Method and device for automatically determining at least one quality parameter characterizing processing of carcass part

Abstract

The invention relates to a method for automatically determining at least one quality parameter characterizing the processing of a carcass part, comprising the following steps: conveying the carcass part as an input product in a conveying direction to a first processing machine (11) by means of a conveying device (10), a first processing machine for performing a first processing step for processing the carcass portion into an output product; acquiring input sensor data of the carcass part by means of at least one sensor device (12) prior to performing the first processing step; classifying partial data related to the machining quality of the first machining step from the input sensor data; determining the at least one quality parameter based on the partial data; and providing the at least one quality parameter for display and / or transmission to a downstream processing machine, the downstream processing machine being adapted to perform at least one subsequent processing step based on the determined quality parameter. The invention also relates to a corresponding device.

Description

Technical Field

[0001] This invention relates to a method for automatically determining at least one quality parameter characterizing the processing of a carcass portion. Furthermore, this invention also relates to an apparatus for automatically determining at least one quality parameter characterizing the processing of a carcass portion. Background Technology

[0002] Such methods and equipment are used for the automated handling and / or processing of slaughtered animals or parts of slaughtered animals, particularly for the processing of slaughtered poultry. Throughout this text, the term "carcass part" is used, and for the sake of brevity, it is used as a synonym for part of an animal carcass, body part, and / or slaughtered animal.

[0003] It is known in the prior art that animal carcasses or carcass portions are detected by sensors before automated processing to determine processing-related parameters, so that subsequent automated processing operations can be adapted to the corresponding animal carcass portion as optimally as possible. For this purpose, sensors are configured, for example, as optical imaging sensors or imaging sensors sensitive to X-ray radiation, to determine parameters such as the external dimensions of the animal carcass portion, the location and distribution of skeletal structures, etc. The processing-related parameters obtained thereby then form the basis for adjusting the corresponding processing operations. In this way, it is possible to automatically start the handling equipment and adapt it to the characteristics of a specific animal carcass portion or product to be processed.

[0004] The drawback of known methods and devices is that they rely solely on sensors to detect inherent physical parameters of the animal carcass to be processed (such as weight, size, external contour, bone position, etc.) and use these parameters as the starting point for subsequent processing operations. They completely disregard the extent to which these parameters affect the corresponding processing operations to be performed. Summary of the Invention

[0005] Therefore, an object of the present invention is to provide a method that ensures the suitability of a carcass portion to be processed for processing is evaluated before performing processing steps. Another object of the present invention is to provide a method that ensures at least one semi-finished product obtained from an animal carcass after processing is evaluated in this manner. Furthermore, an object of the present invention is to provide a corresponding apparatus.

[0006] The aforementioned objective is achieved by the method mentioned at the beginning, which includes the following steps: conveying the carcass portion as an input product along a conveying direction to a first processing machine by means of a conveying device, the first processing machine being used to perform a first processing step for processing the carcass portion into an output product; acquiring input sensor data of the carcass portion by means of at least one sensor device before performing the first processing step; classifying partial data from the input sensor data that are relevant to the processing quality of the first processing step; determining the at least one quality parameter based on the partial data; and providing the at least one quality parameter for display and / or transmission to a downstream processing machine adapted to perform at least one subsequent processing step based on the determined quality parameter. By determining at least one quality parameter, it is possible to initially assess the quality of the animal carcass portion in relation to processing. Therefore, the carcass portion to be processed undergoes an assessment, which is always related to the corresponding processing. In this way, the quality parameter provides variables that give information about the carcass portion regarding its processability and / or suitability for processing.

[0007] An advantageous embodiment of the invention is characterized by classifying partial data using an algorithm based on predetermined selection parameters. This provides the following advantages: the partial data classification is based on a specified algorithm, which can be set by defining the selection parameters for the corresponding processing steps. The advantage of algorithmic classification is that the computational complexity can be kept at a relatively low level, which has a favorable impact on the computational power required. A further advantage is that this algorithmic classification produces reproducible results. Furthermore, no prior training is required. The selection parameters can be, for example, different parts or regions of an animal carcass. For example, if the animal carcass is a poultry carcass, then the selection parameters could be different regions, such as the breast (particularly the left and right breast meat slices), legs, neck, rump, etc.

[0008] Typical carcass portions (especially in the processing of poultry carcasses) include, for example, wings and / or legs. Particularly present are, for example, wingtips (also known as "de-tipped wings"), shoulders, elbows, hips, knees, and common poultry products defined by these joints (such as, for example, leg fillets, anatomically whole legs, and products obtainable through further processing, such as upper legs, cuts of meat, calves, and / or thighs). In the processing of poultry carcasses, for example, forequarters, breastplates, whole cuts, half-cuts, chicken breast slices, or tenderloin are obtained during processing.

[0009] A preferred further improvement of the invention is that the classification of the partial data is performed using a pre-trained neural classification network. This provides the advantage that artificial intelligence is used for classification, eliminating the need for the development and definition of algorithms for classification. Furthermore, such a neural classification network is, in principle, capable of classifying and categorizing even fuzzy or ambiguous input sensor data. This further provides the advantage of producing consistent results related to the quality parameters to be determined, despite large variations in biological products with respect to the input data.

[0010] Another advantageous embodiment of the invention features that the at least one quality parameter is determined by an algorithm based on predetermined quality assessment parameters. Such predetermined quality assessment parameters include, for example, shape, weight, surface characteristics of the product, and the degree of surface damage (such as, for example, bloodstains). The aforementioned advantages of algorithmic classification also apply to the algorithmic determination of at least one quality parameter.

[0011] Another advantageous embodiment of the invention features that the determination of the at least one quality parameter is performed by means of a pre-trained neural quality assessment network. The aforementioned advantages of neural classification networks also apply to neural quality assessment networks used to determine at least one quality parameter.

[0012] Typical quality parameters (especially in poultry processing) include, for example, residues from previous processing steps. For instance, the proportion of feathers remaining on the poultry carcass during the defecation process. Quality parameters can also be the degree of surface damage to the product that results in a quality decrease due to processing. Examples include, for instance, detached skin that exposes the underlying muscle tissue. Bloodstains in the skin and upper tissue areas, and particularly bone and / or joint fractures in the limbs of the animal carcass, are other examples that can be used as quality parameters.

[0013] According to another preferred embodiment of the invention, the quality assessment network comprises a given number of neural subnetworks, each of which is pre-trained to determine at least one of various quality parameters. In other words, these subnetworks advantageously constitute an expert network configured and adapted to determine the corresponding quality parameters.

[0014] According to another preferred embodiment of the invention, the input sensor data includes at least image data of the animal carcass. Advantageously, this enables the acquisition of input sensor data in a non-contact manner. In particular, it is therefore possible to record image data of the carcass while it is being continuously transported.

[0015] Another preferred improvement of the invention is that the sensor device records the image data within the wavelength range of visible light, infrared, ultraviolet light, and / or the spectral range of X-ray radiation. By selecting a specific wavelength range, it is possible to obtain image data representing the optimal starting point for determining the respective quality parameters.

[0016] Another advantageous embodiment of the invention features that the input sensor data at least partially maps the carcass portion in a three-dimensional manner. This provides the advantage of being able to obtain the dimensions of the carcass portion, or at least a portion thereof, and use them for subsequent evaluation.

[0017] According to another preferred embodiment of the invention, the estimated weight and / or carcass size of the output product are determined as at least one of various quality parameters. Advantageously, this determines the core quality parameters that are crucial to reliability and effectiveness.

[0018] According to another preferred embodiment, the quality value assigned to the output product is determined as at least one of various quality parameters, the quality value representing a measure of the expected final quality of the output product after its processing. Advantageously, the quality value is thus used to determine a measure that provides information about the expected quality of the output product after processing. Therefore, the quality value is a measure of the expected final quality after the corresponding processing step.

[0019] Another advantageous embodiment of the invention features a total mass value determined as at least one of the various mass parameters, the total mass value being composed of a given first component related to the estimated weight and / or the carcass size, and a given second component related to the mass value. In other words, the total mass value is determined by a linear combination of the estimated weight and / or carcass size based on the given first and second components. This provides the advantage that the influence of the estimated weight and carcass size can be specified in a customer-specific manner and can be adapted to corresponding needs.

[0020] An advantageous embodiment of the invention is characterized in that the neural classification network is trained by means of the following steps: providing a plurality of animal carcasses; acquiring input sensor data for each of the plurality of carcass portions using the at least one sensor device before performing a search for a first processing step; classifying, by a human trainer, portions of the input sensor data that are relevant to the processing quality of the first processing step from the input sensor data by evaluating the input sensor data; specifying, by the human trainer, classification regions suitable for classifying the relevant portions of the data based on the portions of the data; providing training data for training the neural classification network by inputting the input sensor data as input data and the respective classification regions as target data for the neural classification network; and repeatedly adjusting the weights of the classification network based on the differences between the target data and the output data generated by the neural network. Advantageously, it is therefore possible to pre-train the classification network for the optimal classification of the processing-related portions of the data.

[0021] Another preferred improvement of the invention is that the neural quality network is trained as follows: Multiple animal carcasses are provided; prior to performing the first processing step, input sensor data for each of the multiple carcass portions is acquired using the sensor device; a human trainer classifies partial data from the input sensor data that is relevant to the processing quality of the first processing step by evaluating the input sensor data; the human trainer assigns at least one quality parameter to each of the multiple carcass portions based on the partial data; training data for training the neural quality assessment network is provided by inputting the input sensor data as input data and the respective quality parameters as target data for the neural quality assessment network; the weights of the quality assessment network are repeatedly adjusted based on the difference between the target data and the output data generated by the neural network. Advantageously, it is therefore possible to pre-train the function of the quality assessment network with regard to the optimal determination of the respective quality parameters.

[0022] The aforementioned objective is also achieved by the apparatus mentioned at the beginning, wherein: a conveying device adapted to transport the carcass portion as an input product along a conveying direction to a first processing machine, the first processing machine being adapted to perform a first processing step for processing the carcass portion into an output product; at least one sensor device adapted to acquire input sensor data for searching the carcass portion before performing the first processing step; and a sorting device adapted to sort partial data from the input sensor data that is relevant to the processing quality of the first processing step, wherein the sorting device is configured and adapted to determine the at least one quality parameter based on the partial data and provide the at least one quality parameter for display and / or transmission to a downstream processing machine, the downstream processing machine being adapted to perform at least one subsequent processing step based on the determined quality parameter.

[0023] The advantages achievable through this invention have been described in detail above with respect to the method according to the invention. To avoid repetition, we also refer to the advantages described above with respect to the device according to the invention, which are applied in the same manner to the method claims, with wording substantially similar to that of the method. Therefore, in the following, only selected aspects of the device according to the invention will be commented on separately.

[0024] Another preferred improvement of the invention is that the classification device is configured such that the partial data is classified by an algorithm based on predetermined selection parameters.

[0025] According to another preferred embodiment of the invention, the classification device is configured to perform classification of the partial data by means of a pre-trained neural classification network.

[0026] Another advantageous embodiment of the invention is characterized in that the classification device is configured such that the at least one quality parameter is determined by an algorithm based on predetermined quality assessment parameters.

[0027] According to another preferred embodiment of the invention, the classification device is configured to perform the determination of the at least one quality parameter by means of a pre-trained neural quality assessment network.

[0028] Another advantageous embodiment of the invention is characterized in that the quality assessment network comprises a given number of neural subnetworks, wherein each of the individual neural subnetworks is pre-trained to determine one of the various quality parameters.

[0029] According to another preferred embodiment of the invention, the input sensor data includes at least image data of the carcass portion.

[0030] Another advantageous embodiment of the invention is characterized in that the at least one sensor device is adapted to record the image data in the wavelength range of visible light, infrared, ultraviolet light and / or the spectral range of X-ray radiation.

[0031] According to another preferred embodiment of the invention, the input sensor data of the carcass portion is at least partially mapped in a three-dimensional manner.

[0032] Another preferred improvement of the invention is that the sorting device is configured to determine the estimated weight and / or carcass size of the output product as at least one of various quality parameters.

[0033] According to another preferred embodiment of the invention, the sorting device is configured to determine the quality value assigned to the output product as at least one of various quality parameters, the quality value representing a measure of the expected final quality of the output product after its processing.

[0034] Another advantageous embodiment of the invention is characterized in that the sorting device is configured to determine a total mass value as at least one of various mass parameters, the total mass value being composed of a given first component related to the estimated weight and / or the carcass size and a given second component related to the mass value.

[0035] According to another preferred embodiment, the neural classification network is trained according to the steps mentioned above in this regard.

[0036] Another advantageous embodiment of the invention is characterized in that the neural quality determination network is trained according to the steps mentioned above in this regard. Attached Figure Description

[0037] Other preferred and / or advantageous features and embodiments of the invention can be derived from the dependent claims and the description. Particularly preferred embodiments will be described in detail below with reference to the accompanying drawings: Figure 1 Schematic diagrams of the method and apparatus according to the invention are shown. Figure 2 This diagram illustrates the process of training a neural network to classify partial data. Figure 3 The diagram illustrates the process of training a neural network used to determine quality parameters. Figure 4 The diagram illustrates the relationship between determining quality parameters using algorithms, and Figure 5 This diagram illustrates the relationship between determining quality parameters using a pre-trained neural network. Specific Implementation

[0038] The following will be based on Figure 1 The schematic diagram illustrates, by way of example, the method and apparatus according to the invention. The method and apparatus are configured and adapted to automatically determine at least one quality parameter characterizing the processing of carcass portions.

[0039] The carcass portion is conveyed to the first processing machine 11 in the conveying direction by means of the conveying device 10. Thus, the carcass portion forms the input product of the first processing machine 11, which is adapted to perform a first processing step by means of the first processing step to transport or process the carcass portion as the input product to give an output product.

[0040] Using sensor device 12, input sensor data of the carcass portion is acquired in acquisition step 13 before performing the first processing step. For example, in acquisition step 13, image data is acquired and extracted, which advantageously includes visible light spectrum 14, ultraviolet spectrum 15, infrared spectrum 16, X-ray spectrum 17, and / or depth or 3D information 18. More preferably, the image components containing spectral or depth and 3D information are added by means of addition unit 19 to obtain a total image. Here, the individual components can be weighted according to given specifications to give particular emphasis to features in the image data related to the corresponding processing.

[0041] In the subsequent classification step 20, partial data related to the processing quality in the first processing step is determined from the input data, and preferably generated as object-specific partial data 22. The generation of object-specific partial data 22 also includes, in particular, splitting the input data classified in classification step 20 into smaller parts 21.

[0042] In quality parameter determination step 23, at least one quality parameter is determined based on partial data 22. This at least one quality parameter represents a measure 24 of the expected (final) quality of the output product after its processing. In providing step 25, the at least one determined quality parameter is provided for display and / or transmission to a downstream processing machine adapted to perform at least one subsequent processing step based on the quality parameter.

[0043] Preferably, such as Figure 1 As shown by arrow 26, based on predetermined selection parameters, a partial data classification is performed using an algorithm. "Using an algorithm" means that the partial data is classified using a specified algorithm. For this purpose, selection parameters are specified, and thus optimized and adjusted for the corresponding processing steps. Selection parameters are, for example, different segments or regions of an animal carcass. If the animal carcass is, for example, a poultry carcass, then the selection parameters are, for example, different regions, such as the breast (especially the left and right breast meat slices), legs, neck, rump, etc. More preferably, as... Figure 1As shown by arrow 27, a pre-trained neural classification network is used to classify a portion of the data. More preferably, the at least one quality parameter is determined by an algorithm based on predetermined quality assessment parameters.

[0044] Figure 4 A preferred embodiment of a method 28 for determining quality parameters by means of at least one algorithm is illustrated. For example, object-specific partial data 22 is available as input data for sub-algorithms 29, 30, and 31. By means of these sub-algorithms, a first quality parameter 32, a second quality parameter 33, and an Nth quality parameter 34 are determined, where N is a natural number greater than zero. For example, the first quality parameter 32 is an estimated weight, while the second quality parameter represents the quality of the input product in terms of its expected final quality after processing. The total quality parameter 37 can be determined by linear combination, i.e., by weighting with corresponding weighting factors 35 and 36, based on the first and second quality parameters 32 and 33 described above. The formation of this total quality parameter 37 is not limited to the sum of the two quality parameters 32 and 33 as shown, but includes any desired number of quality parameters as addends. It is also possible that individual quality parameters 31 are not included in the sum but are provided as other quality parameters 34.

[0045] Figure 5 A preferred embodiment of a method 38 for determining quality parameters using at least one pre-trained neural quality assessment network is illustrated by way of example. Specifically, the quality assessment network comprises a given number of neural subnetworks 39, 40, 41, each of which is pre-trained to determine at least one of various quality parameters. For example, object-specific partial data 22 is provided as input data to the neural subnetworks 39, 40, 41.

[0046] Using neural subnetworks 39, 40, and 41, a first quality parameter 32, a second quality parameter 33, and an Nth or other quality parameter 34 are determined, respectively. As described above, the first quality parameter 32 is, for example, an estimated weight, while the second quality parameter represents the expected final quality of the input product after processing. Based on the first quality parameter 32 and the second quality parameter 33, a total quality parameter 37 can be determined by linear combination, i.e., by weighting with corresponding weighting factors 35 and 36. The formation of this total quality parameter 37 is not limited to the sum of the two quality parameters 32 and 33 as shown in the figure, but includes any desired number of quality parameters as additive terms. It is also possible that individual quality parameters 31 are not included in the sum but are provided as other quality parameters 34. Therefore, neural subnetworks 39, 40, and 41 are configured as expert networks suitable for determining the corresponding quality parameters.

[0047] The alternative addition can also provide quality parameters 32, 33, 34, and optionally other quality parameters to another neural network not shown in the figures, which is adapted to determine the total quality parameter 37.

[0048] Preferably, the estimated weight and / or size (i.e., spatial range / size) of the output product is determined both algorithmically and in the determination of at least one of the various quality parameters using a neural network. More preferably, the at least one quality parameter is determined as a quality value assigned to the output product, the quality value representing a measure of the expected final quality of the output product after its processing.

[0049] The aforementioned total mass parameter 37 preferably corresponds to a total mass value, which consists of a given first component related to the estimated weight and / or carcass size and a given second component related to the aforementioned mass value.

[0050] Figure 2 The training of the neural classification network is illustrated schematically. First, multiple animal carcasses are provided, and in acquisition step 13, prior to performing the first processing step, input sensor data for each carcass portion is acquired using the at least one sensor device 12. Then, in classification step 42, the input sensor data is evaluated by a human trainer and broken down into processing-related subsets. In other words, the human trainer performs the classification of subsets from the input sensor data that are relevant to the processing quality in the first processing step by evaluating the input sensor data. Based on these subsets, this specifically specifies classification regions 43 suitable for classifying the relevant subsets.

[0051] In conjunction with classification regions 43, corresponding training data 44 for training the neural classification network is provided from the input sensor data 13. Then, training 45 of the neural classification network is performed by inputting the input sensor data as input data and the respective classification regions 43 as target data for the neural classification network. The weights of the neural network are gradually adjusted by iteratively adjusting the weights of the classification network based on the differences between the target data and the output data generated by the neural network.

[0052] Figure 3 The training of the neural quality network is illustrated schematically. First, in step 13, before performing the first processing step, input sensor data for each of the various carcass parts is acquired by means of the at least one sensor device 12.

[0053] Then, in evaluation step 46, the input sensor data is evaluated by a human trainer, who specifies processing-related quality parameters. The human trainer specifies, based on partial data, a classification region 43 suitable for categorizing the relevant partial data. In other words, the human trainer specifies at least one quality parameter 32, 33, 34 for each carcass portion based on the input sensor data.

[0054] Then, the neural classification network is trained by repeatedly adjusting the weights of the quality assessment network based on the input sensor data as input data and the target data as target data of the neural quality assessment network, as well as the differences between the target data and the output data generated by the neural network.

[0055] All observations made regarding the operation of the method according to the invention also apply to the mode of operation of the device according to the invention, and vice versa. To avoid repetition, some sections discuss only the device or method in detail, but the observations made always apply to both the device and the method.

Claims

1. Method for automatically determining at least one quality parameter characterizing a processing of a carcass part, comprising the following steps: - conveying the carcass part as an input product along a conveying direction by means of a conveying device (11) to a first processing machine (12) for performing a first processing step for processing the carcass part into an output product; - acquiring input sensor data of the carcass part by means of at least one sensor device (12) prior to performing the first processing step; - classifying from the input sensor data partial data relevant in terms of a processing quality of the first processing step; - determining the at least one quality parameter based on the partial data; and - providing the at least one quality parameter for display and / or transmission to a downstream processing machine adapted to perform at least one subsequent processing step based on the determined quality parameter.

2. The method of claim 1, wherein, Classifying the partial data is performed algorithmically based on predetermined selection parameters.

3. The method of claim 1, wherein, Classifying the partial data is performed by means of a pre-trained neural classification network.

4. The method according to any one of claims 1 to 3, characterized in that, Determining the at least one quality parameter is performed algorithmically based on predetermined quality assessment parameters.

5. The method according to any one of claims 1 to 3, characterized in that, Determining the at least one quality parameter is performed by means of a pre-trained neural quality assessment network.

6. The method of claim 5, wherein, The quality assessment network comprises a given number of neural sub-networks (39, 40, 41), each of the individual neural sub-networks (39, 40, 41) being pre-trained to determine at least one of the individual quality parameters, respectively.

7. The method according to any one of claims 1 to 6, characterized in that, The input sensor data comprises at least image data of the carcass part.

8. The method of claim 7, wherein, The sensor device records the image data in a spectral range of visible light, infrared, ultraviolet and / or X-ray radiation.

9. The method according to claim 7 or 8, characterized in that, The input sensor data at least partially maps the carcass part in three dimensions.

10. The method according to any one of claims 4 to 9, characterized in that, Determining an estimated weight and / or carcass dimensions of the output product as at least one of the individual quality parameters.

11. The method according to any one of claims 4 to 10, characterized in that, Determining a quality value assigned to the output product as at least one of the individual quality parameters, the quality value representing a measure of an expected final quality of the output product after its processing.

12. The method of claim 10, wherein, Determining a total quality value (37) as at least one of the individual quality parameters, the total quality value being composed of a given first component related to the estimated weight and / or the carcass dimensions, and a given second component related to the quality value.

13. The method according to any one of claims 3 to 12, characterized in that, The neural classification network is trained by means of the following steps: - providing a plurality of carcass parts; - acquiring input sensor data of each of the plurality of carcass parts by means of the at least one sensor device (12) prior to performing the first processing step (11); - classifying from the input sensor data partial data relevant in terms of a processing quality of the first processing step by a human trainer by evaluating the input sensor data; - assigning by the human trainer a classification region suitable for classifying the relevant partial data based on the partial data. - providing training data for training the neural classification network by inputting the input sensor data as input data and the respective classification region as target data of the neural classification network; - iteratively adjusting the weights of the classification network based on a difference between the target data and output data generated by the neural network.

14. The method according to any one of claims 5 to 13, characterized in that, The neural quality network is trained as follows: - providing a plurality of animal carcasses; - acquiring input sensor data of each of the plurality of carcass portions by means of the sensor device (12) prior to performing the first processing step (11); - classifying, by a human trainer, from the input sensor data, portion data relevant in terms of processing quality of the first processing step by evaluating the input sensor data; - assigning, by the human trainer, at least one quality parameter for each of the plurality of carcass portions based on the portion data; - providing training data for training the neural quality assessment network by inputting the input sensor data as input data and the respective quality parameter as target data of the neural quality assessment network; - iteratively adjusting the weights of the quality assessment network based on a difference between the target data and output data generated by the neural network.

15. Apparatus for automatically determining at least one quality parameter characterizing processing of a carcass portion, comprising: a conveying device (10) adapted to convey the carcass portion as input product in a conveying direction to a first processing machine (11) adapted to perform a first processing step for processing the carcass portion into an output product; at least one sensor device (12) adapted to acquire input sensor data of the carcass portion prior to performing the first processing step; a classification device adapted to classify, from the input sensor data, portion data relevant in terms of processing quality of the first processing step, wherein the classification device is configured and adapted to determine the at least one quality parameter based on the portion data and to provide the at least one quality parameter for display and / or transmission to a downstream processing machine adapted to perform at least one subsequent processing step based on the determined quality parameter.

16. The apparatus of claim 15, wherein, The classification device is configured such that classifying the portion data is performed algorithmically based on predetermined selection parameters.

17. The apparatus of claim 16, wherein, The classification device is configured such that classifying the portion data is performed by means of a pre-trained neural classification network.

18. The apparatus of any one of claims 15-17, wherein, The classification device is configured such that determining the at least one quality parameter is performed algorithmically based on predetermined quality assessment parameters.

19. The apparatus of any one of claims 15-17, wherein, The classification device is configured such that determining the at least one quality parameter is performed by means of a pre-trained neural quality assessment network.

20. The apparatus of claim 19, wherein, The quality assessment network comprises a given number of neural sub-networks (39, 40, 41), wherein each of the respective neural sub-networks is pre-trained to determine one of the respective quality parameters, respectively.

21. The apparatus of any one of claims 15 to 20, wherein, The input sensor data comprises at least image data of the carcass portion.

22. The apparatus of claim 21, wherein, The at least one sensor device (12) is adapted such that the image data is recorded in the wavelength range of visible light, infrared, ultraviolet and / or in the spectral range of X-ray radiation.

23. The apparatus of claim 21 or 22, wherein, The input sensor data at least partially maps the carcass part in three dimensions.

24. The apparatus of any one of claims 19-24, wherein, The classification device is configured to determine an estimated weight and / or carcass dimensions of the output product as at least one of the individual quality parameters.

25. The apparatus of any one of claims 19-24, wherein, The classification device is configured to determine a quality value assigned to the output product as at least one of the individual quality parameters, the quality value representing a measure of the expected final quality of the output product after its processing.

26. The apparatus of claim 25, wherein, The classification device is configured to determine a total quality value (37) as at least one of the individual quality parameters, the total quality value being composed of a given first component related to the estimated weight and / or the carcass dimensions and a given second component related to the quality value.

27. The apparatus of any one of claims 15-26, wherein, The neural classification network is trained according to the method of claim 13.

28. The apparatus of any one of claims 15-27, wherein, The neural quality determination network is trained according to the method of claim 14.

Images