ES2478420A1 - Automated system and procedure for classification of frozen tunas by species (Machine-translation by Google Translate, not legally binding) - Google Patents

Abstract

Automated system and procedure for the classification of frozen tunas by species, which access a scanning zone by a conveyor belt (22). The system comprises: - at least one laser emitter (24) responsible for emitting a laser beam (28) on the tuna (1); - means for capturing images (26) to capture three-dimensional images of the laser beam (28) projected onto the tuna (1); - data processing means for: - detecting the laser beam (28) in each image; - obtaining the gross profile (27) of the tuna from the detected laser beam (28); - generating, from the gross profiles (27) obtained, a three-dimensional point cloud (29) of the tuna; - decompose the three-dimensional point cloud (29) into geometric parameters that correspond to measurements and relationships that define the biometry of the tuna and its species; - estimate the tuna species by means of a decision tree and a comparison of the geometric parameters of the tuna with a series of measures and pattern relationships. (Machine-translation by Google Translate, not legally binding)

Description

DESCRIPTION

Automated system and procedure for classification of frozen tunas by species.

 5

Field of the Invention

The present invention falls within the field of automatic classification systems for frozen tunas through the use of artificial vision.

Background of the invention 10

At present, various fisheries bring their frozen tunas arriving at port facilities, intermediate storage refrigerators or canning factories in bulk, either in warehouses of factory ships, merchants or inside refrigerated containers transported by trucks. The specimens of fish of different sizes (sizes) and species are generally mixed in these 15 containers, and an undetermined percentage of specimens have different quality anomalies, due to friction, blows and, above all, crushing.

In order to optimize the current production processes (separations by species, sizes and specimens with deformities) that are currently carried out manually, and to provide added value to the product, since its quality could be certified and increased the benefit of companies and consequently of the entire value chain, the present invention proposes an automatic system that allows the separation of commercially important frozen tuna species, depending on their species, size / weight or by specimens that have deformities that prevent its subsequent commercialization or specific processing.

Although the applications of artificial vision are not new in their application for fish processing, there are no known specific developments capable of analyzing tuna species. Thus, for example, patent documents US2008137104-A1, 30 EP2559336-A1, CH701341-A2, WO2012008843-A1 and US4934537-A disclose the use of artificial vision techniques to perform fish identification automatically. In them cameras and lighting means are used to obtain images, which are analyzed and different parameters of the fish are obtained, such as biomass, dimensions of the fish (length, thickness), weight and defects or diseases. 35 In most cases the fish are driven by conveyor belts, or they move freely in the water, when identification and classification is performed. However, none of them carries out an identification of the tuna species, only with simpler parameters to identify such as weight, volume and deformities. 40

In fact, the identification of the tuna species is an extremely difficult task, which is carried out manually by visual inspection of operators. Patent document CN102749361-A discloses an automatic method to identify the species of tuna, based on fish chopping, heating of the fish and analysis of the emitted gas, this being a complex, expensive and destructive method, since it requires The fish chop. No methods of identifying tuna species by artificial vision techniques are known, given the enormous similarity between the species.

There are no known methods of automatic identification of tuna species that do not involve destruction of the specimen. Therefore, at present, there are no known automatic methods for identifying tunas that can be used in a productive context where a real-time response is needed.

In addition, the inventions known in the previous patent documents do not allow to solve the existing needs in an industrial frozen tuna processing plant due to not meeting the performance requirements per ton / h, not reaching a minimum percentage of effectiveness or not being able to comply with the environmental working conditions (many times below 0º C) with the consequent 5 technical and operational problems. That is why the selection work is still done manually with the decrease in productivity that this implies. Likewise, this manual work cannot be used for the selection of all commercial tuna species, since in some sizes the species have such a similarity that only a system capable of capturing minimum differences can make that selection. The present invention solves said problems.

Description of the invention

The present invention presents a solution that overcomes the limitations described above, being the first artificial vision system capable of identifying the specific species of frozen tunas. In addition to making a separation of tunas by species, it allows a separation by size and deformities.

With regard to separation by species, the current process is carried out manually under the subjective criteria of the operators. The eye and the human brain are able to distinguish, naturally and automatically, the specimens of a species of tuna and differentiate them from those of others. The present invention is able to automatically identify these differences. For this purpose, several studies of the biology of the species in which the different characteristics of shape and color that they may have were analyzed, in addition to studying their evolution throughout their growth, due to the possibility of significant changes in any of the morphological parameters. As an additional difficulty to working with living beings with notable intraspecific differences, frozen tunas have foreign elements attached to their surface, such as brine and frost, or they may have physical damage and deformities that make it difficult to identify the species 30 of tuna. The present invention allows detecting the presence of foreign elements (brine, frost, physical damage, etc.) adhered to the surface of the fish by acquiring and processing 3D images of the separated specimens to prevent these foreign elements from deceiving the system and produce an error in the identification of the species. 35

With regard to the separation by sizes, currently in some processing plants the specimens are separated manually or by mechanical screening processes only in a few sizes. All those copies weighing more than 20 kilograms are classified as of the same category, due to the impossibility of manipulating them 40 manually due to the weight. The present invention allows an automatic classification with a greater number of categories for sizes due to the differences in performance between them, also allowing the size discrimination to be variable and configurable in real time, depending on the specific needs. The end user can dynamically determine (each time he wishes to configure the system) the size ranges that define each of the categories or sizes in which he wishes to separate the received copies. Thanks to the high resolution of the images with which the vision system works, the present invention is able to differentiate tuna specimens by sizes very precisely (> 98%), improving the classification systems by traditional sizes. In addition, it offers not only a final report with the 50 pieces of tunas from each of the previously set size categories, but also offers the possibility of knowing the actual kilograms of each tuna (without having to weigh them one by one by scale) and consequently the entire item analyzed. Another issue capable of resolving the present invention is that by being

The entire system controlled by a computer can be demanded with great precision to prepare a container with a certain number of pieces or with a certain weight in kilograms, something that is currently difficult to do due to the operation of fish treatment in a industrial plant.

 5

With regard to the separation of specimens with deformities, it should be taken into account that during the processes prior to the reception of the fish some specimens are damaged, due to handling and transport (friction and shock) or overweight crushing due to an accumulation excessive prior to freezing (or cold chain breakage). The present invention allows automatic separation of 10 copies that have certain deformities, configurable by the user. Because tunas have large differences between specimens of the same species and size, these deformities must be very evident, such as the absence of a part of tuna or a deformity that affects an important surface of the fish. The present invention allows to automatically remove 15 from the production chain those elements that have previously been considered of insufficient quality (because they are small in size, due to beating or lacking parts).

A first aspect of the present invention relates to an automated method of classification of frozen tunas by species, where frozen tunas access 20 to a scanning area by a conveyor belt separated at a certain distance from each other. The procedure includes:

 - emit a laser beam (or a coincident double laser beam by means of two laser emitters duly oriented on each side of the conveyor belt) on the tuna as it passes through the scanning area; 25

- capture three-dimensional images of the laser beam projected on the tuna as it passes through the scanning area;

- detect the laser beam in each captured image;

- obtain the raw tuna profile in each image captured from the detected laser beam; 30

 - generate, from the raw profiles obtained in the different images captured, a three-dimensional cloud of tuna points;

 - decompose the three-dimensional cloud of points into a plurality of geometric parameters that correspond to a series of measures and relationships that define the biometry of tuna and its species; 35

 - estimate the tuna species using a decision tree and a comparison of the geometric parameters of tuna with a series of measures and standard relationships.

The laser beam is preferably perpendicular to the conveyor belt, to avoid aberrations in the laser line upon impact on the tuna. In a particular embodiment, the laser beam is red, and preferably with a wavelength of 660 nm.

For the detection of the laser beam the captured images are preferably filtered to detect only the color range used by the laser beam. In a preferred embodiment the three-dimensional images are taken frontally as the conveyor belt 45 moves in an upper and angled position.

To obtain the raw tuna profile in each image, the areas that have volume corresponding to the tuna can be identified and the flat areas corresponding to the conveyor belt removed. fifty

In order to obtain the raw tuna profile in each image, only the central line of the laser beam of greater color intensity can be used, eliminating the rest of the laser beam of less intensity due to the diffusion of the laser beam on the frost of the tuna

frozen.

The geometric parameters can be related to each other and weighted according to their representativeness for the definition of the species.

 5

In a preferred embodiment the images are captured in a cabin protected from outside light and without lighting.

The capture of three-dimensional images is preferably carried out by means of two 3D cameras, arranged on each side of the conveyor belt, in an upper and angled upper position.

The method also estimates, from the three-dimensional point cloud, the volume and weight of tuna. It can also detect deformities present in tuna caused by friction, blows or crushing. The method detects the passage of a tuna through the scanning area to start the image capture process.

A second aspect of the present invention relates to an automated system for classification of frozen tunas by species, where frozen tunas access a scanning area by a conveyor belt separated some 20 distance from each other. The system comprises a scanning module with:

 - at least one laser emitter responsible for emitting a laser beam on the tuna as it passes through the scanning area;

- image capture means configured to capture three-dimensional images of the laser beam projected on the tuna as it passes through the scanned area;

- data processing means configured to:

• detect the laser beam in each captured image;

• obtain the raw tuna profile in each image captured from the detected laser beam; 30

• generate, from the raw profiles obtained in the different images captured, a three-dimensional cloud of tuna points;

• decompose the three-dimensional point cloud into a plurality of geometric parameters that correspond to a series of measures and relationships that define the biometry of tuna and its species; 35

• estimate the tuna species using a decision tree and a comparison of the geometric parameters of tuna with a series of measures and standard relationships.

At least one laser emitter is preferably arranged in such an orientation that the laser beam 40 is perpendicular to the conveyor belt. In a preferred embodiment, the scanning module comprises two laser emitters arranged on each side of the conveyor belt and whose laser beam is perpendicular to the conveyor belt and coincident. The image capture means are arranged frontally to the conveyor belt in an upper and angled position. The imaging means preferably comprise two 3D cameras arranged on each side of the conveyor belt, in an upper and angled position.

The scanning module may comprise an outer shell and protective bands in the tuna inlet and outlet area that define a cabin 50 protected from outside light and without lighting in which images are captured.

The scanning module can comprise presence sensing means responsible for detecting the passage of a tuna through the scanning area and thus activating the means of

Image capture to start the image capture process.

The present invention has been developed based on artificial vision technology, incorporating laser elements for continuous process at high speeds on conveyor belt. The main advantages of the invention are:

- Greater precision in the classification by species, by sizes and detection of specimens with deformities: These improvements suppose a high increase in the price of the product, at the same time that improves the confidence on the part of the clients. It also allows an exhaustive control of the product quality of the suppliers. 10

- Obtaining a higher economic performance for the raw material: Fish is a scarce commodity and it is necessary to optimize its marketing processes. The automation of the classification process implies a high decrease in costs, reducing the low skilled labor involved in the classification process.

- Increase in the processing capacity of the industrial plant: By obtaining a higher speed in the classification process (greater number of kilograms per hour treated), it is possible to optimize the time to serve the orders, as well as the spaces destined to the storage of reception, in which the tunas are less time, being able to allocate them to new processing lines or other services.

- Improvement of the business organization: It provides a very precise estimate of the time to devote to each process, which allows informed decision making in real time.

- Automation of process information management: All the information generated by the system is recorded in memory, 30 allowing:

 Generate a database of historical archives.

 Automatically generate customized reports of the processes performed.

 Possibility of generating coded labels (unit, batch, 35 etc.).

- Implementation of a quality standard: The system can guarantee the quality of the product it processes thanks to its:

- High precision in the detection and automatic separation by species. 40

- High precision in the detection and automatic separation by sizes, thus complying with the legislation on the marketing of minimum sizes.

- High precision in the detection and automatic separation of specimens that present deformities. Four. Five

- Possibility of requiring the production system a specific number of pieces of a species and a certain size (eg 1000 tunas listed between 3 and 5 kilograms) or a container weight or specific specimens. Something that until now is very difficult to achieve due to the operational work in the 50 industrial plants.

Brief description of the drawings

Next, a series of drawings will be described very briefly.

they help to better understand the invention and that they expressly relate to an embodiment of said invention that is presented as a non-limiting example of this.

Figure 1 shows, in side and plan view, the different parts of a manual processing line according to the state of the art. 5

Figure 2 shows, in side and plan view, the processing line with automatic classification of tuna according to the present invention.

Figure 3 shows schematically the different modules distributed in line 10 of tuna processing with automatic classification.

Figure 4 shows the passage of a tuna (schematically represented by a rod-shaped solid) through the scanning module.

 fifteen

Figure 5 shows a general flow chart of the process carried out in the tuna processing line with automatic classification.

Figure 6 shows the obtaining of some specific parameters of tuna used to determine its species. twenty

Figure 7 shows the gross tuna profile obtained by the image collection means.

Figure 8 shows the three-dimensional point cloud obtained after the tuna scanning process.

Detailed description of the invention

The present invention relates to an automated frozen tuna classification system. This classification is carried out in an industrial processing plant of 30 frozen tunas.

Figure 1 shows, in side and plan view, the different parts of a manual processing line, according to the state of the art, where operators located on the distribution belt and responsible for performing the manual classification of frozen fish are appreciated, distributing it in different containers depending on the size, species and deformities found in tuna.

Figure 2 shows, in side and plan view, the processing line with automatic classification (already without operators) of tuna according to the present invention. 40

The frozen tunas 1 are loaded to a tuna receiving module 10, a tuna accumulator tank with variable capacity, depending on the application, with a hopper bottom that can optionally be coupled to a canyon lift and a flat drum rotary, and that allows to collect, orient and distance the specimens of 45 tunas and direct them to the outlet, which flows into the band of the next module, the scanning module 20. In Figure 2, an exit belt of the accumulator 12, an adjustable hatch or hatch 14 and an accelerated and turned belt 16 that allow tuna 1 to be separated at a certain distance towards the scanning module 20, so that it can carry out the individual analysis 50 of the tunas 1 separated from each other. The tuna specimens are automatically oriented and aligned through the adjustable output gate 14 and are deposited one by one on a conveyor belt 22 that directs the tuna to the section of

optical inspection where the different artificial vision devices that allow 3D scanning of tuna are located.

The scanning module 20 is the core of the invention. It is a complete system of detection and analysis by means of artificial vision elements and processing units 5 by microprocessors, capable of capturing images of tunas 1 that move through a scanning conveyor belt 22 and process, in real time, the data obtained, sending the necessary action instructions to the classification system. The scanning module 20 comprises laser liner means and 3-dimensional image capture means. To obtain about 10 homogeneous lighting conditions, the scanning module preferably uses an enclosure where image capture occurs to isolate the module from external lighting conditions. Therefore, the inspection area is preferably a closed cabin (not shown in Figure 2, to simplify the figure), a "dark chamber" protected from outside light and possible splashes of water and other 15 elements that may disturb the inspection. Although the scanning module 20 could also include diffuse lighting means, for example fluorescent lighting where the lighting system is powered by a high frequency electronic ballast to ensure uniformity of the light along the image captured by the cameras. twenty

The scanning module 20 allows the species to be determined from among the target species. In addition, it also allows to determine the size and calculate the weight from the dimensions of the specimen and the presence of significant deformations (crushing). After inspecting and analyzing each specimen, it communicates to a classified module 30 the category of the specimen, so that it can direct the tuna to the appropriate container in each case by automatic actuation of electro-pneumatic gates.

The sorting module 30 comprises a transport system by means of conveyor belts 32 of food rough rubber. In a preferred embodiment, the sorting module 30 is based on a set of electro-pneumatic gates or diverters 34 for the diversion of the tunas towards a collection module 40, formed by a plurality of containers 42 located on one side and at the end of the classification section of the transport system. The classification can be carried out in different ways, for example pneumatically, in that case controlled by elements 35 integrated in an electro-pneumatic frame included as part of the sorting module 30.

An information module 50, implemented for example by means of a large electronic panel installed on the wall or on a support in the upper intermediate part of the sorting tape, shows in real time all the relevant information corresponding to the content of each of the containers 42. This information will indicate to the truck operator the need for container removal, its contents, the speed of the process, number of lots, alarms, etc. You can also indicate the date, time and time elapsed since the beginning of the inspection of the lot. Said module, 45 that obtains the information of the scanning module 20, can also automatically generate customized reports 52 tailored to the user, allowing the generation of reports remotely via "webserver", so that reports 52 can be viewed from any computer connected to your network

 fifty

Figure 3 shows schematically the different modules distributed in the tuna processing line with automatic classification, the core of the invention being the one corresponding to the scanning module 20.

Figure 4 shows in detail the elements of the scanning module 20 and the passage of a tuna 1 (schematically represented in the figure as a parallelepiped-shaped solid) through the inspection area. Tuna is detected by presence sensing means (for example, by means of ultrasound detectors arranged on the sides of the conveyor belt 22, not shown in the figure). 5 The laser emitter 24 is arranged in such a way that the laser beam 28 is substantially perpendicular to the conveyor belt 22 (aberrations in the laser line occur on impacting on the tuna when moving away from the perpendicular orientation). The image capture means 26 used are two equal cameras that are arranged frontally to the march of the conveyor belt 22, in an upper and angled position 10 (preferably at an angle of about 45 ° with respect to the conveyor belt 22) for the capture of the laser beam 28 on the tuna 1. The system is preferably composed of two laser beam emitters 24 that emit two laser beams perpendicular to the scanning tape, one on each side thereof. Said laser emitters 24 are aligned so that only one laser line can be seen when taking a picture. The fact of using two lasers and cameras is to catch as much of the area of the fish that is left against the tape.

Figure 5 shows the general flow chart of the tuna scanning and sorting process: 20

- First, the general application of the system begins 100 and the process begins, starting to operate the tuna conveyor belt, ignition of lasers and optical sensors.

- The tuna passes through an area where presence detecting means 25 (preferably ultrasound) are found which are activated 102 at the passage of the tuna providing a signal that marks the beginning of the process.

- With the previous signal, the image collection means 26 begins with the reading or scanning of the tuna 104. The tuna 1 circulates through a beam of laser 28 of wavelength 660 nm. It has been proven that the system obtains the best results when working on this wavelength of 660 nm, although other similar or other wavelengths such as 540nm (green color) could be used, although in this case worse results would be obtained. 35

- The tuna finishes passing through the area where the ultrasonic sensors are located, which after a programmable “t1” time mark the end of the tuna scanning process 106.

 40

- The image capture means 26 stops scanning and processing of the captured data 110 is performed:

 The activated / deactivated pixel data is processed, generating a three-dimensional point cloud with all the collected data.

 Four. Five

 An application developed specifically for this system collects the point cloud from the optical sensors and delivers it to the processing algorithm.

 The algorithm analyzes this cloud of points by breaking it down into a plurality of parameters that define the biometry of a tuna and therefore its species, such as:

- Volume

- Length: the distance between the beginning of the fish and the end (without considering the tail) is calculated. To have a more robust calculation, polynomial approaches are used to detect the beginning and end of the fish.

- Maximum width. 5

- Maximum height.

- Ratio maximum width vs maximum height.

- Ratio volume vs. length.

- Ratio volume vs. length³: this parameter provides more information.

- Ratio maximum height vs length. 10

- Relationship between the distance from the beginning of the fish to the area of maximum height with respect to the length of the fish.

- MaxDsegment: As shown in Figure 6, maximum distance between the fish profile and the straight line drawn between the maximum diameter point (P1) and the 20% diameter point (P2). fifteen

- Ratio of the location of the previous point with respect to the length of the fish.

- Ratio of lengths L1 vs L2 (see Figure 6).

- Pending the initial part of the fish.

- Pending the final part of the fish. twenty

These parameters are related to each other and are weighted according to their representativeness, extracted through biomorphological studies. For example, for the determination of the species the ratio of the length L1 Vs L2 of the fish and the diameter are twice as important as the ratio of maximum height vs. length.

 The algorithm performs its work and returns to the application the data of volume, species and degree of confidence in the success of the species.

 30

 The application collects that data, calculates the weight through the volume and saves that data in a storage buffer.

- The fish crosses an area where another ultrasonic sensor is located that indicates that the tuna is in the classified zone and with the activation of said sensor 112 the application writes the data previously obtained and calculated in the PLC 116. The PLC manages that information and determines in which container it should classify said tuna.

- The tuna continues on its way through the sorting belt by activating and deactivating 40 the ultrasonic sorting sensors that are in its way and when the PLC detects that the sensor corresponding to the container in which the tuna must be placed has been activated send a gate opening order 118.

- The gate opens 120 through a pneumatic cylinder and waits for a programmable "t2" time until the tuna has fallen into the container. After that time "t2" the door closes.

- With the opening of the door the PLC writes the data on an information screen 122 indicating that there are “n + 1” tunas in said container with a total net weight of 50 containers of “m + x” kilos. Being:

n: number of tunas already in that container.

x: the weight of the tuna you just classified.

m: the sum of the weight of the “n” tunas.

- The process 124 ends.

Figure 7 shows the gross profile 27 of the tuna obtained by both chambers. The upper image shows a “normal” gross profile (frost-free image) and the lower image 5 shows a 27 ’profile with the laser beam more blurred by the uncorrected frost effect (frost image). Although in the lower image the frost effect is not appreciated very significantly, the calculation made with the lower image would be very inaccurate. This figure shows one of the 2000 images per second that the camera can make. From all these raw images, the processing is performed and the 3D point cloud is constructed.

Figure 8 shows the three-dimensional point cloud 29 obtained after scanning. The scanning process for tuna 104 is explained in detail below:

- The scanning process begins with the activation 102 of the presence sensors that detect the passage of tuna 1 through the scanning module 20, in the scanning area.

- The activation of these presence sensors indicates to the image capture means 26 of the scanning module 20 that they should start collecting data. twenty

- Tuna 1 passes through a laser beam 28 with a determined wavelength (660 nm, in the preferred embodiment) at an approximate speed of 1 m / s.

- The image capture means 26, commonly 3D cameras, collect in a region of interest (ROI - Region of Interest) previously defined, the activation or deactivation of a series of pixels that will then be processed by the optical sensor itself generating with that data a three-dimensional cloud of points 29.

- The pixels are activated when they detect the red beam of the laser 28 and the image collection means 26 are provided with a series of filters that prevent detection outside the range of colors other than the chosen 660 nm. 30

- The image capture media make up to 2000 readings per second and are prepared for the capture of tunas up to 1.5 meters in length. In each reading, the raw tuna profile 27 is detected (upper image of Figure 7).

- The optical sensor reads or captures each time a small part of the fish defined by the region of interest, which has the width of the laser line (around 1mm) and the length corresponding to the corresponding width of the conveyor belt. With the 2000 readings of this area of interest that can be made per second, the optical sensor makes the composition of the cloud of points that would determine the biometry of tuna and that will be subsequently processed by the algorithm. 40

- The end of data collection by the optical sensor is marked by the end of detection of tuna 106 by the ultrasonic sensors plus a programmable "t1" time delay.

Regarding the processing of the images, two parts must be distinguished: 45

1. Preprocessing of the cloud images of points obtained. The point cloud obtained needs the following pre-processes for further processing in the algorithm:

a) Removed from the fund. The optical sensors activate pixels based on the reading of the red beam of the laser, therefore they will also activate pixels of the conveyor belt 50 in the area of the region of interest and, therefore, everything that is not tuna must be removed . From the generated point cloud the algorithm preprocess is able to identify the zones that have volume of the flat zones, these being the ones that correspond to the tape. This

The same preprocess also eliminates possible sources of noise that may interfere with the overall volume.

b) Tuna does not always come perfectly aligned, so you have to establish what its axis is and align the points of the cloud. The system can process the tunas it receives up to an angle of ± 30º of deviation. 5

c) The tuna can come both from the front and from the tail and the pre-process will always guide it from the front so that the algorithm is clear that the relationships it is making are consistent.

2. The algorithm process with the previously defined parameters that define the biometrics of tuna. The algorithm decomposes the cloud of points 29 into measures and relationships (the defined parameters) that, through a decision tree and a comparison with curves and patterns, decide the species and the certainty that it has to hit with said sample. This last point is very important because it allows discarding copies that the machine recognizes with a low degree of success. The definition of what is considered acceptable or unacceptable is programmable by the user.

As for the optical sensors used, two 3D CMOS cameras are preferably used, to cover almost the entire fish, especially developed for laser triangulation systems with high triangulation rates, with a high sensitivity CMOS sensor and combined with an algorithm robust for line determination. Each camera has a GigE interface that allows the acquisition of 3D profiles without screen capturers. The CMOS sensor allows a high resolution, with a very small pixel size (5.5 µm x 5.5 µm). In laser triangulation applications, the output of the 3D camera is the line position and the 25 qualification information for each line position value.

Another feature of the system, as can be seen in the explanation of the general flowchart in Figure 5, is that it operates in real time, that is, from the end of the capture of the point cloud until the decision is made. of species, weight and certainty 30 and this is written 116 in the PLC can not pass more than 1 second. To achieve this a wire programming of the different functions has been carried out that take full advantage of each and every processor core by assigning each task or sub-task to a core that is reserved exclusively for this job, always ensuring There is free processing capacity. 35

Likewise, this programming by threads allows the cores to have the capacity to collaborate, that is, when a nucleus has finished its work and it is found that there is another nucleus processing information, this free nucleus can help the occupied nucleus to carry out part of the processes and increase the speed of the response. 40

The present invention further corrects additional problems encountered in the tuna scanning phase:

1. Frost: The tunas arrive at the scan module 20 frozen and, therefore, many of them come with a certain amount of frost. This produces an effect of laser beam scattering. This effect is corrected by making the optical sensors capable of processing the laser beam using only the central line of greater color intensity and allowing the correct reading of the pixels without distortion that are used in the generation of the point cloud. The laser fades and can cause confusion in the camera about the image it takes. The optics used 50 detect the zone of greatest luminosity (the central zone) and eliminate the part of less intensity that is due to the diffusion of the laser beam over the frost. Therefore, the calculation is limited to the central line emitted by the laser and the other dotted lines of the rest of its emission width are discarded.

2. Noise: The system allows to minimize noise and distortions from sunlight, laser brightness in balderas (structural elements of the conveyor belts made of stainless steel and prevent the fish from falling out of the belt while helping to its focus on the path they travel) and other elements. For this, the system has:

to. The correct encapsulation of the scanning module 20 with an outer shell and protective bands in the tuna inlet and outlet area. In addition, these elements inside must be matt to avoid reflections. 10

b. Introduction of a series of filters that only allow reading in the 660nm band by discarding any other light source of color.

3. Laser beam colors: A laser beam preferably of red color is used, with a wavelength of 660nm, since the red color has proven to be the optimum color 15 for the identification of the activated pixels and the one that gives a more consistent reading result.

4.- Angles and distances in optical sensors. The relative position of the elements with respect to the passage of tunas is important to obtain an optimal scan of tuna 20 1, as shown in Figure 4.

From the geometric parameters obtained in the tuna scan, the present invention allows detecting, in addition to the species and volume, the deformities present in the tuna. 25

In general, a defective part is considered to be one that shows significant deformations on its natural form. During the freezing and / or handling of the fish, it can present surface marks produced by the container that contains it or by other adjacent pieces of fish. These marks 30 can be flat (if these are caused by the walls of the container), or bends inwards (caused by other tunas). The defect is considered as a mark if the flat surface or maximum depth is greater than a certain percentage of the total area of the tuna. Normally 10% is considered, so any part that has more than 10% of its surface affected, adding up all types of defects, 35 will be considered as defective.

In this way, the system analyzes the number of tuna surfaces that are crushed and the total accumulated area that is crushed per piece scanned, comparing it with the total area of the tuna and if this crushing exceeds 10% (parameter 40 adjustable by user) considers the piece to be deformed and gives the appropriate signal to the selection actuators.

Claims (29)

1. Automated procedure for classification of frozen tunas by species, where frozen tunas (1) access a scanning area by a conveyor belt (22) separated some distance from each other, characterized in that it comprises:  - emit a laser beam (28) on the tuna (1) as it passes through the scanning area; - capture three-dimensional images of the laser beam (28) projected on the tuna (1) as it passes through the scanning area; - detect the laser beam (28) in each captured image; 10 - obtain the raw profile (27) of the tuna in each image captured from the laser beam (28) detected;  - generate, from the raw profiles (27) obtained in the different images captured, a cloud of three-dimensional points (29) of the tuna;  - decompose the three-dimensional point cloud (29) into a plurality of 15 geometric parameters that correspond to a series of measures and relationships that define the biometry of tuna and its species;  - estimate the tuna species using a decision tree and a comparison of the geometric parameters of tuna with a series of measures and standard relationships.  twenty 2. Method according to claim 1, characterized in that the laser beam (28) is perpendicular to the conveyor belt (22). 3. Method according to any of the preceding claims, characterized in that the laser beam (28) is red. 25 Method according to the preceding claim, characterized in that the wavelength of the laser beam (28) is 660 nm. 5. Method according to any of the preceding claims, characterized in that for the detection of the laser beam (28) the captured images are filtered to detect only the color range used by the laser beam (28). Method according to any of the preceding claims, characterized in that the three-dimensional images are taken frontally to the march of the conveyor belt (22) in an upper and angled position. 7. Method according to any of the preceding claims, characterized in that for obtaining the raw profile (27) of the tuna in each image, the areas that have volume corresponding to the tuna are identified, and the flat areas that correspond to the tuna are eliminated. conveyor belt (22). Method according to any of the preceding claims, characterized in that only the central line of the laser beam (28) of greater color intensity is used to obtain the raw profile (27) of the tuna, eliminating the rest of the image. laser beam (28) of lower intensity due to the diffusion of the laser beam over the frost of frozen tuna (1). Method according to any of the preceding claims, characterized in that the geometric parameters are related to each other and are weighted according to their representativeness for the definition of the species. 10. Method according to any of the preceding claims, characterized in that the images are captured in a cabin protected from outside light and without illumination. Method according to any of the preceding claims, characterized in that the capture of the three-dimensional images is carried out by means of two 3D cameras (26) arranged on each side of the conveyor belt (22), in an upper and angled position. 12. Method according to any of the preceding claims, characterized by comprising estimating, from the three-dimensional point cloud (29), the volume and weight of the tuna. 10 13. Method according to any of the preceding claims, characterized in that it detects the deformities present in the tuna (1) produced by friction, blows or crushing.  fifteen 14. Method according to any of the preceding claims, characterized in that it detects the passage of a tuna (1) through the scanning area to start the image capture process. 15. Automated classification system of frozen tunas by species, where the 20 frozen tunas (1) access a scanning area by a conveyor belt (22) separated at a distance from each other, characterized in that it comprises a scanning module ( 20) with:  - at least one laser emitter (24) responsible for emitting a laser beam (28) on the tuna (1) as it passes through the scanning area; 25 - image collection means (26) configured to capture three-dimensional images of the laser beam (28) projected on the tuna (1) as it passes through the scanning area; - data processing means configured to: • detect the laser beam (28) in each captured image; 30 • obtain the raw tuna profile (27) in each image captured from the laser beam (28) detected; • generate, from the raw profiles (27) obtained in the different captured images, a cloud of three-dimensional points (29) of the tuna; • decompose the three-dimensional point cloud (29) into a plurality of 35 geometric parameters that correspond to a series of measures and relationships that define the biometry of tuna and its species; • estimate the tuna species using a decision tree and a comparison of the geometric parameters of tuna with a series of measures and standard relationships. 40 16. System according to claim 15, characterized in that at least one laser emitter (24) is arranged in such an orientation that the laser beam (28) is perpendicular to the conveyor belt (22).  Four. Five 17. System according to any of claims 15 to 16, characterized in that the scanning module (20) comprises two laser emitters (24) arranged on each side of the conveyor belt (22) and whose laser beam (28) are perpendicular to the conveyor belt (22) and matching.  fifty 18. System according to any of claims 15 to 17, characterized in that the at least laser emitter (24) emits a red laser beam (28). 19. System according to the preceding claim, characterized in that the wavelength of the laser beam (28) is 660 nm. 20. System according to any of claims 15 to 19, characterized in that for the detection of the laser beam (28) the data processing means are configured to filter the captured images to detect only the color range used by the laser beam (28). 21. System according to any of claims 15 to 20, characterized in that the image collection means (26) are arranged frontally to the movement of the conveyor belt (22) in an upper and angled position. 10 22. System according to any of claims 15 to 21, characterized in that for obtaining the raw profile (27) of the tuna in each image the data processing means are configured to identify the areas that have volume corresponding to the tuna, and remove the flat areas that correspond to the conveyor belt 15 (22). 23. System according to any of claims 15 to 22, characterized in that for obtaining the raw profile (27) of the tuna in each image, the data processing means are configured to use only the central line of the laser beam (28) of 20 greater intensity of color, eliminating the rest of the laser beam (28) of lower intensity due to the diffusion of the laser beam on the frost of frozen tuna (1). 24. System according to any of claims 15 to 23, characterized in that the geometric parameters are related to each other and are weighted according to their representativeness for the definition of the species. 25. System according to any of claims 15 to 24, characterized in that the scanning module (20) comprises an outer housing and protective bands in the inlet and outlet area of the tunas that define a cabin protected from light 30 exterior and without lighting in which images are captured. 26. System according to any of claims 15 to 25, characterized in that the image collection means (26) comprise two 3D cameras (26) arranged on each side of the conveyor belt (22), in an upper and angled position. 35 27. System according to any of claims 15 to 26, characterized in that the data processing means are configured to estimate, from the three-dimensional point cloud (29), the volume and weight of the tuna.  40 28. System according to any of claims 15 to 27, characterized in that the data processing means are configured to detect, from the geometric parameters, the deformities present in the tuna (1) produced by friction, blows or crushing.  Four. Five 29. System according to any of claims 15 to 28, characterized in that the scanning module (20) comprises presence sensing means responsible for detecting the passage of a tuna (1) through the scanning area and thus activating the collection means of images (26) to start the image capture process.