Tuna visual recognition and dynamic sorting equipment

Abstract

The application discloses the technical field of aquatic product sorting, and relates to a tuna visual identification and dynamic sorting device, which comprises a clamping type conveying mechanism, a suspension type rotary conveying mechanism, a visual detection system, a sorting mechanism and a plurality of clamping jaw assemblies. The clamping type conveying mechanism is used to realize sequential feeding and equal-interval arrangement of tuna, and eliminate image overlapping from the source. The suspension type rotary conveying mechanism is used in cooperation with the clamping jaw assemblies to realize impact-free transfer of tuna between different stations and guarantee the integrity of the skin of the tuna. The visual detection system is used to effectively extract weak local features and adopt a high-precision identification algorithm that is adaptive to frost interference. The organic combination and collaborative work of the plurality of mechanisms systematically solve the technical problems of low identification precision, discontinuous conveying and poor grading efficiency in the automatic grading process of tuna in the prior art, and realize full-process automation, rhythmization and fine operation from species identification to size sorting.

Description

Technical Field

[0001] This invention relates to the field of aquatic product sorting technology, specifically to a visual recognition and dynamic sorting device for tuna. Background Technology

[0002] As the deep-sea fishing processing industry modernizes, the degree of automation in the processing of tuna, a high-value-added aquatic product, has become a key factor determining the product's freshness, quality, and market pricing power. In the current processing system, species identification (such as bluefin, yellowfin, and bigeye) and size grading (based on body length or weight) constitute the core links in the pre-processing steps.

[0003] In the prior art, a tuna weight sorting device with patent publication number CN221059451U mainly includes a fish box, a first conveying mechanism, a second conveying mechanism, and an air blowing component. Its working process is as follows: tuna enters the first conveying mechanism from the fish box, and after surface moisture is removed by the air blowing component, it enters the second conveying mechanism. The second conveying mechanism is equipped with multiple weight sorting components to classify and sort the tuna according to their weight.

[0004] This existing technology can perform grading and sorting of tuna, but it has the following drawbacks:

[0005] First, the sorting dimension is limited, and it can only achieve weight grading, which cannot meet the actual needs of species identification in tuna processing.

[0006] Secondly, the feeding method lacks an orderly separation mechanism, and tuna are prone to stacking and side by side during transportation, affecting the accuracy of sorting;

[0007] Third, there is a lack of visual recognition systems for frozen tuna, which cannot meet the needs of feature recognition under conditions such as fin and tail removal and frost cover.

[0008] Fourth, the sorting mechanism poses a risk of damaging the fish's skin, making it difficult to meet the stringent requirements for skin integrity in raw-eating tuna.

[0009] Based on this, the present invention designs a tuna visual recognition and dynamic sorting device to solve the above problems. Summary of the Invention

[0010] The purpose of this invention is to provide a visual recognition and dynamic sorting device for tuna to solve the problems mentioned in the background art.

[0011] To achieve the above objectives, the present invention provides the following technical solution:

[0012] A visual recognition and dynamic sorting device for tuna includes:

[0013] The clamping conveyor mechanism is located in the middle of the front side of the equipment and is used to receive tuna loaded by loading and unloading workers. It has opposing sliding platforms on both sides for clamping and fixing the fish. The end platform is equipped with photoelectric sensors and lifts the fish mouth part upward. Together with the gripper assembly and photoelectric sensors, it can accurately grasp the gills and complete the orderly feeding and on-demand conveying of tuna.

[0014] The suspended rotary conveyor mechanism is located at the center of the equipment and at the rear of the clamping conveyor mechanism. It is used to realize the continuous transportation of tuna in the whole machine and to complete the on-demand inspection task in conjunction with the vision inspection system, so as to transport different varieties of tuna to the sorting mechanism.

[0015] Multiple gripper assemblies are provided and connected to the lower part of the suspended rotary conveyor mechanism to achieve stable suspension and transfer by gripping the holes in the gills of the tuna.

[0016] Two visual inspection systems are installed and symmetrically located on both sides of the suspended rotating conveyor mechanism to identify the species of tuna during the conveying process.

[0017] The sorting mechanism consists of multiple units, evenly distributed on both sides of the suspended rotary conveyor, used to sort and transport different varieties of tuna after identification.

[0018] Preferably, the clamping conveying mechanism includes a support frame, on which multiple uniformly and parallelly arranged rollers are rotatably connected, and an upwardly inclined discharge port is provided at the rear end. Multiple opposing side sliding platforms are uniformly arranged along the circumferential direction on the top of the support frame, and each side sliding platform includes a partition. A vertical auxiliary wheel and a power wheel are rotatably connected to the inner side of the partition. A first belt is connected to both the auxiliary wheel and the power wheel. A first motor and a controller are installed on the outer side of the partition, and the first motor is connected to the power wheel through a coupling.

[0019] Preferably, the suspended rotary conveying mechanism includes multiple evenly arranged fixed frames and support legs evenly fixed to the bottom of the fixed frames. The top of the multiple fixed frames is fixed to a sliding track. The sliding track is designed as a ring track structure with an I-shaped cross-section. Sprockets and chain assemblies are provided on both sides of the bottom of the sliding track. The sprockets are connected to a second motor. The top of the gripper assembly is slidably connected to the sliding track and correspondingly connected to the chain assemblies.

[0020] Preferably, four sprockets are symmetrically arranged and rotatably connected at both ends of the bottom of the sliding track. Each sprocket has a first chain, and one of the sprockets is connected to a second motor. The sliding track has four rolling wheels, each of which has a second chain. The gripper assembly is connected to the second chain, and two of the rolling wheels are coaxially fixed with their corresponding sprockets.

[0021] Preferably, the gripper assembly includes a horizontal fixed base and a push rod fixed base fixed to the middle of the bottom of the fixed base. The top of the fixed base is rotatably connected to two symmetrical constraint frames, and the top of each constraint frame is rotatably connected to two symmetrical slide rail wheels, which are rolled to the inner sides of the sliding track. Horizontal auxiliary blocks are provided on the constraint frames corresponding to the positions of the two slide rail wheels, and the auxiliary blocks contact the bottom of the sliding track. A horizontal plate is connected to the bottom of the push rod fixed base through a vertical first push rod. Symmetrical claw plates are rotatably connected to both sides of the horizontal plate, and a second push rod is hinged between the upper part of the claw plate and the top of the horizontal plate.

[0022] Preferably, the visual inspection system includes a gantry and a dark box fixed on the gantry. The dark box is equipped with a supplementary light and a visual recognition module. The dark box is composed of a light shield. The visual recognition module includes an image acquisition unit, an image preprocessing unit, a lightweight target detection unit, an attention mechanism enhancement unit, and a visualization interaction unit. The image acquisition unit includes a camera bracket and a monocular camera installed on the top of the inner side of the dark box.

[0023] Preferably, the sorting mechanism includes a mounting profile and rollers rotatably connected to the top two sides of the mounting profile, and a second belt is provided on both rollers. The end of the mounting profile near the suspended rotary conveyor is provided with an inclined guide channel, and the higher end of the guide channel is located below the suspended rotary conveyor, and the lower end is located at the end of the second belt. A third motor and a right-angle reducer are mounted on the mounting profile, and the right-angle reducer is connected to one end of one of the rollers.

[0024] Compared with the prior art, the beneficial effects of the present invention are:

[0025] 1. This invention achieves sequential feeding and evenly spaced arrangement of tuna through a clamping conveyor mechanism, ensuring that each tuna enters the suspended rotating transport area independently, eliminating image overlap from the source, thereby significantly improving the accuracy and stability of species identification.

[0026] 2. The present invention uses a combination of a suspended rotary conveyor mechanism and a gripper assembly to achieve impact-free transfer of tuna between different workstations, completely avoiding the collision risk in the traditional belt or roller transfer process and ensuring the integrity of the fish's skin.

[0027] 3. This invention is equipped with a visual inspection system, which can effectively extract weak local features and adapt to high-precision recognition algorithms that are not affected by frost, filling the gap in the dedicated dataset for frozen tuna;

[0028] 4. The visual inspection system of the present invention adopts a detection model that integrates attention mechanism and lightweight backbone network, which reduces computing power requirements while ensuring recognition accuracy, and meets the requirements of high-speed continuous operation of factory assembly lines.

[0029] In summary, this invention systematically solves the technical problems of low identification accuracy, discontinuous conveying, and poor grading efficiency in the automated grading process of tuna by organically combining and coordinating multiple mechanisms, and realizes fully automated, rhythmic, and precise operation from species identification to size sorting. Attached Figure Description

[0030] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0031] Figure 1 This is a schematic diagram of the overall structure of the present invention;

[0032] Figure 2 This is a schematic diagram of the clamping and conveying mechanism of the present invention;

[0033] Figure 3 for Figure 2 Schematic diagram of the structure at point A in the middle;

[0034] Figure 4 This is a schematic diagram of the suspended rotary conveying mechanism of the present invention;

[0035] Figure 5 for Figure 4 Schematic diagram of the structure at point B;

[0036] Figure 6 This is a schematic diagram of the side structure of the gripper assembly of the present invention;

[0037] Figure 7 This is a schematic diagram of the front structure of the gripper assembly of the present invention;

[0038] Figure 8 This is a schematic diagram of the structure of the visual inspection system of the present invention;

[0039] Figure 9 for Figure 8 Schematic diagram of the structure at point C;

[0040] Figure 10 This is a schematic diagram of the sorting mechanism of the present invention;

[0041] Figure 11 This is a schematic diagram of image preprocessing according to the present invention;

[0042] Figure 12 This is a diagram of the MobileNetV3 network structure of the present invention;

[0043] Figure 13This is a schematic diagram of the CBAM network structure of the present invention.

[0044] The attached diagram lists the components represented by each number as follows:

[0045] 100-Clamping conveyor mechanism, 101-Discharge port, 102-Support frame, 103-First motor, 104-Controller, 105-Baffle plate, 106-Roller, 107-First belt, 108-Auxiliary wheel, 109-Coupling, 110-Power wheel;

[0046] 200-Suspended rotary conveyor mechanism, 201-Support leg, 202-First chain, 203-Tension wheel, 204-Fixed frame, 205-Sliding rail, 206-Pulley, 207-Rolling wheel, 208-Sprocket, 209-Second motor, 210-Limiting frame, 211-Second chain, 212-Bearing seat, 213-Bearing seat fixing frame;

[0047] 300-Visual inspection system, 301-Monocular camera, 302-Camera bracket, 303-Dark box, 304-Gantry, 305-Light shield;

[0048] 400-Sorting mechanism, 401-Third motor, 402-Right angle reducer, 403-Guideway, 404-Roller, 405-Second belt, 406-Mounting profile;

[0049] 500-Gripper assembly, 501-Slide wheel, 502-Constraint frame, 503-Fixed seat, 504-First push rod, 505-Grip plate, 506-Second push rod, 507-Push rod fixed seat, 508-Pin, 509-Auxiliary block. Detailed Implementation

[0050] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] Please refer to the accompanying drawings. This invention provides a technical solution:

[0052] A tuna visual recognition and dynamic sorting device includes a clamping conveyor mechanism 100, a suspended rotary conveyor mechanism 200, a visual inspection system 300, a sorting mechanism 400, and a gripper assembly 500.

[0053] The clamping conveyor 100 is located in the middle of the front side of the equipment and is used to receive tuna loaded by loading and unloading workers. It has opposing sliding platforms on both sides for clamping and fixing the fish. The end platform is equipped with a photoelectric sensor and lifts the fish mouth part upward. Together with the gripper assembly 500 and the photoelectric sensor, it can accurately grasp the gills of the fish and complete the orderly feeding and on-demand conveying of tuna.

[0054] The suspended rotary conveyor 200 is located at the center of the equipment and at the rear end of the clamping conveyor 100. It is used to realize the continuous transportation of tuna in the whole machine and to work with the vision inspection system 300 to complete the on-demand inspection task, and to transport different varieties of tuna to the sorting mechanism 400.

[0055] Multiple gripper assemblies 500 are provided and connected below the suspended rotary conveyor mechanism 200, used to achieve stable suspension and transfer by gripping the holes in the gills of the tuna.

[0056] Two visual inspection systems 300 are provided and are symmetrically located on both sides of the suspended rotary conveyor mechanism 200, for identifying the species of tuna during the conveying process.

[0057] Multiple sorting mechanisms 400 are provided and evenly distributed on the front and rear sides of the suspended rotary conveyor mechanism 200, for sorting and conveying different varieties of tuna after identification.

[0058] Specifically, the clamping conveying mechanism 100 includes a support frame 102, on which a plurality of uniformly and parallelly arranged rollers 106 are rotatably connected, and a discharge port 101 with an upward tilting and lifting structure is provided at the rear end. The top of the support frame 102 is provided with a plurality of opposing side sliding platforms evenly arranged along the circumferential direction, and the side sliding platform includes a partition 105. A vertical auxiliary wheel 108 and a power wheel 110 are rotatably connected to the inner side of the partition 105. A first belt 107 is connected to both the auxiliary wheel 108 and the power wheel 110. A first motor 103 and a controller 104 are installed on the outer side of the partition 105, and the first motor 103 is connected to the power wheel 110 through a coupling 109.

[0059] The mechanism drives the coupling 109 via a motor, which in turn drives the power wheel 110. The power wheel 110 and the auxiliary wheel 108 drive the first belt 107 to rotate, achieving dynamic friction on both sides. It works in conjunction with the roller 106 to form a clamping and conveying mechanism for the tuna.

[0060] The clamping conveyor mechanism 100 has a main frame made of 304 stainless steel welded together. The platform is 1.5 meters long and the clamping plates on both sides are covered with 3mm thick food-grade wear-resistant rubber (Shore hardness A60). It is driven by two Panasonic MINAS A6 series servo motors (400W each) in conjunction with planetary reducers and ball screws. The clamping force is fed back in real time by a dynamic torque sensor and precisely controlled by a Siemens S7-1200 PLC. It can automatically adjust the range between 50N and 200N according to the width of the fish. The end lifting mechanism is driven by an SMC cylinder with a lifting angle of 15°. It is equipped with Keyence laser sensors and Omron photoelectric sensors to achieve precise positioning of the fish body and gills.

[0061] Specifically, the suspended rotary conveyor mechanism 200 includes multiple evenly arranged fixed frames 204 and support legs 201 evenly fixed to the bottom of the fixed frames 204. The top of the multiple fixed frames 204 is fixed together with a sliding rail 205. The sliding rail 205 is designed as a ring track structure with an I-shaped cross section. The bottom sides of the sliding rail 205 are provided with sprockets 208 and chain groups. The sprockets 208 are connected to a second motor 209. The top of the gripper assembly 500 is slidably connected to the sliding rail 205 and correspondingly connected to the chain group.

[0062] The mechanism uses a second motor 209 as a power source, which drives the sprocket 208 to rotate. The sprocket 208 further drives the chain assembly to run, thereby realizing the drive control of the gripper assembly 500. During the movement of the gripper assembly 500, the sliding rail 205 set above it plays an auxiliary guiding and supporting role through rolling cooperation, ensuring the stability of the gripper assembly 500 operation.

[0063] In addition, the overall structure also realizes the rotation function of the conveyor platform during operation through the sliding cooperation between the slide wheel 501 of the gripper assembly 500 and the chain assembly, thereby meeting the process requirements for tuna to adjust direction or change posture in a suspended state.

[0064] Among them, the sliding track 205 of the suspended rotary conveyor mechanism 200 adopts a ring track structure with a track circumference of 8 meters. Twelve independent gripper assemblies 500 are evenly distributed. The drive motor is a 400W servo motor with a 17-bit absolute encoder to achieve high-precision positioning and speed control (positioning accuracy ±0.5mm, speed range 0-0.5m / s). Each gripper assembly 500 can independently achieve ±180° rotation to adjust the tuna detection posture.

[0065] Specifically, four sprockets 208 are symmetrically arranged and rotatably connected at both ends of the bottom of the sliding track 205. A first chain 202 is provided on the sprockets 208, and one of the sprockets 208 is connected to the second motor 209. Four rolling wheels 207 are provided on the sliding track 205, and a second chain 211 is provided on the four rolling wheels 207. The gripper assembly 500 is connected to the second chain 211, and two of the rolling wheels 207 are coaxially fixed with the corresponding sprockets 208 to realize the movement of the chain assembly.

[0066] Specifically, the gripper assembly 500 includes a horizontal fixed base 503 and a push rod fixed base 507 fixed to the middle of the bottom of the fixed base 503. The top of the fixed base 503 is rotatably connected to two symmetrical constraint frames 502, and the top of each constraint frame 502 is rotatably connected to two symmetrical slide rail wheels 501, which are rolled to the inner sides of the sliding track 205. A horizontal auxiliary block 509 is provided on the constraint frame 502 corresponding to the position of the two slide rail wheels 501, and the auxiliary block 509 contacts the bottom of the sliding track 205. The bottom of the push rod fixed base 507 is connected to a horizontal plate through a vertical first push rod 504. Symmetrical claw plates 505 are rotatably connected to both sides of the horizontal plate, and a second push rod 506 is hinged between the upper part of the claw plate 505 and the top of the horizontal plate.

[0067] The constraint frame 502 in this component is used to limit the range of motion of the moving parts, prevent deviation or shaking, and ensure the stability and reliability of clamping. The slide wheel 501 rolls in contact with the sliding rail 205 and works with the auxiliary block 509 to guide and limit, ensuring that the gripper assembly 500 runs smoothly when moving and rotating. The fixed seat 503 is used to connect the end of the gripper assembly 500 to the front end, and plays a role in bearing and positioning. The opening and closing of the gripper assembly 500 is completed by the second push rods 506 on both sides to achieve flexible clamping and release. The overall structure is compact and the action is coordinated, which can meet the clamping and posture adjustment requirements of tuna in the process of suspended conveying.

[0068] Specifically, the visual inspection system 300 includes a gantry 304 and a dark box 303 fixed on the gantry 304. The dark box 303 is equipped with a supplementary light and a visual recognition module. The dark box 303 is composed of a light shield 305. The visual recognition module includes an image acquisition unit, an image preprocessing unit, a lightweight target detection unit, an attention mechanism enhancement unit, and a visualization interaction unit. The image acquisition unit includes a camera bracket 302 and a monocular camera 301 installed on the top inner side of the dark box 303.

[0069] The vision inspection system 300 is equipped with a 12-megapixel Basler industrial camera with a 12mm fixed-focus lens. The internal dimensions of the dark box are 800mm×600mm×500mm, and the inner wall is coated with a matte black light-absorbing material to eliminate stray light interference. The ring LED fill light has a color temperature of 5500K and an illuminance of 1000 Lux and is driven by a constant current controller to ensure lighting stability. On the software side, it is deployed on an NVIDIA Jetson TX2 embedded development board, running a deep learning model based on TensorRT optimization. The self-developed "FishVision 1.0" software framework integrates an image preprocessing module (saturation adjustment, histogram equalization, gamma correction and improved multi-scale Retinex enhancement algorithm), a MobileNetV3-CBAM lightweight object detection model (approximately 3.2M parameters, inference time ≤15ms), and a visualization interaction unit developed using the Qt framework.

[0070] When the tuna is transported to the position of the visual inspection system 300, a limit signal is triggered, and the suspended rotating conveyor 200 stops moving. At this time, the visual recognition module installed on the gantry 304, under the stable lighting environment formed by the combination of the dark box 303 and the supplementary light, performs image acquisition and recognition of the tuna species, and sends the recognition results to the subsequent sorting mechanism 400 to realize the automatic identification and classification of tuna species. Among them, the dark box 303 is used to isolate the interference of external ambient light, and the supplementary light provides stable lighting conditions to ensure that the acquired frozen tuna images are uniformly illuminated and have clear details.

[0071] The image acquisition unit in this facility acquires image information through a monocular camera 301. The image preprocessing unit is connected to the image acquisition unit and is responsible for enhancing the acquired raw images. Specifically, this includes saturation adjustment, histogram equalization, gamma correction, and improved multi-scale image enhancement algorithms, thereby improving the recognizability of local features of tuna under frost cover.

[0072] The lightweight target detection unit is connected to the image preprocessing unit and uses a lightweight backbone network based on MobileNetV3 as the feature extraction network. This network introduces depthwise separable convolution and inverse residual structures, which effectively reduces the number of model parameters and computational complexity while ensuring detection accuracy, thus meeting the real-time detection requirements of industrial production lines.

[0073] The attention mechanism enhancement unit is embedded in the lightweight object detection unit, and the original SENet module is replaced by CBAM (Convolutional Block Attention Module). CBAM is composed of channel attention mechanism and spatial attention mechanism connected in series. It can enhance effective features in the channel dimension and locate key regions in the spatial dimension, thereby improving the accuracy of fine-grained classification of tuna under finless and tailless conditions and frost cover.

[0074] The visual interactive unit connects to the lightweight target detection unit. Developed based on the Qt framework, it provides functions such as image uploading, real-time capture detection, video stream detection, detection result display, and data statistics. The interface is user-friendly and easy for factory workers to use daily.

[0075] Specifically, the sorting mechanism 400 includes a mounting profile 406 and rollers 404 rotatably connected to the top two sides of the mounting profile 406. A second belt 405 is provided on both rollers 404. An inclined guide 403 is provided at one end of the mounting profile 406 near the suspended rotary conveyor 200. The higher end of the guide 403 is located below the suspended rotary conveyor 200, and the lower end is located at the end of the second belt 405. A third motor 401 and a right-angle reducer 402 are mounted on the mounting profile 406, and the right-angle reducer 402 is connected to one end of one of the rollers 404.

[0076] When the mechanism is in operation, the suspended rotary conveyor 200 transports the tuna to the corresponding guide rail 403 according to the identification result, and moves the tuna along the guide rail 403 to the second belt 405. At the same time, the third motor 401 drives the main shaft, and the main shaft transmits power to the rollers 404 through the right angle reducer 402. The two rollers 404 and the second belt 405 work together to realize the sorting and conveying function of different varieties of tuna, ensuring the smoothness and stability of the sorting process.

[0077] The sorting mechanism 400 is equipped with multiple independent sorting stations, corresponding to yellowfin tuna, bluefin tuna and other / non-conforming categories. Each station is equipped with an independent transverse conveyor unit driven by a 200W servo motor, with a belt width of 300mm and a conveying speed of 0.5m / s. The power is transmitted to the roller group through a right-angle reducer to form a transverse drive circuit. It has no mechanical connection with the main conveyor belt, avoiding cumulative errors and jamming problems.

[0078] When the device of the present invention is in operation, it includes the following steps:

[0079] Step 1: Orderly feeding and clamping conveyor of tuna

[0080] 1. Manual feeding:

[0081] The operator places the tuna to be processed (usually with the fins and tail removed) at the starting end of the clamping conveyor 100.

[0082] 2. Clamping and positioning:

[0083] The control system of the clamping conveyor mechanism 100 is activated, and the first motor 103 starts working. The first motor 103 drives the power wheel 110 and the first belt 107 to rotate through the coupling 109, thereby cooperating with the roller 106 through dynamic friction to apply a uniform clamping force from both sides of the tuna, realizing automatic centering and fixation of the fish.

[0084] 3. Posture adjustment and material feeding:

[0085] The tuna is conveyed forward to the discharge port 101 at the end of the platform while being gripped. The end platform is raised, causing the mouth of the fish to be lifted up and exposing the gill area. After the photoelectric sensor detects that the fish has reached the predetermined gripping position, it sends a signal, and the gripping conveyor mechanism 100 stops moving, waiting for the gripper assembly 500 to pick it up.

[0086] Step 2: Suspended Transfer and Rotary Transport of Tuna

[0087] 1. Gill gripping: The suspended rotary conveyor 200 drives the gripper assembly 500 to the gripping station at the end of the clamping conveyor 100. The first push rod 504 and the second push rod 506 of the gripper assembly 500 work together to drive the claw plate 505 to accurately insert into the hole in the gill part of the tuna and tighten it to achieve stable suspension of the fish.

[0088] 2. Suspended transport and attitude change:

[0089] The second motor 209 of the suspended rotary conveyor 200 drives the sprocket 208 to rotate, which in turn drives the chain assembly, thereby driving the gripper assembly 500 to move along the preset sliding track 205. During transport, the slide wheel 501 above the gripper assembly 500 rolls in cooperation with the sliding track 205, providing guidance and support to ensure stable operation. Through the sliding cooperation between the slide wheel 501 and the chain assembly, the suspended rotary conveyor 200 can control the gripper assembly 500 to rotate ±90° or ±180° during movement, according to the needs of subsequent inspection and sorting, to adjust the posture of the tuna so that its belly or back faces the inspection station.

[0090] Step 3: Visual Recognition and Variety Classification of Tuna

[0091] 1. Transportation Trigger:

[0092] The suspended rotary conveyor 200 transports the suspended tuna to the designated station of the vision inspection system 300; when the tuna triggers the limit signal, the control system issues a command to stop the operation of the suspended rotary conveyor 200.

[0093] 2. Image Acquisition:

[0094] In a stable, interference-free lighting environment provided by the dark box 303 and the internal supplemental light (light intensity constant at 800-1200 Lux, color temperature at 5000-6500 K), the monocular camera 301 acquires images of the tuna.

[0095] The acquired images are 24-bit true-color RGB images with a resolution of no less than 1920×1080 pixels.

[0096] 3. Image Preprocessing: The image preprocessing unit performs enhancement processing on the acquired raw image in the following order to eliminate the effects of frost coverage:

[0097] Saturation adjustment: Increase the color saturation of the image by 15%-25%.

[0098] Histogram equalization: This process equalizes the grayscale distribution of an image, enhancing image contrast.

[0099] Gamma correction: Set the gamma value γ to 0.8-1.2 to perform non-linear tone adjustment and correct image brightness.

[0100] Improved multi-scale image enhancement algorithm: An improved algorithm based on Retinex theory is adopted to perform dynamic range compression and edge enhancement on images at multiple scales (such as small scale 15, medium scale 80, and large scale 250) to highlight the texture and color features of the tuna body surface.

[0101] 4. Lightweight target detection and classification:

[0102] The preprocessed image is input into a lightweight object detection unit. The core of this unit is a lightweight backbone network based on MobileNetV3, with embedded CBAM (Convolutional Block Attention Module).

[0103] Feature extraction: The MobileNetV3 network utilizes depthwise separable convolution and inverse residual structures to extract features from the input image while maintaining detection accuracy.

[0104] Attention Enhancement: The embedded CBAM module first uses a channel attention mechanism to weight each feature channel, enhancing key features (such as fish body color and texture); then, it uses a spatial attention mechanism to locate key discrimination regions on the fish body (such as the middle of the fish body and the back). The computation flow of this module is as follows: the input feature map \(F\) passes through channel attention \(M_c(F)\) and spatial attention \(M_s(F')\) in sequence, and the output enhanced feature map \(F'' = M_s(M_c(F) \otimes F)\otimes (M_c(F) \otimes F)\).

[0105] Classification output: After the above processing, the network outputs the classification results of tuna species, such as "yellowfin tuna", "bluefin tuna", "bigeye tuna", etc.

[0106] 5. Result Interaction and Sending:

[0107] The identification results are displayed in real time on a visual interactive unit interface developed based on the Qt framework, and are automatically entered into the database for statistical analysis. Simultaneously, the vision inspection system 300 uses the tuna species identification result as a sorting signal, sending it to the downstream sorting mechanism 400 control system via industrial Ethernet or a serial communication interface.

[0108] Step 4: Multi-level dynamic sorting of tuna

[0109] 1. Sorting Trigger: After receiving the recognition signal from the vision inspection system 300, the control system of the sorting mechanism 400 begins to prepare to execute the sorting action.

[0110] 2. Lateral conveying and sorting execution:

[0111] When the suspended rotary conveyor 200 transports the identified tuna to the designated sorting station for the corresponding species, the control system issues a command, and the second push rod 506 of the gripper assembly 500 actuates, releasing the gripper plate 505 and allowing the tuna to fall onto the sorting mechanism 400. The third motor 401 at this station immediately starts, driving the roller 404 to rotate via the right-angle reducer 402. The roller 404 works in conjunction with the second belt 405 to quickly and smoothly transport the falling tuna along a direction perpendicular to the main conveyor direction to the corresponding collection container.

[0112] 3. Reset and Cycle: After the sorting action is completed, the sorting mechanism 400 of the station stops running and waits for the next trigger. The suspended rotary conveyor 200 continues to run, transporting the next tuna to be sorted to the next station or repeating the above cycle until all tuna are sorted.

[0113] The present invention provides a visual recognition and multi-level dynamic sorting device and method for tuna, which, through the organic collaboration of various mechanisms and algorithmic innovation, achieves the following significant beneficial effects:

[0114] 1. It achieves non-destructive and orderly feeding of tuna, solving the problems of visual overlap and individual damage at the source.

[0115] To address the shortcomings mentioned in the background art, such as "sorting operations easily causing damage" and "disorderly feeding leading to overlapping visual recognition," this invention achieves a fundamental improvement through a clamping conveyor mechanism 100. This platform employs a side-sliding platform arranged opposite to each other on both sides. A first motor 103 drives a coupling 109 to drive a pulley assembly, which, in conjunction with the roller 106, applies a uniform dynamic friction clamping force to the tuna.

[0116] Compared to the rigid "fish pushers" in existing technologies, this flexible clamping method completely avoids the rigid impact between the push rod and the surface of the fish. At the same time, the lifting structure at the end of the platform, in conjunction with photoelectric sensors, enables precise positioning and grasping of the fish gills, completing the sequential feeding of tuna from a haphazardly piled state to one with equal spacing and uniform posture.

[0117] Experiments show that this method reduces the surface indentation rate of tuna from about 5%-8% in existing technologies to below 0.1%, and ensures that each tuna enters the subsequent detection area independently. This eliminates the misjudgment problem caused by image overlap from a physical structure perspective, laying the foundation for subsequent high-precision identification.

[0118] 2. It achieved shock-free suspended transport and multi-posture adjustment of tuna, significantly reducing transport losses and improving detection flexibility.

[0119] To address the drawbacks of existing technologies, such as low efficiency and transport losses due to layered structures, this invention eliminates the complex transition buffer mechanism and adopts a combination of a suspended rotary conveyor mechanism 200 and a gripper assembly 500. The gripper assembly 500 uses a push rod to drive the grippers to precisely insert into the gill openings for suspension, achieving shock-free transfer of tuna between different workstations and completely avoiding the collision risks associated with traditional belt or roller transport processes.

[0120] This design allows the visual inspection system 300 to obtain the optimal inspection surface by adjusting the fish's posture (such as making its abdomen or back face the camera) without having to place cameras at multiple angles.

[0121] Actual testing showed that this suspended transfer structure reduces the number of times tuna is transferred in the whole machine from 3-5 times in existing technologies to 1 time, reduces the risk of fish collision damage by more than 80%, and improves the adaptability of visual detection by more than 40%.

[0122] 3. Dedicated vision hardware and enhancement algorithms adapted to freezing conditions were developed, significantly improving recognition accuracy under blurred feature states.

[0123] To address the problems of insufficient adaptability to freezing conditions and poor hardware environment adaptability in existing technologies, this invention makes systematic innovations in both hardware and algorithms. On the hardware side, the visual inspection system 300 integrates a darkroom 303 and a constant-illuminance supplementary light (illuminance constant at 800-1200 Lux, color temperature 5000-6500K), effectively isolating it from interference from complex external light in low-temperature, high-humidity, and high-salt environments, and eliminating the impact of frost reflections on image quality.

[0124] In terms of algorithms, the image preprocessing unit adopts a combination of saturation adjustment, histogram equalization, gamma correction and improved multi-scale image enhancement algorithms (small scale 15, medium scale 80, large scale 250), specifically for the restoration of images covered by frost.

[0125] Experimental data show that, in the scenario of identifying frozen tuna (with a 0.5-3mm frost layer on the surface), the image acquisition system constructed in this invention improves the image signal-to-noise ratio (SNR) by more than 12dB and the feature point recognition rate by more than 60% compared with ordinary cameras in open environments.

[0126] 4. It achieves lightweight detection with high precision and low computing power, meeting the high-speed real-time operation requirements of industrial production lines.

[0127] To address the issues of "high computational cost of high-precision models" and "lack of physical constraints in hierarchical methods" in existing technologies, this invention designs a lightweight object detection unit that integrates an attention mechanism. This unit uses MobileNetV3 as its backbone network and, through depthwise separable convolutions and inverse residual structures, compresses the number of model parameters to less than one-tenth of that of traditional networks such as ResNet.

[0128] Meanwhile, the embedded CBAM (Convolutional Block Attention Module) replaces the original SENet module. Through a dual channel and spatial attention mechanism, it improves the fine-grained classification accuracy of tuna with defined tails and blurred features by about 6.5% with only a 2% increase in the number of parameters. In actual testing, the model achieves an inference time of less than 15 milliseconds per frame on industrial-grade embedded devices (such as NVIDIA Jetson TX2) and maintains a stable recognition accuracy of over 97.5%.

[0129] The recognition results are fed back in real time through the visualization unit of the Qt framework, realizing an end-to-end closed loop from image acquisition to sorting instruction issuance. The overall system processing speed can reach 20-25 tails per minute, which fully meets the high-speed operation requirements of the factory assembly line.

[0130] In summary, this invention, through the organic combination of a clamping conveyor mechanism 100, a suspended rotary conveyor mechanism 200, a vision inspection system 300, a gripper assembly 500, and a sorting mechanism 400, and combined with a dedicated image preprocessing and lightweight target detection algorithm, systematically solves the core problems of low recognition accuracy, discontinuous conveying, poor grading efficiency, and easy damage to fish in the automated sorting of frozen tuna in existing technologies. It realizes fully automated, rhythmic, and precise operation from species identification to size sorting, achieving significant technological progress and industrial application value.

[0131] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0132] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A visual recognition and dynamic sorting device for tuna, characterized in that, include: The clamping conveyor (100) is located in the middle of the front side of the equipment and is used to receive tuna loaded by loading and unloading workers. It has opposing side sliding platforms on both sides for clamping and fixing the fish body. The end platform is equipped with a photoelectric sensor and lifts the fish mouth part upward. Together with the gripper assembly (500) and the photoelectric sensor, it can accurately grasp the gills of the fish and complete the orderly feeding and on-demand conveying of tuna. The suspended rotary conveyor (200) is located at the center of the equipment and at the rear end of the clamping conveyor (100). It is used to realize the continuous transportation of tuna in the whole machine and to cooperate with the vision inspection system (300) to complete the on-demand inspection task, and to transport different varieties of tuna to the sorting mechanism (400). Multiple gripper assemblies (500) are provided and connected to the lower part of the suspended rotary conveyor mechanism (200) for stable suspension and transfer by gripping the holes in the gills of the tuna. Two visual inspection systems (300) are provided and symmetrically located on both sides of the suspended rotary conveyor (200) for identifying the species of tuna during the conveying process. The sorting mechanism (400) is provided in multiple units and is evenly distributed on the front and rear sides of the suspended rotary conveyor mechanism (200) for sorting and conveying different varieties of tuna after identification.

2. The tuna visual recognition and dynamic sorting equipment according to claim 1, characterized in that, The clamping conveying mechanism (100) includes a support frame (102), on which multiple uniformly and parallelly arranged rollers (106) are rotatably connected, and an upwardly inclined discharge port (101) is provided at the rear end. Multiple opposing side sliding platforms are uniformly arranged along the circumferential direction on the top of the support frame (102), and each side sliding platform includes a partition (105). A vertical auxiliary wheel (108) and a power wheel (110) are rotatably connected to the inner side of the partition (105). A first belt (107) is connected to both the auxiliary wheel (108) and the power wheel (110). A first motor (103) and a controller (104) are installed on the outer side of the partition (105), and the first motor (103) is connected to the power wheel (110) through a coupling (109).

3. The tuna visual recognition and dynamic sorting equipment according to claim 1, characterized in that, The suspended rotary conveying mechanism (200) includes multiple uniformly arranged fixed frames (204) and support legs (201) uniformly fixed to the bottom of the fixed frames (204). The top of the multiple fixed frames (204) is fixed together with a sliding rail (205). The sliding rail (205) is generally set as a rounded rectangle with an I-shaped cross section. The bottom sides of the sliding rail (205) are provided with sprockets (208) and chain groups. The sprockets (208) are connected to a second motor (209). The top of the gripper assembly (500) is slidably connected to the sliding rail (205) and correspondingly connected to the chain group.

4. The tuna visual recognition and dynamic sorting equipment according to claim 3, characterized in that, The bottom ends of the sliding track (205) are symmetrically arranged and rotatably connected with four sprockets (208). The sprockets (208) are provided with a first chain (202), and one of the sprockets (208) is connected to a second motor (209). The sliding track (205) is provided with four rolling wheels (207), and the four rolling wheels (207) are provided with a second chain (211). The gripper assembly (500) is connected to the second chain (211), and two of the rolling wheels (207) are coaxially fixed with the corresponding sprockets (208).

5. The tuna visual recognition and dynamic sorting equipment according to claim 3, characterized in that, The gripper assembly (500) includes a horizontal fixed base (503) and a push rod fixed base (507) fixed to the middle of the bottom of the fixed base (503). The top of the fixed base (503) is rotatably connected to two symmetrical constraint frames (502), and the top of each constraint frame (502) is rotatably connected to two symmetrical slide rail wheels (501), which are rolled to the inner sides of the sliding track (205). The constraint frame (502) is provided with a horizontal auxiliary block (509) corresponding to the position of the two slide rail wheels (501), and the auxiliary block (509) contacts the bottom of the sliding track (205). The bottom of the push rod fixed base (507) is connected to a horizontal plate through a vertical first push rod (504). Symmetrical claw plates (505) are rotatably connected to both sides of the horizontal plate, and a second push rod (506) is hinged between the upper part of the claw plate (505) and the top of the horizontal plate.

6. The tuna visual recognition and dynamic sorting equipment according to claim 1, characterized in that, The visual inspection system (300) includes a gantry (304) and a dark box (303) fixed on the gantry (304). The dark box (303) is equipped with a supplementary light and a visual recognition module. The dark box (303) is composed of a light shield (305). The visual recognition module includes an image acquisition unit, an image preprocessing unit, a lightweight target detection unit, an attention mechanism enhancement unit, and a visualization interaction unit. The image acquisition unit includes a camera bracket (302) and a monocular camera (301) installed on the top inner side of the dark box (303).

7. The tuna visual recognition and dynamic sorting equipment according to claim 1, characterized in that, The sorting mechanism (400) includes a mounting profile (406) and rollers (404) rotatably connected to the top two sides of the mounting profile (406). A second belt (405) is provided on both rollers (404). An inclined guide channel (403) is provided at one end of the mounting profile (406) near the suspended rotary conveyor (200). The higher end of the guide channel (403) is located below the suspended rotary conveyor (200), and the lower end is located at the end of the second belt (405). A third motor (401) and a right-angle reducer (402) are mounted on the mounting profile (406), and the right-angle reducer (402) is connected to one end of one of the rollers (404).

Images