also evaluate differently.

An objective control using video systems could be another option. In this connection, it should be mentioned that GB-A-2,144,533 (Analytical Instruments Ltd.) describes an image analysis system for determining the fat content in pieces of meat lying on a black substrate. The system has a video camera whose signal output is connected to one comparator 15 input. When the video signal exceeds a threshold applied to the second input of the comparator, the comparator starts a counter whose final value after an image throughput expresses the proportion of light portions in the image, such as fat areas. The counter data is stored in a memory. The threshold is then reduced, and in another image passage the proportion of portions lighter than the substrate is found, ie. the total area of the meat and 20 fat surfaces. The fat area is eventually determined as the difference between the counter values obtained at the first and second image runs.

The system preferably employs a monochrome video camera, but there is also described an embodiment comprising a color video camera with outputs for a red, green and blue video signal. The signals are fed to a comparator with an automatically switchable threshold. The comparator output signal is stored in a red, green, and blue bit plane whose data can be transferred to a computer for further processing.

DK 170845 Pg 3

The system described can, by sufficient difference between the meat and fat surfaces, separate these from each other and in these cases can be used to determine the fat content of meat. However, the system is not able to separate occurrences that differ to a lesser extent, such as fat from light meat.

5 WO 93/21597 A1 discloses an image analysis system for meat by means of which a color video image is recorded of the meat surface. Thresholds determine whether a pixel represents, for example, meat or fat. The system suffers from the same disadvantage as the system described in the above-mentioned publication GB-A-2,144,533, in that it is not able to separate areas that differ less clearly from one another.

GB-A-2,187,281 (Analytical Instruments Ltd.) describes a similar system as in GB-A-2,144,533 for determining the fat content of meat. A red filter, located in one half of the camera's field of view, allows detection of meat and fat areas together. A green filter located in the second half of the image field detects the fat areas. A conveyor belt passes the meat through the image field, so that all meat measured in one half of the image is also measured in the other half.

The video system described provides information on fat content, but no information on PSE occurrences, bloody meat or other quality characteristics of the meat, such as the location of the fat areas on the pieces of meat. In addition, there may be problems synchronizing the images when recording images of meat on a continuous tape.

20 Such systems are therefore completely inadequate for the control of meat in connection with the said ham production.

An unpublished preliminary study of images of pieces of meat with PSE errors and bleeding using a conventional color video camera with red, green and blue bar filters showed that even with color captures it is very difficult to separate and distinguish PSE-25 areas from light meat and bleeding from dark meat. This applies both to visual viewing of the images on a monitor as well as to the usual digital analysis of an image stored in memory. The study also found that the time consumed for a recording and DK 170845 B1 4 analysis cycle is so large that the system cannot be used for continuous objective control of every piece of meat in a ham production.

It is the object of the present invention to provide a plant which, substantially better than the known systems, is able to determine on-line the essential quality properties of ham production of meat pieces, in particular insufficient fat trimming, PSE deposits and bloody areas of the meat.

This object is fulfilled by the system of the present invention, characterized in that it comprises a three-chip CCD color camera which captures a sub-image in the red, green and blue area of each chip and is adapted to record and store corresponding 10 RGB light values derived from the same physical (actual) surface portion of the meat subject, and the data set contains a three-dimensional table that, for each accepted combination of RGB light values, refers to a specific surface type and a program portion that compares stored, related RGB light values with the three-dimensional table embedded in the data set, and assigns pixels in the composite image containing 15 an accepted combination of RGB light values, the table's meat type code.

The system according to the invention is substantially better than the known systems for separating instances which differ in color only slightly, and in addition it has an improved signal processing which makes it suitable for on-line determination. It is even capable of determining quality characteristics while the meat is in motion, taking in parallel 20 and storing three frames and a CCD camera can easily operate with short exposure time.

The system uses a three-chip CCD color camera so arranged that it occupies a color image constructed from RGB light values, in which the associated RGB light values originate from the same actual micro surface of the meat blank, and those on each chip in the camera 25 recorded values are stored in a data set for subsequent analysis. Hereby, an image particularly suitable for analysis is recorded and stored with a very "true color rendering", and the recording and storage is performed in a short time using three chips in the CCD camera.

5 DK 170845 B1 Both insufficient fat trimming and widespread areas of PSE and bleeding can be detected satisfactorily, and a detection rate acceptable for practical conditions (one piece of meat per second or from 1500 to 5000 kg of meat per hour can be worked, depending on meat-type).

5 Using the three-dimensional table, a finer separation between the different surface types is obtained than at the threshold values used in known systems, the total color space being divided into curved, closed surfaces, each enclosing a space for each surface type.

The table allows the utilization of a higher percentage of the stored light values and the 10 light value combinations which have little or no relation to the surface types sought are eliminated immediately, saving data power during the subsequent analysis. A lookup in the table provides direct information on the surface type of a pixel, making the identification process extremely fast. The table therefore allows a quick and accurate determination of the surface type.

15 The surface types can be, for example, fat, PSE or bloody meat. A particular surface type typically constitutes one or a few spaces in the table. The unacceptable combinations of RGB light values are, for example, the background, peripheral areas of the meat or indeterminate areas.

In a preferred embodiment, the camera is arranged to simultaneously record associated RGB light values. In this way, the influence of the movement of the pieces of meat is eliminated by the conveyor. The time taken for recording and storage will thus be shorter.

The data equipment may contain a program portion repeating the comparison and assignment procedure for a substantial portion of the associated RGB light values derived from the surface of a piece of meat, segmenting the composite image area depicting the surface of the piece of meat into regions of different surface types, and then 25 assigns the piece of meat to a quality class for quality control or sorting.

DK 170845 B1 6

The data equipment may include a program portion for unmasking the piece of meat and omitting the edge of the piece of meat.

In addition, there may be a program section that determines whether the image is occupied by one side of the piece of meat, such as a muscle, and then selects the corresponding table.

5 It will also be possible to insert a program part that determines the type of muscle, for example whether it is hip, inner thigh, outer thigh or lump, and then select the corresponding table. It can be important if there are different types of muscle and these are delivered mixed from a conveyor belt or in the batch to be examined.

The camera may be arranged to record a frame in the range of 390 to 440 nm, a 10 frame in the range of 500 to 570 nm, and a frame in the range of 620 to 700 nm. In this connection, the blue region is displaced toward the UV region and the red region toward the NIR region to provide better detection properties. The camera can also operate in the usual ranges of 420-500 nm (B), 500-570 nm (G) and 570-640 nm (R).

The camera used preferably has an electronic shutter whose shutter speed is set to 15 1/500 seconds or less, whereby the resolution and "color fastness" of the recordings are maintained even if the subject is in motion.

The pieces of meat can be illuminated with visible light and / or UV light. Preferably, broad spectrum light is used which covers the entire visible area. The lighting can be continuous or pulsed (flash).

20 Spectral neutral light enhances the ability to detect different types of occurrences. Diffuse lighting reduces shadowing and mirroring of the light sources on the surface of the meat.

Therefore, an illumination arrangement is provided which provides a diffuse, spectrally neutral foreground illumination of the pieces of meat.

7 GB 170845 B1 The meat pieces conveyed during the lighting arrangement may have a lighting of more than 1000 lux, preferably more than 5000 lux, to provide good depth of field and reduce the noise in the electronic image.

The pieces of meat under the illumination arrangement can be illuminated more strongly on the sides than on the top 5 facing the camera, so that the edge area appears to a greater extent with the same strength as the center portion, seen from the camera's point of view.

The lighting arrangement may comprise a downwardly open housing along which walls are provided light emitters, in particular fluorescent lamps, and may contain devices for forced air circulation about 10 light emitting diodes. In addition, means for tempering the circulated air can be found. This stabilizes the light emitters of the light sources, both in terms of spectrum and intensity.

The illumination arrangement may contain devices which regulate the performance and / or color composition of the luminaires to a constant level, preferably the voltage of DC-powered fluorescent lamps, driven by HF chokes.

15 One or more subjects whose surface serves as a reflection and / or color reference may be arranged in the camera's image field, so that the images recorded can be continuously corrected for errors caused by light source instability or the camera. The reference topic may comprise three surface fields, one of which is selected as the reflection and / or color reference based on an assessment of the least possible soiling.

20 The camera and data equipment may be arranged to record and store associated RGB light values in more than 100 levels, such as 250 levels. Using many levels, the system's color resolution ability is improved so that the system is better able to distinguish unwanted occurrences from the "normal" surface types.

The system downstream of the lighting arrangement may comprise a second, corresponding lighting arrangement with camera, and the conveyor may have between the arrangements a strip section arranged to face a camera portion hidden from the first arrangement of a meat piece against the camera in the second arrangement. Checking or sorting is significantly better when examining the pieces of meat on both sides.

Along the conveyor belt arranged according to the image recording equipment may be designated seats S for meat reservoirs with differing properties, and next to each seat may be a baffle or the like, which can be swung over the belt by a sorting unit. At the end of the conveyor may be a reservoir or other conveyor for pieces of meat which do not or only have insignificant properties.

Preferably, the conveyor is arranged to run during image recording. The camera and data equipment may be arranged to record and process images of a new subject at less than five seconds interval, so that the system can follow the normal production speed.

The conveyor may consist of several sections, the speed of which can be continuously controlled independently of one another, so that the pieces of meat are conveyed under the camera at a uniform rate and at the required spacing.

The system may comprise a sensor arranged at the conveyor adapted to trigger the image recording.

The plant may comprise a portion for individually feeding a piece of meat from a storage container to the upstream end of the conveyor.

The invention also relates to a method for on-line determination of quality characteristics of meat pieces for quality control or sorting, which method comprises feeding the meat pieces by means of a conveyor with a band, illuminating the meat pieces by means of an over-band lighting arrangement, recording of three sub-images of the pieces of meat in the red, green and blue area by means of a color camera 25 arranged in connection with the arrangement, storage of the sub-images in the form of RGB light values, analysis of the stored images and determination of the quality characteristics of the meat pieces on the basis of the analysis.

9 DK 170845 B1

The method according to the invention is characterized in that a red, green and blue area image is recorded on each chip in a three-chip CCD color camera, that the corresponding RGB light values derived from the same physical (actual) are recorded and stored. surface portion of the meat subject, that stored associated RGB light values are compared to a three-dimensional table embedded in a data set 5 which, for each accepted combination of RGB light values, refers to a particular surface type and that pixels in the composite image containing an accepted combination of RGB light values is assigned the table's meat type code.

According to a preferred embodiment of the method, associated RGB-10 light values are recorded simultaneously.

The comparison and assignment procedure can be repeated for a substantial portion of the associated RGB light values derived from the surface of a piece of meat, and the area of the composite image depicting the surface of the piece of meat can be segmented into areas with different surface types, after which the piece of meat is assigned to a quality class. 15 for quality control or sorting.

The light emitters of the lighting arrangement can be kept at a constant temperature by forced air circulation about the light emitters.

In a particular embodiment, three additional sub-frames of the meat piece are taken by means of a downstream of the illumination arrangement, second camera having a corresponding illumination arrangement, with a concealed side of the meat piece facing the camera in the second arrangement prior to image recording.

The invention further relates to an apparatus for on-line determination of quality characteristics of meat pieces for the purpose of quality control or sorting by means of the plant concerned. The apparatus is characterized in that it downstream of the illumination arrangement 25 comprises a second, corresponding illumination arrangement with a camera, and that the conveyor between the arrangements has a ribbon section adapted to face a camera portion concealed by the first arrangement of a meat piece towards the camera. the second arrangement.

DK 170845 B1 10

Checking or sorting becomes significantly more secure when meat cuts are examined on both sides using this apparatus.

The reversal can be simply provided by the band section at its upstream end being lowered relative to the band passing under the first arrangement.

The invention is preferably used in the manufacture of a meat product composed of several smaller pieces of meat cooked in mold, intestine or can. Each of the smaller pieces of meat is checked for excess fat, widespread PSE areas and widespread bloody areas while being fed by the tape, and pieces of meat with differing properties are sorted out before the approved pieces of meat are further processed.

The invention can generally be used for examining one piece of meat at a time or for examining several pieces of meat at one time, e.g., a batch of pieces of meat placed in a box conveyed by the conveyor.

The following types of meat may be examined: - boned hams and ends, IS - pigs, sheep and beef trimmings, - pork and beef cuts, - poultry and fish, and - other types of meat.

Above is mentioned in particular sorting and quality control on the basis of fat, PSE and bleeding, but the invention can also be used in connection with other types of occurrences, for example for the detection of severe, tendon, cartilage, bone, jelly or foreign matter, such as plastic or metal parts.

The invention will be explained in more detail below with reference to the drawing, in which fig. 1a shows the front part of a system for detecting meat pieces 25 with different properties and sorting the meat pieces into four different categories from above. 11 FIG. 170845 B1 fig. 1b the rear part of the system, fig. 2a the one shown in FIG. 1a shows a side view of the system in simplified view, fig. 2b is the one shown in FIG. 1b is a side view of the plant, FIG. 3 is a perspective view of a lighting arrangement used in the plant; FIG. 4 spectra of different surface types of meat; S deviations of the surface types from the mean spectrum, fig. 6a-6c examples of locations for dark (M) and bloody (B) meat (outside of hip), fig. 7a-7c are examples of locations for light meat (L) and PSE meat (P) (outer 10 of outer thighs), and Figs. 8a-8c examples of sites for obstacles (H) and fat (F) (outside of hip).

The front portion of the plant (Fig. 1a) serves to separate and individually feed meat pieces to the rear of the plant (Fig. 1b). The front portion comprises a feed hopper 1 capable of holding 15 1000 kg of muscle or meat cuts 2 which are tipped into the hopper 1 from a transport vessel. At the tip of the funnel is a locking device 3, the opening area of which can be adjusted so that the muscles run more or less unobstructed through the lock. Immediately after the locking device, there is a metering wheel 4 with finger-shaped drivers projecting radially from the center of the wheels. The speed of the wheel can be adjusted so that the muscles conveyed by the carriers 20 are released to a subsequent conveyor belt 5 at a specific rate determined by the system's control system. A flap 6 which is freely pivotable about a horizontal shaft, which is arranged above the conveyor belt 5, provides for the alignment of the muscle so that it lies flat on the belt. Along the sides of the conveyor belt 5 are arranged four adjusting members 7, each comprising a baffle hinged on vertical hinge pins and under a counter force of 25 springs which can swing backwards from the shown resting position when a muscle 2a advancing on the conveyor belt bears against the baffle. The two alignment members on the upstream portion of the belt serve mainly to guide the muscles towards the center line of the belt, while the two alignment members located on the downstream portion are intended primarily to rotate the muscles so that they lie longitudinally in the longitudinal direction of the belt.

DK 170845 B1 12

Above the conveyor belt S is a sensor 8 that notifies the control system when a muscle is present under the sensor. Taking into account, inter alia, the sensor signal can control the speed of the tape. It ensures a more even beat of muscle down on a subsequent conveyor belt 9. This belt serves, among other things. as a buffer. The main function of the belt, however, is to rearrange the muscles so that they lie with the long joint across their direction of advance. The band can also be speed controlled and has a sensor that detects when a muscle is transferred to a buffer band 11, shown to the right in FIG. la.

The muscles are transferred one by one from the buffer band 11 to the rear of the abutment (Fig. 1b), being transferred to a conveyor belt 12, shown in Figs. lb's left side. The rear part of the system (fig.

10 lb) serves to record images of the individual pieces of meat and sort them into different categories based on the images analyzed. For recording the images, there is provided a first and a second arrangement 20 and 21, which are constructed in the same way. The construction and operation of the arrangements is particularly apparent from the explanation in connection with FIG.

3. It should be noted that each arrangement comprises a camera 22 and has a sensor 23 which IS detects when a muscle advanced on the underlying band lies in the image field of the corresponding camera 22. The sensor thereby triggers an image capture and image processing sequence in a device connected to the camera.

After joining the arrangement - which takes place with a continuous band 12 - the muscle is moved beyond the downstream end of the conveyor belt and falls onto an intermediate band 13, as is best seen 20 in FIG. 2b. During this fall, the muscle strikes a "bucket" so that it strikes the band 13 with its previous upper side. The side previously lying down against the conveyor belt 12 is now facing upwards so that an image recording of the other side of the muscle can be performed in the second arrangement 21 arranged over a conveyor belt 14 corresponding to the belt 12.

Following inclusion in arrangement 21, the muscle is transferred to yet another conveyor belt 15, from which 25 it can be sorted into one of four muscle categories based on the image recordings. Along one side, the belt has three pivotable baffles 16a, 16b and 16c (Fig. 1b), which can be pivoted by signals from a PLC control unit 40 over the belt 15 so that the further movement of the muscle on the belt is stopped and the muscle is guided beyond the edge of the tape. Here it falls into one of three vessels 17, which are placed next to the respective baffles. The vessels are DK 170845 Bl 13 intended for muscles with insufficient fat depletion, muscles with PSE deposits and muscles with bloody areas. If none of the baffles are pivoted over the band, the muscle continues beyond the downstream end of the band and falls into a vessel 18 intended for muscles without aberrant properties.

The PLC controller 40 ensures that the muscles 2 are dosed from the funnel 1 and fed to the arrangements 20 and 21 at an even rate consistent with the ability of the arrangements to record and process images of the muscles, the unit regulating the opening of the lock 3, and the speed of the wheel 4 and the conveyor belts 5,9,11,12,13,14. The control unit also keeps track of where the muscles passing a sensor 41 (Fig. 1b) are subsequently located in the transport system. Each muscle is assigned a serial number as it passes the sensor 41, and in this way is followed until it is in one of the four vessels 17,18.

The PLC itself receives via a network 50 a signal on the category of the current muscle arriving at the upstream end of the band 15 and serves as mentioned the guide plates 15 16a, 16b, 16c in accordance with the signal (normal, insufficient fat trimming, PSE occurrences or bloody areas). Two sensors 41, 42 are connected to the PLC unit 40 so that it can take the muscles through the arrangements 20 and 21.

The network 50 is operated by a parent data processing unit 51 and, in addition to this and the PLC unit 40, further comprises two image processing units 52 and 53 connected to each camera 22, a terminal 54 for communicating with an operator via a keyboard and monitor, and an alarm box 55 that notifies the operator when the hopper 1 is to be refilled or a vessel 17,18 is full, and a printer 56 for printing labels containing information on the meat collected in one of the vessels 17,18.

Briefly, the parent data processing unit 51 functions in that it is notified by the PLC unit 40 that a muscle is entering one of the arrangements 20 or 21. The data processing unit sends a signal to the respective image processing unit 52 via the network 50 , 53 about going into active position for an image capture cycle. This is started when the respective sensor 23 detects a muscle in the image field of the corresponding DK 170845 B1 14 camera 22, which simultaneously captures three images of the muscle in the blue, green and red area. The images are recorded and processed in the camera-connected unit 52, 53 to find areas of fat, PSE incidence or bleeding. The units each include a "frame grabber" that stores the images taken in the three different wavelength ranges, 5 and a RISC processor directly in connection with the frame grabber for digitizing and processing the images. In addition, various programs with algorithms and tables are available for masking the image, interpreting the pixels (see example 3 below), and dividing the composite image into sections representing each type of surface (usually meat, fat, PSE meat and bloody meat). The result is sent to the data processing unit 51 which sums the areas of the different types of surfaces after images of both sides of the muscle have been recorded and analyzed as indicated.

If the total area of unwanted occurrences (fat, PSE, bloody meat) exceeds some criteria entered in the data processing unit 51, the network 50 is sent to the PLC unit 40 to swing the corresponding baffle 16a, 16b, 16c over the band 15, 15 so that the muscle can fall into the vessel 17 which is intended for muscles with that particular muscle defect. If the criteria are not exceeded, the muscle is allowed to continue unimpeded with the band so that it falls into the vessel 18 at the end of the band, intended for muscles without significant deviations.

With this transport and imaging equipment it is possible to make 20 double-sided determination of muscle surfaces and sort the muscles into four different categories at the rate of one muscle per second. This is equivalent to several batches of 1000 kg of meat per hour. This is satisfactory for production conditions and so far has not been achieved with any of the known systems.

Although the plant cannot completely ensure that muscles with aberrant 25 properties are present in the batch to be used for the production of cooked hams, it can, however, divert muscles with larger or distinct areas of defects, and all muscles are examined, which is a very big step forward compared to the visual inspection used today.

DK 170845 B1

The arrangement of FIG. 3 for illuminating the meat on the ribbon and simultaneously recording images from each of its wavelength ranges are now further described:

The arrangement comprises a steel housing 30 in the form of four vertical walls and four oblique surfaces forming a kind of pyramid. Inside, the house is painted white. Inside the housing is arranged a smaller, translucent housing 31, which also has four vertical walls and four sloping surfaces which together form a pyramid. The housing 31 is made of semi-transparent, light-scattering acrylic sheets. In the space between the steel housing 30 and the translucent housing 31 are fluorescent lamps 32 having a total power consumption of approx. 1500 W. The fluorescent lamps are of the daylight type and are provided with a manually controlled DC voltage.

10 The space between the two houses can be ventilated by fans 33 which suck in cold air from the surroundings. The hot air can evade through the top of the steel housing via openings 34. The fans are thermostatically controlled, so that during operation, it maintains the optimum temperature for the fluorescent lamps and a stable light emission (about 31 ° C). In addition, in the space between the two housings, electric heaters 35 may be provided which ensure that the fluorescent lamps 15 reach their desired operating temperature more quickly at the start of the system.

The pyramid roofs of the houses end in a kind of shaft or box 36 without bottom. In the box is a three-chip CCD color video camera. The camera chips capture images in the blue, green and red areas, dividing the light from the camera lens into the individual chips. Corresponding points (pixels) on the chips receive light from the same part of the muscle surface, and since the 20 image capture occurs simultaneously in the three chips, the composite image is "color-true". The usual single-chip color video cameras and bar filters provide images that are unacceptable in the present context. The camera has a shutter set to 1/1000 sec and it is activated as described above by a sensor 23 that detects when a muscle is inside the camshaft's image field (indicated by the dotted line 37). The recorded image is sent analogously to the image processing unit 52, where it is retained in the frame grabber and processed and analyzed for unwanted occurrences by the RISC processor.

Located in one edge of the image area 37 are three color tiles 38 which have a well-defined color and reflectivity. The tiles are included in the recording and their light / color visuals are used in the DK 170845 B1 16 computer to correct the images, serving as a reference. At the terminal 54, if desired, it can be shown whether the tiles have a deviating light value that is different, which may be instructed by the operator to adjust the DC voltage to the fluorescent lamps so that they again emit the normed light. The adjustment can also be automatic.

5 The inside of the box 36 is black like the conveyor belt 12 to avoid unwanted reflections.

The arrangement provides a very intense and at the same time uniform illumination of muscles in the camera's image field, with little variation over time and from place to place in the image field. Due to the position of the fluorescent lamps, the illumination is strongest on the sides of the muscle and somewhat weaker on the top 10. However, the lighting is of a diffuse nature, so that shadowing and reflections from the light sources are avoided.

As can be seen from the above, the plant is intended for a special use in the production of cooked ham. However, it is clear that the plant is not limited to this use. The system can also be used in other areas, possibly by appropriate design changes.

If desired, the system of FIG. 1-3 are thus used for sorting muscles by color, muscle type and / or shape, as well as less frequent defects such as platelets and DFD occurrences may also be detected and included in the sorting.

It is possible to simplify the system in connection with less demanding tasks. The FIG.

Thus, the portion of the plant shown in 20a may be omitted if the pieces of meat have already been separated from one another in another system or after they have been laid manually on the band 12. Similarly, the sorting part can be omitted, for example if the plant is only used for quality control or the sorting takes place in another system. Furthermore, in many cases, it may be sufficient to simply record images of one side of the piece of meat, so that the reversal band 13 and the subsequent image recording section lapse.

17 DK 170845 B1 Meat pieces also do not need to be transported lying on a belt, but can be found, for example, in a box together with several other pieces of meat, so that several muscles are examined at once. Muscles with defects are detected by the system, which can give an alarm signal and / or by means of, for example, light arrows, can tell an operator to remove S these muscles from the box. If only a quantitative determination of the percentage of a surface type on the meat is desired, this can be carried forward without separation. It is used, for example, in connection with fat determination of trimmings.

The following examples serve to further illustrate the invention:

Example 1 10 Spectra of Different Surface Type On a sample of the four main types of ham muscles, ten spectra of each of the muscle surface types are recorded: fat, obstruction, PSE, light meat, bloody meat, dark meat, normal-colored meat. The muscle surfaces are oxygenated. The spectra of the diffuse surface reflection of the muscle regions in the visible spectral range from 380 to 730 nm are measured with a 15 spectrometer of the Zeiss make. The obtained spectra are shown in FIG. 4 (average reflection). It is not immediately apparent where in the spectral range the best differentiation between the surface types can be achieved.

Example 2

Spectral Statistical Analysis: The results of the spectral analyzes in Example 1 are loaded into the storage of a computer and a calculation is performed using a statistical program to find the wavelengths at which the greatest information is in the spectra. In FIG. Fig. 5 shows the calculated proportional deviation from the average spectrum of the surface types. From the curve (and possibly further calculations) it can be seen that the detection of PSE and bloody surfaces is improved by moving the sensitivity in the blue region to 390-440 nm. Fat and light meat are well separated by the green area (500-570 nm, as in conventional color cameras), while bloody meat and PSE meat are separated from other surface types when the red area is moved to 620-700 nm.

DK 170845 B1 18

Example 3

Provision of sorting table:

Images of different muscle types are captured using a three-color color video camera. The light from the lens is split / filtered so that one chip detects the blue component, another chip the green component and a third chip the red component. The chips thus register respectively. a blue, a green, and a red image. Each image has a resolution of 250 levels. The images are stored in an electronic memory.

One of the three composite color images appears on a monitor on which a trained operator marks the areas of normally colored meat, light meat, PSE meat, etc., finding 10 these areas on the muscle sample. The highlighted areas are stored in memory along with the corresponding light values in the red, green and blue image.

After a mathematical analysis based on a larger number of surface-type muscles, a detection model is formed. It consists of a three-dimensional color space divided into areas, in which normally colored meat is found within certain localities, light meat within others, 15 PSE within others again, etc. The light values in the red image are thereby emitted from an axis, the light values in the green. image from another axis and the light values in the blue image along a third axis. What is special about this is that each surface category constitutes one or a few continua that are not shared by other surface categories. Instead of color spaces, you can also talk about tables or a three-dimensional matrix.

Preferably, such a detection model is formed for each muscle type and, optionally, a model may be provided for each side of the muscle to improve detection security.

Examples of found sites are shown in FIG. 6-8, whereby the given numbers are the light intensity in the blue, green and red spectral range.

The detection model can be used in connection with the system of FIG. 1-3. In experiments, it has been shown that muscles with insufficient fat trimming, PSE deposits and bloody areas can be separated with the necessary safety even if the muscle is obstructed or consists of light or dark meat which it has not hitherto been possible to separate. £ mentioned defects in connection with automatic production plants.

Claims (12)

An installation for on-line determination of quality characteristics of pieces of meat for quality control or sorting, which includes a conveyor with a belt ¢ 12,13,14) for conveying the pieces of meat (2), an illumination arrangement 5 ¢ 20 , 21) for illuminating the pieces of meat, a color camera (22) arranged in connection with the arrangement for recording three sub-images of the meat pieces in the red, green and blue spectral range, and a data equipment (51,52,53) for storing the partial images in form of RGB light values, analysis of the stored sub-images and determination of the quality characteristics of the meat pieces based on the analysis, characterized in that the plant comprises a three-chip 10 CCD color camera which captures a sub-image in the red, green and blue area on each one chip and is designed to record and store associated RGB light values derived from the same physical surface portion of the meat subject and that the data set contains a three-dimensional table, d is for each accepted combination of RGB light values refers to a specific surface type, and a program portion that compares stored associated RGB light values 15 to the three-dimensional table embedded in the data equipment and assigns pixels in the composite image containing an accepted combination of RGB light values, table code for meat type. System according to claim 1, characterized in that the camera (22) is arranged to simultaneously record associated RGB light values. System according to claim 1, characterized in that the lighting arrangement ¢ 20,21) provides a diffuse, spectrally neutral foreground illumination of the pieces of meat. Installation according to claim 1, characterized in that the lighting arrangement comprises a downwardly open housing, along which the walls are arranged light emitters and that the arrangement contains forced air circulation devices about the light emitters. Installation according to claim 1, characterized in that it comprises downstream of the lighting arrangement a second, corresponding lighting arrangement with a camera, and the conveyor 21 DK 170845 B1 between the arrangements has a band section arranged to reverse a camera of the first arrangement. side of a piece of meat towards the camera in the second arrangement. A method for on-line determination of quality characteristics of pieces of meat for quality control or sorting, comprising: feeding the pieces of meat (2) by means of a conveyor with a belt (12,13,14), illuminating the pieces of meat by means of of an over-band illumination arrangement (20,21), recording three sub-images of the pieces of meat in the red, green and blue area by means of a color camera (22) arranged in connection with the arrangement, storing the sub-images in the form of RGB light values, analysis of the stored sub-images and determination of the quality characteristics of the 10 pieces of meat on the basis of the analysis, characterized in that a sub-image in the red, green and blue area is recorded on each chip in a three-chip CCD color camera, and storing associated RGB light values derived from the same physical surface portion of the meat subject that stored related RGB light values are compared to one stored in a data equipment an eight-dimensional table that, for each accepted combination of RGB-15 light values, refers to a specific surface type and that pixels in the composite image containing an accepted combination of RGB light values are assigned to the table's meat type code. Method according to claim 6, characterized in that the comparison and assignment procedure is repeated for a substantial part of the associated RGB light values derived from the surface of a piece of meat and that the area of the composite image depicting the surface of the piece of meat is segmented into areas with different surface types, after which the piece of meat is assigned to a quality class for quality control or sorting. Method according to claim 6, characterized in that associated RGB light values are recorded simultaneously. Method according to claim 6, characterized in that the illumination arrangement of the illumination arrangement is kept at a constant temperature by forced air circulation about the light emitters. DK 170845 B1 22 Method according to claim 6, characterized in that further sub-images are taken of the meat piece by means of a downstream of the lighting arrangement, a second camera with a corresponding lighting arrangement, and that a hidden side of the meat piece for the first arrangement is facing the camera in the other arrangement prior to image 5 recording. Apparatus for on-line determination of quality characteristics of pieces of meat for quality control or sorting by means of the plant according to claim 1, characterized in that the apparatus downstream of the lighting arrangement comprises a second, corresponding lighting arrangement with a camera, and that the conveyor between the arrangements has a ribbon section adapted to face a camera side of the first arrangement hidden side of a piece of meat towards the camera of the second arrangement. Apparatus according to claim 11, characterized in that the belt section (13) is lowered at its upstream end relative to the belt passing during the first arrangement.