JP7271286B2 - Electronic equipment and its control method - Google Patents

Description

The present invention relates to an electronic device and its control method, and more particularly to technology for analyzing an object using a plurality of spectral components.

In recent years, techniques for performing analysis using light have been known. Patent Literature 1 discloses irradiating a cut surface of meat with visible light and extracting bone and fat regions using a plurality of images corresponding to different wavelengths.

JP 2018-066649

However, there is more than one food to be processed in a factory, and even a single piece of meat has different properties depending on the type of beef, chicken, pig, etc. and the production area. The method of Patent Document 1 does not disclose how to identify when the type or production area of meat changes, and when the properties of the food to be analyzed change, high accuracy Analysis may not be possible.

Further, if an attempt is made to perform a more detailed analysis than required as the analysis result, the processing time for the analysis will increase unnecessarily, possibly leading to a decrease in the overall processing efficiency.

SUMMARY OF THE INVENTION In view of the above problems, it is an object of the present invention to provide an apparatus or method that achieves both accuracy and efficiency in analyzing an object using a plurality of spectral components.

In order to solve the above problems, the invention according to claim 1 of the present application performs machine learning using negative teacher data and parameters generated by performing machine learning using positive teacher data. selecting means for selecting one of a plurality of parameters; and generating a plurality of spectra in response to light reflected from the meat to be analyzed using the parameters selected by the selecting means. analysis means for analyzing spectral data indicating spectral intensities of ingredients to identify meat parts contained in the meat; The electronic device is characterized by selecting one of them.

ADVANTAGE OF THE INVENTION According to this invention, when analyzing a target using several spectral components, it becomes possible to aim at both precision and efficiency improvement of analysis.

1 is a block diagram showing the configuration of a system in one embodiment of the present invention; FIG. It is a figure for demonstrating the identification processing of the meat in one embodiment of this invention. It is a figure for demonstrating a lump of meat. 4 is a flowchart of analysis processing in the embodiment of the present invention;

Hereinafter, embodiments of the present invention will be specifically described with reference to the drawings. Although the specific examples described below are examples of the best mode according to the present invention, the present invention is not limited to these specific examples.

First, the configuration of the analysis system according to this embodiment will be described with reference to FIG. FIG. 1 is a block diagram showing the configuration of an analysis system according to this embodiment.

The analysis system according to this embodiment has an electronic device 100 , a measuring device 110 and a display 120 . Electronic device 100 to display 120 are partially or wholly connected via network 130 . Note that the configuration shown in FIG. 1 is only an example, and each component may be configured by distributing it to more devices, or may be configured collectively in one device. Further, the electronic device 100 is communicably connected to a meat processing device (not shown) in the factory, and controls the device. The network 130 includes a LAN (Local Area Network) and the Internet.

Electronic device 100 includes CPU 101 , identifier 102 , database 103 , and operation unit 104 .

The CPU 101 is an arithmetic circuit that controls the overall operation of the electronic device 100 according to a program stored in a memory (not shown). The CPU 101 controls the measuring device 200 and causes the discriminator 102 to analyze the spectrum data acquired from the measuring device 200 . The discriminator 102 is composed of a GPU (Graphics Processing Unit) or an FPGA (field-programmable gate array). In this embodiment, the discriminator 102 is set with discrimination parameters generated by machine learning and stored in the database 103 . This discriminator 102 uses the spectrum data obtained from the measuring device 200 to perform discrimination processing of the part of meat to be analyzed. The database 103 stores identification parameters generated in advance by performing machine learning on a cloud or server. A plurality of types of identification parameters are prepared according to the type of analysis target, analysis level, or teacher data. The operation unit 104 is an input device for accepting user operations, including a character information input device such as a keyboard, a pointing device such as a mouse and a touch panel, buttons, dials, joysticks, touch sensors, and touch pads. It should be noted that the touch panel is an input device that is superimposed on the display 120 and configured in a two-dimensional manner so as to output coordinate information according to the touched position.

The measuring device 110 has a controller 111 , a measuring section 112 and an illumination device 113 . The controller 111 communicates with the CPU 101 and controls the entire measuring device 110 according to instructions from the CPU 101 .

The measuring unit 112 is controlled by the controller 111, measures the signal level of each spectral component from the meat to be analyzed (not shown), and generates spectral data.

Spectral data is not particularly limited as long as it indicates the signal level (spectral intensity) of each of a plurality of spectral components. As spectral data, for example, for the response that occurs when the meat to be analyzed is stimulated, the response intensity (signal level corresponding to the spectral intensity) is measured against the measurement parameter (corresponding to the spectral component) Stored data can be used. The term "stimulus" as used herein includes electromagnetic waves, sound, electromagnetic fields, temperature and humidity.

When the spectral data is spectral data in the ultraviolet, visible, or infrared region, or Raman spectral data, the spectral components can be wavelengths or wavenumbers. Further, when the spectral data is mass spectral data, the spectral components can be mass-to-charge ratios and mass numbers.

Spectral data belong to one of the groups corresponding to the constituent components contained in the meat to be analyzed. The spectral components and spectral intensities differ according to each component of the meat to be analyzed, which is included in the measurement area from which the spectral data was obtained. Therefore, by analyzing the spectral data, the group to which the spectral data belongs can be identified and each spectral data can be attributed to each component.

In this embodiment, the spectral components are tens to hundreds of wavelength components in the wavelength bands of visible light and near-infrared light, and the measurement unit 112 is an image sensor provided with a filter corresponding to each wavelength, Alternatively, it is composed of a line sensor that receives light dispersed by a diffraction grating.

The illumination device 113 has a light source for illuminating the meat to be analyzed. In this embodiment, a halogen lamp that covers near-infrared light and visible light is used, and by irradiating the meat, the intensity of each spectral component reflected by the meat can be increased.

Display 120 displays the identification result of electronic device 100 . As the display 120, for example, a liquid crystal display, an organic EL display, or the like can be used. Display 120 can display image data transmitted over network 130 .

Next, the process of processing meat in this embodiment will be described with reference to FIG. Meat to be processed is transported to a processing plant via a meat wholesaler or a livestock farmer. In the process of meat processing, there are two types of meat: block meat, which is mainly the same part of meat that is solidified, and end meat, which is a lump of meat pieces of various parts that did not enter the block of meat. . In this embodiment, the process of processing meat for making a hamburger steak will be described as an example, but the present embodiment is not limited to this, and can also be applied to sausages and other processed meat foods. Needless to say.

Blocks of meat and cut pieces of meat are each frozen in order to maintain the quality of the meat before proceeding to the meat processing process. In particular, the end meat, which is a collection of pieces of meat from various parts, forms a large mass.

Description will be made below with reference to FIG. First, an analysis using X-rays is performed to detect bones contained in block meat and edge meat, and to confirm whether foreign substances are contained. Analysis is performed while blocks of meat and edge meat are lined up on a belt conveyor, and if there are bones or foreign matter, they are removed in the post-process. X-ray analysis can be performed at a belt conveyor speed of 15 meters/minute or 25 meters/minute.

Next, analysis using a hyperspectral camera with a wide wavelength range that can be irradiated at once, which is called hyperspectral analysis in this embodiment, is performed. In this case, it is assumed that tens to hundreds of outputs of different wavelengths are used, so it is referred to as a hyperspectrum. It is possible to apply the invention.

In the hyperspectral analysis, the discriminator 102 performs discrimination processing on spectral data acquired by the measuring device 110 using discrimination parameters obtained based on machine learning. Then, the position of cartilage, muscle, tendon, or whether or not they are included in the meat is output.

In hyperspectral analysis, meat is placed on a conveyor belt and analyzed in the same manner as in X-ray analysis, so it is desirable to operate the belt conveyor at the same speed as in X-ray analysis. If it takes time for the classifier 102 to output a classification result, it is desirable to install more hyperspectral cameras than X-ray devices or to increase the number of classifiers. After X-ray analysis and hyperspectral analysis, parts such as bones, foreign substances, cartilage, muscles, and tendons are removed, and the block meat and edge meat after removing these are minced to form hamburgers. and proceed.

Next, the state of X-ray analysis and hyperspectral analysis will be described with reference to FIG. 2(b). In this embodiment, the X-ray analysis and hyperspectral analysis of the meat 201 shall be performed on the belt conveyor 202 . As shown in FIG. 2(b), meat 201, which is cut meat or block meat, flows on a belt conveyor 202, and is first analyzed by an X-ray analyzer 203. A spectral analysis is performed. The moving direction of the belt conveyor 202 is the direction of the arrow shown in FIG. 2(b). Thus, each analysis is performed by applying light waves from above the meat. In hyperspectral analysis, the light reaches only near the surface of the object to be analyzed, compared to X-ray analysis. Therefore, the results of hyperspectral analysis are limited to information about a shallow range from the irradiation surface compared to the results of X-ray analysis. Therefore, in the hyperspectral analysis, it is desirable to slice the meat to be analyzed into 2-cm or 5-cm slices. Alternatively, for example, the meat 201 may be sliced at a thickness of 10 cm or 8 cm, and both sides of the meat 201 may be analyzed. Based on the results of the analysis, the meat 201 is classified into meat that can proceed to the normal mincing process as it is (pattern A), meat that requires further human judgment (pattern B), and normal minced meat as it is. It is classified as meat that should not proceed to the process of (C pattern). In the case of block meat, since it is possible to grasp which part of the meat is at the stage of making it into a block meat, X-ray analysis and hyperspectral analysis are not performed on block meat, and only the end meat is subjected to X-ray analysis and hyperspectral analysis. It may be subject to spectrum analysis.

The illumination device 113 is provided with a pair of light sources that irradiate light having directivity so as to sandwich the measurement unit 112, and the irradiation directions of the respective light sources are inclined in opposite directions with respect to the vertical direction. By irradiating light at different angles, it is possible to irradiate a part that is shadowed by the irradiation of one light source with the other light source, and by collecting the light irradiated from both light sources in one place, It is possible to irradiate the object to be analyzed with more light.

Next, the lump of meat called edge meat in this embodiment will be described with reference to FIGS. 3(a) and (b). An example of block meat is shown in Fig.3 (a), and an example of edge meat is shown in FIG.3(b). FIG. 3(a) shows a block of meat containing lean meat and fat. In a block of meat, the positional relationship of each part is known, so for example, it may be possible to determine that cartilage is not contained in the lean meat and fatty meat blocks shown in FIG. 3(a) without analysis. be. As shown in FIG. 3(b), in the edge meat, the finely cut parts form lumps regardless of the positional relationship of the original parts.

The meat analysis process in this embodiment will be described with reference to FIG. FIG. 4 shows a flowchart of meat analysis processing, which corresponds to the X-ray analysis and hyperspectral analysis in FIG.

This processing is realized by the CPU 101 processing according to a program stored in a memory (not shown). In FIG. 4, steps are omitted and indicated as S.

In S401, the CPU 101 confirms the set menu. The menu includes the type of meat, the place of production, the condition (whether block meat or cut meat), the type and level of the process performed after analysis, the corresponding menu (for hamburgers, for shredded meat, etc.), and the like. Such menu settings may be based on user input or according to a preset schedule.

In S402, the CPU 101 performs initial setting according to the set menu. First, the CPU 101 selects one of the identification parameters obtained by machine learning and stored in the database 103 according to the menu, and sets it in the identifier 102 . Even if meat is a single word, identification parameters are different for beef, pork, and chicken, and furthermore, even for the same beef, it is necessary to use different identification parameters depending on the type and production area.

Moreover, even for meat of the same type and origin, different identification parameters may be used according to the difference in the process applied to the meat after analysis. For example, the allowable range of tendons, muscles, or cartilage contained in the meat differs depending on whether the meat after analysis is finely minced or coarsely minced. Alternatively, since the accuracy of analysis required differs between a block of meat of which parts are known in advance and a cut meat of which parts are mixed, different identification parameters may be used accordingly.

Also, the spectral components to be noted may be changed depending on the type, place of production, or condition of the meat. In other words, among the spectral data obtained by the measurement unit 112, spectral components to be noted may be specified according to a set menu, or unnecessary spectral components that are not considered important in identification may be prevented from being used. . Alternatively, the spectral components to be weighted as judgment materials may be changed according to the part to be identified, such as tendon, muscle, or cartilage.

Further, the CPU 101 may adjust the light amount of the irradiation device 103 according to the driving speed of the belt conveyor 202 or the distance from the lighting device 113 to the meat 201 to be analyzed. Since the light emitted from the irradiation device 103 has heat, the meat may be spoiled if the amount of light is too large. Conversely, if the amount of light is too small, each spectral component reaching the measurement unit 112 will become small, making it difficult to perform highly accurate analysis. Therefore, the slower the driving speed of the belt conveyor 202, the longer the meat 201 is illuminated by the illumination device 113, and the more susceptible it is to heat. Alternatively, the closer the distance between the light source of the irradiation device 113 and the meat 201 is, the more the meat 201 is affected by heat from the light source. In this way, the CPU 101 adjusts the light amount of the lighting device 113 based on the driving speed of the belt conveyor 202 or the distance from the lighting device 113 to the meat 201 to be analyzed, according to data measured or tested in advance. .

The distance from the illumination device 113 to the meat 201 to be analyzed may be measured and input by the user in advance, or may be estimated from the distance between the two light sources of the illumination device 113 and the inclination of the light sources with respect to the vertical direction. You may For example, a motor for changing the inclination of the two light sources of the illumination device 113 is mounted, and image data is generated from spectral data obtained by the image sensor of the measurement unit 112 . In this image data, the inclination of the light sources may be automatically adjusted so that the positions of the light emitted from the two light sources match.

Also, the amount of light of the illumination device 113 may be adjusted according to the type of analysis target. Depending on the degree of freezing and the type of foodstuffs to be analyzed, the degree of heat effect caused by the light emitted from the light source differs. Therefore, the amount of light of the lighting device 113 may be adjusted according to the heat resistance of the food material to be analyzed.

Also, hard muscle is easier to distinguish from lean meat and fat than relatively soft muscle. Therefore, if soft muscles do not need to be identified, they are more likely to be successfully identified even if the power of each spectral component is lower than if soft muscles need to be identified. In this way, when the required accuracy of analysis can be suppressed, the light intensity of the illumination device 113 may be reduced.

Further, a slicer may be provided for slicing the food material to be placed on the belt conveyor 202, and the thickness of slicing by this slicer may be changed according to the type of food material.

In S403, the CPU 101 performs calibration processing. The spectral data obtained from the measurement unit 112 may vary due to the brightness and lights in the factory, or the lighting used in other lines. Therefore, based on spectral data obtained by photographing the belt conveyor 202 of a certain color, white balance processing is performed to adjust the gain applied to each spectral component, and calibration processing is performed on the spectral data. Then, when this calibration process is completed, the user is notified that preparations for X-ray analysis and hyperspectral analysis are complete.

In S404, the CPU 101 causes the X-ray analysis device 203 to perform X-ray analysis on the meat 201 conveyed by the belt conveyor 202, and detects foreign substances such as metals and bones contained in the meat. A well-known method may be used for this X-ray analysis.

In S405, the CPU 101 proceeds to S406 if foreign matter is detected as a result of the X-ray analysis of S404, and proceeds to S407 if not detected.

In S406, the CPU 101 stores the analysis data DXn in a memory (not shown). n is a number for identification given to each end meat or block meat. DXn may indicate only information about the presence or absence of bones or foreign substances, may include coordinate information indicating where bones or foreign substances are present, or may be image data indicating whether bones or foreign substances are included. . If bones are included, information on the presence or absence of bones may be included in DXn in the case of block meat, and information including the position of bones may be included in the case of edge meat. In the case of block meat, if there is bone, most of the mass is likely to be bone, but in the case of end meat, there is a high possibility that part of the mass is bone. Therefore, in the case of block meat, it is possible to quickly grasp whether the block itself can be used for processing by indicating the presence or absence of bones. On the other hand, in the case of the end meat, the position of the bone is indicated so that the bone can be removed in a later process. Further, DXn may indicate the ratio or area of bones included. Also, the results of the X-ray analysis may be displayed on the display 120 before analysis by the hyperspectral camera. In addition, blocks of meat or ribs determined to have a bone percentage (within a piece of meat) or size greater than or equal to a threshold may be discarded or used for other processing without hyperspectral analysis. .

In S407, the CPU 101 performs hyperspectral analysis on the meat 201 that has undergone the X-ray analysis. The meat 201 on the belt conveyor 202 is irradiated with light from the illumination device 113, and the measurement unit 112 that receives the reflected light from the meat 201 generates signal levels corresponding to each of several tens to hundreds of wavelength components. and output as spectral data. It is assumed that this spectral data has a signal level at each two-dimensional coordinate corresponding to the image sensor for each spectral component. That is, if the two-dimensional coordinates corresponding to the image sensor are (x, y), the spectral component is (u), and the signal level is (v), the spectral data can be represented by (x, y, u, v). can.

Using a cloud or a server in advance, machine learning by deep learning such as a deep convolutional neural network is performed using a group of spectral data for meat whose parts have been discriminated as teacher data, and a learning model for the classifier 102 is generated. By storing discrimination parameters corresponding to this learning model in the database 103 of the electronic device 100, the discriminator 102 analyzes the spectral data obtained from the measuring device 100 to determine which part of meat is which. It becomes possible to identify whether it corresponds to the part.

Teacher data used in machine learning includes positive data and negative data. Since the cartilage, muscle and tendon are removed, the spectral data for the meat in which the portions corresponding to these cartilage, muscle and tendon have already been determined are taken as positive data. On the contrary, since the cartilage, muscle, and tendon are removed, the spectral data for meat from which portions corresponding to lean meat and fat other than these have already been determined are treated as negative data. By performing machine learning using positive data, the identifier 102 can identify cartilage, muscle, and tendon parts. By performing machine learning using negative data, the discriminator 102 can now discriminate sites that are highly likely not to contain cartilage, muscle, and tendon.

It is not highly likely that image data and spectral data are accumulated for cartilage, muscle, and tendon that have been targeted for removal so far. Therefore, it may take time to prepare a sufficient amount of positive data to identify them. Therefore, we first use a learning model that performs machine learning using negative data using spectral data corresponding to lean meat and fat, which are relatively easy to prepare. Then, after sufficient spectral data corresponding to cartilage, muscle, and tendon have been collected, the learning model is switched to a learning model based on positive data, or to a learning model using both negative data and positive data. If there is no need to distinguish between cartilage, muscle, and tendon, only learning models based on negative data may be used.

Note that the classifier 102 identifies a part for each coordinate of spectrum data, but there is a possibility that the identification result is erroneous. Therefore, the CPU 101 groups adjacent coordinates that are determined to be the same part, and if the size of the grouped area is less than a threshold value, determines that there is a high possibility that the identification result is erroneous. may Alternatively, the degree of dispersion on two-dimensional coordinates may be discriminated for each part, and the identification result may be adopted only for areas where the discrimination result is determined to be dense.

Furthermore, as described with reference to FIG. 3, since the positional relationship of each part of the block of meat is known, the spectral data may be machine-learned not in units of coordinates but in units of areas containing a plurality of coordinates. By doing so, the discriminator 102 can discriminate the parts by considering the relative positional relationship of each part. On the other hand, since the edge meat is a lump regardless of the positional relationship of the original part, it is desirable to perform machine learning in units of coordinates or extremely narrow units as an area.

In S408, the CPU 101 acquires hyperspectral analysis data from the discriminator 102 and sets it as DMn. n is a number for identification given to each end meat or block meat. DMn includes data indicating the size and ratio of each part, data indicating the position of each part, and data indicating whether or not cartilage, muscle, and tendon are larger than a predetermined size.

If the cartilage, muscle, or tendon is included in a predetermined amount or more, for example, 20% or more of the area of the edge meat, the meat is recorded as NG in the data, and is determined as the C pattern meat described above. Conversely, if it is determined that cartilage, muscle, and tendon are not found, the meat is determined to be meat of pattern A described above. DMn includes information indicating the area of each part. For example, in this analysis, the operator wants to know information about cartilage, muscles, and tendons. It is also possible to obtain information only on the ratio. In this way, by obtaining detailed information about necessary information and only general information about other information, the processing load on the discriminator 102 can be reduced and the analysis time can be shortened. Then, the CPU 101 causes the display 120 to display the result of the X-ray analysis and the result of the hyperspectral analysis.

In S409, the CPU 101 advances to S411 if the meat 201 is meat with no visible cartilage, muscle, or tendon and is clearly problem-free (A-pattern meat); otherwise, the CPU 101 advances to S410.

In S410, the CPU 101 proceeds to S412 if the meat 201 contains more than a predetermined number of cartilages, muscles, and tendons, and if the meat is clearly problematic (C pattern meat), otherwise proceeds to S413. proceed to At this time, if it is not known whether it is the C pattern or not (is it the B pattern), the process proceeds to S413. For example, if 20% or more of cartilage and muscle are included, it is determined to be a C pattern. , S413. In this way, when the threshold value for judging pattern C is clearly not exceeded, by proceeding to S413, which is different from pattern C, the amount of meat not used for hamburger preparation can be reduced.

In S411, the CPU 101 proceeds to the processing step A, which is mincing processing for making the meat n into hamburgers. Here, the process of making hamburger steak is taken as an example of the processing process A, but any process may be used as long as the analyzed meat is processed for the dish or cooking set in the menu in advance.

In S412, the CPU 101 advances the meat n to the processing process C other than the hamburger making process. For example, if it is clearly determined to be NG in S410, mince the edge meat more finely than usual, use it for processed products that require other textures, use it for soup stock, etc. You may make it

In S413, the CPU 101 has the meat n re-analyzed by a human, and re-determines whether the A pattern is the C pattern based on the human re-analysis result.

In S414, the human determination result obtained in S413 and the spectral data for the meat are added to teacher data for strengthening the learning model of the discriminator 102. FIG. Machine learning is considered to be insufficient for the meat that has proceeded to S413. Therefore, the determination result of S413 and its spectrum data are transmitted to the cloud or server and used as teacher data for enhancing machine learning.

The processing of S404 to S414 is repeated while the meat 201 is flowing on the belt conveyor 202, and when the operator gives an instruction to end the hyperspectral analysis, the flow chart of FIG. 4 ends.

As described above, according to the present invention, the parameters used for analysis are changed according to the type and production area of the food to be analyzed, or the cooking method. By doing so, it is possible to achieve both accuracy and efficiency of analysis when analyzing an object using a plurality of spectral components.

(Other embodiments)
The present invention is also realized by executing the following processing. That is, the software (program) that realizes the functions of the above-described embodiments is supplied to a system or device via a network or various recording media, and the computer (or CPU, MPU, etc.) of the system or device reads the program code. This is the process to be executed. In this case, the program and the recording medium storing the program constitute the present invention.

Claims (10)

Select one of a plurality of parameters, including parameters generated by performing machine learning using positive teacher data and parameters generated by performing machine learning using negative teacher data a selection means to
Using the parameters selected by the selection means, the light reflected from the meat to be analyzed is received and the spectral data indicating the spectral intensity of a plurality of spectral components is analyzed to determine the meat part contained in the meat. having analytical means to identify
The electronic device, wherein the selection means selects one of the plurality of parameters according to the part of meat to be identified . The analysis means outputs the analysis result according to at least one of the area of the region identified as the same site and the degree of dispersion of the locations identified as the same site. Item 1. The electronic device according to item 1 . 3. The electronic device according to claim 1 , wherein said selection means further selects one of said plurality of parameters according to the type of said meat. 4. The electronic device according to claim 3 , wherein said selection means further selects one of said plurality of parameters according to the production area of said meat. 5. The analyzing means identifies at least one of cartilage, tendon, and muscle contained in the meat to be analyzed by analyzing the spectral data. 1. The electronic device according to item 1. If the meat to be analyzed is a piece of meat that is a collection of a plurality of pieces of meat, the analysis means performs analysis based on the spectral data for each coordinate, and if the meat to be analyzed is block meat, 6. The electronic device according to any one of claims 1 to 5, wherein analysis is performed based on spectral data for each area including a plurality of coordinates. 7. The electronic device according to any one of claims 1 to 6 , wherein the analysis means analyzes by focusing on a specific spectral component according to the part of the meat. 8. The electronic device according to any one of claims 1 to 7 , wherein at least one of the type of analysis target, the required accuracy of analysis, and the processing to be performed after analysis is set by a user. device. illumination means for illuminating the meat to be analyzed;
generating means for receiving light reflected from the meat to be analyzed irradiated by the lighting means and generating the spectrum data;
An analysis system comprising the electronic device according to any one of claims 1 to 8 . Select one of a plurality of parameters, including parameters generated by performing machine learning using positive teacher data and parameters generated by performing machine learning using negative teacher data a selection step to
An analysis step of analyzing spectral data indicating spectral intensities of a plurality of spectral components upon receiving light reflected from the meat to be analyzed using the selected parameters to identify portions of meat contained in the meat. has
A control method for an electronic device, wherein in the selecting step, one of the plurality of parameters is selected according to the part of meat to be identified .

Images