A pig carcass processing procedure integrated intelligent device for a slaughtering line

Abstract

The present application relates to live pig slaughtering detection technical field, specifically to a kind of live pig carcass processing procedure integration intelligent device of slaughtering line.It includes:carcass orientation stabilizing component is used to guide carcass to carry out posture calibration, label reading code ware first reads RFID label to obtain carcass information, then multi-spectrum metal detection module adopts three-frequency combination strategy to carry out needle detection to carcass, then vision quarantine module collects carcass image, subsequently central control module takes the reading of RFID label as starting point, uniformly arranges the working time sequence of each module, fuses multi-source real-time information to obtain quality grading conclusion, generates spray printing information according to quarantine result and quality grading conclusion, finally digital code spraying module is executed spray printing on carcass skin.The present application is aimed at high-speed, continuous live pig slaughtering assembly line scene, carries out depth fusion and creative reconstruction to each procedure, realizes the automation, continuous and data of whole process of post-slaughtering processing, significantly improves efficiency.

Description

Technical Field

[0001] This invention relates to the field of pig slaughtering and testing technology, specifically to an integrated intelligent device for the pig carcass processing process in a slaughter line. Background Technology

[0002] With the continuous improvement of meat consumption levels and increasingly stringent food safety regulations, large-scale pig slaughtering has become the industry mainstream. Modern slaughterhouses typically have a single-line capacity of 300-600 pigs per hour, with some ultra-large-scale enterprises even exceeding 800 pigs per hour. However, the post-slaughter processing procedures for pigs are complex, mainly including: carcass quarantine (according to official veterinary inspection standards), detection of metal foreign objects such as broken needles, quality grading, affixing of quarantine stamps, and displaying traceability labels. Therefore, in the context of high-speed, continuous pig slaughtering production lines, how to improve the processing efficiency of each post-slaughter process and reduce labor costs while ensuring food safety has always been a core issue that urgently needs to be addressed in this field. For example, Chinese patent CN202310362601 proposes an X-ray-based carcass foreign object detection system and method, which sets up a vertical detection channel by scanning the carcass with an X-ray machine after it has traveled a long transportation distance, thereby realizing the acquisition of X-ray carcass images and foreign object detection; Chinese patent CN201510051845 proposes a multi-dimensional, refined, and intelligent carcass meat grading method on an automated pig slaughtering line, which integrates historical feeding information before slaughter and actual carcass measurement information after slaughter for phased evaluation to obtain the carcass meat grade; Chinese patent CN202323153920 proposes a device for automatically stamping metal stamps on the surface of livestock carcasses, which automatically stamps the surface of livestock carcasses with synchronously moving heated metal stamps.

[0003] As can be seen, existing technologies or conventional approaches in this field often provide "point-based" solutions to individual problems (such as detecting only metals or automating only stamping). Currently, most slaughterhouses still employ a multi-station, decentralized operation model, where each process is independent. This results in post-slaughter processing of pigs still relying heavily on manual intervention, data lag, and low efficiency. This is because simply connecting and integrating these independent process devices would inevitably lead to systemic imbalances in actual industrial production, presenting significant technical difficulties. Specifically:

[0004] 1. Spatial Layout Conflicts and Efficiency Bottlenecks in Physical Integration: Large equipment such as X-ray machines have rigid requirements for safety distances and radiation protection, and their large size occupies a lot of space. If independent equipment such as visual inspection chambers and automatic hot stamping / coding devices are arranged adjacent to each other in physical location and logically exchange information, it will form an extremely long production line. The physical distance between processes directly translates into long waiting times for carcass transfer, making it impossible to achieve truly streamlined and high-speed processing. This not only increases site costs but may also lead to operational and maintenance difficulties due to the crowded layout.

[0005] 2. The challenge of real-time collaborative control of multi-source heterogeneous sensors: On a production line where the carcass moves continuously, the sensing, processing, and triggering logic of multiple independent devices will amplify any slight delay in any link on the long production line, leading to serious misalignment of actions (such as incorrect hot stamping of the carcass), rendering the entire automated process meaningless, and even causing batch chaos.

[0006] 3. Limitations and risks of detection methods cannot be reconciled: Existing technologies each have their own limitations. While existing X-ray machines can penetrate and detect metal, they pose radiation safety concerns. Traditional metal detection is susceptible to interference from the "product effect" and individual carcass differences, and is insensitive to broken stainless steel needles. On the one hand, broken stainless steel needles (such as austenitic stainless steel) are weakly magnetic or even non-magnetic metals with low conductivity, which is easily overlooked. However, pig carcasses have high water content (approximately 60-75%) and high salt content, and their conductivity is inherent, resulting in "product effect" interference. Traditional metal detection cannot distinguish between carcass variations and genuine broken needles, leading to false alarms or missed detections. On the other hand, there are significant individual differences in carcasses; differences in electrical properties between carcasses of different weights, fatness, and locations cause severe baseline drift in interference signals.

[0007] 4. High-contrast and feature-enhanced imaging requirements of carcasses are difficult to meet in high-speed movement and low-light environments: Carcasses move continuously on the assembly line, while cameras use conventional continuous white light illumination and exposure. The result is likely to be either clear images but slow acquisition speed, forming a bottleneck, or fast acquisition but insufficient image quality to support fine grading. For example, during the exposure time, the carcass moves relative to the camera, resulting in image blur (motion blur). In low-light environments, if you want to prevent motion blur with a very short exposure time, you will inevitably be limited by the brightness of the continuous light source, which is difficult to meet.

[0008] 5. Data fragmentation leads to decision-making lag and broken traceability chains: In the existing multi-station distributed operation mode, data at each stage becomes isolated. Quality grading decisions (such as Chinese patent CN201510051845) must wait for all data on the lengthy production line to be aggregated offline before they can be made, and cannot be completed in real time as the carcass flows through, resulting in serious decision-making lag. Consequently, subsequent identification (stamping / coding) actions cannot immediately carry the correct grading and traceability information. The traceability chain has "breakpoints" or "delays." Broken needle detection is usually performed in the downstream segmentation stage. When a broken needle is found, because the data is not linked and bound in real time at the carcass level, it is impossible to quickly and accurately trace the problematic carcass, leading to the spread of the problem.

[0009] Therefore, the simple connection of existing technologies cannot solve the fundamental contradiction that multiple post-slaughter processing steps cannot effectively coordinate in the high-speed, continuous pig slaughtering production line scenario due to multiple intertwined technical problems such as "spatial conflict, control loss, detection limitations, imaging contradictions, and data fragmentation". Summary of the Invention

[0010] To address the aforementioned technical issues, this invention proposes an integrated intelligent device for the processing of pig carcasses in a slaughter line. This device aims to deeply integrate and creatively reconstruct various processes within a high-speed, continuous pig slaughtering production line, achieving automation, continuity, and data-driven processing throughout the entire post-slaughter process. This significantly improves efficiency and provides key equipment support for the high-quality development of the slaughtering industry.

[0011] In a first aspect, this application provides an integrated intelligent device for the processing of pig carcasses in a slaughter line, including a guide rail arranged along the direction of the pig carcass conveyor line, a plurality of carrying hooks arranged on the guide rail, the carcass being suspended on the carrying hooks and moving continuously with the guide rail, and each carrying hook being equipped with an RFID tag containing carcass information. The device includes:

[0012] The main cabin frame is designed as a semi-enclosed structure covering the length of the carcass, with guide rails passing through the inside of the main cabin frame;

[0013] The carcass guidance and stabilization assembly is used to guide the carcass through attitude calibration and into the main cabin frame along the guide rails;

[0014] A tag reader, installed at the entrance of the main cabin frame, is used to read RFID tags on the carrying pole hooks to obtain carcass information;

[0015] The multi-spectral metal detection module is installed on both sides of the upstream of the main cabin frame and is used to detect broken needles in the carcass using a three-frequency combination strategy.

[0016] The visual quarantine module, installed downstream of the main cabin frame, is used to collect carcass images and identify quarantine results and fat thickness based on the carcass images and carcass information.

[0017] The central control module is used to uniformly arrange the working sequence of each module according to the speed of the conveyor line, starting from the reading of RFID tags. It is also used to collect carcass information, carcass images, quarantine results and fat thickness into multi-source real-time information and fuse the multi-source real-time information to obtain quality grading conclusions. The central control module is also used to generate quarantine inspection stamp patterns and traceability QR codes based on the carcass quarantine results and quality grading conclusions.

[0018] The digital inkjet printing module, installed downstream of the main cabin frame, is used to print quarantine inspection stamps and traceability QR codes on the carcass skin that has passed the quality grading conclusion.

[0019] The label reader, multi-spectral metal detection module, visual quarantine module, and digital inkjet printing module are all connected to the central control module.

[0020] In some embodiments, the multi-spectral metal detection module includes a non-metallic channel partition, a balanced coil assembly, a multi-spectral transmitter, and a signal processor. The non-metallic channel partition is installed on both sides upstream of the main cabin frame to form a detection channel for the body to pass through. The balanced coil assembly includes one transmitting coil and two receiving coils. The transmitting coil is connected to the multi-spectral transmitter, and the two receiving coils are connected to the signal processor. Both the multi-spectral transmitter and the signal processor are connected to the central control module. The multi-spectral metal detection module is used to perform the following steps:

[0021] When the working sequence arranged by the central control module is reached, the multi-spectral transmitter drives the transmitting coil to synchronously transmit a composite electromagnetic wave with three characteristic frequencies: low frequency for suppressing product effects, medium frequency as the main foreign object detection frequency band, and high frequency for enhancing the identification of weak foreign object signals.

[0022] The signal processor acquires the composite output signal from the two receiving coils and decouples the composite output signal to separate the low-frequency component, the intermediate-frequency component, and the high-frequency component.

[0023] The signal processor uses a pre-trained benchmark model based on the low-frequency components to perform dynamic background compensation on the intermediate-frequency and high-frequency components, and determines the broken needle detection result based on the compensated intermediate-frequency and high-frequency components.

[0024] In some embodiments, the signal processor is used to perform the following steps during the pre-training of the baseline model:

[0025] Collect the composite output signal of two receiving coils when several groups of carcasses without metal foreign objects pass through the detection channel at a constant speed;

[0026] Each composite output signal is decoupled to separate the corresponding low-frequency component, mid-frequency component and high-frequency component;

[0027] Calculate the phase and differential signal of the low-frequency, mid-frequency and high-frequency components of each composite output signal to obtain several sets of sample data;

[0028] The baseline model was pre-trained using sample data to obtain a baseline model with low-frequency components as input and the mid-frequency and high-frequency components as input and the background phase and background differential signal of the body without metal foreign objects as output.

[0029] In some embodiments, when the signal processor performs dynamic background compensation for intermediate frequency and high frequency components based on a pre-trained baseline model using low-frequency components, it performs the following steps:

[0030] Calculate the phase and differential signal of the low-frequency component of the current composite output signal, input them into the pre-trained benchmark model, and output the predicted background phase and background differential signal of the mid-frequency and high-frequency components.

[0031] Calculate the actual phase of the intermediate frequency component and the high frequency component of the current composite output signal, as well as the actual differential signal;

[0032] The difference between the actual phase of the intermediate frequency (IF) component and the high frequency (HF) component of the current composite output signal and the actual differential signal and the predicted background phase and background differential signal of the IF component and the high frequency component is calculated to obtain the compensated phase and differential signal of the IF component and the high frequency component.

[0033] In some embodiments, a light-blocking curtain is provided in the middle of the main cabin frame. The light-blocking curtain and the main cabin frame enclose a dark box-like downstream space to serve as an image acquisition channel. The visual quarantine module includes an active lighting component and a distributed camera group. The active lighting component provides lighting around the body. Both the distributed camera group and the active lighting component are connected to the central control module. The visual quarantine module is used to perform the following steps:

[0034] When the working sequence arranged by the central control module is reached, the active fill light component triggers flashes of different wavelengths in a preset order and sets the flash time to a very short range. At the same time, the exposure time of the distributed camera group is set to match the flash time, so as to realize the strobe linkage shooting between the active fill light component and the distributed camera group to obtain the body image.

[0035] In some embodiments, the flash wavelengths triggered by the active lighting component include white light, blue-violet light, and green light.

[0036] In some embodiments, a skin pretreatment module installed at the entrance of the main cabin frame is also included. The skin pretreatment module is located before the multi-spectral metal detection module. The skin pretreatment module includes a bamboo tube assembly and a high-pressure blower. A scraper is provided at the head of the bamboo tube assembly. The length of the bamboo tube assembly is set so that the scraper contacts the attitude-calibrated carcass skin. The working position of the scraper matches the printing height range of the digital inkjet module. The high-pressure blower is connected to the central control module. When the working sequence arranged by the central control module is reached, the high-pressure blower performs high-speed blowing on the attitude-calibrated carcass skin.

[0037] In some embodiments, the digital inkjet printing module is located on the other side of the distributed camera group. The digital inkjet printing module includes a printhead telescopic assembly and an integrated printhead. The printhead telescopic assembly consists of a longitudinal moving screw and a vertical moving screw. The integrated printhead is mounted on the longitudinal moving screw to achieve telescopic movement toward or away from the carcass skin. The longitudinal moving screw, the vertical moving screw, and the integrated printhead are all connected to the central control module. When the working sequence arranged by the central control module is reached, the printhead telescopic assembly drives the integrated printhead to perform high-speed printing of quarantine inspection stamp patterns and traceability QR codes on the carcass skin with a quality grading conclusion of qualified.

[0038] In some embodiments, the carcass guiding and stabilizing assembly includes a first horizontal straight rod and a second horizontal straight rod arranged parallel to the guide rail. The first horizontal straight rod and the second horizontal straight rod are located on both sides of the front leg of the carcass and are parallel in the horizontal direction. A first inclined guide rod and a second inclined guide rod are respectively provided at the ends of the first horizontal straight rod and the second horizontal straight rod facing the entrance of the main cabin frame. The first inclined guide rod and the second inclined guide rod are arranged in opposite directions so that the distance between the first inclined guide rod and the second inclined guide rod gradually decreases along the forward direction of the carcass.

[0039] In some embodiments, the body guiding and stabilizing assembly further includes a third horizontal rod and a fourth horizontal rod arranged parallel to the first horizontal rod. The third horizontal rod and the fourth horizontal rod are located at the middle of the body and the buttocks of the body, respectively, and are parallel in the vertical direction. A third tilting guide rod and a fourth tilting guide rod are respectively provided at the ends of the third horizontal rod and the fourth horizontal rod facing the entrance of the main cabin frame. The third tilting guide rod and the fourth tilting guide rod are both arranged in the same direction as the first tilting guide rod, and the tilting arc of the third tilting guide rod and the fourth tilting guide rod is greater than the tilting arc of the first tilting guide rod.

[0040] The beneficial technical effects of the present invention include at least the following:

[0041] 1. This application presents an integrated intelligent device for the processing of pig carcasses in a slaughtering line. Designed for high-speed, continuous pig slaughtering production lines, it deeply integrates and creatively restructures each process. This application compactly integrates multi-spectral metal detection, visual quarantine, central control decision-making, and digital coding modules into a single main frame. The coupling design of each module is based on its physical characteristics, fundamentally resolving the physical space constraints and redundant layout problems caused by simply connecting large equipment in series, thus greatly saving space and improving production cycle time. In terms of collaborative control, this application uses RFID tags as the sole data link and the central control module as the scheduling core. Based on the conveyor line speed, the working sequence of each module is uniformly arranged to ensure precise connection between "broken needle detection → visual quarantine → graded decision-making → coding execution" within the short few seconds of carcass passage. This effectively overcomes the problems of control synchronization failure and action misalignment caused by multi-source heterogeneous sensors. In terms of detection methods, this application uses multi-spectral metal detection instead of X-ray machines, and combines a three-frequency strategy to effectively address interference from broken stainless steel needles and the "product effect" of carcasses, balancing safety and high sensitivity. Simultaneously, the visual quarantine module acquires images in real time, and the central control module instantly integrates multi-source information for quality grading, resulting in near-zero delay in decision-making. Finally, all this real-time data and grading conclusions are aggregated by the central control module, driving the digital inkjet printing module to instantly print quarantine inspection stamps and traceability QR codes on qualified carcasses, forming a real-time data closed loop from detection and decision-making to labeling. This completely breaks down data silos and achieves a complete and immediate traceability chain. Therefore, this application reconstructs previously discrete multi-process equipment into a collaborative intelligent entity, realizing the automation, continuity, and datafication of the entire post-slaughter processing process, significantly improving efficiency and food safety assurance levels, and providing key equipment support for the high-quality development of the slaughtering industry.

[0042] 2. This application replaces the existing broken needle detection scheme that uses a bulky and radiation-risk X-ray machine for transmission imaging with a multi-spectral metal detection module based on electromagnetic induction disturbance. X-rays identify foreign objects through material attenuation imaging, while the multi-spectral metal detection module detects them by disturbing the alternating electromagnetic field with metal. This directly solves the problems of spatial antagonism and radiation safety. Considering that stainless steel broken needles are the main metallic foreign objects on pig carcasses in slaughter lines, and that they are weakly magnetic or even non-magnetic metals with low conductivity, and that the pork carcass itself generates "product effect" interference, this application creatively adopts a three-frequency combination strategy to detect broken needles on the carcass. This achieves high sensitivity, high reliability, and rapid online detection of small stainless steel broken needles under the interference of the carcass product effect, eliminating the need for further detection in the subsequent cutting process. Specifically, existing technologies struggle to detect weak metallic foreign object characteristic signals amidst extremely strong biological background noise. When the background and foreign object signals overlap in the frequency domain, those skilled in the art may employ general filtering. However, while suppressing the background, general filtering inevitably damages the useful foreign object signal. Therefore, this application specifically utilizes low-frequency features for prediction, then predicts the product effect interference signal, i.e., the "background component," from the original mid / high-frequency signal. This achieves "source separation," enabling better active suppression of interference signals, preserving foreign object signal components distinct from the background features, and thus better protecting weak foreign object signals. Especially in the mid and high-frequency bands, foreign object characteristics (such as phase abrupt changes) are more completely preserved, significantly improving the detection sensitivity for small-sized and weakly magnetic metallic foreign objects.

[0043] 3. In existing technologies, due to significant individual differences in carcasses—different weights, fat percentages, and water content—the electrical characteristics of these carcasses vary, leading to severe baseline drift in interference signals. Fixed filtering thresholds cannot perfectly adapt to all individuals, resulting in incomplete compensation (high residual noise) or overcompensation (signal attenuation). Therefore, the core idea of ​​this application is to address the problem of complex background interference and carcass-specific signal baseline variations. During the offline pre-training phase, the benchmark model does not record the characteristics of each frequency band in isolation, but rather focuses on learning a stable mapping relationship between "low-frequency features" and "mid- and high-frequency features." By learning from a large number of samples, it intrinsically incorporates the range of interference signal variations in normal carcasses. This benchmark model encapsulates the typical characteristics of the "product effect" of pig carcasses. During online detection, the multi-spectral metal detection module measures the current low-frequency characteristics and uses this learned mapping relationship to predict the most likely mid-frequency and high-frequency background characteristics corresponding to the current carcass. This allows for dynamic and adaptive background subtraction within the current mixed mid-frequency and high-frequency signal; that is, dynamic background compensation is based on real-time matching of the current low-frequency signal, rather than a fixed value. Therefore, it can automatically adapt to different carcasses without requiring manual sensitivity adjustments for each pig, ensuring consistent and stable detection, providing a reliable signal baseline for subsequent foreign object signal detection, and significantly reducing the false alarm rate.

[0044] 4. This application abandons the inefficient continuous illumination method used in existing technologies, which is prone to heat accumulation and motion blur. At the physical level, it solves the problem of image clarity for fast-moving objects by using LED strobe to precisely synchronize with the camera shutter. By combining light sources of different wavelengths to enhance the contrast of different tissue features (such as lesions, blood vessels, and fat), it improves the visual detection effect and thus significantly improves the accuracy of subsequent quarantine result identification.

[0045] 5. The skin pretreatment module designed in this application adopts a physical combination of "contact mechanical scraping + non-contact high-pressure air drying". It sets up an active cleaning checkpoint in a very short time before the carcass enters the precision inspection area, automatically giving the carcass surface a repeatable and high-quality basic state, thereby ensuring the reliability and accuracy of all subsequent electromagnetic sensing, optical and printing processes.

[0046] Other features and advantages of the present invention will be disclosed in detail in the following detailed description and accompanying drawings. Attached Figure Description

[0047] The invention will be further described below with reference to the accompanying drawings:

[0048] Figure 1 This is a schematic diagram of the overall structure of the integrated intelligent device for pig carcass processing in a slaughter line according to Embodiment 1 of the present invention.

[0049] Figure 2This is an internal top view of the integrated intelligent device for pig carcass processing in a slaughter line according to Embodiment 1 of the present invention.

[0050] Figure 3 This is a schematic diagram of the overall structure of the upstream part of the main cabin frame in Embodiment 1 of the present invention.

[0051] Figure 4 This is a schematic diagram of the overall structure downstream of the main cabin frame in Embodiment 1 of the present invention.

[0052] Figure 5 This is a partial structural diagram of the downstream of the main cabin frame in Embodiment 1 of the present invention.

[0053] Figure 6 This is an internal top view of the integrated intelligent device for pig carcass processing in a slaughter line according to Embodiment 2 of the present invention.

[0054] Figure 7 This is a partial structural diagram of the upstream part of the main cabin frame in Embodiment 2 of the present invention.

[0055] The components include: 1. Guide rail, 11. Carrying pole hook, 12. RFID tag, 2. Tag reader, 3. Main cabin frame, 31. Blackout curtain, 4. Carcass guiding and stabilizing assembly, 41. First horizontal straight rod, 42. Second horizontal straight rod, 43. First inclined guide rod, 44. Second inclined guide rod, 45. Third horizontal straight rod, 46. Fourth horizontal straight rod, 47. Third inclined guide rod, 48. Fourth inclined guide rod, 5. Central control module, 6. Non-metallic channel partition, 71. Active lighting assembly, 72. Distributed camera group, 8. Digital inkjet printing module, 81. Longitudinal moving screw, 82. Vertical moving screw, 83. Integrated nozzle, 9. Skin pretreatment module, 91. High-pressure blower, 92. Bamboo joint tube assembly, 921. Scraper. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be explained and described below with reference to the accompanying drawings. However, the following embodiments are only preferred embodiments of the present invention and not all of them. Other embodiments obtained by those skilled in the art based on the embodiments in the implementation methods without creative effort are all within the protection scope of the present invention.

[0057] In the following description, terms such as “inner,” “outer,” “upper,” “lower,” “left,” and “right” are used only to indicate orientation or positional relationship for the convenience of describing the embodiments and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention.

[0058] Example 1:

[0059] Please see the appendix Figure 1 , Figure 1 A schematic diagram of the integrated intelligent device for pig carcass processing in a slaughter line, according to one embodiment of this specification, is shown.

[0060] like Figure 1 As shown, the integrated intelligent device for pig carcass processing in the slaughter line includes a guide rail 1 arranged along the direction of the pig carcass conveyor line. Several carrying hooks 11 are installed on the guide rail 1. The carcass is suspended on the carrying hooks 11 and moves continuously with the guide rail 1. Each carrying hook 11 is equipped with an RFID tag 12 containing carcass information. This integrated intelligent device for pig carcass processing in the slaughter line can include at least a main compartment frame 3, a carcass guiding and stabilizing component 4, a tag reader 2, a central control module 5, a multi-spectral metal detection module, a visual quarantine module, and a digital inkjet printing module 8. The tag reader 2, multi-spectral metal detection module, visual quarantine module, and digital inkjet printing module 8 are all connected to the central control module 5. The main compartment frame 3 is designed as a semi-enclosed structure covering the length of the carcass. The guide rail 1 passes through the interior of the main compartment frame 3. The carcass guiding and stabilizing component 4 guides the carcass to perform posture calibration and enter the main compartment frame 3 along the guide rail 1.

[0061] The spreader hook 11 is the lifting device that supports the torso. The torso length refers to the length of the torso from the front legs to the rear legs when suspended on the spreader hook 11. The supporting legs of the main cabin frame 3 are installed on the ground base.

[0062] It is understandable that this embodiment cleverly integrates core functional modules such as the carcass guiding and stabilizing component 4, the multi-spectral metal detection module, the visual quarantine module, and the digital coding module 8 into a single main cabin frame 3 through an innovative system architecture, which greatly saves space and significantly improves the overall production cycle of the pig slaughtering and processing process.

[0063] In this embodiment, before entering the main cabin frame 3, the carcass first passes through the carcass guiding and stabilizing component 4. The function of the carcass guiding and stabilizing component 4 is to physically guide and confine the swaying carcass that moves continuously with the guide rail 1 within a parallel track, so that the carcass suspended on the carrying pole hook 11 runs stably. At the same time, it calibrates the carcass to an approximately horizontal unfolding posture before entering the main cabin frame 3, providing a stable and uniform field of vision and execution surface for subsequent processes.

[0064] Specifically, in this embodiment, the body guiding and stabilizing assembly 4 includes a first horizontal straight rod 41 and a second horizontal straight rod 42 arranged parallel to the guide rail 1. The first horizontal straight rod 41 and the second horizontal straight rod 42 are located on both sides of the front leg of the body and are parallel in the horizontal direction. The first horizontal straight rod 41 and the second horizontal straight rod 42 are respectively provided with a first inclined guide rod 43 and a second inclined guide rod 44 at the end facing the entrance of the main cabin frame 3. The first inclined guide rod 43 and the second inclined guide rod 44 are arranged in opposite directions so that the distance between the first inclined guide rod 43 and the second inclined guide rod 44 gradually decreases along the forward direction of the body, so that the first inclined guide rod 43 and the second inclined guide rod 44 are combined in a "trumpet mouth" shape.

[0065] In this embodiment, a combination of a trumpet-shaped first inclined guide rod 43 and a second inclined guide rod 44, along with a horizontal straight rod, is used. The inclined guide rods provide progressive convergence guidance, smoothly guiding the swaying carcass onto a predetermined track, avoiding impact; the horizontal straight rod provides a stable running track. This is a flexible "guide-constraint" process, rather than an instantaneous rigid collision, causing no damage to the carcass and offering better tolerance.

[0066] Furthermore, in this embodiment, the body guiding and stabilizing assembly 4 also includes a third horizontal rod 45 and a fourth horizontal rod 46 arranged parallel to the first horizontal rod 41. The third horizontal rod 45 and the fourth horizontal rod 46 are located at the middle of the body and the buttocks of the body, respectively, and are parallel in the vertical direction. The ends of the third horizontal rod 45 and the fourth horizontal rod 46 facing the entrance of the main cabin frame 3 are respectively provided with a third tilting guide rod 47 and a fourth tilting guide rod 48. The third tilting guide rod 47 and the fourth tilting guide rod 48 are both arranged in the same direction as the first tilting guide rod 43. The tilting arc of the third tilting guide rod 47 and the fourth tilting guide rod 48 is greater than the tilting arc of the first tilting guide rod 43.

[0067] This embodiment features multiple sets of guide rods with varying heights and inclination angles (e.g., for the forelegs, midsection, and hips). This enables coordinated guidance and support for multiple key areas along the entire length of the torso. In particular, the guide rods for the midsection and hips have a larger inclination angle, which more effectively expands and guides the rear half of the torso (which is heavy and prone to swaying) to a horizontal position, ensuring the stability of the entire torso, not just the hook points.

[0068] It is understood that the body guiding and stabilizing component 4 in this embodiment, through the carefully calculated geometry and spatial layout of the guide rod (e.g.), utilizes its own kinetic energy and contact with the guide rod during the movement of the body into the main cabin to naturally and forcibly guide and constrain it from a disordered swinging state to a stable, uniform standard posture (i.e., the cross-section is approximately horizontally unfolded) into a standard posture that is conducive to subsequent processing.

[0069] In this embodiment, the tag reader 2 is installed at the entrance of the main cabin frame 3. The height of the tag reader 2 is the same as the height of the RFID tag 12 fixed on the carrying pole hook 11, and it is used to read the RFID tag 12 on the carrying pole hook 11 to obtain the body information.

[0070] The RFID (Radio Frequency Identification) tag is a wireless radio frequency identification tag. The structural design of the tag reader 2 is similar to that of existing technologies, and will not be described in detail in this embodiment.

[0071] Before the carcass enters the main cabin frame 3, the tag reader 2 reads the carcass information stored in the RFID tag 12 fixed at the hanging position of the carrying pole hook 11. The carcass information may include at least: the unique ID and location of the corresponding carcass (to provide stable coordinates for subsequent multi-spectral metal detection, visual quarantine and digital coding), the breeding source information of the carcass, and the weighing data upstream of the slaughtering conveyor line.

[0072] Understandably, the operation of the tag reader 2 is the key starting point for transforming the moving physical body into a data stream carrier. It serves as the unique key for associating, binding, calling, and storing all subsequent detection data (images, signals, results), and also provides a unified, real-time spatiotemporal reference coordinate for all subsequent stages, ensuring from the source that the data throughout the entire process is not misaligned or lost.

[0073] In this embodiment, the central control module 5 is used to uniformly arrange the working sequence of the multi-spectral metal detection module, visual quarantine module and digital inkjet printing module 8 according to the conveyor line speed, starting from the reading of RFID tag 12.

[0074] The working sequence of the central control module 5 can be calculated based on the conveyor line speed and the carcass position information read by RFID. The specific calculation method is a conventional technique for those skilled in the art, and will not be described in detail here. The central control module 5 coordinates the working rhythm of each module according to the conveyor line speed, ensuring that within 3-5 seconds of a single carcass passing through, each module seamlessly connects in the order of "broken needle detection → visual inspection → quality grading decision → inkjet coding execution", forming a complete closed loop of "perception-decision-execution-traceability".

[0075] It is understood that in this embodiment, the entire device uses the uniquely identified RFID tag 12 as the data link and the central control module 5 as the scheduling core to precisely synchronize and close the loop of the three links of perception, decision-making and execution in time and space. Within the physical space of the compactly coupled integrated main cabin frame 3, multiple pig carcass processing procedures are reconstructed into an organically coordinated and efficient production line through the collaborative design of hardware and software.

[0076] While existing X-ray machines can penetrate and detect metal objects, they suffer from spatial layout conflicts, efficiency bottlenecks, and radiation safety issues. Traditional metal detection methods are susceptible to interference from the "product effect" and individual carcass variations, and are insensitive to broken stainless steel needles. On one hand, broken stainless steel needles (such as those made of austenitic stainless steel) are weakly magnetic or even non-magnetic metals with low conductivity, easily overlooked. However, pig carcasses have high moisture content (approximately 60-75%) and high salt content, inherently possessing conductivity that generates the "product effect," making it difficult for traditional metal detection to distinguish between carcass variations and genuine broken needles, leading to false alarms or missed detections. On the other hand, significant individual variations in carcasses—different weights, body fat percentages, and locations—result in varying electrical properties that cause severe baseline drift in interference signals.

[0077] To address this, this embodiment creatively designs a multi-spectral metal detection module based on multi-spectral electromagnetic induction to solve the problem of high-sensitivity, high-reliability, and rapid online detection of extremely small stainless steel broken needles on a specific detection object such as pig carcasses (high moisture, high salt content, and irregular shape), while overcoming the inherent defects of X-ray equipment in terms of space occupation, radiation protection, and cost.

[0078] As attached Figure 2 and attached Figure 3 As shown, in this embodiment, the multi-spectral metal detection module is installed on both sides upstream of the main cabin frame 3, and is used to perform broken needle detection on the carcass using a three-frequency combination strategy.

[0079] Specifically, in this embodiment, the multi-spectral metal detection module includes a non-metallic channel partition 6, a balanced coil group, a multi-spectral transmitter, and a signal processor. The non-metallic channel partition 6 is installed on both sides upstream of the main cabin frame 3 to form a detection channel for the carcass to pass through. The balanced coil group (not shown in the figures) includes one transmitting coil and two receiving coils. The transmitting coil is connected to the multi-spectral transmitter, and the two receiving coils are connected to the signal processor. Both the multi-spectral transmitter and the signal processor are connected to the central control module 5 to receive the current carcass ID from the tag reader 2 and the working timing sequence arranged by the central control module 5. The multi-spectral metal detection module is used to perform the following steps:

[0080] S41, when the working sequence arranged by the central control module 5 is reached, the multi-spectral transmitter drives the transmitting coil to synchronously transmit a composite electromagnetic wave with three characteristic frequencies: low frequency for suppressing product effects, medium frequency as the main detection frequency band for foreign objects, and high frequency for enhancing the identification of weak foreign object signals.

[0081] S42, the signal processor acquires the composite output signal of the two receiving coils and decouples the composite output signal to separate the low-frequency component, the intermediate-frequency component and the high-frequency component;

[0082] S43, the signal processor uses a pre-trained benchmark model based on the low-frequency component to perform dynamic background compensation on the intermediate-frequency and high-frequency components, and determines the broken needle detection result based on the compensated intermediate-frequency and high-frequency components.

[0083] Furthermore, the multi-spectral metal detection module also includes an audible and visual alarm, which is installed on the top outer side of the main cabin frame 3. The audible and visual alarm is connected to the signal processor. When the broken needle detection result indicates the presence of a broken needle or foreign object, the signal processor transmits the broken needle detection result to the central control module 5 to record the corresponding body ID and triggers the audible and visual alarm to prompt manual re-inspection.

[0084] The non-metallic channel partition 6 is made of high-density plastic, nylon, and other non-metallic materials. It is installed on the main cabin frame via connectors, forming a relatively stable detection channel and constituting a "Faraday cage." This actively isolates the environmental electromagnetic noise from external frequency converters, motors, wireless equipment, and other production line components, preventing external electromagnetic waves from coupling into the receiving coil. This is a prerequisite for ensuring the accurate extraction of subsequent weak signals. Furthermore, the balance coil assembly frame is made of non-metallic materials (such as polytetrafluoroethylene or high-strength engineering plastics) to avoid introducing electromagnetic interference from the balance coil assembly frame itself.

[0085] The balanced coil assembly structure is a current sensing technology. Balanced coil technology in metal detectors achieves metal detection through electromagnetic induction. Its core structure consists of a transmitting coil and two receiving coils. The transmitting coil generates the initial electromagnetic field, and the transmitting and receiving coils are distributed horizontally or in a specific geometric relationship on both sides of the detection channel through which the body passes, to sense changes in the electromagnetic field. The transmitting coil is connected to a multi-spectral transmitter. For example, the multi-spectral transmitter can use a high-precision crystal oscillator or DDS (Direct Digital Synthesis) technology to generate the excitation signal, with a frequency stability better than ±50ppm to eliminate the influence of frequency drift on phase detection accuracy. When a high-frequency alternating current is applied to the transmitting coil, a high-frequency alternating magnetic field is generated around it. The two receiving coils are symmetrically distributed equidistant from the transmitting coil and connected to the input terminals of the signal processor. When no metal object passes through the detection channel, the alternating magnetic field generated by the transmitting coil induces electromotive forces of equal magnitude and opposite direction in the two receiving coils. These two electromotive forces theoretically cancel each other out or maintain a stable baseline, and the output signal of the signal processor is zero or in a balanced state. When a metal object enters the detection channel, the alternating magnetic field induces eddy currents inside the object. According to Lenz's law, the magnetic field generated by the eddy currents opposes the change in the original magnetic field, thus altering its distribution. This change causes the induced electromotive forces in the two receiving coils to become unequal, resulting in a voltage difference at the input of the signal processor, which outputs a level signal related to the size and properties of the metal object.

[0086] Furthermore, this embodiment can also pre-analyze the magnetic field distribution of the transmitting coil in the detection channel through finite element simulation, analyze the key parameters affecting the differential induced electromotive force under equilibrium conditions, thereby optimizing the relative positions of the equilibrium coil group and the body transport, and making the long axis of the broken needle cut the magnetic field lines perpendicularly when passing through. According to Faraday's law of electromagnetic induction, the rate of change of the metal cutting the magnetic field lines is the largest at this time, and the generated disturbance signal is the strongest, thereby ensuring that the detection sensitivity is in the optimal state.

[0087] The implementation method of the multi-spectral metal detection module in this example is as follows:

[0088] First, the multi-spectral metal detection module in this embodiment targets metallic foreign objects such as broken stainless steel needles, and employs the following three-frequency combination strategy:

[0089] 1. Low-frequency (e.g., 60-80kHz) signals have longer wavelengths and better penetration, reflecting the macroscopic electrical properties (permeability, conductivity) of the carcass (high moisture, high salt content). This frequency band signal is used as a background reference and the basis for dynamic compensation to suppress product effects, penetrate carcass tissue, and overcome electrical signal interference caused by the high moisture and high salt content of the pig carcass itself.

[0090] 2. The intermediate frequency (e.g., 150-200kHz) is used as the main detection frequency band for weakly magnetic metallic foreign objects such as broken stainless steel needles. In this intermediate frequency band, the eddy current effect of broken stainless steel needles is significant, while the response of the body tissue is relatively weakened, thereby maximizing the signal-to-noise ratio.

[0091] 3. High frequencies (such as 400-600kHz) are more sensitive to the edge and tip effects of tiny objects, which can enhance the ability to identify weak foreign object signals such as stainless steel, especially when the broken needle is not completely perpendicular to the magnetic field lines, thus improving the detection sensitivity of extremely fine broken needles.

[0092] Secondly, the composite output signal is decoupled to separate the low-frequency component, the mid-frequency component, and the high-frequency component;

[0093] Finally, by pre-training the baseline model of a normal carcass through "product learning", and then predicting the product effect interference through the low-frequency component, and performing dynamic background compensation on the mid-frequency and high-frequency components to filter the product effect interference of the carcass itself, the product effect interference of the carcass is significantly improved in the subsequent differentiation of the carcass product effect interference signal and the stainless steel broken needle foreign object signal.

[0094] Understandably, in this embodiment, the multi-spectral metal detection module based on electromagnetic induction disturbance replaces the existing broken needle detection scheme that uses a bulky and radiation-risk X-ray machine for transmission imaging. X-rays identify foreign objects through material attenuation imaging, while the multi-spectral metal detection module detects them by disturbing the alternating electromagnetic field with metals. This directly solves the problems of spatial antagonism and radiation safety. Furthermore, considering that stainless steel broken needles are the main metallic foreign objects on pig carcasses in slaughter lines, and are weakly magnetic or even non-magnetic metals with low conductivity, and that pork carcasses have high water and salt content, inherently possessing a certain conductivity, which can generate "product effect" interference, this embodiment creatively employs a three-frequency combination strategy for broken needle detection on carcasses. This achieves high sensitivity, high reliability, and rapid online detection of small stainless steel broken needles under the interference of the carcass product effect, eliminating the need for further detection in subsequent cutting stages. Specifically, existing technologies struggle to detect weak metallic foreign object characteristic signals amidst extremely strong biological background noise. When the background and foreign object signals overlap in the frequency domain, those skilled in the art might employ general filtering. However, while suppressing the background, general filtering inevitably damages the useful foreign object signal. Therefore, this embodiment specifically utilizes low-frequency features for prediction, then predicts the product effect interference signal, i.e., the "background component," from the original mid / high-frequency signal. This achieves "source separation," theoretically enabling better active suppression of interference signals, preserving foreign object signal components different from the background features, and thus better protecting weak foreign object signals. Especially in the mid and high-frequency bands, foreign object characteristics (such as phase abrupt changes) are more completely preserved, significantly improving the detection sensitivity for small-sized and weakly magnetic metallic foreign objects.

[0095] Furthermore, in this embodiment, the signal processor performs the following steps during the pre-training of the baseline model:

[0096] S4311 collects the composite output signals Va and Vb of two receiving coils (RX1 and RX2) when several groups of bodies without metal foreign objects pass through the detection channel at a constant speed.

[0097] Among them, the carcasses without metal foreign objects are a sufficient number of fresh carcasses of different weights and fatnesses, without metal foreign objects.

[0098] S4312 decouples each set of composite output signals Va and Vb, separating the corresponding low-frequency component, mid-frequency component and high-frequency component.

[0099] The operation of decoupling the composite output signal to separate the low-frequency, mid-frequency and high-frequency components is similar to the decoupling operation in the prior art, and is usually achieved by Fast Fourier Transform (FFT) or digital filter bank. This embodiment will not elaborate on this.

[0100] S4313 calculates the phase and differential signal of the low-frequency, mid-frequency and high-frequency components of each composite output signal to obtain several sets of sample data.

[0101] The phase calculation for low-frequency, mid-frequency, and high-frequency components refers to independently calculating the phase offset relative to the excitation signal of the transmitting coil for each frequency band component of a single receiving coil. The differential signal calculation for low-frequency, mid-frequency, and high-frequency components refers to the signal difference between the two receiving coils. In the absence of metallic foreign objects, this differential signal value should be stable near a baseline.

[0102] Understandably, in this embodiment, phase is a key parameter for distinguishing different materials (biological tissue vs. metal). The phase change caused by metal (especially weakly magnetic austenitic stainless steel) differs significantly from the phase change caused by high-moisture tissue. Another dimension of differential signal processing involves subtracting (e.g., Va-Vb) the complete signals or components of the same frequency band from the two receiving coils in the time or frequency domain. This is primarily used to suppress common-mode interference and amplify local magnetic field asymmetry caused by metallic foreign objects. These two methods are parallel and complementary feature extraction techniques.

[0103] S4314 uses sample data to pre-train the benchmark model, resulting in a benchmark model with low-frequency components as input and the mid-frequency and high-frequency components of the predicted carcass without metal foreign objects as input and the background phase and background differential signal.

[0104] The pre-trained baseline model can be a lookup table, such as fast matching based on "nearest neighbor search in feature space", or a linear formula, such as direct prediction based on "regression model". This embodiment does not limit the specific model.

[0105] Understandably, in existing technologies, due to significant individual differences in carcasses—different weights, fat percentages, and water content—the electrical characteristics vary, leading to severe baseline drift in interference signals. Fixed filtering thresholds cannot perfectly adapt to all individuals, resulting in incomplete compensation (high residual noise) or overcompensation (signal attenuation). Therefore, the core idea of ​​this embodiment is to address the complex background interference and carcass-specific signal baseline variations. During the offline pre-training phase, the benchmark model does not record the characteristics of each frequency band in isolation, but rather focuses on learning a stable mapping relationship between "low-frequency features" and "mid- and high-frequency features." By learning from a large number of samples, it intrinsically incorporates the range of interference signal variations in normal carcasses. This benchmark model encapsulates the typical characteristics of the "product effect" of pig carcasses. During online detection, the multi-spectral metal detection module measures the current low-frequency characteristics and uses this learned mapping relationship to predict the most likely mid-frequency and high-frequency background characteristics corresponding to the current carcass. This allows for dynamic and adaptive background subtraction within the current mixed mid-frequency and high-frequency signal; that is, dynamic background compensation is based on real-time matching of the current low-frequency signal, rather than a fixed value. Therefore, it can automatically adapt to different carcasses without requiring manual sensitivity adjustments for each pig, ensuring consistent and stable detection, providing a reliable signal baseline for subsequent foreign object signal detection, and significantly reducing the false alarm rate.

[0106] Specifically, in S43, when the signal processor performs dynamic background compensation for the intermediate frequency and high frequency components based on the low-frequency components using a pre-trained benchmark model, it performs the following steps:

[0107] S4321, calculate the phase and differential signal of the low-frequency component S_low_now of the current composite output signal, input it into the pre-trained benchmark model, and output the background phase and background differential signal of the predicted mid-frequency component B_mid_pred and high-frequency component B_high_pred.

[0108] S4322, calculate the actual phase of the intermediate frequency component S_mid_now and the high frequency component S_high_now of the current composite output signal, as well as the actual differential signal;

[0109] S4323 calculates the difference between the actual phase and the actual differential signal of the intermediate frequency (IF) and high frequency (HF) components of the current composite output signal and the predicted background phase and background differential signal of the IF and HF components, to obtain the phase and differential signal of the compensated IF component S_mid_clean and HF component S_high_clean.

[0110] For example, taking a benchmark model based on fast matching using "nearest neighbor search in feature space" as an example, the implementation of the pre-trained benchmark model in S4321 is as follows: When a new carcass passes through the detection channel, the signal processor extracts the phase and differential signals of its low-frequency components in real time, forming a query vector F_query_low=[Φ_low_now, D_low_now]. Then, in the pre-trained benchmark model, algorithms such as Euclidean distance or Mahalanobis distance are used to quickly search for the K historical samples that are closest to the query vector F_query_low. Then, the phase and difference signals of the intermediate frequency (IF) and high frequency (HF) components corresponding to the K nearest neighbor samples are averaged (or weighted averaged, with weights inversely proportional to distance) to obtain the predicted background phase and background difference signal of the IF component (B_mid_pred = mean (Φ_mid_neighbors, D_mid_neighbors)) and the predicted background phase and background difference signal of the HF component (B_high_pred = mean (Φ_high_neighbors, D_high_neighbors)). These predicted B_mid_pred and B_high_pred are the most likely IF and HF background features of the current carcass inferred by the baseline model. Subsequent dynamic background compensation involves comparing and eliminating the actually measured IF and HF features with these predicted IF and HF background features.

[0111] For example, taking a baseline model based on direct prediction using a "regression model" as an example, the pre-trained baseline model is implemented as follows: In the pre-training stage, the low-frequency features [Φ_low_i, D_low_i] of each sample data are used as input variables X, and the mid-frequency features [Φ_mid_i, D_mid_i] and high-frequency features [Φ_high_i, D_high_i] of the same sample data are used as output variables (Y_mid, Y_high), respectively. A regression model f_mid (such as linear regression, support vector regression SVR, or a very lightweight neural network, etc., which is not limited in this embodiment) is trained using all sample data (X_i, Y_mid_i), such that f_mid(X) ≈ Y_mid. Similarly, another regression model f_high is trained to predict high-frequency features. Then, in the online prediction stage, when a new carcass passes through the detection channel, the signal processor extracts the phase and differential signals of its low-frequency components in real time, that is, the low-frequency features X_now=[Φ_low_now, D_low_now] of the current carcass. Then, X_now is directly input into the pre-trained regression model to obtain the predicted background phase and background difference signal of the mid-frequency component (B_mid_pred=f_mid(X_now)) and the predicted background phase and background difference signal of the high-frequency component (B_high_pred=f_high(X_now)).

[0112] After steps S4321-S4323, the interference from the body itself in S_mid_clean and S_high_clean has been basically filtered out. If there are still foreign object signals, it is likely caused by metal foreign objects such as broken stainless steel needles.

[0113] In existing technologies, the signal characteristics of broken stainless steel needles are weak. For example, austenitic stainless steel is weakly magnetic and has a weak eddy current effect, which is easily ignored. To address this, this embodiment cleverly utilizes the inherent correlation of the electromagnetic response of the carcass itself, transforming "background suppression" from a general filtering problem into a "personalized" signal prediction and subtraction problem. It creatively employs multi-frequency (mid-frequency main detection, high-frequency auxiliary) and multi-feature (phase, differential) fusion analysis. Phase features are sensitive to metals, differential signals can suppress common-mode interference, and multi-feature fusion improves the reliability of the decision. Through dynamic compensation algorithms, interference signals from the carcass itself are filtered out, significantly improving the detection sensitivity of small-sized and weakly magnetic metallic foreign objects on pig carcasses, accurately distinguishing product signals from broken stainless steel needle signals, and greatly reducing the risk of missed / false detections.

[0114] Specifically, in S43, the broken needle detection result is determined based on the compensated intermediate frequency component and high frequency component. This can be achieved as follows:

[0115] S4331, Establish a foreign object feature database: Through experiments, standard foreign object samples of known size and material (such as broken stainless steel needles) are embedded in specific parts of a simulated or real carcass, and the samples are passed through a detection channel to collect relevant signals. The abnormal feature patterns that distinguish the signals (especially the phase and differential signals in the mid- and high-frequency bands) from the reference model when foreign objects are present are analyzed and extracted, such as phase abrupt changes at specific frequencies, peak values ​​of differential signals, and energy of differential signals, thus forming a foreign object feature database.

[0116] S4332, Feature Extraction: Quantize the phase and differential signals of the compensated mid-frequency component S_mid_clean and high-frequency component S_high_clean.

[0117] 1. Phase abrupt change value extraction: Set a short-time analysis window that matches the physical size of the broken needle and the conveying speed (e.g., the time corresponding to the carcass moving 2-3 cm). Within the window, calculate the first-order difference (i.e., the phase change between adjacent sampling points) of the phase sequences S_mid_clean and S_high_clean, and find the maximum absolute value. This yields two eigenvalues: ΔΦ_mid_max (maximum phase abrupt change in the mid-frequency range) and ΔΦ_high_max (maximum phase abrupt change in the high-frequency range). When a metallic foreign object passes through a magnetic field, it causes a rapid and drastic phase jump; this value is used as a key indicator.

[0118] 2. Differential Signal Feature Extraction: Calculate specific waveform parameters of the compensated intermediate frequency (IF) and high frequency (HF) differential signals within a short time window, such as kurtosis (describing the sharpness of the waveform; the signal is sharper when metal passes through). Calculate the energy (i.e., the sum of squares of the amplitudes at sampling points) of the compensated IF and HF differential signals within the short time window. If the residual differential signal energy is significant after background compensation, it indicates the possible presence of an asymmetric disturbance source (such as metal).

[0119] 3. The extracted feature values ​​(phase change value, differential signal kurtosis, differential signal energy) are combined into a feature vector to represent the state of the medium and high frequency signals at the current detection time.

[0120] S4333, Feature Comparison and Fusion Decision: The extracted feature vector is compared with the stainless steel broken needle features stored in the foreign object feature database. A decision algorithm (e.g., comparing each individual feature in the feature vector with a preset individual threshold, calculating the Euclidean or Mahalanobis distance between the current feature vector and the center of typical broken needle features in the foreign object feature database, using a lightweight classifier, etc.) is used for fusion analysis. If the matching degree between the fusion analysis result and the foreign object feature database exceeds the set confidence level, it is determined that "a broken needle foreign object has been detected." The broken needle detection result is bound to the unique ID of the current carcass and uploaded to the central control module 5 in real time, triggering subsequent processes (such as audible and visual prompts, traffic diversion). If the matching degree does not exceed the set confidence level, it is determined that "no broken needle foreign object has been detected," and the data is uploaded to the central control module 5 in real time for recording, and the process continues.

[0121] It is understandable that this embodiment abandons the simple decision-making method of relying on a single signal amplitude threshold in traditional metal detection. Instead, it constructs a space based on multi-dimensional features of "phase change - differential peak - signal energy" and uses pattern recognition methods to perform comprehensive matching and decision-making in this space, thereby achieving highly specific and reliable identification of the specific target of stainless steel broken needles.

[0122] As attached Figure 2 and attached Figure 4 As shown, the visual quarantine module is installed downstream of the main cabin frame 3. It is used to collect carcass images and identify quarantine results and fat thickness based on the carcass images and carcass information.

[0123] The identification of quarantine results is based on existing technology. By acquiring image data of head and hoof detection points, red viscera detection points, white viscera detection points, and other key quarantine points associated with the carcass ID, the embedded existing AI algorithm performs fusion analysis to identify suspected disease symptoms and key grading indicators such as backfat thickness. This embodiment will not elaborate further on this aspect.

[0124] Understandably, existing technologies struggle to meet the high-contrast, feature-enhanced imaging requirements of carcasses in high-speed, low-light environments. When a carcass moves continuously on an assembly line, and the camera uses conventional continuous white light illumination and exposure, the result is likely to be either a clear image but a slow acquisition speed creating a bottleneck, or fast acquisition but insufficient image quality to support fine grading. For example, during the exposure time, the carcass may shift relative to the camera, leading to image blur (motion blur). In low-light environments, preventing motion blur with a very short exposure time is inevitably limited by the brightness of the continuous light source, making it difficult to achieve the desired effect.

[0125] Therefore, this embodiment constructs a visual quarantine module on a high-speed, continuously operating slaughter line, which features "controllable lighting environment, image strobe-linked shooting, and enhanced characteristic spectrum," to replace traditional manual visual inspection or existing continuous white light photography quarantine. This provides a high-quality image data foundation for the identification of carcass diseases and key quality indicators (fat thickness).

[0126] Specifically, in this embodiment, a light-blocking curtain 31 is provided in the middle of the main cabin frame 3. The light-blocking curtain 31 and the main cabin frame 3 enclose a dark box-like downstream space to serve as an image acquisition channel. The visual inspection module includes an active lighting component 71 and a distributed camera group 72. The active lighting component 71 provides lighting around the body, and both the distributed camera group 72 and the active lighting component 71 are connected to the central control module 5. When the working sequence arranged by the central control module 5 is reached, the active lighting component 71 triggers flashes of different wavelengths in a preset order, and sets the flash time to a very short range. At the same time, the exposure time of the distributed camera group 72 is set to match the flash time, realizing the stroboscopic linkage shooting of the active lighting component 71 and the distributed camera group 72 to obtain the body image.

[0127] Understandably, the lighting in existing slaughterhouses is complex and variable, with severe shadows and water stain reflections, resulting in low contrast and high noise in the acquired images, directly and seriously affecting the accuracy of automatic identification. Therefore, this embodiment proactively optimizes the imaging environment rather than adapting to harsh conditions. By installing a light-shielding curtain 31 in the middle of the main cabin frame 3, forming a semi-enclosed light-shielding imaging channel with the main cabin frame 3, external interference light is isolated to a certain extent.

[0128] For example, as shown in the attached document Figure 2 and attached Figure 4 As shown, the active lighting component 71 surrounding the torso can include a four-LED array installed at the four corners of the imaging area, forming an illumination environment that approximates an integrating sphere. Light illuminates the torso from multiple angles, filling in the shadows and making the illumination on the torso surface (especially complex curved surfaces) uniform and consistent, thus avoiding, to some extent, the loss or misjudgment of local features caused by shadows.

[0129] In this embodiment, the flash duration (pulse width) of the LED array is controlled to an extremely short time (microseconds), and the camera's exposure time is set to perfectly match or slightly longer than the flash duration. During this extremely short flash, the movement distance of the subject is negligible (e.g., at a speed of 0.5 m / s and a flash duration of 100 microseconds, the movement distance is only 0.05 mm), thus "freezing" it in the image and obtaining a clear image without motion blur. Furthermore, under pulse drive, the LED array can emit peak light intensity several times or even tens of times higher than its rated continuous power in an extremely short time. This allows the sensor to receive sufficient photons within the microsecond-level exposure time, thereby generating a bright, low-noise image. Simultaneously, due to the extremely short exposure time, the amount of ambient stray light entering the sensor is negligible, further highlighting the subject under pulsed light illumination and achieving extremely high background contrast.

[0130] The distributed camera group 72 refers to cameras arranged in an array, such as 1*2 in the vertical direction, mounted on the main cabin frame facing the carcass cross-section, for acquiring high-definition images of the carcass cross-section. A 10-20% overlap area can be set between the fields of view of each camera in the distributed camera group 72, so that the overall field of view covers the cross-sectional area of ​​the carcass.

[0131] Preferably, in this embodiment, the flash wavelength triggered by the active illumination component 71 may include narrowband light sources such as white light (380-780nm), blue-violet light (440-450nm), and green light (550-560nm). This embodiment utilizes the selective interaction between narrowband light and biological tissue to actively "stain" or enhance target features (for example, hemoglobin has strong absorption of blue-violet light, so under blue-violet light irradiation, congested and hemorrhagic lesions will appear darker in contrast; fat and connective tissue have different green light reflection characteristics, which helps to clearly present the backfat layering interface), thereby directly improving the contrast between diseased tissue and normal tissue at the image level. By sequentially triggering flashes of different wavelengths and stroboscopic linkage for shooting, it is equivalent to taking multiple "feature-enhanced" thematic photos of the same carcass, providing raw image data with higher discriminative power than ordinary RGB images for subsequent identification.

[0132] Understandably, this embodiment abandons the inefficient, heat-accumulating, and motion-blurring continuous illumination methods found in existing technologies. At the physical level, it solves the problem of image clarity for fast-moving objects by precisely synchronizing LED strobe with the camera shutter. Furthermore, by combining light sources of different wavelengths to enhance the contrast of different tissue features (such as lesions, blood vessels, and fat), it improves visual detection performance, thereby significantly enhancing the accuracy of subsequent quarantine result identification.

[0133] Specifically, in this embodiment, the central control module 5 is also used to combine carcass information, carcass images, quarantine results and fat thickness into multi-source real-time information, and to fuse the multi-source real-time information to determine the quality grading conclusion of the carcass in real time.

[0134] It is understood that the central control module 5 in this embodiment incorporates an existing algorithm model for quality grading. The difference lies in that the quality grading in this embodiment is not based on historical breeding information or limited indicators lagging behind on lengthy production lines, but rather on multi-source real-time information with virtually no delay, based on the hardware solution of the shortest processing line provided in this embodiment. The embedded quality grading algorithm model fuses and calculates this multi-dimensional, heterogeneous real-time data, instantly outputting the quality grading result, achieving real-time and accurate grading on high-speed production lines with almost no decision-making delay.

[0135] Specifically, in this embodiment, the central control module 5 is also used to generate a quarantine inspection stamp pattern and a traceability QR code according to regulations based on the quarantine results and quality grading conclusions of the carcass.

[0136] In this embodiment, the inkjet-printed content of the traceability QR code is directly linked to the entire process data of the ID carcass, achieving "one pig, one code; one code for traceability." The specific operation of generating the inkjet-printed information is similar to existing technologies, and will not be described in detail in this embodiment. The difference from existing technologies lies in the fact that, in this embodiment, the inkjet-printed information is a direct output of the real-time decision-making results from the central control module 5, based on the hardware solution of the shortest process line provided in this embodiment. The inkjet printing action of the digital inkjet module 8 is the natural end of the decision-making pipeline, forming a seamless real-time closed loop for detecting the identifier.

[0137] Among them, the central control module 5 is connected to a remote cloud server, and data can be uploaded, stored, processed and analyzed.

[0138] As attached Figure 2 and attached Figure 5 As shown, the digital inkjet printing module 8 is installed downstream of the main cabin frame 3 and is used to print quarantine inspection stamp patterns and traceability QR codes on the carcass skin that has passed the quality grading conclusion.

[0139] Among them, the digital inkjet printing module 8 is located on the other side of the distributed camera group 72. This layout realizes the spatial separation and sequential connection of detection and marking, and shares the light-shielding and stable environment downstream of the main cabin, avoiding mutual interference.

[0140] Specifically, in this embodiment, the digital inkjet printing module 8 includes a printhead telescopic assembly and an integrated printhead 83. The printhead telescopic assembly consists of a longitudinal moving screw 81 and a vertical moving screw 82. The integrated printhead 83 is mounted on the longitudinal moving screw to achieve telescopic movement toward or away from the carcass skin. The longitudinal moving screw 81, the vertical moving screw 82, and the integrated printhead 83 are all connected to the central control module 5. When the working sequence arranged by the central control module 5 is reached, the printhead telescopic assembly drives the integrated printhead 83 to perform high-speed printing of quarantine inspection stamp patterns and traceability QR codes on the carcass skin with a quality grading conclusion of qualified.

[0141] The vertical moving screw 82 is perpendicular to the ground base and is installed on the main cabin frame on the back side of the torso via a connector. Due to individual differences in the torso (length, build), the optimal coding area (such as the buttocks or specific positions on the back) varies in the vertical direction. The vertical screw can drive the printhead to rise and fall to accurately position it to the preset coding start position of the current torso. The longitudinal moving screw 81 is connected to the vertical moving screw 82 to form a cross-shaped (T-shaped) structure. The integrated printhead 83 is fixedly installed on the longitudinal moving screw, realizing the longitudinal extension and vertical movement of the integrated printhead 83. When not in operation, the integrated printhead 83 retracts into the side wall of the frame to reduce the width of the main cabin frame 3 and further reduce the size of the device. The integrated printhead 83 is a prior art technology, using a high-resolution on-demand inkjet (DOD) printhead, such as a piezoelectric printhead, using food-grade edible ink. Its droplet size and frequency can be precisely programmed to meet the printing needs of high-definition patterns and QR codes. The integrated printhead 83 is interconnected with the central control module 5. It performs digital inkjet printing based on the dynamic information such as the quarantine inspection stamp pattern and traceability QR code transmitted by the central control module 5. The longitudinal moving screw 81 and the vertical moving screw 82 are driven by servo motors or stepper motors under the unified control of the central control module 5. The inkjet logic and data of the integrated printhead 83 (including the graphic information of the quarantine inspection stamp and traceability QR code) are also sent down in real time by the central control module 5.

[0142] It is understandable that the digital inkjet printing device in this embodiment, through its ingenious two-dimensional kinematic design, deeply embedded compact layout, and highly coordinated synchronous control logic with the central control module 5, successfully overcomes the technical challenge of achieving high-quality online inkjet printing on a high-speed, continuous pig slaughtering line with significant individual differences. The digital inkjet printing device in this embodiment is not only an marking tool, but also a key actuator that "materializes" all the results of preceding intelligent inspections onto each carcass. It represents the ultimate embodiment of integrated equipment achieving a fully automated and data-driven closed loop of the "inspection-grading-marking" process.

[0143] On the other hand, the implementation process of the integrated intelligent device for the pig carcass processing step in the slaughter line in this embodiment is as follows:

[0144] S1, the carcass moves continuously with the guide rail 1 and first enters the carcass guiding and stabilizing component 4, which calibrates the carcass into a stable posture with an approximately horizontally unfolded cross-section;

[0145] S2, before the carcass enters the main cabin frame 3, the tag reader 2 reads the RFID tag 12 on the carrying pole hook 11 to obtain the carcass information and transmits it to the central control module 5, and the carcass moves continuously into the main cabin frame 3;

[0146] S3, the central control module 5 starts with the reading of RFID tag 12 and arranges the working sequence of multi-spectral metal detection module, visual quarantine module and digital inkjet module 8 according to the conveyor line speed;

[0147] S4, when the working sequence arranged by the central control module 5 is reached, the carcass passes through the detection channel of the multi-spectral metal detection module. The multi-spectral metal detection module uses a three-frequency combination strategy to detect broken needles on the carcass.

[0148] S5, when the working sequence arranged by the central control module 5 is reached, the carcass continues to enter the image acquisition channel along the guide rail 1. The active supplementary lighting component 71 in the visual quarantine module triggers flashes of different wavelengths in a preset order and sets the flash time to be within a very short range. At the same time, the exposure time of the distributed camera group 72 is set to match the flash time, so as to realize the stroboscopic linkage shooting of the active supplementary lighting component 71 and the distributed camera group 72, obtain the carcass image, and then obtain the quarantine result and fat thickness based on the carcass image and carcass information recognition.

[0149] S6, the central control module 5 combines carcass information, carcass images, quarantine results and fat thickness into multi-source real-time information and fuses the multi-source real-time information to obtain a quality grading conclusion;

[0150] S7, the central control module 5 generates the quarantine inspection stamp pattern and traceability QR code corresponding to the carcass ID based on the carcass quarantine results and quality grading conclusions, and transmits them to the digital inkjet module 8;

[0151] S8, when the working sequence arranged by the central control module 5 is reached, the digital inkjet module 8 performs the printing of quarantine inspection stamp patterns and traceability QR codes on the carcass skin with a quality grading conclusion of qualified.

[0152] S9. For a carcass whose quality grading conclusion is unqualified (such as the detection of disease or broken needle metal foreign objects), after leaving the main cabin frame 3, the carcass is diverted to a special processing area for unqualified products according to the carcass ID corresponding to its RFID tag 12 on the subsequent conveyor line.

[0153] S10, all process data of each carcass is uploaded to the central database.

[0154] Understandably, the entire integrated device uses RFID unique identification as the data link and the central control module 5 as the scheduling core. Within the physical space of the integrated main cabin frame 3, multiple processes are cleverly coordinated into an automated production line. The carcass guiding and stabilizing component 4 provides attitude reference for subsequent inspection; the multi-spectral metal detection module and visual quarantine module provide quality and safety data; the quality grading model embedded in the central control module 5 integrates multi-source real-time information to make decisions; and the digital inkjet coding module 8 executes the marking task based on the decision results. The working rhythm of all modules is uniformly coordinated by the central control module 5 according to the conveyor line speed, ensuring seamless connection within the few seconds of a single carcass passing through, forming a complete closed loop of "perception-decision-execution-traceability".

[0155] Example 2:

[0156] This embodiment only applies to comparisons with... Figure 1-5 The differences between the two embodiments will be described in the following descriptions. The technical concepts of the remaining designs are similar to those of the first embodiment, and will not be repeated here.

[0157] As attached Figure 6 and attached Figure 7 As shown, specifically in this embodiment, the integrated device also includes a skin pretreatment module 9 installed at the entrance of the main cabin frame 3. The skin pretreatment module 9 is located before the multi-spectral metal detection module. The skin pretreatment module 9 includes a bamboo tube assembly 92 and a high-pressure blowing pipe 91. The head of the bamboo tube assembly 92 is provided with a scraper 921. The length of the bamboo tube assembly 92 is set so that the scraper 921 contacts the carcass skin after attitude calibration. The working position of the scraper 921 matches the printing height range of the digital inkjet module 8. The high-pressure blowing pipe 91 is connected to the central control module 5. When the working sequence arranged by the central control module 5 is reached, the high-pressure blowing pipe 91 performs high-speed blowing on the carcass skin after attitude calibration.

[0158] The bamboo-joint tube assembly 92 is used to physically scrape the carcass after its posture has been calibrated, as the carcass moves continuously. This mechanical force removes water droplets, blood foam, hair residue, and other dirt from the carcass surface. The support portion of the bamboo-joint tube assembly 92 features a multi-segment adjustable design similar to bamboo joints, allowing the position and angle of the scraper 921 to be finely adjusted in three-dimensional space to adapt to the surface contours of carcasses of different sizes. This ensures that the scraper 921 maintains the most effective contact pressure and angle with the carcass skin, preventing damage from excessive pressure or ineffective cleaning from insufficient pressure. The scraper 921 at the head of the bamboo-joint tube assembly 92 is made of a material that combines wear resistance, a certain degree of elasticity, and food contact safety (such as food-grade silicone or specific engineering plastics), effectively removing adhesive dirt without snagging or tearing the pigskin. The working position of the scraper 921 is matched with the printing height range of the digital coding module 8. This means that the scraper 921 pre-cleans the surface area of ​​the carcass that will be coded later, providing a good foundation for high-quality coding.

[0159] The high-pressure blower 91 may include an air source (air compressor), filter, pressure regulating valve, solenoid valve and end air knife or nozzle array, which is similar to the prior art. This embodiment will not elaborate on this. All control elements (especially solenoid valves) are connected to the central control module 5.

[0160] Understandably, in this embodiment, the scraper 921 solves the problem of "stickiness" (contaminant adhesion), and the high-pressure air blowing solves the problems of "floating" (particles, water film) and "wetness" (drying), forming a complementary process.

[0161] In this embodiment, the skin pretreatment module 9 is executed after the carcass guidance and stabilization component 4 stabilizes the carcass posture in step S3, and before the multi-spectral metal detection module performs broken needle detection in step S4. The skin pretreatment module 9 is used to perform the following steps:

[0162] After the carcass is calibrated by the guiding and stabilizing module, it enters the pre-processing station. The surface of the carcass comes into contact with the scraper 921 of the bamboo tube assembly 92, which is positioned in a pre-set location. The scraper 921 uses mechanical force to scrape away sticky deposits (such as water droplets, blood foam, and hair residue). Almost simultaneously with or immediately after the scraping, when the carcass reaches the precise position arranged by the central control module 5 according to the RFID timing sequence, the central control module 5 triggers a command, and the high-pressure blowing pipe 91 instantly forms a high-speed, uniform high-pressure airflow, covering the surface of the carcass to be treated. Its functions are twofold: first, to blow away contaminants that have been loosened but not yet detached by the scraper 921; and second, to powerfully disperse and evaporate the free water film and water droplets on the surface of the carcass.

[0163] At this point, the carcass, having completed the "scraping-blowing" co-processing, has achieved a relatively clean and dry state on its key surfaces (especially the area to be marked). The carcass then continues to move along guide rail 1 to undergo a series of subsequent processes, including multi-spectral metal detection, visual quarantine, and digital marking.

[0164] It is understandable that the skin pretreatment module 9 in this embodiment adopts a physical combination of "contact mechanical scraping + non-contact high-pressure air drying" to set up an active cleaning checkpoint in a very short time before the carcass enters the precision inspection area, automatically giving the carcass surface a repeatable, high-quality basic state, thereby ensuring the reliability and accuracy of all subsequent electromagnetic sensing, optical and printing processes.

[0165] The above description is merely a preferred embodiment disclosed in this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of protection involved in this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-disclosed concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this disclosure.

[0166] Furthermore, while the operations are described in a specific order, this should not be construed as requiring these operations to be performed in the specific order shown or in a sequential order. In certain environments, multitasking and parallel processing may be advantageous. Similarly, while several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of this disclosure. Certain features described in the context of individual embodiments may also be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment may also be implemented individually or in any suitable sub-combination in multiple embodiments.

Claims

1. An integrated intelligent device for processing pig carcasses in a slaughter line, comprising a guide rail arranged along the direction of the pig carcass conveyor line, a plurality of carrying hooks arranged on the guide rail, and the carcass suspended on the carrying hooks and moving continuously with the guide rail, characterized in that, Each carrying pole hook is equipped with an RFID tag containing carcass information. The device includes: The main cabin frame is designed as a semi-enclosed structure covering the length of the carcass, with guide rails passing through the inside of the main cabin frame; The carcass guidance and stabilization assembly is used to guide the carcass through attitude calibration and into the main cabin frame along the guide rails; A tag reader, installed at the entrance of the main cabin frame, is used to read RFID tags on the carrying pole hooks to obtain carcass information; The multi-spectral metal detection module is installed on both sides of the upstream of the main cabin frame and is used to detect broken needles in the carcass using a three-frequency combination strategy. The visual quarantine module, installed downstream of the main cabin frame, is used to collect carcass images and identify quarantine results and fat thickness based on the carcass images and carcass information. The central control module is used to uniformly arrange the working sequence of each module according to the speed of the conveyor line, starting from the reading of RFID tags. It is also used to collect carcass information, carcass images, quarantine results and fat thickness into multi-source real-time information and fuse the multi-source real-time information to obtain quality grading conclusions. The central control module is also used to generate quarantine inspection stamp patterns and traceability QR codes based on the carcass quarantine results and quality grading conclusions. The digital inkjet printing module, installed downstream of the main cabin frame, is used to print quarantine inspection stamps and traceability QR codes on the carcass skin that has passed the quality grading conclusion. The label reader, multi-spectral metal detection module, visual quarantine module, and digital inkjet printing module are all connected to the central control module.

2. The integrated intelligent device for pig carcass processing in a slaughter line as described in claim 1, characterized in that, The multi-spectral metal detection module includes a non-metallic channel bulkhead, a balanced coil assembly, a multi-spectral transmitter, and a signal processor. The non-metallic channel bulkhead is installed on both sides upstream of the main cabin frame to form a detection channel for the body to pass through. The balanced coil assembly includes one transmitting coil and two receiving coils. The transmitting coil is connected to the multi-spectral transmitter, and the two receiving coils are connected to the signal processor. Both the multi-spectral transmitter and the signal processor are connected to the central control module. The multi-spectral metal detection module is used to perform the following steps: When the working sequence arranged by the central control module is reached, the multi-spectral transmitter drives the transmitting coil to synchronously transmit a composite electromagnetic wave with three characteristic frequencies: low frequency for suppressing product effects, medium frequency as the main foreign object detection frequency band, and high frequency for enhancing the identification of weak foreign object signals. The signal processor acquires the composite output signal from the two receiving coils and decouples the composite output signal to separate the low-frequency component, the intermediate-frequency component, and the high-frequency component. The signal processor uses a pre-trained benchmark model based on the low-frequency components to perform dynamic background compensation on the intermediate-frequency and high-frequency components, and determines the broken needle detection result based on the compensated intermediate-frequency and high-frequency components.

3. The integrated intelligent device for pig carcass processing in a slaughter line as described in claim 2, characterized in that, The signal processor is used to perform the following steps during the pre-training of the baseline model: Collect the composite output signal of two receiving coils when several groups of carcasses without metal foreign objects pass through the detection channel at a constant speed; Each composite output signal is decoupled to separate the corresponding low-frequency component, mid-frequency component and high-frequency component; Calculate the phase and differential signal of the low-frequency, mid-frequency and high-frequency components of each composite output signal to obtain several sets of sample data; The baseline model was pre-trained using sample data to obtain a baseline model with low-frequency components as input and the mid-frequency and high-frequency components as input and the background phase and background differential signal of the body without metal foreign objects as output.

4. The integrated intelligent device for pig carcass processing in a slaughter line as described in claim 3, characterized in that, When the signal processor performs dynamic background compensation for mid-frequency and high-frequency components based on a pre-trained benchmark model using low-frequency components, it performs the following steps: Calculate the phase and differential signal of the low-frequency component of the current composite output signal, input them into the pre-trained benchmark model, and output the predicted background phase and background differential signal of the mid-frequency and high-frequency components. Calculate the actual phase of the intermediate frequency component and the high frequency component of the current composite output signal, as well as the actual differential signal; The difference between the actual phase of the intermediate frequency (IF) component and the high frequency (HF) component of the current composite output signal and the actual differential signal and the predicted background phase and background differential signal of the IF component and the high frequency component is calculated to obtain the compensated phase and differential signal of the IF component and the high frequency component.

5. The integrated intelligent device for pig carcass processing in a slaughter line as described in claim 1, characterized in that, A blackout curtain is installed in the middle of the main cabin frame. The blackout curtain and the main cabin frame enclose a dark box-like downstream space to serve as an image acquisition channel. The visual quarantine module includes an active lighting component and a distributed camera group. The active lighting component provides lighting around the body. Both the distributed camera group and the active lighting component are connected to the central control module. The visual quarantine module is used to perform the following steps: When the working sequence arranged by the central control module is reached, the active fill light component triggers flashes of different wavelengths in a preset order and sets the flash time to a very short range. At the same time, the exposure time of the distributed camera group is set to match the flash time, so as to realize the strobe linkage shooting between the active fill light component and the distributed camera group to obtain the body image.

6. The integrated intelligent device for pig carcass processing in a slaughter line as described in claim 5, characterized in that, The active fill light component triggers flash wavelengths including white light, blue-violet light, and green light.

7. The integrated intelligent device for pig carcass processing in a slaughter line as described in claim 1, characterized in that, It also includes a skin pretreatment module installed at the entrance of the main cabin frame. The skin pretreatment module is located before the multi-spectral metal detection module. The skin pretreatment module includes a bamboo tube assembly and a high-pressure blower. The head of the bamboo tube assembly is equipped with a scraper. The length of the bamboo tube assembly is set so that the scraper contacts the carcass skin after attitude calibration. The working position of the scraper matches the printing height range of the digital inkjet module. The high-pressure blower is connected to the central control module. When the working sequence arranged by the central control module is reached, the high-pressure blower performs high-speed blowing on the carcass skin after attitude calibration.

8. The integrated intelligent device for pig carcass processing in a slaughter line as described in claim 5, characterized in that, The digital inkjet printing module is located on the other side of the distributed camera group. The digital inkjet printing module includes a printhead telescopic assembly and an integrated printhead. The printhead telescopic assembly consists of a longitudinal moving screw and a vertical moving screw. The integrated printhead is installed on the longitudinal moving screw to achieve telescopic movement toward or away from the carcass skin. The longitudinal moving screw, the vertical moving screw, and the integrated printhead are all connected to the central control module. When the working sequence arranged by the central control module is reached, the printhead telescopic assembly drives the integrated printhead to perform high-speed printing of quarantine inspection stamp patterns and traceability QR codes on the carcass skin with a quality grading conclusion of qualified.

9. The integrated intelligent device for pig carcass processing in a slaughter line as described in claim 1, characterized in that, The hull guiding and stabilizing assembly includes a first horizontal straight rod and a second horizontal straight rod arranged parallel to the guide rail. The first horizontal straight rod and the second horizontal straight rod are located on both sides of the front leg of the hull and are parallel in the horizontal direction. A first inclined guide rod and a second inclined guide rod are respectively provided at the ends of the first horizontal straight rod and the second horizontal straight rod facing the entrance of the main cabin frame. The first inclined guide rod and the second inclined guide rod are arranged in opposite directions so that the distance between the first inclined guide rod and the second inclined guide rod gradually decreases along the forward direction of the hull.

10. The integrated intelligent device for pig carcass processing in a slaughter line as described in claim 9, characterized in that, The body guiding and stabilizing assembly also includes a third horizontal rod and a fourth horizontal rod arranged parallel to the first horizontal rod. The third horizontal rod and the fourth horizontal rod are located at the middle of the body and the buttocks of the body, respectively, and are parallel in the vertical direction. The ends of the third horizontal rod and the fourth horizontal rod facing the entrance of the main cabin frame are respectively provided with a third tilting guide rod and a fourth tilting guide rod. The third tilting guide rod and the fourth tilting guide rod are both arranged in the same direction as the first tilting guide rod, and the tilting arc of the third tilting guide rod and the fourth tilting guide rod is greater than the tilting arc of the first tilting guide rod.

Images