Image-based processing of poultry parts

Abstract

The disclosure relates to a system and method of image-based processing of a poultry part, comprising: at an imaging station, imaging the poultry part to obtain at least one image; using at least one predetermined algorithm applied to the at least one image, automatically determining at least one parameter regarding the poultry part; and, at a distance from the imaging station, processing the poultry part in dependence of the at least one parameter.

Description

[0001] P138341PC00

[0002] Title: Image-based processing of poultry parts

[0003] FIELD

[0004] The invention relates to a method of image-based processing of a poultry part, particularly a poultry extremity part such as a leg or wing, as well as to a system configured to carry out the method.

[0005] BACKGROUND

[0006] Image-based processing of poultry parts is known as such. For example, WO2024 / 155189A1 discloses acquiring an image of a poultry leg and determining, based on the acquired image, whether to operate a system according to either a first mode for making a first preparatory incision longitudinally of the poultry leg, or according to a second mode for making a second, different, preparatory incision longitudinally of the poultry leg.

[0007] There is a general desire for further improvements in this area, in particular with respect to versatility, reliability and anatomical precision.

[0008] SUMMARY

[0009] According to an aspect, a system is provided for processing a poultry extremity part, such as a poultry leg or wing, having at least one bone surrounded by meat. The system comprises: an imaging station for obtaining an image of the poultry part; a processing station for processing the poultry part; a conveyor for conveying the poultry part along a conveyance path coincident with the imaging station and the processing station; a bypass mechanism associated with the processing station, arranged for selectively being in a first mode or a second mode, wherein the bypass mechanism in the first mode is configured for having the poultry part processed by the processing station, and wherein the bypass mechanism in the second mode is configured for having the poultry part bypass the processing station to refrainfrom processing the poultry part by the processing station; and a control unit configured for receiving the image, and, based on the image, controlling the bypass mechanism according to selectively the first mode or the second mode. Hence, the processing station can be selectively bypassed, depending on the image, such as in dependence of a parameter that is determined from the image. It will be appreciated that the processing station can be bypassed by temporarily disabling the processing station, and / or by adjusting a conveyance path for the poultry part, such as to circumvent the processing station or divert the poultry part to another processing station. The system may be configured to carry out a method as described herein.

[0010] The control unit may for example automatically determine at least one parameter regarding the poultry part. The parameter may for example be indicative of a fracture of the bone, such as whether or not the bone is fractured, and / or, in case the bone is fractured, a location of the fracture. The parameter may additionally or alternatively be indicative of a joint location, such as of a knee joint that articulates a thigh portion and a drum portion of the poultry leg or an elbow joint that articulates an upper and lower wing part; a size of the poultry part; an orientation of the poultry part; a weight of the poultry part; a presence of the poultry part; a type of the poultry part, such as whether the poultry part is a leg, leg part, leg quarter, wing, wing section, or saddle; a state of the poultry part; etc.

[0011] Optionally, the control unit is configured for determining whether or not the bone is fractured, and controlling the bypass mechanism according to the first mode in case the bone is determined to be not fractured and control the bypass mechanism according to the second mode in case the bone is determined to be fractured, or vice versa.

[0012] Optionally, the control unit is configured for determining a bone fracture location of a fracture of the bone, and controlling the bypass mechanism according to the first mode in case the bone fracture location is determined to be within a first set of predetermined bone fracture locationsand control the bypass mechanism according to the second mode in case the bone fracture location is determined to be within a second set of bone fracture locations. In some instances, despite there being a bone fracture, incising, and deboning, may nonetheless yield acceptable results in practice for fracture locations that lie with a predetermined range.

[0013] Optionally, the processing station is or includes a cutting station arranged for making a preparatory incision longitudinally of the bone. The cutting station may hence be bypassed, to refrain from making the longitudinal incision, for example in case the bone of the poultry part is fractured. The preparatory incision may facilitate deboning. In case the bone is fractured, for example, it may be desirable to refrain deboning, to instead obtain a substantially intact, undeboned, poultry part, e.g. an undeboned leg or wing, or an undeboned thigh portion, drum portion, or wing section. Also, for example, it may be desirable to refrain from incising and deboning relatively small poultry legs, poultry legs with abnormal proportions, and / or poultry legs with an abnormal orientation with respect to the conveyor.

[0014] The poultry part particularly is, or includes, a poultry leg. The bone may hence be a drum bone or thigh bone for example. The poultry leg may for instance be a drum portion of the poultry leg, separated from a thigh portion. The poultry leg may alternatively be a thigh portion of the poultry leg, separated from a drum portion. The poultry leg may also include a thigh portion and drum portion attached to each other. The poultry leg may also include a thigh portion and drum portion attached to each other, wherein a back portion is attached to the thigh portion. The processing station may for example be configured for making the preparatory incision longitudinally along the drum portion or along the thigh portion. The processing station for example comprises a J-cut device configured for making a J-cut along the drum portion and the thigh portion.The poultry part may also be, or include, a poultry wing or part thereof. The bone may hence be a wing bone, such as the humerus, radius and / or ulna.

[0015] Optionally, the processing station is or includes a deboning station arranged for deboning the poultry part. The processing station may for example comprise a device for extracting meat from the bone, such as a scraper device. It will be appreciated that processing station may include a cutting station as well as a deboning station.

[0016] Optionally, the poultry extremity part is or includes a poultry leg or poultry wing having a first portion and a second portion articulated by a joint and surrounded by meat.

[0017] Optionally, the processing station is configured for separating the first portion from the second portion.

[0018] Optionally, the control unit is configured for determining, based on the image, whether or not a first bone of the first portion is fractured, and control the bypass mechanism according to the first mode in case the first bone is determined to be not fractured and control the bypass mechanism according to the second mode in case the first bone is determined to be fractured, or vice versa. Hence, in case the first bone is determined to be fractured, it may be desirable to separate the first portion from the second portion. The second portion may for instance be deboned, while the first portion can be obtained as a substantially intact product. If the first bone is determined to be not fractured, it may be desirable to leave the first portion and the second portion attached to each other, to be jointly processed further, e.g. deboned.

[0019] Optionally, the control unit is configured for determining, based on the image, whether or not the first portion at a side opposite to the joint has a further poultry portion attached to it, and control the bypass mechanism according to the first mode in case the first portion is determined to have the further poultry portion attached to it and control the bypass mechanismaccording to the second mode in case the first portion is determined to not have a further poultry portion attached to it, or vice versa.

[0020] Optionally, the poultry part is or includes a poultry leg having a thigh portion and a drum portion articulated by a joint and surrounded by meat.

[0021] Optionally, the processing station is configured for separating the thigh portion from the drum portion. Hence, depending on the image, the control unit can control the bypass mechanism to separate the drum portion for the thigh portion, e.g. severing the knee joint, or to leave the drum portion and thigh portion attached to each other.

[0022] Optionally, the control unit is configured for determining, based on the image, whether or not a thigh bone of the thigh portion is fractured, and control the bypass mechanism according to the first mode in case the thigh bone is determined to be not fractured and control the bypass mechanism according to the second mode in case the thigh bone is determined to be fractured. Hence, in case the thigh bone is determined to be fractured, it may be desirable to separate the thigh portion from the drum portion. The drum portion may for instance be deboned, while the thigh portion can be obtained as a substantially intact product. If the thigh bone is determined to be not fractured, it may be desirable to leave the thigh portion and the drum portion attached to each other, to be jointly processed further, e.g. deboned.

[0023] Optionally, the control unit is configured for determining, based on the image, whether or not the thigh portion has a poultry back portion attached to it, and control the bypass mechanism according to the first mode in case the thigh portion is determined to have the poultry back portion attached to it and control the bypass mechanism according to the second mode in case the thigh portion is determined to not have a poultry back portion attached to it.

[0024] Optionally, the processing station is or includes an unloading station arranged for unloading the poultry part from the conveyor. Hence, dependingon the image, the control unit can control the bypass mechanism to unload or not unload the poultry part from the conveyor. For example, it may be desirable to unload poultry parts with abnormalities, such as bone fractures, and / or abnormal proportions. Alternatively, it may be desirable to unload poultry parts without abnormalities.

[0025] Optionally, the processing station is or includes a reloading station arranged for reloading the poultry part to another conveyor.

[0026] Optionally, the imaging station includes an x-ray imaging device and the image is an x-ray image.

[0027] Another aspect provides a method of image-based processing of a poultry extremity part such as a poultry leg or poultry wing. The method comprises: at an imaging station, imaging the poultry part to obtain at least one image; using at least one predetermined algorithm applied to the at least one image automatically determining at least one parameter regarding the poultry part; and, at a distance from the imaging station, processing the poultry part in dependence of the at least one parameter.

[0028] A further aspect provides a system for image-based processing of a poultry part, configured to carry out the method as described herein. The system comprises: the imaging station; a parameter determining unit operatively connected to the imaging station and configured to carry out the automatic determining of the at least one parameter; and at least one processing unit operatively connected to the parameter determining unit and configured to carry out the processing of the poultry part in dependence of the at least one parameter.

[0029] A further aspect provides the parameter determining unit of the system as described herein.

[0030] A further aspect provides a computer readable storage medium storing data that, when executed by a computer, causes the computer to carry out the automatic determining of the at least one parameter in the method as described herein.The automatic determining uses at least one predetermined algorithm applied to the at least one image. As further explained elsewhere herein, it has been found that various improvements in versatility, reliability and anatomical precision of the processing can thus be provided. To facilitate the use of the at least one predetermined algorithm, the parameter determining unit preferably comprises a memory storing data representative of the at least one predetermined algorithm.

[0031] Optionally, the imaging is or comprises X-ray imaging. Optionally, the at least one image is or comprises at least one X-ray image. In this way, advantageously, internal structures of the poultry part can be discernable in the at least one image. In particular, higher density structures such as bones may be discerned with respect to lower density structures such as meat and skin without requiring said structures to be exposed. Alternatively or additionally, the imaging may be or comprise one or more of: visible light imaging, ultrasound imaging, magnetic resonance imaging, and optical coherence imaging. Correspondingly, the at least one image may be or comprise one or more of: at least one ultrasound image, at least one magnetic resonance image, and at least one optical coherence image. More generally, the imaging may be configured to show one or more internal structures of the poultry part, and / or the at least one image may show one or more internal structures of the poultry part. Examples of internal structures of a poultry part include: bones, in particular a drum bone, a thigh bone, a wing bone, a patella; joints, in particular a knee joint, hip joint or elbow joint; meat; and sinew. Preferably, if the poultry part is or includes a poultry leg, the at least one image shows at least the knee joint, at least part of the thigh bone and at least part of the drum bone.

[0032] Optionally, the at least one parameter comprises at least one joint parameter, in particular at least one joint position parameter. In this way, the processing of the poultry part can be adapted depending on the at least one joint position parameter. The at least one joint parameter preferablyincludes at least one leg joint parameter, such as at least one knee joint parameter, in particular at least one knee joint position parameter; or an wing joint parameter, such as at least one elbow joint parameter, in particular at least one elbow joint position parameter.

[0033] Optionally, the at least one parameter comprises at least one bone fracture parameter. In this way, the processing of the poultry part can be adapted depending on the at least one bone fracture parameter. Examples of possible bone fracture parameters include: a parameter representing a likelihood or conclusion of any bone fracture being present in the poultry part; and a parameter representing a position of a bone fracture in the poultry part.

[0034] Optionally, the at least one parameter comprises at least one orientation parameter. In this way, the processing of the poultry part can be adapted depending on the at least one orientation parameter. Examples of possible orientation parameters include: a parameter representing a likelihood or conclusion of the poultry part having a predetermined orientation; and a parameter representing an actual orientation of the poultry part.

[0035] Optionally, the at least one parameter comprises at least one weight parameter. In this way, the processing of the poultry leg can be adapted depending on the at least one weight parameter. Example of a possible weight parameter include: a parameter representing a weight of the whole poultry part; and a parameter representing a weight of a predetermined part of the poultry part, for example a thigh section or a drum section of a poultry leg.

[0036] Optionally, the at least one parameter comprises at least one size parameter. Possible examples of such a size parameter include: a leg length, an thigh length, a drum length, a wing length, a wingtip length, a midwing length, upper wing length, a leg width, an thigh width, and a drum width, a wing width, a wingtip width, a midwing width, an upper wing width, etc. In this way, one or more size parameters can be taken into account in theprocessing. For example, parts of different sizes may be cut differently and / or may be processed using different processing units.

[0037] Optionally, the at least one parameter comprises at least one product type parameter. It can for example be determined what type of poultry part is present in the image, e.g. whether the poultry part is or includes a poultry leg, leg quarter, drum portion, thigh portion, poultry wing, saddle, etc.

[0038] Optionally, the parameter determining unit comprises a graphics processing unit. In this way, the application of the algorithm to the at least one image can be performed particularly efficiently and quickly.

[0039] Optionally, the automatic determining of the at least one parameter comprises automatically extracting and / or classifying one or more image patches from the at least one image. In this way, a modular patch-wise algorithmic approach can be utilized, enabling improvements in efficiency and reliability.

[0040] Optionally, the automatic determining of the at least one parameter comprises automatically fitting a curve based on the at least one image. The curve preferably corresponds to a center line of the part, in particular of one or more bones of the poultry part. In this way, the curve can be taken into account in one or more algorithmic steps, for example in the automatic extraction and / or classification of the one or more image patches.

[0041] Optionally, the automatic determining of the at least one parameter comprises automatically determining at least one gradient from the at least one image. In this way, analysis of locations and directions of edges such as bone edges in the at least one image can be facilitated. In turn, such analyses can facilitate further algorithmic steps such as the curve fitting.

[0042] Optionally, the automatic determining of the at least one parameter comprises automatically determining a set of line segments based on the at least one image. In this way, further algorithmic steps can be facilitated. For example, line segments linking opposite bone edges can be used as references for the curve fitting.Optionally, the automatic determining of the at least one parameter comprises automatically calculating at least one radiation attenuation parameter from the at least one image. In this way, estimation of one or more volume and / or mass parameters can be facilitated, for example.

[0043] Optionally, the automatic determining of the at least one parameter comprises automatically applying a filter to data relating to the at least one image. In this way, further algorithmic steps can be applied selectively to filtered data, e.g. facilitating suppression of irregularities and / or amplifying relevant characteristics.

[0044] Optionally, the at least one predetermined algorithm comprises at least one convolution kernel. In this way, filtering and other algorithmic steps can be applied particularly efficiently with respect to the at least one image.

[0045] Optionally, the at least one predetermined algorithm comprises at least one classification algorithm. In this way, the automatic determining can utilize distinctions among different predefined classes, for example a broken bone class, an unbroken bone class, and a joint class. Alternative sets of predefined classes can be utilized in relation the determination of an orientation of the poultry part, or for the detection of a poultry part type, for example.

[0046] Optionally, the at least one predetermined algorithm comprises at least one regression algorithm. In this way, the automatic determining can utilize predetermined, e.g. empirically or theoretically determined, relationships among different parameters.

[0047] Optionally, the at least one predetermined algorithm comprises at least one trained model. In this way, historical empirical data, e.g. comprising historical images annotated by experts, can be utilized particularly effectively in the automatic determining.

[0048] Optionally, the at least one predetermined algorithm comprises at least one heuristic algorithm. In this way, domain knowledge from subject-matter experts can be utilized particularly efficiently and robustly in the automatic determining.

[0049] Optionally, the system further comprises a conveyor. In this way, poultry parts can be imaged and processed in a sequential manner, in particular at subsequent stations.

[0050] Optionally, the poultry part is imaged by the imaging station without interrupting conveyance or changing a conveyance speed of the poultry part by the conveyor.

[0051] Optionally, the imaging station comprises a line-scan camera. Alternatively, or additionally, the imaging station may comprise an area-scan camera.

[0052] Optionally, the method comprises conveying the poultry part along the imaging station prior to the processing of the poultry part. Optionally, the conveyor is configured to convey the poultry part along the imaging station prior to the processing of the poultry part. In this way, the imaging can be performed as part of an automatic sequential handling of poultry parts.

[0053] Optionally, the processing of the poultry part in dependence of the at least one parameter comprises conveying the poultry part along a conveying path that depends on the at least one parameter. Optionally, the conveyor is configured to convey the poultry part along a conveying path that depends on the at least one parameter. In this way, parallel arrangements of processing units can be utilized, enabling additional versatility and efficiency of the image-based processing.

[0054] Optionally, the processing of the poultry part in dependence of the at least one parameter comprises applying to the poultry part an orientation adjustment that depends on the at least one parameter. Optionally, the at least one processing unit comprises an orientation adjustment unit configured to apply to the poultry part an orientation adjustment that depends on the at least one parameter. In this way, for example, subsequentpoultry parts can be made to conform to a substantially same orientation to facilitate further processing.

[0055] Optionally, the processing of the poultry part in dependence of the at least one parameter comprises cutting the poultry part along a cutting path that depends on the at least one parameter. Optionally, the at least one processing unit comprises at least one cutting unit configured to cut the poultry part along a cutting path that depends on the at least one parameter. In this way, image-based cutting can be performed, e.g. to obtain specific cuts from the poultry part and / or to facilitate further processing that may include for example pulling, scraping and / or further cutting.

[0056] Optionally, the processing of the poultry part in dependence of the at least one parameter comprises causing the poultry part to visit or not visit a predetermined processing station in dependence of the at least one parameter. In this way, one or more processing steps can be performed selectively, in particular in combination with parallel arrangements of processing units as mentioned elsewhere herein.

[0057] Optionally, the processing of the poultry part in dependence of the at least one parameter comprises causing the poultry part to be retained on or released from a conveyor in dependence of the at least one parameter. In this way, the image-based processing can include image-based sorting, grading, and / or discarding, for example.

[0058] It shall be appreciated that aspects and options described herein can be variously combined, as can be understood particularly well with reference to the below detailed description.

[0059] DETAILED DESCRIPTION

[0060] In the following, the invention will be explained further using examples of embodiments and drawings. The drawings are schematic and merely show examples. In the drawings, corresponding elements are provided with corresponding reference signs. In the drawings:Fig. 1 shows a flow chart of a method of image-based processing of a poultry leg;

[0061] Fig. 2 shows a diagram of a system for image-based processing of a poultry leg;

[0062] Fig. 3A shows a top view of an assembly for image-based processing of a poultry leg, comprising stations arranged along a conveyor;

[0063] Fig. 3B shows a variation of Fig. 3A wherein the conveyor is forked to provide different conveyor paths;

[0064] Figs. 4A-C each show a same X-ray image of a poultry leg overlaid with geometric indications relating to algorithmic steps for the automatic determining of at least one parameter regarding the poultry leg; and Figs. 5A-C show annotated X-ray images of poultry legs taken with the poultry leg in different orientations with respect to the imaging station.

[0065] For the purpose of understanding the invention, examples of a method 100 and a system 200 are explained for a situation in which the poultry part is a poultry leg. The poultry leg particularly has a thigh portion and a drum portion articulated by a knee joint. It will however be appreciated that the method 100 and system 200 described herein can equally be applied to other poultry extremity parts as well, for example to a poultry wing, a leg with part of a poultry back attached thereto, a drum portion, a thigh portion, wing sections, a poultry saddle in which two legs are attached via a back portion, etc.

[0066] Fig. 1 generally illustrates a method 100 of image-based processing of a poultry leg. The method 100 comprises: presenting the poultry leg to an imaging station 201, and, at the imaging station 201, imaging 101 the poultry leg to obtain at least one image; using at least one predetermined algorithm applied to the at least one image, automatically determining 102 at least one parameter regarding the poultry leg; and, at a distance from the imaging station, processing 103 the poultry leg in dependence of the at least one parameter.Fig. 2 generally illustrates a system 200 for image-based processing of a poultry leg, for example configured to carry out the method 100 as described herein. The system 200 comprises: an imaging station 201; a control unit 202 having a parameter determining unit operatively connected to the imaging 201 station. The control unit 202 is configured to carry out the automatic determining 102 of the at least one parameter. At least one processing station 204 is operatively connected to the control unit 202 and configured to carry out the processing 103 of the poultry leg in dependence of the at least one parameter. The system 200 in this example also comprises a bypass mechanism 206. The bypass mechanism 206 is controlled by the control unit 202 for operating according to a first mode or a second mode. In the first mode, the bypass mechanism 206 enables the poultry leg to be processed by the processing station 204. In the second mode, the bypass mechanism 206 causes the poultry part to bypass the processing station 204. The bypass mechanism 206 may for example temporarily disable the processing station 204, or provide an alternative conveyance path that circumvents the processing station 204. In this example, the bypass mechanism 206 is arranged to temporarily disable the processing station 204 in the second mode so that the poultry leg bypasses the processing station 204 and not processed by the processing station 204.

[0067] In a first embodiment of the method 100 and the system 200, the at least one parameter comprises parameters representative of where in a specific X-ray image of the at least one image bone tissue is present, e.g. more likely than not or according to another criterion. In this first embodiment, the at least one predetermined algorithm defines, and the automatic determining 102 comprises, at least the following subsequent steps 1.1 to 1.4.

[0068] In step 1.1, the X-ray image and / or an image derived therefrom is divided into a set of strips, each strip extending mainly transverse to a predetermined expected direction in which the poultry leg mainly extends in the image.In step 1.2, for each strip, an image intensity level is determined for each position of a series of positions along the length of the strip. The image intensity levels are preferably determined based on pixel values of the image, wherein either higher or lower pixel values may correspond to higher intensity. In turn, the pixel values are preferably based on X-ray brightness. The determining of the intensity levels may comprise filtering and / or thresholding. The intensity levels in each strip comprise at least two different intensity levels.

[0069] In step 1.3, for each strip, a pair of inflection positions along the length of the strip is determined representative of inflections of the determined intensity levels, such that the intensity levels determined for positions within the range defined by the inflection positions mainly exceed the intensity levels determined for positions outside said range.

[0070] In step 1.4, the parameters representative of where in the X-ray image bone tissue is present are determined based on the set of pairs of inflection positions determined for the set of strips. For example, first inflection positions may be interpolated across strips to determine a first bone boundary and second inflection positions may be interpolated across strips to determine a second bone boundary, wherein in each strip, relative to a common origin position, the second inflection position is beyond the first inflection position.

[0071] In the first embodiment, preferably, the at least one predetermined algorithm comprises at least one classification algorithm. Specifically, in this case, the at least two different intensity levels are an example of two different classes, so that the step of determining the intensity levels, optionally together with preceding steps, is an example of a classification algorithm being executed.

[0072] In the first embodiment, preferably, the at least one predetermined algorithm comprises at least one regression algorithm. Specifically, in this case, the parameters representative of where in the X-ray image bone tissueis present are an example of likelihood parameters determined from data of the image, so that the step of determining the parameters representative of where in the X-ray image bone tissue is present, optionally together with preceding steps, is an example of a regression algorithm being executed.

[0073] In the first embodiment, the optional step of filtering and / or thresholding may use one or more convolution kernels, for example.

[0074] By processing the poultry leg in dependence of one or more of the parameters representative of where in the X-ray image bone tissue is present, in particular as in the first embodiment, various improvements in versatility, reliability and anatomical precision of the processing can be provided. For example, automated cutting may be performed more precisely for a larger anatomical variety of poultry legs.

[0075] In a second embodiment of the method 100 and the system 200, the at least one parameter comprises parameters representative of where in a specific X-ray image of the at least one image bone tissue is present, e.g. more likely than not or according to another criterion. In this second embodiment, the at least one predetermined algorithm defines, and the automatic determining 102 comprises, at least the following subsequent steps 2.1 to 2.4.

[0076] In step 2.1, the X-ray image and / or an image derived therefrom is filtered using a convolution kernel to determine spatial derivatives of the image intensity across the image, e.g. at each pixel, wherein the derivative has both a magnitude and a direction relative to the image. Preferably, the Sobel operator or a variant thereof is used as the convolution kernel. The Sobel operator is known as such for determining respective intensity derivative components for x- and y-directions in an image, wherein a magnitude of the derivative can then be determined as the length of the hypotenuse of the two components, and the direction of the derivative can be determined as the arc tangent of the two components.In step 2.2, positions in the image having gradients with a magnitude larger than a predetermined threshold and a direction within a predetermined range are selected as possible bone edge positions.

[0077] In step 2.3, for some or all of the possible bone edge positions, one corresponding possible bone edge position is determined, in particular searched, having a similar magnitude (i.e. the difference in magnitudes is within a predetermined range that includes zero) and a substantially opposite direction (i.e. the difference in directions is within a predetermined range that includes 180 degrees). If multiple positions satisfy the criteria for similar magnitude and substantially opposite direction, the spatially nearest position may be selected as the one corresponding possible bone edge position. The result of step 2.3 thus is or includes a set of determined pairs of possible bone edge positions. Each of the pairs may be represented as a respective line segment LS overlaid on the image, e.g. as shown in Fig. 4A as illustrative example. Additional selection criteria may be applied to the set of pairs, e.g. to reject one or more determined pairs as relatively unsuitable for further use. For clarity of the drawing, a relatively small number of line segments LS is shown in Fig. 4A, wherein it shall be appreciated that additional such line segments may be determined, e.g. at intermittent positions between the shown line segments LS.

[0078] In step 2.4, a curve C is fitted so as to intersect some or all of the aforementioned line segments LS at central positions along the line segments LS. In particular, a so-called Bezier-type curve-fitting algorithm may be used. The resulting curve C is then considered to represent and / or approximate a central axis of bones of the poultry leg, so that the parameters defining the curve can be regarded as parameters representative of where in the X-ray image bone tissue is present. Additionally or alternatively, the parameters defining the line segments LS, i.e. the set of pairs of possible bone edge positions, can be regarded as parameters representative of where in the X-ray image bone tissue is present.In the second embodiment, preferably, the at least one predetermined algorithm comprises at least one convolution kernel. Specifically, in this case, a convolution kernel may be used to determine spatial derivatives of intensity as described.

[0079] In the second embodiment, preferably, the at least one predetermined algorithm comprises at least one classification algorithm. Specifically, in this case, possible bone edge positions may be classified as belonging or not belonging to a particular pair, and such pairs may be classified as suitable or not suitable for further use.

[0080] In the second embodiment, preferably, the at least one predetermined algorithm comprises at least one regression algorithm. Specifically, in this case, a regression algorithm may form part or all of the curve-fitting algorithm.

[0081] In the second embodiment, preferably, the at least one predetermined algorithm comprises at least one heuristic algorithm. Specifically, in this case, selections in steps 2.2 and 2.3 may be implemented using heuristics, e.g. based on expert knowledge, experimental data, etc.

[0082] In the second embodiment, preferably, the automatic determining of the at least one parameter comprises automatically fitting a curve based on the at least one image. Specifically, in this case, such a curve C may be fitted as describedin step 2.4.

[0083] In the second embodiment, preferably, the automatic determining of the at least one parameter comprises automatically determining at least one gradient from the at least one image. Specifically, in this case, gradients may be determined as describedin step 2.1.

[0084] In the second embodiment, preferably, the automatic determining of the at least one parameter comprises automatically determining a set of line segments based on the at least one image. Specifically, in this case, line segments LS representing pairs of possible bone edge positions may be determined as in step 2.3.By processing the poultry leg in dependence of one or more of the parameters representative of where in the X-ray image bone tissue is present, in particular as in the second embodiment, various improvements in versatility, reliability and anatomical precision of the processing can be provided. For example, automated cutting may be performed more precisely for a larger anatomical variety of poultry legs.

[0085] In a third embodiment of the method 100 and the system 200, the at least one parameter comprises parameters representative of where in a specific X-ray image of the at least one image one or more predefined bone interruption features such as fractures and / or joints are present, e.g. more likely than not or according to another criterion. In this third embodiment, the at least one predetermined algorithm defines, and the automatic determining 102 comprises, at least the following subsequent steps 3.1 to 3.3.

[0086] In step 3.1, a curve C representing and / or approximating a central axis of bones of the poultry leg is used to define a set of mutually overlapping patches P distributed along the curve C. The curve C may be determined using steps 2.1 to 2.4 described above, for example. The curve C may alternatively be determined using steps 1.1 to 1.4 described above, for example. The classified output from steps 1.1 to 1.4 can for instance be used to create a set of points along a center of the bone in a similar way as in steps 2.1 to 2.4. The curve C may also be determined using another method known in the art.

[0087] In Fig. 4B, as illustration, such a series of patches P is shown using white-line rectangles, of which for clarity of the drawing only some patches P are provided with a reference sign P. Sizes and mutual distances of the patches may be predetermined and / or may be determined using one or more parameters defining the curve C, for example, in particular such that areas of the image expected to show bone are substantially fully covered by the series of patches P.In step 3.2, some or all of the patches P are classified with respect to the presence of interruption features such as fractures and / or joints in the respective patch P. Thereto, a classification algorithm may be used that applies to the patch P a classification model trained using a dataset comprising annotated patches. Examples of possible class labels relating to bone interruption features include: ‘broken’, ‘unbroken’, ‘knee joint’, ‘thigh joint’, and ‘ankle joint’. To facilitate the classification, the relevant patch P is preferably first extracted from the image so as to not burden the classification by data from outside the patch. The classification algorithm may include one or more refinement steps in which the model outputs for multiple of the separate patches P are analyzed together taking into account the relative positions of the patches P. In particular, such refinement steps may be designed to promote a coherent and consistent result in terms of classifications across the set of patches P in view of their relative positions and known general anatomical characteristics of poultry legs, such as general relative positions of different joints.

[0088] In step 3.3, the results from step 3.2 are translated into parameters representative of positions and associated types of bone interruption features in the image. For example, as illustrated in Fig. 4C, the parameters may define one or more levels and / or coordinates of a knee joint KJ and one or more levels and / or coordinates of a thigh joint TJ. With continued reference to Fig. 4C, it shall be appreciated that in the present context a joint such as a thigh joint TJ may be defined as the relevant end of an associated bone such as a femur bone, even if no articulation is present there anymore, e.g. due to a preceding butchering step. The parameters may additionally define whether or not any fractures are present, and, if so, their number and position(s), e.g. in terms of a level and / or coordinate.

[0089] In the third embodiment, preferably, the at least one predetermined algorithm comprises at least one classification algorithm. Specifically, in thiscase, patches may be classified with respect to bone interruption features as describedin step 3.2.

[0090] In the third embodiment, preferably, the at least one predetermined algorithm comprises at least one trained model. Specifically, in this case, a trained classification model may be used as described in step 3.2.

[0091] In the third embodiment, preferably, the at least one parameter comprises at least one leg joint position parameter. Specifically, in this case, one or more levels and / or coordinates of leg joints may be determined as describedin step 3.3.

[0092] In the third embodiment, preferably, the at least one parameter comprises at least one leg bone fracture parameter. Specifically, in this case, one or more levels and / or coordinates of bone fractures may be determined as described in step 3.3, in particular if any bone fractures are determined to be present.

[0093] In the third embodiment, preferably, the at least one parameter comprises at least one leg size parameter. Specifically, in this case, levels and / or coordinates of leg joints as described in step 3.3 may be representative of leg sizes, e.g. if defined or considered relative to each other and / or to some common reference that may be defined based on a known holding position of the poultry leg relative to the imaging station 201.

[0094] In the third embodiment, preferably, the automatic determining of the at least one parameter comprises automatically extracting and / or classifying one or more image patches from the at least one image. Specifically, in this case, patches P may be extracted and classified as describedin step 3.2.

[0095] By processing the poultry leg in dependence of one or more of the parameters representative of where in a specific X-ray image of the at least one image one or more predefined bone interruption features such as fractures and / or joints are present, in particular as in the third embodiment, various improvements in versatility, reliability and anatomical precision ofthe processing can be provided. For example, automated cutting may be performed more precisely for a larger anatomical variety of poultry legs.

[0096] In a fourth embodiment of the method 100 and the system 200, the at least one parameter comprises one or more parameters representative of an orientation of the poultry leg, for example whether or not the poultry leg is shown in a correct orientation for processing, e.g. more likely than not or according to another criterion. In this fourth embodiment, the at least one predetermined algorithm defines, and the automatic determining 102 comprises, at least the following step 4.1.

[0097] In step 4.1, an X-ray image of the at least one image is classified with respect to the orientation of the poultry leg relative to the imaging station 201 at the time of taking the image. Thereto, a classification algorithm may be used that applies to the image a classification model trained using a dataset comprising annotated images. Examples of possible class labels relating to poultry leg orientation include: ‘orientation correct’ (see e.g. Fig. 5A as illustrative example); ‘orientation incorrect by rotation of about 180 degrees about longitudinal axis of leg’ (see e.g. Fig. 5B as illustrative example); and ‘orientation incorrect by rotation of about 90 degrees about longitudinal axis of leg’ (see e.g. Fig. 5C as illustrative example). The application of the model may return a likelihood level for each predefined class, wherein the highest likelihood level among the classes may be used as a confidence level for the classification. If the confidence level is high, e.g. at or above a predefined threshold, the respective class may be selected as parameter representative of whether or not the poultry leg is shown in a correct orientation for processing. If the confidence level is low, e.g. below the predefined threshold, the parameter representative of whether or not the poultry leg is shown in a correct orientation for processing may be determined as ‘orientation possibly incorrect’, for example.

[0098] In the fourth embodiment, preferably, the at least one predetermined algorithm comprises at least one trained model. Specifically,in this case, a trained classification model may be used as described in step 4.1.

[0099] In the fourth embodiment, preferably, the at least one parameter comprises at least one leg orientation parameter. Specifically, in this case, the determined class with respect to the orientation of the poultry leg relative to the imaging station 201 at the time of taking the image as described in step 4.1 is an example of a leg orientation parameter.

[0100] In the fourth embodiment, preferably, the processing of the poultry leg in dependence of the at least one parameter comprises applying to the poultry leg an orientation adjustment that depends on the at least one parameter, and / or the at least one processing unit comprises an orientation adjustment unit configured to apply to the poultry leg an orientation adjustment that depends on the at least one parameter. Specifically, in this case, such an orientation adjustment may be applied in dependence of the determined class as described in step 4.1. In particular, the orientation adjustment may serve to correct, after the image is taken, a determined incorrectness of the orientation at the time of taking the image.

[0101] By processing the poultry leg in dependence of one or more parameters representative of whether or not the poultry leg is shown in a correct orientation for processing, in particular as in the fourth embodiment, various improvements in versatility, reliability and anatomical precision of the processing can be provided. For example, if the orientation is incorrect or possibly incorrect, corrective action may be taken automatically and / or by a worker in response to an automatically generated indication. Alternatively or additionally, poultry legs may be routed differently for processing depending on their determined orientation parameter, for example.

[0102] In a fifth embodiment of the method 100 and the system 200, the at least one parameter comprises one or more parameters representative of an estimated weight of the poultry leg. In this fourth embodiment, the at leastone predetermined algorithm defines, and the automatic determining 102 comprises, at least the following step 5.1.

[0103] In step 5.1, a weight of the poultry leg is estimated based on an X-ray image of the at least one image. Thereto, a regression algorithm may be used that applies to the image a regression model trained and / or designed using a dataset comprising images of poultry legs with known weights. In particular, such a model may reflect an estimated linear or other relationship between a total attenuation p as determined from the X-ray image of the poultry leg and a weight of the same poultry leg. Although such a relationship may not be fully deterministic, a usable estimated relationship may nevertheless be determined, e.g. using a regression approach.

[0104] In the fifth embodiment, preferably, the at least one predetermined algorithm comprises at least one trained model. Specifically, in this case, a trained regression model may be used as described in step 5.1.

[0105] In the fifth embodiment, preferably, the at least one parameter comprises at least one leg weight parameter. Specifically, in this case, an estimated leg weight may be determined as described in step 5.1.

[0106] In the fifth embodiment, preferably, the automatic determining of the at least one parameter comprises automatically calculating at least one radiation attenuation parameter from the at least one image. Specifically, in this case, a total attenuation p is determined from the X-ray image as describedin step 5.1.

[0107] By processing the poultry leg in dependence of one or more parameters representative of an estimated weight of the poultry leg, in particular as in the fifth embodiment, various improvements in versatility, reliability and anatomical precision of the processing can be provided. For example, depending on the estimated weight, possibly together with other parameters as described herein, the poultry leg may be processed to obtain different poultry products such as a whole-leg product, a thigh product, a drumstick product, etc.In the described embodiments, preferably, the processing of the poultry leg in dependence of the at least one parameter comprises conveying the poultry leg along a conveying path that depends on the at least one parameter, and / or the system 200 comprises a conveyor 205 configured to convey the poultry leg, in particular along the imaging station 201 prior to the processing of the poultry leg and / or along a conveying path that depends on the at least one parameter. In the described embodiments, preferably, the processing of the poultry leg in dependence of the at least one parameter comprises causing the poultry leg to visit or not visit a predetermined processing station 204a, 204b in dependence of the at least one parameter. As an illustration, Fig. 3A shows a conveyor 205, and a bypass mechanism 406 for providing different conveyor paths to allow the poultry leg to be conveyed along different processing stations 204a, 204b, in particular in dependence of the at least one parameter as determined using the imaging station 201. As another illustration, Fig. 3B shows the conveyor 205, and the bypass mechanism 206 in a forked configuration providing different conveyor paths to allow the poultry leg to be conveyed along different processing stations 204a, 204b, in particular in dependence of the at least one parameter as determined using the imaging station 201 arranged upstream of the fork. Different processing stations 204a, 204b may comprise processing units of different types and / or configurations, for example.

[0108] In the described embodiments, preferably, the processing of the poultry leg in dependence of the at least one parameter comprises cutting the poultry leg along a cutting path CP that depends on the at least one parameter, and / or the at least one processing unit comprises at least one cutting unit configured to cut the poultry leg along a cutting path CP that depends on the at least one parameter. An example of a possible cutting path CP is shown in Fig. 4C, here passing between thigh and drum bones.

[0109] In the described embodiments, preferably, the processing of the poultry leg in dependence of the at least one parameter comprises causing thepoultry leg to be retained on or released from a conveyor 205 in dependence of the at least one parameter.

[0110] In the described embodiments, preferably, the imaging is or comprises X-ray imaging, and / or the at least one image is or comprises at least one X-ray image.

[0111] In the described embodiments, preferably, the automatic determining of the at least one parameter comprises automatically applying a filter to data relating to the at least one image.

[0112] In the described embodiments, preferably, the parameter determining unit comprises a graphics processing unit.

[0113] In the described embodiments, preferably, the parameter determining unit comprises a memory storing data representative of the at least one predetermined algorithm. The memory is preferably implemented as a computer readable storage medium. More generally, a computer readable storage medium may store data that, when executed by a computer, causes the computer to carry out the automatic determining of the at least one parameter according to one or more of the described embodiments. When a computer is provided with such a storage medium, it may form a parameter determining unit of the system 200.

[0114] Although the invention has been explained herein using examples of embodiments and drawings, these do not limit the scope of the invention as determined by the claims. Within said scope, many variations are possible. For example, parameters determined in any embodiment as described herein may be used as inputs for the determining of further parameters is other embodiments as described herein.LIST OF REFERENCE SIGNS 100. Method

[0115] 101. Imaging poultry leg 102. Determining parameter 103. Processing poultry leg 200. System

[0116] 201. Imaging station

[0117] 202. C ontr ol unit

[0118] 204. Processing station 205. Conveyor

[0119] C. Curve

[0120] CP. Cutting path

[0121] KJ. Knee joint

[0122] LS. Line segment

[0123] P. Patch

[0124] TJ. Thigh joint

Claims

28Claims1. A system for processing a poultry extremity part, such as a leg or wing, having at least one bone surrounded by meat, the system comprising:an imaging station for obtaining an image of the poultry extremity part;a processing station for processing the poultry extremity part; a conveyor for conveying the poultry extremity part along a conveyance path coincident with the imaging station and the processing station;a bypass mechanism associated with the processing station, arranged for selectively being in a first mode or a second mode, wherein the bypass mechanism in the first mode is configured for having the poultry extremity part processed by the processing station, and wherein the bypass mechanism in the second mode is configured for having the poultry extremity part bypass the processing station to refrain from processing the poultry extremity part by the processing station; anda control unit configured for receiving the image, and, based on the image, controlling the bypass mechanism according to selectively the first mode or the second mode.

2. The system according to claim 1, wherein the control unit is configured for determining, based on the image, whether or not the bone is fractured, and controlling the bypass mechanism according to the first mode in case the bone is determined to be not fractured and control the bypass mechanism according to the second mode in case the bone is determined to be fractured.

3. The system according to claim 1 or 2, wherein the processing station is or includes a cutting station arranged for making a preparatory incision longitudinally of the bone.

4. The system according to any of the preceding claims, wherein the processing station is or includes a deboning station arranged for deboning the poultry extremity part.

5. The system according to any of the preceding claims, wherein the poultry extremity part is, or includes, a poultry leg or poultry wing having a first portion and a second portion articulated by a joint and surrounded by meat.

6. The system according to claim 5, wherein the processing station is configured for separating the first portion from the second portion.

7. The system according to claim 6, wherein the control unit is configured for determining, based on the image, whether or not a first bone of the first portion is fractured, and control the bypass mechanism according to the first mode in case the first bone is determined to be not fractured and control the bypass mechanism according to the second mode in case the first bone is determined to be fractured.

8. The system according to claim 6 or 7, wherein the control unit is configured for determining, based on the image, whether or not the first portion at a side opposite to the joint has a further poultry portion attached to it, and control the bypass mechanism according to the first mode in case the first portion is determined to have the further poultry portion attached to it and control the bypass mechanism according to the second mode in case the first portion is determined to not have a further poultry portion attached to it.

9. The system according to any of the preceding claims, wherein the processing station is or includes an unloading or reloading station arranged for unloading the poultry extremity part from the conveyor or reloading the poultry extremity part to another conveyor.

10. The system according to any of the preceding claims, wherein the imaging station includes an x-ray imaging device and the image is an x-ray image.

11. A method of image-based processing of a poultry extremity part such as a poultry leg or wing, comprising:at an imaging station, imaging the poultry extremity part to obtain at least one image;using at least one predetermined algorithm applied to the at least one image, automatically determining at least one parameter regarding the poultry extremity part; and,at a distance from the imaging station, processing the poultry extremity part in dependence of the at least one parameter.

12. The method according to claim 11, wherein the at least one parameter comprises at least one joint position parameter.

13. The method according to claim 11 or 12, wherein the at least one parameter comprises at least one bone fracture parameter.

14. The method according to any of claims 11-13, wherein the at least one parameter comprises at least one orientation parameter.

15. The method according to any of claims 11-14, wherein the at least one parameter comprises at least one weight parameter.

16. The method according to any of claims 11-15, wherein the at least one parameter comprises at least one size parameter.

17. The method according to any of claims 11-15, wherein the at least one parameter comprises at least one product type parameter.

18. The method according to any of claims 11-17, comprising conveying the poultry extremity part along the imaging station prior to the processing of the poultry extremity part.

19. The method according to any of claims 11-18, wherein the processing of the poultry part in dependence of the at least one parameter comprises conveying the poultry part along a conveying path that depends on the at least one parameter.

20. The method according to any of claims 11-19, wherein the processing of the poultry extremity part in dependence of the at least one parameter comprises applying to the poultry extremity part an orientation adjustment that depends on the at least one parameter.

21. The method according to any of claims 11-20, wherein the processing of the poultry extremity part in dependence of the at least one parameter comprises cutting the poultry extremity part along a cutting path that depends on the at least one parameter.

22. The method according to any of claims 11-21, wherein the processing of the poultry extremity part in dependence of the at least one parameter comprises causing the poultry extremity part to visit or not visit a32predetermined processing station in dependence of the at least one parameter.

23. The method according to any of claims 11-22, wherein the processing of the poultry extremity part in dependence of the at least one parameter comprises causing the poultry extremity part to be retained on or released from a conveyor in dependence of the at least one parameter.

24. The method according to any of the preceding claims, wherein the imaging is or comprises X-ray imaging, and / or wherein the at least one image is or comprises at least one X-ray image.

25. The method according to any of claims 11-24, wherein the automatic determining of the at least one parameter comprises automatically extracting and / or classifying one or more image patches from the at least one image.

26. The method according to any of claims 11-25, wherein the automatic determining of the at least one parameter comprises automatically fitting a curve based on the at least one image.

27. The method according to any of claims 11-26, wherein the automatic determining of the at least one parameter comprises automatically determining at least one gradient from the at least one image.

28. The method according to any of claims 11-27, wherein the automatic determining of the at least one parameter comprises automatically determining a set of line segments based on the at least one image.3329. The method according to any of claims 11-28, wherein the automatic determining of the at least one parameter comprises automatically calculating at least one radiation attenuation parameter from the at least one image.

30. The method according to any of claims 11-29, wherein the at least one parameter comprises parameters representative of where in a specific X-ray image of the at least one image bone tissue is present,wherein the X-ray image and / or an image derived therefrom is divided into a set of strips, each strip extending mainly transverse to a predetermined expected direction in which the poultry extremity part mainly extends in the image,wherein, for each strip, an image intensity level is determined for each position of a series of positions along the length of the strip,wherein, for each strip, a pair of inflection positions along the length of the strip is determined representative of inflections of the determined intensity levels, such that the intensity levels determined for positions within the range defined by the inflection positions mainly exceed the intensity levels determined for positions outside said range,wherein the parameters representative of where in the X-ray image bone tissue is present are determined based on the set of pairs of inflection positions determined for the set of strips.

31. The method according to claim 30, comprising determining a center between each pair of inflection positions, and fitting a curve so as to intersect some or all of the aforementioned centers, in particular such that the curve represents and / or approximates a central axis of bones of the poultry extremity part.3432. The method according to any of claims 11-31, wherein the at least one parameter comprises parameters representative of where in a specific X-ray image of the at least one image bone tissue is present,wherein the X-ray image and / or an image derived therefrom is filtered using a convolution kernel to determine spatial derivatives of the image intensity across the image, wherein the derivative has both a magnitude and a direction relative to the image,wherein positions in the image having gradients with a magnitude larger than a predetermined threshold and a direction within a predetermined range are selected as possible bone edge positions,wherein, for some or all of the possible bone edge positions, one corresponding possible bone edge position is determined, in particular searched, having a similar magnitude and a substantially opposite direction, so as to determine a set of pairs of possible bone edge positions representable as line segments overlaid on the image.

33. The method according to claim 32, wherein a curve is fitted so as to intersect some or all of the aforementioned line segments at central positions along the line segments, in particular such that the curve represents and / or approximates a central axis of bones of the poultry extremity part.

34. The method according to any of claims 11-33, wherein the at least one parameter comprises parameters representative of where in a specific X-ray image of the at least one image one or more predefined bone interruption features such as fractures and / or joints are present,wherein a curve representing and / or approximating a central axis of bones of the poultry extremity part, preferably the curve of claim 31 or 33, is used to define a set of mutually overlapping patches distributed along the curve,35wherein some or all of the patches are classified with respect to the presence of interruption features such as fractures and / or joints in the respective patch,wherein the results of the classification are translated into parameters representative of positions and associated types of bone interruption features in the image.

35. The method according to any of claims 11-34, wherein the at least one parameter comprises one or more parameters representative of whether or not the poultry extremity part is shown in a correct orientation for processing,wherein an X-ray image of the at least one image is classified with respect to the orientation of the poultry extremity part relative to the imaging station at the time of taking the image.

36. The method according to any of claims 11-35, wherein the at least one parameter comprises one or more parameters representative of an estimated weight of the poultry extremity part,wherein a weight of the poultry extremity part is estimated based on an X-ray image of the at least one image.

37. A system for image-based processing of a poultry extremity part, configured to carry out the method according to any of claims 11-36, the system comprising:the imaging station;a parameter determining unit operatively connected to the imaging station and configured to carry out the automatic determining of the at least one parameter; and36at least one processing unit operatively connected to the parameter determining unit and configured to carry out the processing of the poultry extremity part in dependence of the at least one parameter.

38. The system according to claim 37, further comprising a conveyor configured to convey the poultry extremity part:along the imaging station prior to the processing of the poultry extremity part; and / oralong a conveying path that depends on the at least one parameter.

39. The system according to claim 37 or 38, wherein the at least one processing unit comprises an orientation adjustment unit configured to apply to the poultry extremity part an orientation adjustment that depends on the at least one parameter.

40. The system according to any of claims 37 - 39, wherein the at least one processing unit comprises at least one cutting unit configured to cut the poultry extremity part along a cutting path that depends on the at least one parameter.

41. The system according to any of claims 37 - 40, wherein the parameter determining unit comprises a graphics processing unit.

42. The system according to any of claims 37 - 41, wherein the parameter determining unit comprises a memory storing data representative of the at least one predetermined algorithm.

43. The parameter determining unit of the system according to any of claims 37 -42.3744. Computer readable storage medium storing data that, when executed by a computer, causes the computer to carry out the automatic determining of the at least one parameter in the method according to any of claims 11 - 36.

Images